+ All Categories
Home > Documents > Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 ·...

Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 ·...

Date post: 01-Jun-2020
Category:
Upload: others
View: 2 times
Download: 0 times
Share this document with a friend
158
Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis, Diagnostic Tools and Accommodation A Dissertation Presented By Naiqian Zhi to The Department of Bioengineering in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the field of Bioengineering, Motor control Northeastern University Boston, Massachusetts May 2016
Transcript
Page 1: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Quantitative Assessment of Micrographia

and Tremor in Static Handwriting Samples Analysis, Diagnostic Tools and Accommodation

A Dissertation Presented

By

Naiqian Zhi

to

The Department of Bioengineering

in partial fulfillment of the requirements

for the degree of

Doctor of Philosophy

in the field of

Bioengineering, Motor control

Northeastern University

Boston, Massachusetts

May 2016

Page 2: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

ii

ABSTRACT

The two most common movement disorders, Essential Tremor (ET) and Parkinson’s

disease (PD), affect about 11 million people only in the US. While PD and ET can be

assessed through clinical tests, these tests remain relatively subjective, require expertise

to implement, and may not always render reliable results when initial symptoms are

subtle. An alternative to this is to study handwriting of subjects, since PD and ET

strongly affect handwriting – a precision task. A large pool of static –existing–

handwriting samples provides rich information regarding symptomatic effects of PD and

ET and their progression. However, detection of subtle yet relevant changes in

handwriting as a manifestation of symptomatic progression or therapeutic response in PD

and ET is quite challenging. A computerized toolkit based on quantitative analysis of

static handwriting samples would be valuable as it could be used to supplement and

support clinical assessments, help monitor symptomatic changes, and potentially link this

monitoring to PD or ET. Especially, if it could detect relevant changes in handwriting

morphology, thus enhancing the resolution of detection procedure.

In this dissertation, innovative computerized metrics are presented, by which static

handwriting samples can be analyzed to decide whether studied conditions affect the

samples. Specifically, these metrics are tested and validated in their ability to measure (a)

micrographia effects by comparing normal writing samples with symptomatic ones

Page 3: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

iii

collected from PD patients, and (b) tremors by comparing unaffected writing samples

with those affected by artificially induced tremor on healthy subjects. Results suggest that

both sets of metrics are sensitive enough to detect and discern changes with specificity. A

two-week self-administrated handwriting therapeutic exercise has been performed by PD

subjects with micrographia, to investigate influence of this exercise on handwriting

samples by means of developed quantitative metrics. A home-based micrographia/tremor

monitoring mode using a designed Graphical User Interface (GUI) infrastructure is

presented next, within which developed quantitative metrics are integrated. Such GUI

applications offer strong promise for remotely tracking large at-risk and symptomatic

populations, while allowing it to be expanded upon via existing tools that instead use

digitizing tablets and perform dynamic analysis of hand movements.

Dissertation Advisors: Rifat Sipahi, Andrew Gouldstone, Beverly Kris Jaeger, Charles

A. DiMarzio and Jose Martinez-Lorenzo.

Page 4: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

iv

ACKNOWLEDGEMENTS

My deepest gratitude is to my advisor, Dr. Rifat Sipahi. I have been fortunate to have an

advisor who gave me the freedom to explore the research direction on my own, and at the

same time the guidance and practical advice to recover when my research strategies were

not appropriate. His patience and support helped me overcome many crisis situations and

finish this dissertation. All of these made me know how to do research in future.

My co-advisor, Dr. Andrew Gouldstone, has been always there to listen and give advice.

I am deeply grateful to him for teaching me how to question thoughts and express ideas.

I am also thankful to him for many thought-provoking comments on my views and

helping me understand and focus my ideas.

My co-advisor, Dr. Beverly Kris. Jaeger, for her insightful discussions that helped me

learn and apply, and even teach others how to properly perform statistical analysis. I am

also thankful to her for guiding me on how to create proper experimental protocols for

human subjects testing, and encouraging the use of correct grammar and consistent

notation in my writings and carefully reading and commenting on countless revisions of

the manuscripts we prepared together.

Dr. Samuel Frank, one of co-PIs of the NSF funded project that I worked during my Ph.D

study, brought his professional medical advices and constructive comments at different

Page 5: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

v

stages of my research. I am grateful to him for holding me to a high research standard by

his clinical point of views in neurology.

I also would like to thank the other two members of my PhD committee, Dr. Charles A.

DiMarzio and Dr. Jose Martinez-Lorenzo for their helpful comments and suggestions for

my research in general.

I would like to acknowledge Ms. Beverly Ribaudo for numerous email conversations on

related topics of the therapeutic handwriting exercise that helped me obtain my

knowledge from a Parkinson patient perspective. I am grateful to her for sharing her

handwriting samples on her blog and providing me insights as to how to improve my

experimental protocol.

I also appreciate two undergraduate students, Hye Ryong Sin and Daniel Croitoru,

making their efforts on a side work to support the assessment metrics introduced in this

dissertation.

Most importantly, none of this would have been possible without the participation of

subjects in this research. I would like to acknowledge several local support group

facilitators, their permission to access and recruit the Parkinson’s patients during the

group meetings. I would also like to express my heart-felt gratitude to those subjects for

providing their writing samples.

Finally, I appreciate the financial support from US National Science Foundation (Award

CBET 1133992) that funded parts of the research discussed in this dissertation.

Any opinions in this dissertation are those of the author and do not necessarily represent

the viewpoints of US National Science Foundation.

Page 6: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

vi

TABLE OF CONTENTS

Abstract ii

Acknowledgements iv

List of Figures ix

List of Tables xii

Chapter 1 Introduction 1

1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Overview of Research Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.4 Outline of the Dissertation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Chapter 2 Introduction of Movement Disorders 7

2.1 Background of movement disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2 Literature Reviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2.1 Parkinson’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2.2 Essential Tremor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.2.3 Handwriting Issues for People with Parkinson’s D isease and

Essential Tremor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.2.4 Available Engineering Solutions . . . . . . . . . . . . . . . . . . . . . 14

2.3 Problem Statement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.4 Significance & Innovation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Chapter 3 Development and Evaluation of Metrics for Micrographic

Handwriting 24

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.2 Development of Metrics for Signature/String Analysis . . . . . . . . . . . . . . . . . . 25

3.2.1 Global Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.2.2 Local Metric-Pixel Density Variation Metric . . . . . . . . . . 32

3.2.3 Local Metric-Number of Peaks in Horizontal Projection Profile . . . 33

3.3 Evaluation of Metrics for Signature/String Analysis . . . . . . . . . . . . . . . 36

3.3.1 Subjects and Study Samples . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3.3.2 Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Page 7: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

vii

3.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

3.4 Development of Metrics for Single Character Analysis . . . . . . . . . . . . . . . 67

3.4.1 Skeletal Points Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.4.2 Polar Distribution for Angles and Distances Metric . . . . . . . . . . . . . 70

3.5 Evaluation of Metrics for Single Character Analysis . . . . . . . . . . . 71

3.5.1 Subjects and Study Samples . . . . . . . . . . . . . . . . . . . . . . . . . 71

3.5.2 Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3.5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3.5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

Chapter 4 Development and Evaluation of Metrics for Tremulous Handwriting 78

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

4.2 Development of Metrics for Single Character Analysis . . . . . . . . . . . . . 79

4.2.1 Critical Point Distance Metric . . . . . . . . . . . . . . . . . . . . . . . . . 79

4.2.2 Segmental Curvature Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4.2.3 Horizontal/Vertical Projection Profile Peak Metric . . . . . . . . . . . . 84

4.3 Artificially Induced Tremor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

4.4 Evaluation of Metrics for Single Character Analysis . . . . . . . . . . . . . . 88

4.4.1 Subjects and Study Samples . . . . . . . . . . . . . . . . . . . . . . . . . . 88

4.4.2 Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

4.4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

4.4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

Chapter 5 Quantitative Assessment of a Therapeutic Exercise in Temporally

Mitigating Micrographia Associated with Parkinson’s Disease 94

5.1 Introduction of Amplified Air Writing (AAW) . . . . . . . . . . . . . . 94

5.2 AAW Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

5.3 Quantitative Assessment Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

5.3.1 Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

5.3.2 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

5.3.3 Evaluation Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

5.3.4 Metric Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

5.4 Test Results from PD Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

5.5 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

Chapter 6 Design of Graphic User Interface (GUI) for Monitoring

Micrographic and/or Tremulous Handwriting 114

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

6.2 Infrastructure and Application of GUI for Monitoring Micrographia

Associated Parkinson’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

6.2.1 Proposed Micrographia Monitoring Framework . . . . . . . . . . . . . 115

6.2.2 Application of the GUI for Monitoring Micrographia for a Case

Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

Page 8: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

viii

6.3 Infrastructure and Application of GUI for Identifying Tremulous Writing

Associated Essential Tremor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

6.3.1 Proposed Tremor Identifying Framework . . . . . . . . . . . . . . . . . . . . . 122

6.3.2 Application of the GUI for Identifying Tremulous Writing for a

Case Study . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

6.4 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

Chapter 7 Conclusions and Future Work 130

7.1 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

7.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

Bibliography 133

Page 9: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

ix

LIST OF FIGURES

2.1 Illustration of dopamine losses associated with Parkinson’s disease . . . . . . . 9

2.2 Handwriting samples of PD and ET affected subjects from two individual

studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.3 Kinesia motor assessment system integrates accelerometers and

gyroscopes in a compact patient-worn unit to capture kinematic movement

disorder features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.4 Archimedes spiral drawing test results among normal, PD and ET

subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.5 Traditional patient-clinician evaluation mode vs. New home-based

monitoring and evaluation mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.1 Example of 1st signature image (Sig 11) in (a) within Figure3.8 . . . . . . . . . 28

3.2 Visualization of quantifiable ink deposit metric implemented on 1st

signature image (Sig 11) in (a) within Figure3.8 . . . . . . . . . . . . . . . . . . . . 29

3.3 Original Horizontal/Vertical Projection Profiles of Subject 06: Signature H

(1st signature is asymptomatic) vs. Signature S (5th signature symptomatic) . 30

3.4 Visualization of areas under HPP/VPP curves (blue colored areas) for 1st

signature image (Sig 11) in (a) within Figure 3.8 . . . . . . . . . . . . . . . . 30

3.5 Resized Horizontal/Vertical Projection Profiles Subject 06 (with fixed

aspect ratio): Signature H (1st signature is asymptomatic) vs. Signature S

(5th signature symptomatic) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.6 Visualization of Pixel Density Variation Metric implemented on 5th

signature image (Sig 15) in (a) within Figure3.8 . . . . . . . . . . . . . . . . . . . 33

3.7 Simple geometrical samples illustrate the relationship between the peak

occurrences in HPP profile and the unique and spaced pixel strokes in

samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.8 Historical signature samples of Subjects 01-10 (a)–(j) in chronological

order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.9 Historical signature samples from Subjects 11-12 in chronological order

[panel (a) and (b)] and artificially generated control signature samples

[panel (c)] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.10 Pre-Processing procedures on example of 1st signature image (Sig11):

step (a) – (d) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.11 Metric values for Subjects 01-10 and control samples from Figure 3.9c . . . 44

3.12 Metric values for Subjects 11-12 and control samples from Figure 3.9c . . . 45

3.13 Flowchart showing the selection and generation procedures of samples:

OH, OS, AH, AS for Subjects 01-12 . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

Page 10: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

x

3.14 HPP profile for a sample with Original Healthy (OH) and Original

Symptomatic (OS) of Subject 02, and the creation of Artificially-sized

Symptomatic (AS) and Artificially-sized Healthy (AH) samples as

described in Figure 3.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.15 The NP-HPP metric differences obtained between Data Set 1 and Data Set

3 for Subjects01-10 where PD onset is known . . . . . . . . . . . . . . . . . . . . . 52

3.16 The NP-HPP metric differences obtained between Data Set 1 and Data Set

3 for Subjects 11-12 where PD onset is unknown . . . . . . . . . . . . . . . . . . 53

3.17 NP-HPP metric results for PD Subject 05 (Data set 1) and size-matched

synthetic writing created with Microsoft Word (Data set 4) . . . . . . . . . . . 58

3.18 NP-HPP metric results for synthetic writing data in Microsoft Word using

various fonts and stroke styles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

3.19 NP-HPP metric values for Sample patient handwriting (original data set),

along with width-matched synthetic writing AH (resize of OH), and

Microsoft Word with “best-fit” combination of font style and stroke style

(baseline data set) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.20 Normalized NP-HPP metric values, where each data point in Figure 3.19

is divided by the width of the corresponding sample . . . . . . . . . . . . . . . . . 62

3.21 Standard Cursive Alphabet Table with Marked Skeletal Points (SP) in red

dots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.22 Example of implementing the Skeletal (Critical) Points Distance metric on

the capital letter “K”, from the standard Cursive Alphabet table . . . . . . . . 69

3.23 Example analysis for letter “c” using Polar Distribution for Angles and

Distances (PDAD) metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.1 Pixilation and identification of Critical Points, and Critical Points’ spatial

distributions of pixels with tremor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4.2 Segmental curvature metric applied with a circle . . . . . . . . . . . . . 83

4.3 Flowchart of curvature analysis for a letter affected by artificially induced

tremor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

4.4 Model of EMS machine (ProM-555) and specified placement of four

surface electrodes in the anterior forearm muscle . . . . . . . . . . . . . . . . . . 86

4.5 Wacom Intuos4 tablet and NeuroGlyphics computer software for dynamic

data observation and analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

4.6 Examples of unaffected and AIT affected sample collected from three

individuals of the healthy subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

4.7 Examples of a sample image after thinning processing step . . . . . . . . . . . 89

5.1 Snapshot from a video filmed AAW exercise had been practiced by a PD

patient and handwriting samples before and after AAW . . . . . . . . . . . . . 97

5.2 Typical writing sample from Subjects with mild UPDRS group: Pre-AAW

vs. Post-AAW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

5.3 Typical writing sample from Subjects with Severe UPDRS group: Pre-

AAW vs. Post-AAW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

5.4 Non-improvement percentage of tasks summary for Subjects 1-13: over 14

days . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

Page 11: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

xi

5.5 Non-improvement percentage of tasks summary for Subjects 1-13: day-to-

day at least half the time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

6.1 New home-based micrographia/tremor characterization, monitoring and

evaluation mode using the GUI infrastructure . . . . . . . . . . . . . . . . . . . . . 116

6.2 Snapshot of GUI applications for micrographic handwriting on computer

for one of sample image from Subject 06 . . . . . . . . . . . . . . . . . . . . . . . . 117

6.3 Resulting images shown for original uploaded sample and pre-processed

sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

6.4 Snapshot of numerical results of GUI applications (micrographic

handwriting) for the signature sample case on the computer . . . . . . . . . 120

6.5 Snapshot of visualized results of GUI applications (micrographic

handwriting) for the signature sample case on the computer . . . . . . . . . 121

6.6 Snapshot of GUI applications for tremulous handwriting on computer for

one of sample image from Subject 05 . . . . . . . . . . . . . . . . . . . . . . . 123

6.7 Resulting images shown for original uploaded sample and pre-processed

character sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

6.8 Snapshot of numerical results of GUI applications (tremulous handwriting)

for the character sample case on the computer . . . . . . . . . . . . . . . . . . . . . 125

6.9 Snapshot of numerical results of GUI applications (tremulous handwriting)

for the character sample case on the computer . . . . . . . . . . . . . . . . . . . . . 126

6.10 Computer aided diagnostic support system for the most common

movement disorders characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

Page 12: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

xii

LIST OF TABLES

3.1 Summary of metrics described in Chapter 3 used for monitoring

micrographic handwriting associated PD symptom variations . . . . . . . . . . 26

3.2 Demographic information and clinical characteristics of study

participants from subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3.3 Computed density ratios Rij in signature samples marked in Figure 3.8

with rectangular frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.4 NP-HPP metric results for Subjects 01-12 between different sample groups . 53

3.5 Summary (A): ANOVA single factor for Subjects 01-10 and Summary (B):

Subjects 01-12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.6 Paired t-tests between groups with Bonferroni correction for Subjects 01-10 . 57

3.7 Paired t-tests between groups with Bonferroni correction for Subjects 01-12 . 57

3.8 Metrics results from Subjects 01-12 by implementing skeletal point metric

and polar distribution for angles and distances metric . . . . . . . . . . . . . . . . . . 74

3.9 Paired and independent t-tests results on metric values obtained from

Subject 01-12 by implementing developed single character metrics . . . . . . . . 74

4.1 Nine different critical points have been identified and defined . . . . . . . . . . . . 80

4.2 Averaged metric results for tremulous writing analysis between “EMS-On

“and “EMS-Off” sample groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

4.3 Paired- and independent- tests results for metric values from Subject 01-12

between “EMS-On “and “EMS-Off” sample groups . . . . . . . . . . . . . . . 91

5.1 Demographic and Clinical Characteristics of Study Participants . . . . . . . . . . 99

5.2 Resulting difference values for all subjects from 1 to 13 and metrics

results evaluation over 14 days . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5.3 Resulting difference values for all subjects from 1 to 13 and metrics

results calculation: day-to-day at least half the time . . . . . . . . . . . . . . . . . 108

5.4 Statistical Analysis: p values for metrics used to quantify improvement . . . 111

5.5 Relationship between different tasks and corresponding statistical analysis

results from each metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

Page 13: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 1. Introduction 1

Chapter 1

Introduction

1.1 Motivation

Movement disorders are pervasive, in particular among the people over the age of 50,

which affect specifically the fine motor skills associated with precision grips [1].

Movement disorders are the most common in patients with Essential Tremors (ET),

Parkinson’s disease (PD), and at a certain degree in people with Dystonia, and Ataxia. In

this dissertation, we focus on ET and PD, which impact a large population of about 11

million only in the US [1-3]. ET presents itself in people in the form of uncontrolled

shaking, most notably during muscle usage [4-5], while PD causes four main symptoms:

rest tremor, slowness of movement, stiffness or rigidity of the limbs, and impaired

balance and coordination [6-8]. The presence of above-mentioned motor symptoms for

both conditions can have a range of effects, from embarrassment and discomfort to

severely impaired daily activities including handwriting.

Currently, there are no wildly accepted laboratory tests to determine the existence or

extent of PD and ET [9-10]. The efficacy of any intervention is generally determined via

detection of changes in symptoms [11-12]. The gold standard in this regard is a physical

examination administered by a neurologist, who may score symptomatic severity using

Page 14: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 1. Introduction 2

clinical rating scales [13-16]. While this process is quite reliable, established, and has

been modified over years of experience with clinicians including recent improvements

with MDS-UPDRS [17], it remains relatively subjective and there is certainly potential to

improve its resolution to capture symptomatic responses to therapy, which can be subtle

[11-12]. The requirement of a clinical visit also limits this process.

Analysis of handwriting could improve the understanding of symptom variations [14, 18],

which is a part of typical clinical tests for movement disorders, since issues with legible

handwriting are common problems for those with ET and PD [7-9]. In a typical clinical

setting, micrographia can be subjectively identified from drawings of repeated loops

and/or letters, while tremor is measured through deviations from drawings of simple

geometries, such as Archimedes’ spirals [12, 14, 19-21].

On the basis of the above rationale, using an analytical engineering approach, the

handwriting of people with ET or PD has been studied with data acquisition devices

(tablets) using embedded multi sensors and cameras to characterize kinetics (forces) and

kinematics (movement profiles) of writing in observable and measurable ways [22-30].

Although positive results have been reported, this practice has not been fully developed

and is not in widespread use, since the key methods require specialized devices and

equipment. Also, the methods rather focus on a few specific letters, but not on words,

sentences, or signatures, nor on their key geometric and morphological features where

there is rich information available, and mostly require trained personnel to be involved in

the testing and analyses. This leads to limited usage and reliability in the clinic. Moreover,

many studies focus on collecting samples dynamically with time stamps. This not only

Page 15: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 1. Introduction 3

adds equipment complexity but it could also bias subjects outside their natural settings

(‘white coat’ syndrome).

Therefore, the development of simple, objective and quantifiable tests on “static images

of handwriting” which patients could potentially carry out themselves at home, or which

clinicians could utilize to ultimately make more informed decisions pertaining to the

severity of symptoms would be clinically valuable. Such self-administered assessments

would be even more valuable if (i) they require little training on the part of patients and

(ii) results or data could be readily sent to doctors or experts, or analyzed by trained

clinicians, suggesting a more inter-connected data-driven patient-clinician interface.

1.2 Overview of Research Aims

Although various measurement techniques are investigated with the aim to objectively

quantify motor symptoms variations caused by PD and/or ET progression as discussed in

Section 1.1, development of a computerized toolkit to quantify static handwriting tests for

experts and non-experts would be a valuable assistance method for clinical assessment,

be more suitable for long term symptomatic progression monitoring in natural settings for

large population screening, as well as intervention efficacy monitoring. In this

dissertation, we propose the following specific research aims:

Aim 1. Develop a set of quantifiable metrics for static handwriting analysis to

indicate handwriting changes associated with symptom variations in ET and

PD.

Page 16: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 1. Introduction 4

Aim 2. Establish an objective set of measurements to evaluate the efficacy of

short-term handwriting therapeutic exercise for people with PD -Amplified Air

Writing (AAW).

Aim 3. Establish a graphical user interface for handwriting tests that can be

used as a tool to assess symptomatic variations associated with ET and PD.

1.3 Contributions

As motivated in the above section, and in order to complete our research aims, the

major contributions of this dissertation are listed as follows:

Two sets of quantitative metrics were proposed and developed to assess

handwriting changes related movement disorders through static writing

samples: one set is implemented for micrographia monitoring in both

signature/string and single character analyses, and the other set is used for

evaluation of hand tremor through single character writing samples.

Two different human subjects groups, symptomatic PD affected individuals

with micrographia and healthy individuals without movement disorders,

participated in this study to provide their handwriting samples for metric

validation.

A self-administrative short-term (two weeks) therapeutic handwriting exercise

protocol, amplified air writing (AAW), was designed and tested by PD

affected subjects at their own home. The results are presented and illustrated

for studying the efficacy of this therapeutic exercise, and how it can be

quantified by developed metrics.

Page 17: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 1. Introduction 5

A novel infrastructure and application of graphic user interface for monitoring

handwriting changes associated with PD and ET is introduced including

details of the framework and its showcasing over realistic handwriting

samples.

Results obtained in this dissertation call for further infrastructural developments in

conjunction with data-driven assessment tools aligned with rapidly growing tele-

monitoring technology aimed at efficient delivery of healthcare beyond existing

practice. Furthermore, this work would be useful for remotely tracking of large at-risk

and symptomatic populations. Finally, the studied framework can be further

supplemented with digitizing tablets, to perform analysis of dynamic hand

movements. All these utilities would then feed into medical doctors in a clinical

setting in term of statistical inferences, to enhance the decision making process in

diagnosis, assessment of symptomatic variations, and evaluation of interventions in

ET and/or PD.

1.4 Outline of the Dissertation

This dissertation starts with introducing background on movement disorders, especially

with a focus on PD and ET and related existing studies in Chapter 2. The development

and evaluation of metrics for micrographic handwriting related to PD are explained, and

validated with real PD affected subjects’ historical signature writing samples in Chapter 3.

The design and validation of the metrics for tremulous handwriting related to ET are

presented and discussed in Chapter 4. Chapter 5 studies the utilization of metrics

designed for micrographic handwriting to validate the efficacy of a short term therapeutic

Page 18: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 1. Introduction 6

exercise in temporally mitigating PD-related micrographia effects in writing. Two

graphic user interface tools for monitoring micrographia and tremulous handwriting are

introduced separately with case studies in Chapter 6. In the last Chapter 7, the dissertation

is concluded with future research directions.

Page 19: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 7

Chapter 2

Introduction of Movement Disorders

2.1 Background of Movement Disorders

Movement disorders are referred as those neurological disorders that adversely affect the

speed, fluency, ease and quality of people’s movements. Only in United States, ET and

PD affect respectively 10 million people and 1 million individuals [3-4]. As the aging

population grows, e.g., in US, France, Japan, Italy, Germany, and many other countries, it

is expected that the prevalence rates and incidence rates for both ET and PD may even

increase in the next decade. This increase can be further amplified by a population of

about 20% of diagnosed patients corresponding to underestimated cases of people who

are never diagnosed [31-32]. The growing influence of ET and PD on the aging

population already requires increased healthcare services and added financial costs, and it

calls for inputs from engineers to address various grand challenges associated with them,

such as objective quantitative symptom progression indicators, and also development of

devices to manage daily life tasks including writing and dressing tasks [10, 14].

Although there is no cure for either ET or PD, interventions such as medication, surgery,

as well as therapeutic exercise, with various cost levels exist for ET and PD [1-14].

Pharmacological treatment and brain surgery can significantly alleviate symptoms,

Page 20: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 8

especially at the early stage of the conditions [5-6, 33-39]. However, the effectiveness of

such treatments also varies, and incorrect prescriptions and treatments can cause delay of

improvement, decreased patient compliance for further medication, and varying levels of

depression [40]. Patient intolerance of drugs can also be another issue [40].

2.2 Literature Reviews

2.2.1 Parkinson’s Disease

Parkinson’s disease (PD) is a movement disorder that is both chronic and progressive,

thus the condition will continue and might get worse over time [41]. PD is caused by the

death of neurons in the brain, especially within the area called substantia nigra. Within

this area of the brain melanin-containing cells are dying. When people are healthy, these

cells produce a neurotransmitter called dopamine, a chemical messenger that regulates

movement by assisting with the transmission of electrochemical signals. When these cells

within the substantia nigra degenerate, they no longer produce dopamine (see Figure 2.1).

The lack of this neurotransmitter leads to the loss of control of movement in many PD

patients [42]. Additional research also looks at a protein called alpha-synuclien which is

mainly found in neurons. This protein abnormally clumps together to form large masses

called Lewy Bodies which are believed to be the cause for many of the non-motor

symptoms presented by people with PD [1].

Page 21: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 9

Figure 2.1 Illustration of dopamine losses associated with Parkinson’s disease.

Motor and non-motor symptoms are two major categories of symptoms for PD. Motor

symptoms are those involving the loss of control of the movement of the body. Four

primary motor symptoms associated with PD [11] are: 1) Rest tremor. A shaking or

oscillating movement that usually occurs when the muscles are relaxed. Patients with this

symptom often have difficulty with repetitive movements or daily tasks. 2) Rigidity.

Muscles in a patient with PD lose the regularity of contracting and relaxing, since they

are always stiff. This symptom can be very uncomfortable for the patient and at times

even painful. 3) Postural instability. This symptom causes instability when standing

upright and dangerous sways backwards [41]. 4) Slowness of movement, called

Bradykinesia, which may cause people with PD to develop a slow, shuffling walk or a

stooped posture. Eventually, patients may lose their ability to start and keep moving

which is called akinesia which means patients will not be able to move at all. Any

symptom that does not involve movement is considered to be a non-motor symptom.

There are many different non-motor symptoms in this category: in early stages of the

conditions, patients may experience a loss of sense of smell, REM behavior disorder,

Page 22: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 10

mood disorders, and low blood pressure when standing up; with the progression of the

condition, other symptoms like weight gain and loss, fatigue, depression, dementia, and

other cognitive issues can also develop [1, 11, 41]. However, each patient may manifest

different symptoms, and may experience varied rates of progression as well as severity of

symptoms.

Rating scales such as the Unified Parkinson’s Disease Rating Scale (UPDRS), are used

for neurologists to determine the severity and present state of people affected by PD. The

most commonly used URDRS includes both self- and clinical-evaluation modes, in which

symptoms are reported on a scale from 0 (normal) to 4 (severe) [17]. In early state of PD,

medication can mitigate the symptoms effectively and regular exercise can help maintain

and improve the patient’s mobility. In the middle stage of PD (moderate degree),

medications can still help; however, they may wear off between doses and also start to

cause side effects. In the late stages of PD (severe degree), since medications may not

adequately suppress the symptoms, brain surgeries including Deep Brain Stimulation

(DBS) have been recommended by clinicians. Cognitive problems such as hallucinations

or delusions can become noticeable, and patients may need someone to look after them to

avoid unfortunate falling, and other dangerous issues.

Currently, there are many different options to treat the associated symptoms [1, 11, 16]:

1) Medication: there are five different classes of prescription drugs; Carbidopa/Levodopa

therapy, dopamine agonists, anticholinergics, MAO-B inhibitors, and COMT inhibitors.

However, with the incorrect dosage, adverse effects may develop. 2) Surgical option with

DBS: an electrode is placed into brain and an impulse generator is placed under the

collarbone. These impulses disrupt the electrical signals that cause the impaired

Page 23: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 11

movement, which have been programed by a medical doctor specifically. A controller

allows patients to turn the device on and off and to check the battery of the impulse

generator. 3) Physical therapy and exercise, including exercises like dancing, yoga, Tai

Chi, as well as bodyweight-supported treadmills, and other physical therapies such as one

of the prevalent programs called LSVT BIG and LOUD, to alleviate walking and vocal

related symptoms with repetitive exercise of high intensity and increasing complexity

[15, 43-48].

2.2.2 Essential Tremor

Essential Tremor (ET) is a neurological condition that causes oscillation of affected

muscles on the patients. More than 50% of existing ET cases link to genetic mutation,

since the genetic mutation that causes ET is an autosomal dominant disorder; if a

defective gene is passed on from one parent then the child will develop this condition

[49]. Another risk is age. ET is much more common among people over the age of 40 [2].

The symptoms of ET involve rhythmic trembling of the hand, head, voice, legs or trunk,

which begin gradually and worsen over time. These tremors only occur when the subjects

tries to use the affected muscle. For most people with ET, tremor usually starts in the

hands and spreads to the whole body as the condition worsens. These tremors can also

become more intense due to emotional stress, fatigue, caffeine, or extreme temperatures

[4]. These symptoms may vary from day to day; also the severity of tremor may vary

from morning to night. The mechanism of hourly and daily variability of ET severity,

however, remains unclear [50], and represents an important source of potentially spurious

Page 24: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 12

or unreliable outcomes in clinical trials designed to assess efficacy of a therapeutic

intervention.

At present, ET symptoms can be assessed by a neurologist in a clinical setting, where a

patient is observed for presence and severity of postural tremor, voice tremor, or kinetic

tremor via a series of specified tests [5]. Tremor severity variation may be measured by

rating scales or they could be more objectively assessed with a quantitative motor

assessment system (QMAS), see one example in Figure 2.3 [23]. Clinicians commonly

evaluate postural and kinetic tremor of ET patients using The Essential Tremor Rating

Assessment Scale (TETRAS) developed by the Tremor Research Group (TRG) [41, 46],

and the non-invasive home based self-administered assessment systems such as QMAS

are with limited use of patients for symptomatic monitoring. The limitations of current

tests include difficult to monitor and measure the symptom variability for each patient

through a day for a long period in clinical setting.

Several different strategies for reducing symptoms of ET exist in the research literature

and in clinical practice. Although ET affects millions of people worldwide, none of the

medications currently used for treatment were developed specifically for this purpose [2].

Pharmacological treatments include drugs also used for high blood pressure, seizure, or

anxiety. In addition, each patient responds differently to treatment, and in some cases

patients fail to tolerate the drug. For cases in which intervention using pharmaceuticals is

no longer feasible, surgery including thalalotomy or DBS can be successful [2, 3].

Physical therapy, including yoga and low-impact exercise, may also be effective in

reducing symptoms [4, 16].

Page 25: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 13

2.2.3 Handwriting Issues for People with Parkinson’s Disease and Essential Tremor

Figure 2.2 Handwriting samples of PD and ET affected subjects from two individual

studies [53-54].

Handwriting is a complex motor skill, which involves the interplay of arm and hand

muscles, as well as precise hand-eye coordination. This is why handwriting difficulties

may be one of the earliest symptoms presented in patients with PD and other movement

disorders, like ET [2, 34-35]. This complex motor skill, when neurologically affected and

degenerated, may manifest itself in various handwriting patterns that can be very

different from those written by healthy subjects. For instance, up to 75% of the people

with PD are unable to sustain normal-sized writing for more than a few letters. This

Page 26: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 14

clinical symptom, in which the handwriting becomes smaller, is called micrographia [55].

If this shrinkage is progressive from left to right then the condition is called progressive

micrographia; or if shrinkage is consistent across the handwriting then it is called

consistent micrographia [21]. Most people with PD usually show micrographia in their

handwriting, but no tremulous strokes. However, people with ET may show handwriting

with jitters and sometimes pretend to produce larger size samples but do not write

typically while shrinking the size of handwriting, see examples in Figure 2.2 [53-54]. In

addition, micrographia can manifest in the early stages of PD, so this presents also an

opportunity for reduced error in diagnosis if an objective assessment can be developed.

2.2.4 Available Engineering Solutions

Deterioration of handwriting is associated with some of the earliest symptoms of PD [27-

28]. Specific characteristics of handwriting can be used to distinguish PD from ET. Thus,

a more feasible and quick measure with minimal experimental bias is handwriting in a

natural setting, which can be easily implemented in larger populations. Although both PD

and ET patients commonly present with hand tremors and compromised legibility of

handwriting, only PD patients exhibit micrographia, a unique characteristic of PD [13,

27-28]. Writing is a complex motor task, and the precise mechanism of why

micrographia is generated in individuals with PD has not been established [20-21, 56-60].

One theory proposes that recursive distorted visual feedback may play a part in

handwriting size reductions for PD patients [57]. In fact, the diagnostic potential of

dysgraphia –the debilitation of handwriting motions– has already been recognized. There

have been explorations of electronic pen-pad combinations where dynamic data for

writing includes tip pressure, velocity, acceleration, and other kinetic and kinematic

Page 27: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 15

parameters in observable and measurable ways [26-30, 61-64]. But the problem is not

solved yet.

Several attempts have been made at automatic diagnosis and quantifying movement

disorders by using other motion tasks, most of them are clinical based. The Objective

Parkinson’s Disease Measurement (OPDM) System [25, 51] can extract the motor score

of a patient by putting a combination of multiple sensors on the subject’s sternum, wrists,

ankles, and sacrum. After the subject performs several standardized tests, all motion data

are processed to calculate a single score to indicate the mobility level of subject. Several

similar systems like the Kinesia motor assessment system [51], Motus Movement

Monitor [26] also exists. All these existing systems need to precisely place several

sensors on patients’ body by trained clinician or perform the test on specific location with

required equipment. Also, the cost of diagnosis and tracking systems is another concern

for used in large population. Thus, these systems for now only have limited application

mainly in the clinic. These solutions therefore call for developing better assessment

systems, which should be a low-cost, non-invasive, home based, self-administrated and

equipment-free computer aided diagnostic tool. The most prevailing solution is that

establishing a graphic user interface tool for static handwriting tests used for

distinguishing the symptomatic characteristics of ET or PD, as well as for long term

symptom tracking and treatment efficacy monitoring.

Page 28: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 16

Figure 2.3 Kinesia motor assessment system [51] integrates accelerometers and

gyroscopes in a compact patient-worn unit to capture kinematic movement disorder

features.

2.3 Problem Statement

Until now, the major challenges from an engineering point of view in this field include

the development of novel diagnostic techniques that can be used to distinguish ET and

PD, and the monitoring of symptom progression. Several research groups are currently

working in this endeavor. For example, researchers perform vocal testing/recording

during telephone conversations, study kinematic movements of body through sensors by

smart phones, analyze handwriting on digitized tablets, and with take-home monitoring

devices [19, 22-30]. Majority of these tests rely on real-time data, that is, data must be

recorded with time stamps, which means patients may need to have access to certain

equipment, or devices, which may also inevitably contribute to white coat syndrome.

Such procedures can be time-consuming, and financially costly, and required specialized

medical personnel. With the existing limitations, there are no globally agreed upon

techniques by which the symptoms of ET and/or PD can be detected as early as possible,

and the golden standard for diagnosis and symptom progression monitoring for both PD

Page 29: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 17

and ET are still subjective [74-75]. For example, a common misdiagnosis is to attribute

PD symptoms to the normal aging process, or to ET. In fact, diagnostic error rates of 43%

in undiagnosed conditions and 28% in misdiagnosis have been found [10-11]. This may

ultimately prevent patients from accessing early intervention in the initial stages of PD,

such as some medications and training therapies, which could otherwise improve their

quality of life. Furthermore, incorrect diagnoses may lead to undesirable medical

consequences due to taking inappropriate or contraindicated treatment [12]. Thus, an

objective and reliable assessment of symptoms, linked to PD or ET, is required.

Analysis of handwriting is one solution for this goal, since issues with legible

handwriting is a common problem for both people with ET and PD, and a series of

simple handwriting/drawing tests are often used to distinguish between these two

handwriting symptoms in clinical: ET often results in large, shaky handwriting samples,

while Parkinson’s writing typically starts out normal, then gets smaller and smaller,

squeezed and squeezed. Typically, handwriting samples and/or certain drawing tasks are

observed and recognized visually by neurologists. The limitations of current testing are

linked with the idea that patients are fully aware of the test implications. This may alter

the authenticity of the sample or reduce sensitivity to subtle improvements that could

present themselves over a long time. Also, it is difficult to monitor and measure the

symptom fluctuations for each patient throughout a day or for a long period to improve

the efficacy of clinical tests and treatments. Furthermore, the diagnostic results fully

depend on the experience of the neurologists. Although there are a number of researchers

who reported treatment efficacy in previous literature, there has been less correlative

work performed between clinical diagnoses and quantifiable daily symptomology

Page 30: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 18

measurements. In addition, the severity of the symptoms is diagnosed by a neurologist in

a clinic, via expert-level yet subjective recognition of symptoms. This method of

diagnosis adds clinical inertia, in that patients must wait to learn whether or not they are

improving or deteriorating, beyond their own recognition.

Consequently, a more feasible and rapidly acquired measure of handwriting with minimal

experimental bias would be valuable. Analysis of historical static images has this

potential [76-80], and can be easily implemented in larger populations. Because both

micrographia and tremors exhibit shapes and curvatures beyond those observed in

asymptomatic writing, and are different from one another, this option offers the potential

to be quantified, and thus can objectively and automatically assess static writing samples.

2.4 Significance & Innovation

Significance – Worldwide the number of people living with PD is roughly 10 million,

while ET is about 8 times more common than PD. However many more are unfortunately

not diagnosed. Thus, the growing influence of such conditions on the aging population

already requires increased healthcare services and financial burden [9-10]. Furthermore,

healthcare management for the increased risk of these two major conditions following

rapid progression with incorrect diagnosis and treatment should also be noticed, including

leading to dangerous falls, losing ability to move, associated severe psychological

conditions, and additional need for partner caring [31]. Hence, these two most common

movement disorders as a whole pose an important concern for the world.

At present, ET and PD symptoms can be assessed by a neurologist in regular clinical

visits between several month time slots, where a patient is observed for presence and

Page 31: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 19

severity of certain motor symptoms via a series of specified tests including a series of

simple drawing objects, see example in Figure 2.4 [81]. However, the severity of

patients’ conditions may vary between two regular clinical visits: from month to month,

day to day, even morning to night [50]. The limitations of current tests are difficult to

quantitatively monitor and measure the symptom variability for each patient through a

day for a long period in clinic setting. However, determination of circadian condition

fluctuation is critically valuable, since it can assist specialists in this field to adjust their

treatment for each individual more accurately and efficiently.

Figure 2.4 Archimedes spiral drawing test results among normal, PD and ET subjects

[81].

Thus, to bridge the above described critical barriers, we propose an approach to develop

and validate quantitative metrics by which symptomatic initiation and progression can be

assessed, and thus a clinical based or a home based self-administered assessment tool for

individuals with ET or PD can be made available. We also design and study the efficacy

of AAW exercise for people with PD. The success of this dissertation has the following

impacts:

Page 32: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 20

Identifying the objective quantitative metrics for static handwriting analysis of

people with PD or ET: establishing of such metric sets will not only provide

more objective information to support clinical decisions, but will also provide an

opportunity for rapid and low cost clinical diagnosis and treatment follow-up

(Aim 1).

Validation of AAW exercise will benefit many PD subjects with micrographic

handwriting issues in their daily lives (Aim 2), with potentials of this contribution

becoming a systematic exercise protocol in the future.

Toward a symptomatic progression monitoring tool through automated via

routinely measuring and analyzing the subjects’ daily symptomatic variations on

handwriting, which will assist neurologists to identify fluctuations in motor

symptoms at hand for ET or PD affected individuals (Aim 3).

The above listed impacts have many repercussions as well:

Enabling the efficacy analysis for a wide variety of ET or PD treatment in both

short and long terms, such as medicine, surgery, physical therapy.

Altering the traditional patient-clinician evaluation mode to save more healthcare

resources in ET and PD community, as illustrated in Figure 2.5.

Providing more reliable assessment results. Since whole assessment is carried out

in the subject’s familiar environment, the data are more ecologically valid and the

inconvenience and potential risks associated with a clinic visit are reduced.

Reducing embarrassment for people with ET or PD in public and supporting them

to pursue better quality of their daily living.

Page 33: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 21

Figure 2.5 Traditional patient-clinician evaluation mode (a) vs. new home-based

monitoring and evaluation mode (b) [23].

Innovation – The majority of efforts in researching, treating, and managing movement

disorders, including ET and PD, can be credited to medical, pharmaceutical, and

therapeutic origins. The role of engineers in this field has typically been limited to

mechanical design of affordances or tools to reduce the difficulty associated with

performing some activities of daily living. However, an opportunity has arisen to develop

engineering-based tools with the potential to aid in symptomatology analysis and

improvement. This dissertation describes the promising findings of identifying and

applying quantitative metrics to one of the fundamental activities affected by both ET and

PD –handwriting. In many symptomatic subjects, handwriting clarity and quality is

compromised due to motor symptoms. Accordingly, handwriting tasks are commonly

among the battery of clinical tests undertaken to observe progression of PD and ET, or

assess positive response to treatment interventions. Here we concern the resolution of

these tests. Across several medical fields, it is clinically valuable to develop tools that can

objectively detect small relevant changes in symptoms, and moreover in a reliable

Page 34: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 22

manner. Our preliminary data strongly indicates that such a tool can be developed for

both micrographia of PD affected and tremulous handwriting of ET affected individuals.

We propose writing activities that will enable objective, quantifiable, and automated

analysis of tremor and micrographia symptoms in handwriting. The analysis will be

performed on static writing samples on paper that are digitized and subjected to high

quality image processing. Note that this approach is different from existing dynamic

measurements that use an instrumented pen and electronic surface to ascertain pen

pressure and velocity, and other relevant kinetics and kinematics. Our success will enable

an ET or PD affected subject to submit any number of historical writing samples to a

clinic via email or post, or potentially apply the analysis on a home computer without the

need of a special instrumentation.

This work is innovative for several reasons.

First, analysis of tremor and micrographia in handwriting is significantly more difficult

than spirals and lines, as baseline writing maneuvers vary widely between individuals and

pen stroke formations may sometimes be similar to the dynamics of motor conditions of

the hand [61]. The reason for not following the tradition to dynamically analyze

handwriting but instead using static samples along with is to uncover particular geometric

aspects of handwriting that can be evaluated for deviation from standard legibility, and

abundant of historical data that can be analyzed only through static means. The static and

enduring nature of the samples also permits analysis of signatures prior to diagnosis,

providing an opportunity to track, predict, and rapidly diagnose early motor symptoms

and the progression.

Page 35: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 2. Introduction of Movement Disorders 23

Second, AAW has been first time quantitatively validated and optimized through

developed metrics here. This will provide a guideline for people with PD in their

handwriting exercise, and may shed light for other researchers in other similar principle

body/motion exercises.

Third, this home based self-administered infrastructural framework with GUI tool for

monitoring handwriting changes associated with ET and PD motor symptom, in itself, is

novel: we propose ET or PD motor symptom, specifically handwriting, that will be

enable objective, quantifiable, and automatically analyzed through self-administrated

static handwriting assessment. Next, the current gold standard for testing the severity of

ET/PD symptoms is diagnosed by a neurologist in a clinic, via expert-level yet subjective

recognition of symptoms. Our success may ultimately change the existing patient-

clinician evaluation mode, and propose a reliable and accurate self-administrated home

based monitoring and evaluation tool. Moreover, this enables several types of

applications including rapid adjustment for more precise medication level, long-term

evaluation and prediction.

Page 36: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

24

Chapter 3

Development and Evaluation of Metrics for

Micrographic Handwriting

3.1 Introduction

An alternative to the above-mentioned methods of quantitatively assessing movement

disorders is through the analysis of handwriting in subjects’ natural settings, which can be

easily implemented in larger populations with minimal experimental bias. Detection of

early-stage micrographia as a manifestation of Parkinson’s disease (PD) is challenging as

symptoms may be subtle. It is valuable if a computerized toolkit can be developed for

quantitatively assessing subtle yet relevant micrographia changes linked to PD symptom

variations trough static writing images, as it may be provide support and supplement for

clinical decision.

Indeed published work already advocated that handwriting can be analyzed toward

understanding symptom variations [18-20, 59-60, 82]. Building upon this knowledge

pool, in this chapter, several metrics sensitive to the characteristics of micrographia were

developed, with minimal sensitivity to confounding handwriting artifacts. These metrics

capture not only the global spatial changes but also the local spatial variations for

Page 37: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

25

signature or string analysis, such as character size-reduction, ink utilization, ink pixel

distribution on both horizontal and vertical directions, and pixel density within a sample

from left to right. As a supplement for string analysis, a set of metrics has also been

developed for assessing single characters, which are recorded as widely adopted practice

in clinical setting.

These two sets of metrics are introduced here to objectively evaluate digitized historical

signatures images of twelve different individuals corresponding to healthy and

symptomatic PD conditions, and control signatures sample that had been artificially

reduced in size with an averaged decrease rate, for comparison purposes. Metric analyses

of samples from ten of the twelve individuals for which clinical diagnosis time is known

show informative variations when applied to static signature samples obtained before and

after PD onset.

In summary, we present two sets of specific metrics for monitoring micrographia related

Parkinson’s symptom variation in this chapter: metrics for signature/string analysis

(Section 3.2) and metrics for single character analysis (Section 3.4). The schema table of

metrics is showed in Table 3.1. Study population and samples for those metrics are

described in Section 3.3.1 and Section 3.5.1 separately. Results obtained in experiments

are presented and discussed in Section 3.3 and 3.5 as well.

3.2 Development of Metrics for Signature/String Analysis

As described in Chapter 2, up to 75% of people with Parkinson’s disease are unable to

sustain normal-sized writing for more than a few letters; this symptom is known as

micrographia. It is almost unique to PD [57-60], usually shown as one of first early signs.

Page 38: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

26

In the clinic, subjective observations of handwriting changes for PD affected subjects’

symptomatic progression have been widely used based on UPDRS.

Quantitative Metrics for Micrographic Handwriting

Metric

Type

Signature/String Analysis Single Character

Analysis Global Metrics Local Metrics

Metric

Name

Sizing Metrics

(Area, Weight, Height)

Pixel Density

Variation metric

Skeletal Points

Metric

Quantifiable Ink Deposit

Metric

Number of Peaks in

Horizontal Projection

Profile (NP-HPP)

Metric

Polar Distribution for

Angles and Distances

Metric

Resized Horizontal/Vertical

Projection Profile

(HPP/VPP) Difference

Metric

Table 3.1 Summary of metrics described in Chapter 3 used for monitoring micrographic

handwriting associated PD symptom variations.

While the limitations of this have been listed in Chapter 2, here, we propose a set of new

metrics and also adopted- and modified-metrics from existing studies [83] by which one

can quantitatively analyze handwriting of individuals, with specific focus on subjects

with micrographia. We also borrow already suggested metrics [18-20], mainly related to

global size changes (height, width, area of sample). Using all these metrics, one could

reliably compare string patterns of subjects, which can actually be tied to progression of

micrographia. Specific string (such as signature) refers to one’s name written in cursive.

Next, two types of metrics were developed by author:

Page 39: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

27

(a) Global metrics, which are expected to be affected by fundamental changes in

successive signature dimensions (scaling) [59-60] as observed especially in

consistent micrographia [82] and

(b) Local metrics describing the longitudinal compression of handwriting in a

single line from left to right (distortion and shrinkage), as is present especially in

progressive micrographia [82].

Note that historical signatures are available across a range of sizes. Therefore, any useful

metric would need to differentiate between micrographia symptoms and artifacts such as

arbitrary or synthetic scaling (as seen in Figure 3.9c), distortion, or size-reduction during

data processing/collection.

3.2.1 Global Metrics

Sizing Metrics: Following the approach introduced by previous research groups to study

micrographia, at first, we find the boundary box of each signature sample. Next, we use

this cropped signature image to find its height (as measured by maximum vertical stroke)

and width (as measured by total horizontal distance traveled by pen) in pixels, and then

calculate the overall signature area as height multiplied by width, see Figure 3.1. This

metric is used by Teulings et al. [58], Kim et al. [82] and Shukla et al. [20]. In [58],

Teulings measured the height and width of loop patterns (like “lllll”) while in [82, 20],

the authors measured the size of each letter written by subjects, which are used as a

metric to assess micrographia. Here, the signature sizing metrics are automatically

computed in MATLAB, and can be used to study the change in the signature scale

changings for each subject, which may help us establish a link to understanding the

Page 40: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

28

progression of micrographia. We also hypothesize that while signature area can serve as

an indicator of overall writing shrinkage, it cannot necessarily differentiate micrographia

symptoms from artificially linearly resized signature samples.

Figure 3.1 Example of 1st signature image (Sig 11) in (a) within Figure 3.8 (a), the original

sample image from [59].

Quantifiable Ink Deposit Metric: Since after the pre-processing procedure (see Section

3.3.2), a signature trajectory is represented in MATLAB by a combination of pixels with

the logic value of 1, we can calculate the total ink utilization as measured by the number

of pixels in a sample. With the onset and progression of PD conditions, subjects may be

affected by many factors, and thus they may not complete their signatures as they

otherwise would. For example, they may not draw a big loop or circle, or may even omit

some specific letters, and all of these will lead to decreased ink deposits in their

signatures. This metric can therefore tell us how much effort a subject makes when

producing a signature. A metric analogous to this one was suggested by Teulings [18] but

was called as “stroke length” which is equivalent to counting the ink registrations in a

sample.

Page 41: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

29

Figure 3.2 Visualization of quantifiable ink deposit metric implemented on 1st signature

image (Sig 11) in (a) within Figure 3.8(a), the original sample image from [59].

Resized Horizontal/Vertical Projection Profile (HPP/VPP) Difference Metric: Studying

ink distribution is another promising direction, especially in the security and forgery

detection literature [83]. To apply this method, we define the function f(x,y) on the plane

(x,y), where, at the point (x0,y0) (x,y), f(x0,y0) = 1 whenever a pixel in a digitized

image exists, else f is zero. Then the horizontal projection profile (HPP) and vertical

projection profile (VPP) curves are calculated respectively as follows

𝑉𝑃𝑃 (𝑥) = ∑ 𝐹(𝑥, 𝑦)𝑦𝑚𝑎𝑥𝑦=0 (3.1)

𝐻𝑃𝑃 (𝑦) = ∑ 𝐹(𝑥, 𝑦)𝑥𝑚𝑎𝑥𝑥=0 (3.2)

which represent the sum of all pixels in a vertical (or horizontal) strip at each horizontal

(or vertical) location, plotted along the horizontal (or vertical) [83].

Page 42: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

30

Figure 3.3 Original Horizontal/Vertical Projection Profiles of Subject 06: Signature H (1st

signature is asymptomatic) vs. Signature S (5th

signature -symptomatic).

The areas under a HPP curve or a VPP curve are identical as this area is indicative of

pixel accumulation in a sample, which corresponds to how much ink is being used to

produce that sample.

Figure 3.4 Visualization of areas under HPP/VPP curves (blue colored areas) for 1st

signature image (Sig 11) in (a) within Figure 3.8(a), the original sample image from [59].

We expect that distortions due to micrographia would ostensibly alter the shape of HPP

(or VPP) curves. However, if a signature is artificially linearly reduced in size (as in

Figure 3.9c) the shape, or profile, of these curves would remain unchanged. For these

reasons we hypothesize that the area under the HPP (or VPP) curve serves as a

Page 43: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

31

differentiator between affected and unaffected subjects. Details in calculating this metric

are explained next.

Area Variation under “resized” HPP and VPP Curves: Within each subject, we computed

the differences in areas formed under HPP (or VPP) curves, across signatures Sigij with i

being the subject number and j being the sample number. The first sample of a subject

Sigi1 is used as the baseline for studying the changes in the remaining samples Sigi2, …,

SigiN, for each i. First of all, each signature Sigi2, …, SigiN is scaled horizontally to fit its

HPP (or vertically to fit its VPP) curve to that of Sigi1 while keeping sample aspect ratio

fixed. This normalization is performed to ensure that the areas under the curves are scaled

comparably. Next, the areas formed under the resized HPP (or VPP) curves are computed,

and the differences in area between Sigi1 and each one of the samples, including Sigi1, are

calculated as

Sigi1 – Sigi1 = 0, Sigi1 – Sigi2, … Sigi1 – SigiN (3.3)

Here, the zero difference is the first data point representing the baseline Sigi1 being

compared to itself. In light of the performed steps, this new metric is called resized HPP

(or VPP) area difference. Since this metric quantifies the differences in area, it is plotted

as an absolute value in order to simplify its presentation.

Page 44: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

32

Figure 3.5 Resized Horizontal/Vertical Projection Profiles Subject 06 (with fixed aspect

ratio): Signature H (1st signature is asymptomatic) vs. Signature S (5

th signature -

symptomatic).

3.2.2 Local Metric-Pixel Density Variation Metric

Since micrographic effects can be progressive as samples are produced especially in

“progressive micrographia”, such effects can be detected based on how various

characteristics change locally from left to right within a sample. One such novel metric to

measure compression/distortion features in signature samples is described here, namely,

pixel density variation (PDV).

Calculation of PDV metric is performed as follows. Each signature is first split into

several cells of identical width. Cell height is determined by the upper and lower

boundaries of an ink deposit, and this multiplied by the width between the lateral

boundaries computes the cell area Aijk, in pixels with i being the subject number, j being

the sample number, and k being the cell number in the sample counting from left to right.

The quantity of ink pixels Pijk is measured in each cell, and divided by area to calculate

pixel density

ρijk = Pijk

Aijk (3.4)

Page 45: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

33

The linear fit of pixel density ijk plotted along the cell locations k = 1…K, with K being

the total number of cells in the sample, is reported as pixel density variation for that

signature sample. For instance, when the linear fit has a positive slope mij = mij (ijk) > 0

based on the density readings, this indicates that the density in this signature sample is

increasing on average from left to right.

Figure 3.6 Visualization of Pixel Density Variation Metric implemented on 5th signature

image (Sig 15) in (a) within Figure 3.8, the original sample image from [59].

3.2.3 Local metric-Number of Peaks in Horizontal Projection Profile (NP-HPP)

The Number of Peaks in Horizontal Projection Profile (NP-HPP) metric is another novel

metric developed in this work, aimed to address the morphological distortions in

handwriting samples relative to non-symptomatic writing. The metric evaluates the pixel

distribution profile of a given sample from left to right, and computes the number of

peaks in this profile. The distribution, known as horizontal projection profile (HPP) is

representative of sample morphology along horizontal direction (which is also the writing

Page 46: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

34

direction when sample produced), while the peak occurrences in this distribution indicate

the presence of unique and spaced pixel strokes, which is the novel contribution of this

dissertation. The fewer of these occurrences, the fewer the number of such unique strokes,

and more connected and densely grouped the stroke patterns are; see a representative

image in Figure 3.7. This is similar to handwriting patterns being horizontally stretched

and vertically pressed, which has elements analogous to micrographic effects [20, 59-60].

Further, this analysis is similar to approaches used in crystallography, particularly for

structural interrogation via X-ray diffraction [84].

As explained in the previous section, the focus of this metric is to capture variations in

morphological aspects of static handwriting samples. We investigate below whether or

not the NP-HPP metric can correlate with micrographia symptoms with statistical

reliability in the analysis of subjects’ static handwriting samples.

In a digital setting, HPP is obtained by counting the number of pixels with ink

registrations, in each vertical column of the sample at hand. This generates a HPP curve

distributed over the horizontal axis of the sample. With this, mathematically HPP, which

is only a function of x is formulated as before in Equation 3.2.

Page 47: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

35

Figure 3.7 Simple geometrical samples illustrate the relationship between the peak

occurrences in HPP profile and the unique and spaced pixel strokes in samples.

The HPP metric has been applied in the signature security literature [83]. However, the

following observations and mathematical analysis have not thus far been performed on

HPP to the best of our knowledge. Once the HPP curve is available, the number of peaks

in this curve is counted. This count is strongly related to the intensity of unique stroke

movements, and is proportional to the presence of spaces between such strokes, see

Figure 3.7. Mathematically speaking, the peak occurrences can be interpreted as the

local/global maxima points of HPP(x), which are in the set of singularity points of HPP.

We propose that such strokes and spacing described above can be produced easily and

more deliberately when micrographia symptoms are not present, consistent with medical

knowledge that micrographia leads to cramped writing and muscle rigidity, thus not

allowing such deliberate strokes [19, 85-86]. Therefore, we hypothesize that the result of

NP-HPP metric will be much larger in samples of healthy subjects than in samples

affected by micrographia.

Page 48: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

36

3.3 Evaluation of Metrics for Signature/String Analysis

3.3.1 Subjects and Study Samples

Table 3.2 Demographic information and clinical characteristics of study participants from

Subjects: 01-05 & 07-10.

Page 49: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

37

For the purpose of this work, signature refers to one’s name written in cursive, and word

refers to any single string, such as a first name, middle name, or surname. Historical

signature samples of twelve different individuals with PD (referred to here as Subjects

01-12) were analyzed1.

The study was approved by the Institutional Review Board (IRB #13-04-03). Nine of

them (Subjects 01-05 & 07-10) were recruited through local support groups and 2013

World Parkinson Congress. Data including relevant study information (Table 3.2) were

collected from these subjects under approved IRB. All recruited subjects signed a written

informed consent. Inclusion criteria were (a) subjects’ clinical diagnosis of PD and (b)

micrographia. For Subject 06 and Subjects 11-12, samples were downloaded directly

from published literature without the need for IRB clearance due to the publicly available

nature of the data (Figures 3.8-3.9) [59-60].

From Subjects 01-12, we received 4 to 8 signature samples dating back to times before

and after clinical diagnosis of PD. Moreover, the signatures for Subjects 01-10 include

both healthy and symptomatic samples, along with the information about when the first

signature after clinical diagnosis of PD was obtained. Diagnosis information for Subjects

11-12 is not available, and hence they are addressed separately. For purposes of

comparison, the control sample sequence depicted in Figure 3.9(c) corresponds to the

mean shrinkage rate observed in samples at an average of 5.62% successive linear

artificial shrinkage of Signature 1 (Sig11) of Subject 1. All the samples of Subjects 01-12

in Figures 3.8-3.9 are presented in chronological order from top to bottom, and whenever

1 Signature images were resized to conserve space. Analyses were performed on the original size of

images.

Page 50: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

38

Figure 3.8 Historical signature samples of Subjects 01-10 (a)–(j) in chronological order.

In order to protect subject’s privacy, parts of the signature images were pixelated. The

labeling order (a)-(j) does not necessarily match with the labeling order 01-10 used in

data analysis. An arrow is used to mark the PD onset based on clinical assessment.

Except for pixel density variation, all samples are used in metric calculations. For pixel

density variation, only samples marked with rectangular frames next to them are used in

statistical analysis, with green dashed frames corresponding to “early recordings” and

solid red frames corresponding to “recent recordings”. Subject 06 is excerpted from [59];

Subjects 01-05 and 07-10 were collected under approved IRB.

Page 51: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

39

Figure 3.9 Historical signature samples from Subjects 11-12 in chronological order [panel

(a) and (b)] and artificially generated control signature samples [panel (c)]. Subjects 11 &

12 are excerpted respectively from [59] and [60]. Onset of PD is unknown for Subjects

11-12. All samples are pixelated to respect subjects’ privacy.

available, the timing of data recording is provided next to each sample along with a red

arrow indicating the timing of clinical diagnosis.

3.3.2 Pre-processing

Prior to calculating the metrics for the signatures at hand, standard preparatory digital

processing tools were applied on the samples in their original sizes, following established

methodology [87-93]. As such, there are four basic pre-processing steps, see an example

in Figure 3.10:

(a) Conversion from gray-color to a binary (black and white) signature image

The color of ink has no significance in this study; however, the signatures still need to be

in a comparable color-density format. For this, we take a snapshot of all signature

samples in Figure 3.8, and convert these samples to binary images where all colors are

Page 52: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

40

mapped to black which is represented by 0, and areas without any trace of ink are made

white with a logical value of 1. This is, the signature part of an image is represented by 0

and the background part of the image is assigned the value of 1 in the initial uptake

matrix.

(b) Inverting the binary signature image to a white signal on a black background

Next, the pixels forming the signature are represented by a value of 1 and the background

area is represented by a 0 value. This conversion makes further coding much easier.

(c) Filtering out stray dots, deposits and visual noise

We applied the filtering command of MATLAB to remove the extra dots and random

noise in the signature image, which may undesirably deteriorate the accuracy of the

results. For instance, the ‘noisy’ parts of the 3rd

signature image in Figure 3.10 can be

removed in this step.

(d) Identifying the rectangular boundary of the signature image and accordingly cropping

the margins

The boundary box of the signature is identified according to the maximal vertical and

horizontal strokes. To eliminate the blank areas form the sides of the images, they were

cropped out, leaving only the signature in the working box region.

In general, we implemented the pre-processing procedures (a) – (d) to all the signature

images in a MATLAB computer program developed by the author to eliminate

nonessential marks and capture relevant writing characteristics from the signature

foundation.

Page 53: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

41

Figure 3.10 Pre-Processing steps (a) – (d) on example of 1st signature image (Sig11) in

Figure 3.8 (a), the original sample image from [59].

3.3.3 Results

Feature Normalization: To remove the individual dependency, we normalize the obtained

metrics before analyzing the results. This normalization step brings all the metrics in

between 0 to 1, such that

Page 54: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

42

Fnorm =test (F)

max(F) (3.5)

where test(F) and max(F) are the current tested sample values and maximum values of

each metric calculated for a particular signature sample.

Analysis of Metrics on Signature Samples:

It is important to note that some metrics are meaningful on an absolute scale, while others

are scaled with respect to the largest value in that metric to represent metric changes as

percentages (see Figures 3.11 & 3.12). For instance, 0.4 in a metric value indicates that

the metric value is 40% of the maximum value that the metric takes across all the samples

j of the corresponding subject i. This would imply a 60% drop from the maximum to that

metric value.

Across all samples, the metrics pertaining to area, ink deposit, height, width and HPP (or

VPP) area are plotted. Notice that HPP area and VPP area metrics represent the total

amount of ink utilized in a sample (hence the similarity of numerical readings below);

these metrics are different from the one described as resized HPP (or VPP) area

difference. As can be seen in Figures 3.11 and 3.12 the metric values for area, ink deposit,

height, width and HPP (or VPP) area respectively show an average decrease of 54.26%,

52.91%, 36.79%, 35.43% and 52.29% for Subjects 01-10; 66.69%, 57.67%, 44.93%,

40.29% and 57.71% for Subjects 11and 12. On the other hand, some of the metrics under

consideration also show a decrease for the “artificially reduced” signature samples in

Figure 3.9 (c). Metric results for these samples in area, ink deposit, height, width and

HPP (or VPP) area present respectively an average drop of 54.89%, 54.97%, 32.50%,

33.17% and 54.99%, which introduces an anticipated confounding factor.

Page 55: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

43

By design, the above issue can be addressed by resized HPP (or VPP) area difference and

pixel density variation (PDV) metrics. For example, large fluctuations are observed in

resized HPP and resized VPP area difference metrics for Subjects 01-12 with maximum

changes across subjects averaging respectively 1523.08 and 2628.46 units, while the

control value set remains relatively flat, as expected, only showing negligible average

changes of 28 and 51 units2, providing supporting evidence that this metric can serve as

an appropriate differentiator of micrographic writing in symptomatic samples. Detailed

discussions on PDV metric are provided next in a separate subsection.

2 Notice that, ideally this variation must be zero, since the control samples simply reflect linear shrinkage of

Sig11. Nevertheless negligible variations in these metrics arise due to small imperfections in image quality

and resizing.

Page 56: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

44

Figure 3.11 Metric values for Subjects 01-10 and control samples from Figure 3.9(c) (see

text for definitions and normalization procedures). The x-axis represents the signature

samples in chronological order. A red dashed vertical line is used to mark the clinical

diagnosis of PD as onset. Samples are horizontally aligned to easily present before and

after onset.

Page 57: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

45

Figure 3.12 The y-axis represents the corresponding metric values (see text for

definitions and normalization procedures) for Subjects 11-12 and control samples from

Figure 3.9(c). The x-axis represents the signature samples in progressive chronology;

consistent with Figure 3.9. PD onset for Subjects 11-12 is unknown.

Page 58: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

46

Detailed Analysis on PDV Metric:

The PDV metric presents 104.68% changes in samples from Subjects 01-10 and 132.55%

changes for Subjects 11-12 compared to only 8.18% changes in the artificially reduced

signatures (see also footnote 2).

To statistically analyze the density variation metric, we divide the two-string signatures

into a total of 6 cells (K = 6); 3 for the first name and 3 for the surname, and those with

three strings (first, middle, and surname) are divided into a total of 9 cells (K = 9). We

next compute and analyze the ratio between density value in Cell#2 (ij2) and the density

value ijK in the last Kth cell. Here we have respectively K=6 in two-word signatures

(corresponding to Cell#6) and K=9 in three-word signatures (corresponding to Cell#9).

Mathematically, this ratio is given as follows:

Rij =ρij2

ρijK (3.6)

The rationale behind the above formulation is two-fold. First, since leading capital letters

of samples are relatively larger in Cell#1, the cells containing them were not used; Cell#2

is used instead to prevent confounding the representative statistics. Second, since

handwriting shrinks from left to right due to progressive micrographia, it makes sense to

investigate how shrinkage compares within a contiguous writing sample.

Density ratios Rij are calculated next for both “earlier recordings” (Group 1) and “recent

recordings” (Group 2) for Subjects 01-10; see the markings in Figure 3.8. For statistical

analysis of the metrics, independent inter-subject t-tests and dependent intra-subject t-

Page 59: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

47

Density Ratio Rij between the last cell and Cell #2

Subject i

(Sample ij)

Group 1: Early

Recordings

Group 2: Recent

Recordings

Subject 01 (S11 vs S17) 1.033 1.212

Subject 01 (S12 vs S18) 0.887 1.297

Subject 02 (S21 vs S24) 0.550 1.430

Subject 02 (S22 vs S25) 0.613

613

1.122

Subject 03 (S31 vs S33) 0.778 1.351

Subject 03 (S32 vs S34) 0.927 1.144

Subject 04 (S41 vs S43) 1.374 1.640

Subject 04 (S42 vs S44) 1.618 1.553

Subject 05 (S51 vs S53) 1.183 1.526

Subject 05 (S52 vs S54) 1.082 1.130

Subject 06 (S61 vs S63) 0.825 1.505

Subject 06 (S62 vs S64) 1.059 1.259

Subject 07 (S71 vs S75) 1.110 1.138

Subject 07 (S72 vs S76) 1.023 1.169

Subject 08 (S81 vs S83) 1.421 2.846

Subject 08 (S82 vs S84) 1.463 1.592

Subject 09 (S91 vs S93) 1.010 1.161

Subject 09 (S92 vs S94) 1.365 1.209

Subject 10 (S10,1 vs S10,5) 0.840 1.209

Subject 10 (S10,2 vs S10,6) 0.868 1.298

Average 1.0515 1.3896

Paired t-test* p = 0.00024 < 0.001

Independent t-test* P = 0.00191 < 0.002

* Denotes statistical significance between Group 1 and Group 2.

Table 3.3 Computed density ratios Rij in signature samples marked in Figure 3.8 with

rectangular frames.

tests are performed3. Specifically, we find that the density ratio Rij in (Equation 3.5)

shows significant differences in both paired and independent t-tests, across Subjects 01-

3 Data from Subjects 11-12 are omitted from statistical analysis since it is unknown when PD has been

diagnosed for each of these individuals.

Page 60: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

48

10 and their signatures, p<0.001 and p<0.002, respectively (Table 3.3), with power of the

study being 99.39% following the analysis from [94-95].

Detailed Analysis on NP-HPP Metric:

Study of size, style, and micrographia effects on NP-HPP metric: Sample dimensions

(height, width, and spacing) may vary regardless of the presence of micrographia, and

this factor may affect the HPP curve, and ultimately the NP-HPP values. Similarly, stroke

style (e.g., italic, cursive) and thickness of the stroke (bold vs. normal font) can

contribute to this value. Hence, it is necessary to investigate how NP-HPP is related to

such features.

Once this is understood, it will then be possible to utilize the NP-HPP metric to study

whether or not samples with micrographia can be discerned from healthy samples, and

even from artificially size-reduced samples not related to any symptomatic root cause.

For this purpose, four sets of data were generated:

Data Set 1: NP-HPP is calculated for all the samples available. Moreover,

samples with the oldest records are labeled as Original Healthy (OH). For

Subjects 01-10, these are the samples collected before PD has been diagnosed,

however for Subjects 11-12 (analyzed separately and in aggregate with the other

subjects), we do not know their exact clinical states, and hence it is assumed that

the oldest records of these subjects correspond to relatively healthier states. Next,

the signature samples with the smallest width were labeled as Original

Symptomatic (OS). Since micrographia causes shrinkage in samples, OS

Page 61: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

49

corresponds to the most micrographic sample. For all the subjects, we definitely

know that OS samples were recorded after clinical diagnosis of PD.

Data Set 2: Several artificially-resized samples are generated to isolate and

investigate size effects. For this, we artificially increased the size of OS sample to

match its width with the width of OH, keeping the aspect ratio fixed. This yielded

the Artificially amplified Symptomatic sample, denoted by AS. Next, we

artificially shrunk OH to match its width with the width of the selected OS

sample, keeping again the aspect ratio fixed. This created an Artificially shrunken

Healthy sample, labeled as AH; see Figures 3.13-3.14.

Data Set 3: The OH sample of each subject is resized to match its width with each

sample available from the subject, keeping the aspect ratio fixed. In other words,

multiple AH samples are produced, where each such sample has matching width

with one of the samples of the subject. These samples will be used to study size

effects only as they are not affected by micrographia since their origin is the OH

sample.

Data Set 4: Microsoft Word®-simulated writing samples matching with the

widths of the samples in Data Set 1 were produced. Since simulated samples are

not affected by micrographia, they are used to conduct baseline testing and

sensitivity analyses, specifically regarding how the NP-HPP metric was affected

purely due to size effects. This study helps (a) to evaluate NP-HPP against

symptomatic samples affected both by micrographia and size reduction, and (b) to

understand the effects of italic, normal, and bold fonts, as well as various (serif

and sans-serif) font styles, including Gigi, Calibri, Palatino Linotype, Edwardian

Page 62: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

50

Script, Bradley Hand ITC and Freestyle Script to eliminate the style effect as a

factor. The Microsoft Word®-simulated writing samples were produced by

reducing the font size, converting the sample to an image, and then resizing the

image to match its width with those of the samples in Data Set 1. These

treatments carefully prepare the synthetic samples to serve as baseline for

comparison with realistic samples of the subjects.

Figure 3.13 Flowchart showing the selection and generation procedures of samples: OH,

OS, AH, AS for Subjects 01-12.

In all analyses, samples are treated in their original dimensions, but are presented in

figures in smaller size (maintaining aspect ratio) to fit within page margin requirements.

An example calculation of the NP-HPP metric is depicted in Figure 3.14. Data Set 1 is

used to plot the NP-HPP metric. Data Set 2 is used to determine whether the NP-HPP

metric is different among the four sets OH, OS, AH, and AS. For this, Analysis of

Variance (ANOVA) with single-factor tests was conducted. If significant differences

existed among these four sets, post-hoc tests with Bonferroni Correction were used to

Page 63: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

51

adjust critical p value in multiple pairwise comparisons, and pairwise t-tests were

performed between OH vs. OS, OH vs. AS, and OS vs. AH to investigate whether or not

statistically significant differences existed within any of the comparisons. Since we wish

to study variations with respect to an “original-sized” sample, comparisons between

artificially generated AS and AH are not considered. Data Set 3 and Data Set 4 are used

to study size effects on NP-HPP metric, and to compare the metric values with those in

Data Set 1 to determine the difference of metric values caused solely by symptomatic

micrographia effects.

Figure 3.14 A sample with Original Healthy (OH) and Original Symptomatic (OS) of

Subject 05, and the creation of Artificially-sized Symptomatic (AS) and Artificially-sized

Healthy (AH) samples as described in Figure 3.13. Samples OH and OS are treated in

their original sizes. All these samples are reduced in size here to fit them within margins.

HPP profiles and the peak points are marked on the second row, similar to Figure 3.7.

See statistical analysis on Tables 3.4-3.5. All samples are pixelated partially to respect

subject’s privacy.

NP-HPP metric is sensitive to micrographia independent of size effects: NP-HPP is

calculated over Data Set 1 refers to the samples in Figure 3.8 and 3.9. Note in this figure

that sample numbers were intentionally randomized as (a)-(j) to prevent directly

Page 64: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

52

matching them with subjects’ information. Next, the difference in metric value is

investigated between Data Set 1 and Data Set 3, where metric values in Data Set 1 may

be affected by size and micrographia, and in Data Set 3 the metric value may change

solely due to size effects since this data is generated from a healthy sample. A positive

difference in the metric value indicates that metric values in Data Set 1 drop more

dramatically compared to those obtained from Data Set 3. In other words, a positive

difference is indicative of micrographic effects. As observed in Figure 3.15-3.16, this is

indeed the case for almost all data points available. Note in figures that the horizontal

axis corresponds to sample number, from older to more recent recordings, with the

understanding that not all sample numbers for all the subjects have the same time stamps.

Figure 3.15 The NP-HPP metric differences obtained between Data Set 1 and Data Set 3

for Subjects 01-10 where PD onset is known (data horizontally aligned). Positive value of

the vertical axis indicates that NP-HPP metric differentiates micrographia effects,

regardless of changes in sample size.

Page 65: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

53

Figure 3.16 The NP-HPP metric differences obtained between Data Set 1 and Data Set 3

for Subjects 11-12 where PD onset is unknown. Positive value of the vertical axis

indicates that NP-HPP metric differentiates micrographia effects, regardless of changes in

sample size.

Number of Peaks in HPP

Subject # OH OS AH AS

01 38 24 30 28

02 22 13 19 16

03 29 11 25 14

04 25 14 21 16

05 30 22 27 25

06 38 22 30 29

07 29 18 26 20

08 30 15 25 23

09 27 18 23 22

10 31 18 22 20

11 19 8 20 10

12 45 14 33 24

Mean 30.25 16.42 25.08 20.58

Table 3.4 NP-HPP metric results for Subjects 01-12 between different sample groups.

NP-HPP metric values are significantly different in pair-wise comparisons between

size-matched most healthy states and most symptomatic states of subjects: NP-HPP

metric values on Data Set 2 are reported in Table 3.4 Statistical analysis of these values

through ANOVA showed that there was significant difference among the 4 groups of

samples as seen on Table 3.5 (p < 1E-05 with the power of the study being 99.97% for

Page 66: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

54

tests both on Subjects 01-10 and Subjects 01-12). With Bonferroni Correction, the

adjusted critical p value becomes α = 0.05/4 = 0.0125, and based on our expectation of

which direction the metric value will tend to change, one-tail tests are performed, which

indicate significant differences in paired t-tests between OH and OS, OH and AS, OS and

AH in one-tail tests (p < 0.0125); see the details on Table 3.6 for Subjects 01-10, and

Table 3.7 for Subjects 01-12.

Statistical results support our hypothesis that the NP-HPP metric is effective: (i)

differentiate healthy/healthier signature samples from symptomatic micrographic samples

(OH vs. OS); (ii) readily discern healthy signature samples from like-sized micrographic

samples (OH vs. AS); (iii) identify the differences between symptomatic micrographic

signatures and like-sized healthy signature samples (AH vs. OS).

NP-HPP metric for Microsoft Word simulated samples compare similarly with

those of size matched healthy samples: We next performed another analysis using Data

Set 1 and Data Set 4, as outlined in Figure 3.17. Here, we took the four samples of

Subject 05 as a benchmark and selected a similar font style in Microsoft Word with a

comparable font size, in order to “simulate” the four samples. One of the authors,

considered healthy, produced samples that were matched in size with the writing samples

of Subject 05. These newly generated healthy samples were treated as “control” in the

analysis. At first glance, one may perceive this to introduce experimental bias however as

observed in Figure 3.17 (column 4) the sample produced by the author does not present

any obvious distortion, and also almost identically matches in NP-HPP metrics with

respect to the samples generated by Microsoft-Word (Figure 3.17, columns 2-3).

Ultimately, this “control” sample is deemed minimally-biased / unbiased. Results in

Page 67: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

55

Figure 3.17 show that, while keeping size-reduction ratios consistent across the four

columns, the number of peaks in HPP in the samples of Subject 05 decreased 76%

compared with only about 47% decrease in simulated samples. Results indicate that the

NP-HPP metric values for simulated samples are comparable to those in size-matched

healthy samples, and reduction in the metric values among size-matched samples is

exacerbated in the presence of micrographia.

Different styles in Microsoft Word simulated samples do not render statistically

significant differences in NP-HPP metric values: In order to test sensitivity to various

styles, we selected six varied font styles (Gigi vs. Calibri vs. Palatino Linotype vs.

Edwardian Script vs. Bradley Hand ITC vs. Freestyle Script) and three stroke styles

(Normal vs. Bold vs. Italic) to investigate how such styles affect the NP-HPP metric.

Figure 3.18 illustrates that the cursive handwriting style (Gigi) shows a larger NP-HPP

metric value than the printed (Calibri) style. Cursive handwriting style is also more

sensitive to variations in NP-HPP metric than the printed style, and under the same size

reduction ratio, the cursive Gigi-styled sample led to a decrease of 38%, and that of

Calibri caused a decrease of 29% in the metric value. Similar findings are also seen for

the other font pairs: Palatino Linotype, Edwardian Script, Bradley Hand ITC and

Freestyle Script. We note that this is compared to the 76% reduction in the metric value

found with symptomatic signatures in Figure 3.17.

Further, in the Gigi font style, normal samples and bold samples lead to a decrease at

similar rates, respectively at 38% and 39% in the NP-HPP metric, whereas this decrease

was at 52% using the italic stroke style. In Calibri font style, the normal and bold samples

correspond to decreases at 29% and 27%, respectively, and the italic samples encounter

Page 68: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

56

SUMMARY (A)

Groups Count Sum Average Variance

Col 1, OH 10 299 29.9 25.43333

Col 2, OS 10 175 17.5 18.27778

Col 3, AH 10 248 24.8 13.28889

Col 4, AS 10 213 21.3 26.01111

Source of

Variation SS df MS F P-value F critical

Between Groups 834.275 3 278.0917 13.40021 4.99E-06 2.866266

Within Groups 747.1 36 20.75278

Total 1581.375 39

SUMMARY (B)

Groups Count Sum Average Variance

Col 1, OH 12 363 30.25 52.2045

Col 2, OS 12 197 16.4167 22.9924

Col 3, AH 12 301 25.0833 18.9924

Col 4, AS 12 247 20.5833 32.9924

Source of

Variation SS df MS F P-value

F

critical

Between Groups 1272.67 3 424.222 13.3422 2.5E-06 2.81647

Within Groups 1399 44 31.7955

Total 2671.67 47

Table 3.5 Summary (A): ANOVA single factor for Subjects 01-10 and (B): Subjects 01-12.

Page 69: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

57

OH vs. OS OH vs. AS OS vs. AH AH vs. AS

P value (one-tail) 4.6E-07 4.8E-06 1.3E-05 3E-03

Bonferroni Correction

with adjusted

α = 0.0125

TRUE TRUE TRUE N/A

Table 3.6 Paired t-tests between groups with Bonferroni correction for Subjects 01-10.

OH vs. OS OH vs. AS OS vs. AH AH vs. AS

P value (one-tail) 4.6E-06 6.7E-06 1.5E-05 6.7E-04

Bonferroni Correction

with adjusted

α = 0.0125

TRUE TRUE TRUE N/A

Table 3.7 Paired t-tests between groups with Bonferroni correction for Subjects 01-12.

the greatest amount of decrease in the metric value at 34%. Similar findings are also

found for the other four font styles: Palatino Linotype, Edwardian Script, Bradley Hand

ITC and Freestyle Script. While some variations in metric values manifested due to size

and style effects, paired t-tests between normal vs bold, normal vs italic, and bold vs

italic rendered that it is not possible to claim that these values present statistical

significance in Figure 3.18. All these results indicate that, despite some variability due to

different styles, Word®-simulated samples have certain degree of robustness in metric

values.

Page 70: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

58

Figure 3.17 Sample PD patient handwriting (Data Set 1), along with size-matched

synthetic writing created with Microsoft Word (Data Set 4), and by one of the authors

considered healthy/control. Samples in each row have the same width. For the column

“Control”, each sample was separately generated and then resized to match the sample

width in the same row with “PD Patient Sample” row. Information in this figure

demonstrates that the reduction in the NP-HPP metric is much more pronounced when it

is combined with micrographic effects (PD patient samples) than when it is affected

solely by size (Word®-simulated and Control samples). Portions of the images have been

pixelated to respect subjects' privacy.

Microsoft Word simulated samples can help evaluate micrographia in the absence

of healthy samples: We plot the NP-HPP metric values with respect to the samples at

hand. Here, the samples are chronologically sorted and numbered, such as smaller sample

numbers correspond to older samples; see Figure 3.19. In this figure, in separate panels

for each subject, we denote the metric values corresponding to Data Set 1 as “original”

(blue data points). In each panel, Data Set 3 is also plotted (labeled “resized,” data points

in red) demonstrating metric value changes due to only size effects. Moreover, “baseline”

data points in green are plotted, where these data points are the NP-HPP metric values

Page 71: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

59

corresponding to the best-fit Microsoft Word simulated size-like samples (Data Set 4),

also consistent with Figures 3.17-3.18.

In light of the above discussions, note that in all the panels, for a given sample number,

the corresponding NP-HPP metric is calculated on samples with identical widths but

different origins (either healthy, symptomatic, or word processor simulated). Finally, we

scale the data points in Figure 3.19 by the width of the samples. Since a wider sample is

likely to produce larger NP-HPP metric values, with this scaling, we aim to remove such

effects, normalizing the metric independent of sample width size, see Figure 3.20. This

plot, consistent with Figure 3.19, demonstrates that the NP-HPP metric presents sufficient

differential between healthy and symptomatic samples.

As to how one can interpret the information in Figures 3.19-3.20, we note that the most

critical observation is that the metric values of “resize” and “baseline” data points (size

effect only) are almost always larger than the “original” data points (size and

micrographia effects combined). This positive difference, consistent with Figure 3.7, is

indicative of how the NP-HPP metric captures micrographic effects, regardless of size

effects.

In this section, we investigated how the NP-HPP metric systematically varies in original

samples obtained from subjects before and after their clinically diagnosed PD onset, as

well as in artificially produced samples in various sizes that are free of micrographia. The

aim was to determine whether the proposed NP-HPP metric can discern micrographic

effects in samples regardless of their size changes.

Page 72: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

60

Figure 3.18 Creation of synthetic writing data in Microsoft Word using various fonts and

stroke styles to demonstrate how NP-HPP metric is affected. All samples are treated in

their original sizes and are presented here in smaller size for convenience. In each

column, the sample on top was produced in Microsoft Word and next converted to an

image. The subsequent samples in this column were generated by resizing this image to

match its width with that of the sample in “PD Patient Sample” row. A side study (not

reported here) in which each sample is directly created in Microsoft Word with

appropriate font size and then converted to an image indicated that NP-HPP metric was

only negligibly different. Portions of the images have been pixelated to respect subjects'

privacy.

Based on the analyses conducted, we conclude that indeed the NP-HPP metric shows

consistency and analogy regarding its sensitivity to handwriting and to micrographic

symptoms, and is much less sensitive to only style or size effects, making it an

Page 73: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

61

appropriate and effective metric for studying static handwritten samples for the purposes

of detecting micrographia, irrespective of sample size/style effects.

Figure 3.19 Sample patient handwriting (original data set), along with width-matched

synthetic writing AH (resize of OH), and Microsoft Word with “best-fit” combination of

font style and stroke style (baseline data set). The x-axis represents the signature samples

in chronological order; the y-axis represents the NP-HPP metric value. Almost all the

original data points are below the baseline data showing clear evidence of micrographic

effects. Samples corresponding to the same x value have the same width. In all the

analysis, distortion was avoided by respecting sample aspect ratio.

With further analyses on the selected fonts and stroke styles in Microsoft Word®-

simulated samples, we find out that italic stroke styles and cursive writing styles

Page 74: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

62

produced with Microsoft Word® are the most sensitive styles to the NP-HPP metric. This

finding is interesting, as it aligns with medical information regarding micrographia.

Figure 3.20 Normalized NP-HPP metric values, where each data point in Figure 3.19 is

divided by the width of the corresponding sample. This normalization removes width-size

effects enabling transparent comparison across different samples. Similar observations as

in Figure 3.19 are obtained where most of the resized data points are above the baseline

data, and the original data points are below the baseline data showing clear evidence of

micrographic effects. Samples corresponding to the same x value have the same width. In

all the analysis, distortion was avoided by respecting sample aspect ratio.

It is known that micrographia effects are more prevalent in cursive writing compared to

writing in block letters [20], and this can be explained mainly due to the fact that

Page 75: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

63

micrographia effects become more pronounced in continuous writing as the subject does

not (or cannot) make any pauses in between strokes. Furthermore, for comparative and

control purposes, we investigated ways to create Microsoft Word®-simulated samples of

the corresponding original handwriting samples, with the aim to understand the size

effects of NP-HPP metric in these simulated samples. While some of the style and size

elements somewhat affected the metric of interest in some simulated samples, there were

no statistically significant differences in metric values. Metric values of those samples,

however, differed noticeably from the width-matched original samples of the subjects.

This approach can thus be implemented in future studies to create subject-specific data

with which metric variations can be calculated for comparison with the original data, with

the ultimate objective of detecting micrographic patterns regardless of sample size and

sample variability, and even in the absence of asymptomatic samples.

Overall, the results indicate that the NP-HPP metric can be fully automated, and using

healthy subject data and or word processors, one can generate baseline data that can be

used to compare the metric values with those obtained from original subject-specific data

for unmasking micrographic effects in the original data.

3.3.4 Discussion

The effectiveness of several metrics is presented over actual PD affected subjects’ static

signature samples, including some metrics already studied in the literature regarding size

effects, namely, signature area, and the width and height of certain letters or writing

samples.

Page 76: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

64

Signature samples of subjects including timing of clinical diagnosis have been excerpted

from published work and/or collected through IRB-approved research. It is assumed in

this study that all signatures were produced voluntarily, without any

intentional/unintentional distortions and/or haste induced by the participants. This

reasonable assumption was made owing to consistency observed in handwriting samples.

Another assumption is that subjects were provided sufficient amount of space to produce

signatures in their most natural forms, without having to compress or alter their

handwriting due to lack of space. Clinical diagnosis of PD and onset were also assumed

based on published work and/or subjects’ consented disclosure.

The proof-of-concept methodology presented here serves to establish the developed

metrics and present opportunities for applying them. Specifically, the local metric pixel

density variation can pinpoint shrinkage in samples pertaining to progressive

micrographia, while global metrics directly associated with size effects can help

characterize consistent micrographia. NP-HPP metric has been demonstrated can detect

micrographia regardless of size and style effects on writing samples. These metrics can

thus be used in a combined manner in future studies to study and/or differentiate various

characteristics of micrographia as well.

Furthermore, the proposed metrics are calculated based on static images of subjects’

signatures or handwriting samples. While the nature of using static images does not

consider subjects’ hand kinematics thus differing from dysgraphia analysis, this also

presents a number of opportunities, including use of a large pool of historical data, and

collection of new data in subjects’ natural environments with minimal bias and

Page 77: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

65

infrastructural requirements. Further opportunities lie in fusing proposed metrics with

those of dysgraphia to enhance the characterization of PD symptomatology.

It is noted that factors such as gender [96], age [96], motor asymmetry [97] and

medication [37] as well as environmental conditions under which data is collected can all

impact PD or its characterization [96]. Moreover, pen grip, pen pressure, and even

elements of socio-economic status may affect handwriting, although large scale

normative data is still needed to make stronger claims on factors such as age, gender, and

occupation [61]. This study has been conducted with a sufficient number of subjects

consistent with published work [63], and the statistical power of pixel density variation

metric analysis was ~99%, yet, further research is needed to investigate the sensitivity of

the studied metrics with respect to the aforementioned factors in a broader study

population. On the other hand, it is promising to observe that metric values show trends

across samples that are consistent with findings and expectations in medical practice.

Specifically, one can anticipate that subjects’ conditions after diagnosis, whether or not

they are treated, will not recover fully to their healthy states, partly because PD cannot be

completely cured, and it is known to be progressive. This is consistent in readings of

metric values across subjects, where pre-diagnosis metric values differ from those of

post-diagnosis.

Consistency in metric variations is also promising as it demonstrates that the studied

metrics present a certain degree of robustness to environmental factors unknown to us in

terms of the data recording and transmission process, and subjects’ medication history.

Moreover, based on the data available from the subjects regarding their current

medication, we assume that medication neither improves nor negatively impacts subjects'

Page 78: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

66

handwriting, see Table 3.2 - row 8. Nevertheless, since we do not know all the details of

subjects’ physiological conditions, it was critical to represent this properly in the data

analysis. For this reason, we have classified the existing samples into two main groups,

called “earlier recording” and “recent recording”, instead of “healthy” versus

“symptomatic”. Moreover, a side study, which is not reported in the previous section,

was conducted where data was re-grouped as “the first signature” versus “the last

signature” across Subjects 01-12. In this setting, both paired and independent t-tests on

the density ratio Rij also showed significant differences (p = 0.00067 < 0.001 and p =

0.00077 < 0.001), with the power of the study being ~99%.

In particular, a measure called pixel density variation showed statistically significant

differences (p < 0.05) between two comparison groups of remote signature recordings:

earlier versus recent, based on independent and paired t-tests analyses on a total of 40

signature samples.

Consistency in NP-HPP metric variations is also promising (p = 4.99E-06 for ANOVA

test on Subjects 01-10 with a statistical power of 99.97% and p = 2.5E-06 for ANOVA

test on Subjects 01-12 with a statistical power of 99.98%) as this demonstrates that the

studied metrics present a degree of robustness in light of environmental factors unknown

to us in terms of the data recording and transmission process, and the details of subjects’

medication history. Nevertheless, since we cannot know all details of the subjects’

physiological conditions, this remains to be studied further following the framework

presented here.

Page 79: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

67

3.4 Development of Metrics for Single Character Analysis

Previous section demonstrated that analysis of static writing samples holds promise for

detecting micrographia. Specifically, micrographia has been observed as not only the

global size shrinkage over time as writing becomes progressively smaller, but also as

local morphology distortions associated with pixel density changes within a phrase from

left to right in a sample [76]. Further investigation of local morphology changes on single

character may lead to supplemental metrics for string/signature analysis and also provide

enriched morphological features for single character analysis.

3.4.1 Skeletal Points Metric

Skeletal points are the key characteristics for forming a specific shape including the

cursive alphabet. This was recognized in handwriting analysis in [61]. Indeed, without

skeletal points, one could only generate horizontal and vertical lines. With this

observation, we propose that turning points known as critical points, they are indicators

of how well and how intricately a letter is generated by a subject.

Detection of skeletal points from a digitized handwriting sample can be done following

either one of the two options: 1) fully manual-where an operator selects these points by

clicking on the image, or 2) semi-automatic, where an algorithm will automatically pre-

select several potential skeletal points, then the study personnel can select the final

confirmed skeletal points capturing the exact morphology of characters. We have

performed tests with both options and found that there is no significant changes

difference between them, except slight variation in numerical values. Therefore, we opted

to only report the first option here.

Page 80: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

68

Using the standard cursive alphabet table as baseline, we first identify the critical points

for each letter. They are usually defined as turning points, intersection points, edge points,

etc. in handwriting; see the Standard Cursive Alphabet Table with marked Skeletal Points

(SCATSP) in Figure 3.21.

Notice that by passing ink through defined skeletal points, a specific letter shape can be

created. However, for people with micrographia, some of these points in their cursive

writing may overlap or even be missing due to shrinkage in handwriting. Thus, this

metric was designed to detect these changes by using the SCATSP samples as the

baseline to confirm the skeletal points on tested sample. The metric is calculated as

follows. We mark the skeletal points on a test sample and compute the summation of the

distances between each two skeletal points divided then by the total number of skeletal

points for each character. This allows us to discover the local morphological distortion

and shrinkage for each letter, which can be correlated to micrographia.

In order to mark each critical point more accurately, we thin the analyzed letter first –as

described above and then we apply the following procedure to quantitatively detect the

Critical Point Distance metric:

1. First, check the number N of specific types of critical points for the particular letter

under investigation, as identified in Figure 3.21, see Standard Critical Points for

Cursive Alphabet.

2. Then, mark the position of each critical point for the letter at hand e.g. letter “K” in

Figure 3.22, in MATLAB.

3. Next, calculate the distance between the critical points in a pair-wise fashion, and

obtain the sum of all these distances D = ∑ diNi=1 .

Page 81: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

69

4. Repeat steps 1-3 at least five times, in order to reduce errors due to variability in

selecting the critical points. Then calculate the average of D.

5. Finally, divide the average of D from step 4 by the number of critical points, D/N,

and plot all D/N values for each letter sample at hand.

Figure 3.21 Standard Cursive Alphabet Table [99] with Marked Skeletal Points (SP) in

red dots.

Figure 3.22 Example of implementing the Skeletal (Critical) Points Distance metric on

the capital letter “K”, from the standard Cursive Alphabet table.

Page 82: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

70

3.4.2 Polar Distribution for Angles and Distances Metric

Using what is called the “radar” principle, another metric has been developed to observe

and analyze the polar distribution variations of both angle and distance of pixels in

samples affected by micrographia.

Figure 3.23 Example analyses for letter “c” using Polar Distribution for Angles and

Distances (PDAD) metric. Comparing two samples, the values of mean and standard

deviation for both angle and distance distribution decrease consistently with micrographic

effect. Sample 2 is a symptomatic entry.

The zero degree point is defined as the bottom horizontal boundary of the pre-processed

sample image, and the radar will detect the angle distribution for existing ink deposits

Page 83: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

71

from 0 to 90 degrees (the left vertical boundary of the metric sample image). Similarly,

the origin for calculating the distance is defined as the left bottom corner of the pre-

processed sample image. The ‘radar scan’ from the origin to the distal boundary is the

distance distribution in polar coordinates. In order to quantify the changes in the

generated histograms, we compare their mean and standard deviation values. Refer to the

example analysis in Figure 3.23.

3.5 Evaluation of Metrics for Single Character Analysis

3.5.1 Subjects and Study Samples

To obtain handwriting samples for our study, subjects from qualified study groups were

recruited: Symptomatic subjects (n = 12) with Parkinson’s disease and self-reported

micrographic handwriting issues as disclosed in their subject questionnaires. Nine of

these subjects provided their historical signature samples under #13-04-03 and 3 of

subjects’ samples were downloaded directly from published literature without the need

for IRB clearance due to the publicly available nature of the data; see Figures 3.8 & 3.9.

As noted, the PD affected subjects with self-reported micrographia provided their

historical signature samples with reported chronology. As a standard procedure,

characters that are analyzed are the first letters in the first name and last name, and

denoted as the “beginning letter” and the “end letter”4. The reason for this choice is as

follows. It is known that micrographia causes shrinkage in writing in either one of the

two ways, namely, consistent micrographia (consistent scaling of size over the space) and

4 This was the most appropriate choice given the samples at hand. When appropriate, one can also select the

last character of last name in the samples as long as these characters are easily distinguishable.

Page 84: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

72

progressive micrographia (shrinkage of handwriting from left to right). Therefore, it is of

interest how letters in a sample vary with respect to each other.

3.5.2 Pre-processing

Prior to calculating metrics for the static writing images, initial digital pre-processing was

performed on the samples in their original sizes by following the methods suggested in

the security literature [87-93]. There are four required steps are same as the steps we

introduced in Section 3.3.2. After these pre-processing steps were executed in MATLAB,

the noise and irrelevant information have been extracted and the processed sample

images are ready for further metric implementation.

3.5.3 Results

The above-described metrics for micrographic handwriting are applied to the selected

letters for the first sample and the last sample in every subject. Then, the metric results

are systematically compared as percentage changes within each subject: the “early

recordings” (first sample) vs. “recent recordings” (last sample), as well as “beginning

letters” vs. “end letters”, under the assumption that “early recordings” correspond to less

micrographic effects and “recent recordings” are strongly affected by micrographia, as

well as “end letters” are affected more by progressive micrographia than consistent

micrographia compared to “beginning letters”, which are less affected by cramped

micrographic writing.

Results indeed show correlation with symptom variations: the SP metric value, and the

mean and standard deviation values in PDAD metric for angle distribution, as well as the

Page 85: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

73

mean and standard deviation value in PDAD metric for distance distribution decreased

consistently from early to recent recordings. By observing the results summarized on

Table 3.8, both SP and PDAD metrics show noticeable declining trend (negative

percentage changes) in the samples from the 12 subjects in both “beginning (letters)” and

“end (letters)” columns.

Moreover, when comparing the results for the two columns of “beginning letter” and

“end letter” from each subject on Table 3.8, a measureable difference is also found in

most subjects across both SP and PSAD metric results: beginning letters show less of

micrographic effects (less progressive micrographia). However, Subjects 11 & 12 show

similar decline in percentage values for both the “beginning” and “end” letters on SP

metric and mean values for both the distance and angle of PDAD metric, which may shed

light on those two subjects’ handwriting, may present consistent micrographia. These

metrics have the potential to be used as standards to identify progressive micrographia

from consistent micrographia.

Paired t-test and independent t-test are performed on the actual values of the metrics, not

the percentages. The tests are conducted on the selected letters, which are the first letter

from signature samples of Subjects 01-12. This corresponds to analyzing 126 characters,

with statistical results as displayed on Table 3.9. Except the standard deviation values of

PDAD-Angle metric, all the other metrics present statistically significant differences.

These observations indicate that SP and PSAD metrics have the potential to identify

micrographic patterns from static handwriting samples.

Page 86: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

74

Table 3.8 Metrics results from Subjects 01-12 by implementing skeletal point metric and

polar distribution for angles and distances metric.

*The Percentage Change values were calculated using the resulting metric values from the selected letters

in the first signature in “early recordings” and the selected letters in the last signature in “recent

recordings”. All values shown on Table 2 are negative, signifying a consistent declining trend. For subjects

1-12, the selected “Beginning” letters are “L”, “D”,“B”, “G”, “B”, “F”, “M”, “J”, “A”, “G”, “P” and

“R” ; for subjects 1-12, the selected “End “letters are “G”, “D”,“R”. “M”, “B”, “L”,”W” “B”, “F”,

“E”, “A”, and “M”.

Table 3.9 Paired and independent t-tests results on metric values obtained from Subject

01-12 by implementing developed single character metrics.

Page 87: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

75

3.5.4 Discussion

Improvement by treatment for PD subjects can be dramatic and rapid in some cases.

However, subtle changes can also occur which may not be easily detected via

conventional diagnostic rating scales commonly used by neurologists5. In clinical settings,

the lack of detection of these subtle morphological changes shown on character tests can

lead to erroneous adjustments to or cessation of treatment, and can hinder patient

compliance. From the patients’ perspective, continuing ineffective treatment can delay

improvement, or even worsen the condition. Analysis of written characters could

improve the understanding of symptom variations, and it has been already widely used by

clinicians.

Although we only have tested 12 subjects and 126 writing character samples, the

statistical results are promising for both skeletal points metric and polar distribution for

angles and distances metric. Further expanded study population is suggested to

investigate these metric results and their correlation with symptomatic subjects’

evaluation results by clinical rating scale. Also, both metrics have been designed

specifically to capture the morphological changes with micrographic handwriting; we

believe they can be easily adopted in clinical test.

3.6 Conclusion

In line with the above rationale illustrated in Chapter 2, using an analytical engineering

approach, the handwriting of people with PD has been studied in several observable and

5 Currently, symptoms are usually assessed by a neurologist’s observation of several tasks in a clinical

setting, using The Unified Parkinson Disease Rating Scale (UPDRS) for PD [17].

Page 88: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

76

measurable ways. Several drawbacks of current approaches mentioned in Section 2.1 lead

to limited usage and reliability in the clinic. Analysis of historical static images has the

potential for monitoring micrographia symptom variations on PD subjects with minimal

experiment bias, and can be easily implemented in larger populations, because

micrographia exhibit shapes and curvatures beyond those observed in asymptomatic

writing.

In this chapter, we introduce a more effective assessment method for motor symptoms

found in PD, namely an automated, objective analysis of static handwriting samples,

which can be easily generated under subjects’ natural and comfortable conditions

Furthermore, calculation of the these metric can be easily performed using static images

of handwriting samples. While this framework does not incorporate subjects’ hand

kinematics, thereby differing from published studies on dysgraphia analysis, use of NP-

HPP metric yet presents a number of opportunities, such as use of a large pool of

historical data, and collection of new data in subjects’ natural environments with minimal

bias and infrastructural requirements. Other opportunities are anticipated in settings

where both static and kinematic analysis tools can be tailored together to even further

strengthen statistical inferences of micrographia, and hence PD symptomatology.

Currently, the set of metrics for monitoring micrographic handwriting presented in this

dissertation are effectively executed by importing sample images via MATLAB routines.

This analytical study provides quantitative, consistent and replicable results. In the future,

this expertise can be easily transformed into a graphical user interface, which can be

adapted by medical staff with minimal training. Such utility hence has the potential to

develop into a readily administered computerized toolkit to support clinical practice,

Page 89: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 3. Development and Evaluation of Metrics for Micrographic Handwriting

77

supplement digital tablet data, provide a data-based archiving mechanism, be applied in

controlled experiments to characterize micrographia, and assess relevant changes that

may result from treatments and interventions intended to mitigate PD symptoms.

Finally, a side work of this study not reported here has not shown that writing samples for

such tests are significantly affected by typical data transferal format (e.g., fax, digital

photography), so symptomatic individuals could be instructed to send samples to an

appropriate expert for analysis via one of several easy ways.

Page 90: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

78

Chapter 4

Development and Evaluation of Metrics for

Tremulous Handwriting

4.1 Introduction

Essential Tremor (ET) manifests itself in people in the form of uncontrolled shaking,

most notably during skeletal muscle usage. This disorder, which is currently incurable,

can have a range of effects on motor and non-motor symptoms. Severity of ET is

commonly assessed through a set of clinical tests, using The Essential Tremor Rating

Assessment Scale (TETRAS); one of such tasks involves handwriting/drawing [3, 13].

However, the subjective analyses of which limit resolution and availability.

Since tremor presents as distorted jitters of writing segments beyond typical construction

of characters, it has the potential to be quantified, and thus objectively and automatically

assessed. To address this, we are pursuing objective and computerized metrics to assess

and quantify the extent of tremor in pre-existing static writing samples. In this

dissertation, we tested these metrics by comparing unaffected writing samples with

writing affected by artificially-induced tremor.

Page 91: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

79

This chapter starts with introducing the quantitative metric design for single character

analysis in Section 4.2, and then is followed by introducing artificially tremor generated

by Electrical Muscle Stimulation (EMS) device in Section 4.3. Section 4.4 presents the

experimental protocol and the outcomes of the human subject experiments, which are

further concluded in Section 4.5.

4.2 Development of Metrics for Single Character Analysis

Like another common movement disorder we discussed in Chapter 3, handwriting is one

of motor skills has been neurologically affected and degenerated for people with ET.

There is evidence that the deterioration of handwriting is associated with some of the

earliest symptoms. Key differences between PD and ET make this goal feasible.

Specifically, ET patients exhibit handwriting with oscillations and usually produce

larger-sized writing samples [27, 53-54]. However, tremulous strokes are less common in

PD. In a typical clinical setting, tremor is measured through deviations from drawings of

simple geometries,; however, these tasks are not naturally linked to people’s handwriting

features. Patients may hide their real conditions through perform unfamiliar tasks. Here,

we propose and introduce a set of objective, quantifiable and automated metrics for

studying tremor in handwriting.

4.2.1 Critical Point Metric

The notion of critical points for handwritten letters has been defined and presented by

Chakravarthy, et al. [95]. These points are characterized by the number and direction of

Page 92: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

80

line segments that connect them as seen in Figure 4.1. Although the cited work is not

focused on neurological disorders, it has the potential to be applied here.

Definition of 9 Types of Critical Points

Dot Point (D) 1 situation, no ink pixel exists in the adjacent zone*

End Point (E) 8 situations, only 1 ink pixel exists in the adjacent zone

Bump Point (B) 8 situations, 2 ink pixels exist in the adjacent zone (45

degree)

Cusp Point (C) 8 situations, 2 ink pixels exist in the adjacent zone (135

degree)

Perpendicular Point (P)

24 situations, 2 or 3 ink pixels exist in the adjacent zone (90 degree)

T Point (T) 8 situations, 3 ink pixels exist in the adjacent zone

Cross Point (X) 2 situations, 4 ink pixels exist in the adjacent zone

Y Point (Y) 4 situations, 3 ink pixels exist in the adjacent zone

Star Point (S) several situations, >=5 ink pixels exist in the adjacent zone

*adjacent zone is defined as the 8 neighboring pixels of a text pixel

Table 4.1 Nine different critical points have been identified and defined

Page 93: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

81

We identified and categorized 9 types of geometric critical points in handwriting at the

pixel level. For this, the relationship between a pixel and its 8 neighbors has been

explored and described; see details on Table 4.1. These pixels describe the locations

within a character in which lines meet, intersect, and cross under different conditions.

These conditions can readily be converted to quantitative descriptions of pixel

arrangements (Figure 4.1). We explored the usage of these critical points to identify

tremor in handwriting. We hypothesize that with more severe tremor, the occurrence of

these defined critical points will increase by showing wiggling on writing characters.

In order to automatically identify and mark these critical points on the sample image in

MATLAB, every pixel within the ink trajectory of the character has been screened one by

one to find out the specific relationship among the test point and its 8 adjacent

neighboring pixels. If the test pixel meets any one of the critical points definition, the

algorithm will selected this pixel and marked it with different color symbols on character.

After screening procedure, the spatial distribution of critical points will show on the

writing sample, and with their identified locations, numbers and types, which can then be

measured and computed automatically. For example in Figure 4.1, we analyzed writing

samples (Letter “C”) of a healthy individual, with and without artificially induced tremor

at hand, and noted where and which critical points arose. We found that with more severe

tremor, the occurrence of critical points would increase, as they are associated with

reciprocating motion. These preliminary results support our hypothesis, see details in

Section 4.4.

Page 94: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

82

Figure 4.1 (A) Critical Points can be pixelated (B) and thus identified. (C) For ET, we

have begun to explore Critical Points’ spatial distributions of pixels and show that their

occurrence increases with tremor.

4.2.2 Segmental Curvature Metric

The line segments between endpoints and/or intersection points can be used for

segmental curvature analysis with the premise that under ideal conditions such segments

are straight or at least smooth with a relatively low cumulative value of curvature, see an

example of circle and its curvature analysis in Figure 4.2. We hypothesized that with the

increasing tremor severity, the curvature analysis will show more varied results among

each pixel within the trajectory.

Inspired by [18], we can similarly identify the endpoints and intersection points in

MATLAB automatically. Once a letter is split into segments between these points, here

Page 95: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

83

we use a piecewise cubic spline interpolation to create splines passing through the pixel

points from an endpoint to an intersection point, and between intersection points. The

splines can then be used to quantify local curvature by using the Eq. 4.1.

Figure 4.2 (A) Original Circle with radius of 1, (B) using spline interpolation to pass

those blue dots marked points and (C) calculate curvature for every pixel in the fitting

curve of the test circle, and the plot shows stable numerical results for curvature analysis

(around 1).

Given a plane curve as y=f(x), the curvature 𝜌 can be calculated as

ρ =y''

(1+y'2)3/2 (4.1)

where 𝑦′and 𝑦′′ respectively are the first and second derivatives of y.

Page 96: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

84

Steps to analyze this data include: (a) stitching of segmental curvature on a single plot,

and (b) analysis of curvature distribution by comparing the corresponding mean and

standard deviation values.

Figure 4.3 Flowchart of curvature analysis for a letter affected by artificially induced

tremor. Preliminary results show that tremor alters the distribution of curvature with

increased mean and standard deviation.

4.2.3 Horizontal and Vertical Projection Profile Peak Metric

Horizontal and vertical projection profiles (HPP and VPP) describe the ink distribution of

a sample along respected directions [83], which have been explained in Section 3.2 with

full of details.

Once the HPP and VPP curves are available, the number of peaks in both curves can be

counted in MATLAB. In some respects, this count is related to the intensity of unique

stroke movements, and is proportional to the presence of multiple gaps between such

Page 97: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

85

strokes. We propose that such strokes and space between them are altered when ET

symptoms are present. The rationale behind this is that this metric has already showed

promise in the analysis of micrographia in Chapter 3.

Metric Value =The total number of peaks in both HPP and VPP

(Height+Width) (4.2)

Different from our previous metric calculation presented in Chapter 3, the metric here

calculates the ratio of total peaks in both HPP and VPP with respect to the summation of

height and width at this time on tremor-affected samples, since tremor presented in ET

may affect both directional pixel distributions.

4.3 Artificially Induced Tremor (AIT)

We have developed and adopted a safe way of inducing comparable hand and wrist

tremors via a commercially available Electrical Muscle Stimulation (EMS) device

(ProM-555-Promed Specialties, Huntingdon Valley, PA).

EMS device is commonly used in physical therapy and muscle rehabilitation. Briefly, in

EMS a micro direct current (DC) pulse is sent between two easily removable adhesive

and conductive surface electrodes attached on a patient’s skin in the target region. By

placing the surface electrodes in relation to specified muscle fibers and varying the pulse

parameters, repeatable and representative muscle contractions can be induced in subjects.

Specifically, these parameters include the pulse width (μs), amplitude (mA), and

frequency range (Hz), which can be adjusted to emulate typical tremor frequencies found

in ET affected individuals. Under expert monitoring and testing, the settings of ProM-555

used in our experiments have been established, when the surface electrodes are placed in

Page 98: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

86

appropriate areas of the inner forearm as in Figure 4.4, a finger and wrist oscillation can

be induced to emulate movements caused by tremor.

Parameter settings of ProM-555 used in our experiments:

Pulse Rate: 8-Hz

Pulse Width: 140µs

Pulse Amplitude: 20~30 (depends on the person & the battery)

Figure 4.4 (A) Model of EMS machine (ProM-555) used in this work [100], (B)

specified placement of four surface electrodes in the anterior forearm muscle image [101]

marked by red boxes.

We have two channels on our ProM-555 and we noticed better results when we placed

those pads on the marked muscle groups in the palm of the hand and inner arm. This

Page 99: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

87

made subject harder to resist it and simulate hand tremor during the writing like ET

patients.

This method was reviewed and approved by Northeastern University Institutional Review

Board, IRB #13-04-03, and was used to generate artificial tremor in handwriting of

asymptomatic “healthy” subjects without neurological disorders, allowing direct

comparison of affected and unaffected writing using our objective assessment metrics.

Figure 4.5 (A) Wacom Intuos 4 Pen Tablet with Inking Pen [102]; (B) the active surface

tab presents an assortment of tasks in NeuroGlyphics software [103]; (C) A example of

static handwriting data sheet collected during experiment; and (D) Dynamic data

analyzed by NeuroGlyphics software and show an example of the Fast Fourier

Transforms (FFT) result which displayed AIT system rendered a dominant vibration

frequency at 12 Hz during handwriting experiment.

In terms of data collection, effects of induced hand tremor by EMS device can be

recorded on data sheets, or dynamically using a Wacom Intuos4 tablet with an ink pen on

the data sheet. Here, NeuroGlyphics computer software [103], which has been used by

Page 100: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

88

other studies for hand tremor analysis, is adopted for dynamic data observation and

analysis; see an example in Figure 4.5.

4.4 Evaluation of Metrics for Single Character Analysis

4.4.1 Subjects and Study Samples

To obtain handwriting samples for our tremor study, qualified healthy subjects were

selected with no self-reported cardiac and neurological issues as disclosed in their subject

questionnaires under IRB #13-04-03 at Northeastern University. Prior to the experiments,

all participants signed a written informed consent. The experiments are carried out with

12 subjects among whom there were 3 females and 9 males with ages ranging from 18 to

29 years.

Figure 4.6 Examples of unaffected (first row) and AIT affected (second row) sample in

part (A) to (C) collected from three individuals of the healthy subjects.

Page 101: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

89

As we do not intend to compare the severity of writing-based symptoms across

individuals, age-matching was neither required nor pursued. Healthy subjects (n = 12)

who produced handwriting samples with and without artificially induced tremor (AIT)

are studied in this dissertation. A sample set is depicted in Figure 4.6.

4.4.2 Pre-processing

Besides the four basic preprocessing steps to eliminate unessential information and noise

(see details in Section 3.3.2), an additional methodical pre-processing step, thinning, was

used for metrics when quantifying tremor in handwriting. This procedure removes some

of the pixels so that the penmanship of the sample scales to a minimally connected stroke

and is hence prepared for further comparison; see an example of writing sample after

thinning procedure in Figure 4.7.

Figure 4.7 Examples of a sample image (bottom image) after thinning processing step,

the top image is the preprocessed sample image showed in Figure 3.10 after “step (d)”.

The original sample image is from [59].

4.4.3 Results

As noted, a set of healthy subjects participated in the study, and their unaffected and AIT-

affected alphabet handwriting samples were studied by implementing the described three

Page 102: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

90

metrics for tremulous handwriting, and all metric values from Subjects 01-12 were used

to perform statistical analysis.

Metric Results for Tremulous Writing Analysis between Two Sample

Groups

Metric

Name

Critical Points

(CP) Metric

Segmental

Curvature (SC)

Metric

HPP &VPP

Peak (P)

Metric

Subject# Result Mean St. Dev. Result

*Average Percentage

Change Over The Whole

Alphabet Samples

(+ %)

1 24.18 28.66 110.22 26.49

2 35.48 34.46 82.43 31.50

3 33.60 32.80 109.78 29.58

4 30.86 37.47 119.37 25.19

5 29.05 31.85 128.55 24.88

6 22.54 38.95 94.35 24.42

7 28.18 40.92 132.45 26.96

8 21.19 34.93 71.95 24.20

9 20.55 37.87 62.45 18.04

10 30.90 32.39 65.59 22.27

11 29.30 30.88 78.02 31.47

12 22.69 26.43 94.49 19.74

Mean 27.38 33.97 95.80 25.40

Table 4.2 Averaged metric results for tremulous writing analysis between “EMS-On “and

“EMS-Off” sample groups.

*The Percentage Change values were calculated using the resulting metric values from

the cursive alphabet letters in “EMS-ON” writing samples and “EMS-OFF” writing

samples from cursive alphabet recording from all 12 subjects, and the Averaged

Percentage Change values were calculating by finding the mean change vaule among 26

alphabet letters. All values shown on Table 3 are positive, signifying a consistent

showing the effect of tremor caused by EMS on cursive alphabet letters.

Page 103: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

91

On Table 4.2, key findings of the healthy subject set were those with tremor, the

occurrences of critical points consistently and significantly increased with the averaged

percentage among 12 subjects as 27.38%, as they are associated with reciprocating

motion. This motion under tremor resulted in an increase in both mean and standard

deviation values of overall segment curvature measures as well, with the averaged

percentages for all participants as 33.97% and 95.80% separately. The resulting values

for peak numbers in HPP and VPP metric also increased in samples with tremor in

averaged 25.40% from Subject 01-12.

Further, results bolstered by statistical analysis (both paired and independent t-tests p

values are less than 10-6

) show significant differences between writing effects of the

unaffected sample group and the AIT-affected group by both paired and independent t-

tests, as seen on Table 4.3.

Statistical Analysis of Metric Values for All Tremulous Writing

Samples from all participants between Two Sample Groups

Metric

Name CP Only SC (Mean) Only SC (St. Dev.) Only P Only

Paired

t-test

Independent

t-test

+ <1E-06

Table 4.3 Paired- and independent- tests results for metric values from Subject 01-12

between “EMS-On “and “EMS-Off” sample groups.

+When p value for t-test is extremely small which is less than 10-6

, we will only report it

as “<1E-06”.

Page 104: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

92

4.4.4 Discussion

The analyses presented in this chapter show that all the metrics appear to clearly

distinguish tremulous writing from unaffected writing samples. Through observing the

averaged percentage changes over the whole alphabet character samples across Subjects

1-12, we also found that although the EMS settings are consistent, the average percentage

changes of each metric are slightly varied among the 12 subjects on Table 4.2. The

reasons for this are possibly due to different body (muscle) weights, different tolerance to

EMS system, penmanship, as well as whether or not the subjects put any voluntary forces

against the tremors. While all these potential questions need to be addressed in future

studies, favorable statistical results clearly demonstrate that the proposed metrics are

indeed robust against these factors, and have the potential for use in a broad range of

conditions.

4.5 Conclusion

The limitations of current clinical ET symptom assessment methods are linked with the

notion that they are not particularly natural; subjects are fully aware of the test

implications by incorporating with specified equipment.

In this chapter, we introduce a set of metrics that can be used as an assessment tool to

objectively analyze static handwriting samples affected by hand tremors. Such samples

can be easily generated under natural and comfortable conditions, and moreover the

process of calculating the metrics can be automated. The results showed that our metrics

appear to clearly distinguish tremulous writing from unaffected writing.

Page 105: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 4. Development and Evaluation of Metrics for Tremulous Handwriting

93

Two key considerations are: (1) while metrics are still relative, and require comparison to

be effective, currently it appears possible to generate more absolute scales based on more

extensive datasets; (2) while the test itself can be performed on samples generated in any

setting, including at home, we are not suggesting that the analyses be performed by

symptomatic subjects themselves.

Finally, our supplemental work not reported in this dissertation has shown that writing

samples for such tests are not significantly affected by typical data transferal methods

(e.g., reproduction, fax, digital photography), so symptomatic subjects could be instructed

to send samples to an appropriate expert for analysis via one of several easy ways.

Further directions include consideration of a larger set of samples, efforts in creating

metrics that can be assessed in absolute scales, and integration of the metrics to clinical

practice as a toolkit for assisting diagnosis.

Page 106: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Associated with Parkinson’s Disease 94

Chapter 5

Quantitative Assessment of a Therapeutic

Exercise in Temporally Mitigating

Micrographia Associated with Parkinson’s

Disease

5.1. Introduction of Amplified Air Writing (AAW)

Currently there is no cure for PD. Accordingly; practiced treatments are focused on

symptom mitigation. These include deep brain stimulation (DBS), medication, and

therapeutic exercises. One of the prevalent forms of the latter is the LSVT BIG® program

[103], in which PD patients participate in repetitive full-body exercises of high intensity

and increasing complexity. The general notion behind this therapy is that the neural

degeneration in PD causes deficiency in speed-amplitude regulation, which subsequently

leads to a ‘shrinkage’ or diminishment of various movements [62]. Such movements

include physical motor tasks, but also muscular actions that control speech [65-66].

Intentional repetitive, high-amplitude actions using the whole body, with external

Page 107: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 95

feedback, have been shown to partially restore regulation in the short term in physical

movements [67, 69]. Similarly, an analogous therapy known as LSVT LOUD® has been

shown to mitigate vocal symptoms in the short and long term [70].

Several PD patients have explored task-specific amplitude-based therapy for writing. That

is to say, instead of pursuing full-body exercises with external feedback, subjects have

shown that exaggerated arm movement, prior to engaging in a writing task, can improve

micrographia symptoms [71]. In this training, while standing up or in a seated position,

one moves his/her dominant arm in the air with large strokes and models writing sentences,

numbers, and the alphabet; hence the term Amplified Air Writing (AAW). Quasi-external

feedback in exercise comes from standard instructions, e.g., to extend arms fully at the

elbow.

The efficacy of AAW has been noted anecdotally in several clinics and physical therapy

programs around the world [72-73]. However, to the best of our knowledge, no scientific

study on AAW has been reported in the literature. One reason for this is a lack of

quantifiable metrics for micrographia. Currently UPDRS is widely used for self- and

clinical-evaluation, in which symptoms presented on handwriting are reported on a scale

of 0 (normal) to 4 (severe) [17]. This is valuable for communication and diagnosis, but a

clear opportunity for more quantitative assessment exists. To address this, we describe

relevant and replicable metrics based on quantitative analysis for investigating

micrographia progression changes of subjects (presented in Chapter 3) through their

writing samples collected from short-term home-based therapeutic exercises.

In this chapter, a few metrics discussed previously in Chapter 3 are used to measure

changes in symptomatic handwriting, after AAW exercises. In section 5.2, we introduce

Page 108: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 96

the AAW protocol we designed for investigaiting the efficacy of this type therapeutic

exercise for improving PD affected subjects’ handwriting performance, detailed

quantitative assessment methods are discussed in Section 5.3. Experiment results from PD

subjects are presented in Section 5.4, and a brief summary and discussion concludes this

chapter in Section 5.5.

5.2 AAW protocol

After obtaining written consent, participants are asked to practice the following protocol 3

times per day, targeting morning, noon and evening of the day, 7 days per week, for a total

of 2 weeks.

For each session, subjects were asked to hold a ‘large’ implement (e.g., TV remote,

spatula, etc.) like a pen in a self-selected grip posture using the dominant hand, stretched

the arm on the dominant side and repeated the following set of exercises with giant strokes

(at least 2 feet) in the air in the vertical direction at their own pace:

(1) write a sentence- “The quick brown fox jumps over the lazy dog.”,

(2) compose the cursive lower case alphabet,

(3) write the numbers from 0 to 9,

(4) print subject’s name, and then make a signature.

Subjects need to complete a total of 3 sets of exercise back to back, in all three time

sessions: morning, noon and evening. AAW exercise (3 sessions) takes about 30 to 45

minutes per day total.

Page 109: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 97

Figure 5.1 (A) Snapshot from a video filmed AAW exercise had been practiced by Mrs.

Beverly Ribaudo (a PD patient), she posted her pre- and post- AAW exercise handwriting

image (B), which was downloaded directly from parkinsonHumor.com in its original

scale without any further image processing. Images are extracted from [71].

Page 110: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 98

5.3 Quantitative Assessment Method

5.3.1 Subjects

The experiment was approved by the Institutional Review Board at Northeastern

University (IRB Protocol Number: # 13-04-03). Experiments are announced through

flyers that were distributed through local Parkinson support groups. The experiments

were open to volunteering PD subjects who have handwriting issues with sufficient

literacy level in English. $50 compensation was paid to subjects for their two weeks take

home participation in the experiments. Prior to the experiments, a subject questionnaire

and written consent were obtained before the start of the AAW exercise program, and

subjects were excluded from the study if there were no self-reported micrographic

handwriting issues disclosed in this questionnaire. The AAW exercise study are included

with 14 people with Parkinson’s disease from the US and Australia, additional details are

reported on Table 5.1.

Based on their self-reported UPDRS score for handwriting issues, the fourteen subjects

were separated into two functional sets:

Subjects 1 to 8, describing their handwriting as slightly slow or small (UPDRS =

1-2), named as mildly affected (UPDRS) group;

and Subjects 9 to 14, describing their handwriting as moderately affected

(UPDRS = 3) with some illegible words, named as moderately affected (UPDRS)

group.

Page 111: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 99

Table 5.1 Demographic and Clinical Characteristics of Study Participants.

According to their returned data collection sheets, Subject 1-4, 6-12 and 14 complete

two-week AAW exercise, two of subjects completed only part of it (Subject 5 completed

7 days and Subject 13 completed 10 days separately). Since the quality of returned data

collection sheets from Subject 14 is poor for analysis, we ignore this subject’s all samples

for analysis in this chapter. For eleven subjects who completed the two-week AAW

exercise, all 14 days results are used here for metric evaluation, while for the rest two

subjects (Subjects 5 & 13), only completed date samples are used for metric calculation

in this chapter. Totally, 13 of 14 subjects’ writing samples are included here.

Page 112: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 100

Some other analyses are also performed to investigate any differences from AAW,

between these two sets with different UPDRS severities, and results from 13 PD subjects

are presented in Section 5.4.

5.3.2 Data Acquisition

Experimental results were recorded on paper by using data collection sheets. On the

sheets, subjects were asked to write the same handwriting contents as those practiced in

AAW, but on a typical handwriting scale. Subjects performed this task twice a day:

before the second AAW exercise around noon (‘pre-AAW’), and immediately after the

second AAW (‘post-AAW’). A sample of pre-AAW and post-AAW data collection

sheets upon written permission of a subject is shown in Figure. 5.2 and Figure 5.3.

5.3.3 Evaluation Metrics

To prepare for analysis, all handwritten data sheets were scanned into electronic form

using identical quality settings. Next, electronic data was pre-processed to capture

relevant writing characteristics and eliminate nonessential marks from the handwriting

sample foundation, see more in Section 3.3.2. Samples were then evaluated by the

following metrics presented in Chapter 3 to detect micrographia [76, 79]: Area (with

Height and Width), Ink Deposit, Pixel Density Variation, and Word Width in relation to

Space Width.

Page 113: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 101

Figure 5.2 Typical writing sample from Subjects with mild UPDRS group: Pre-AAW

(top) vs. Post-AAW (bottom).

Page 114: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 102

Figure 5.3 Typical writing sample from Subjects with moderate UPDRS group: Pre-

AAW (top) vs. Post-AAW (bottom).

Page 115: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 103

Four global metrics are used to quantify the size change(s) in handwriting; one of the

critical indicators of the changes induced by AAW. Here, a fundamental Area metric is

applied to samples by calculating the product of the maximum vertical stroke and total

horizontal length. Also, the number of ink pixels forming the ink trajectory created in each

handwriting sample is measured by the Ink Deposit metric. Word Width calculates the

total horizontal span of the nine words within the sentence sample, and this metric is

compared to Space Width, which quantifies the total space placed between these nine

words within the sentence sample. The two latter metrics differentiate between changes in

writing size itself, or formation of handwriting, and are thus considered together as

descriptors of improvement.

One local Metric- Pixel Density Variation- is an indicator of changes induced by AAW

exercise on the progressive shrinkage of writing in micrographia. The handwriting

samples are separated into cells according to distinct strings, such as words or names [19].

Then these word-string cells are split into several sub-cells of identical width for each cell.

Sub-cell height is determined by upper and lower boundaries of ink deposit, and this

multiplied by width between the lateral boundaries that gives the cell area, in pixels. The

quantity of ink pixels was measured for each sub-cell, and divided by area to calculate

pixel density. The linear fit of the associated density plot from sub-cell to sub-cell is

reported as Density Variation for that handwriting sample. If the linear fit has a positive

slope, it indicates that the density in this sample is, on average, increasing from left to

right.

Using these methods, each subject generated 22 total metric results: seven for the sentence,

five each for cursive alphabet, printed name, and signature.

Page 116: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 104

5.3.4 Metric Analysis

Differences between pre-AAW and post-AAW were determined for each metric.

Moreover, to compare these difference values within and across subjects, normalization

was applied to metric data computed as:

Metric Normalization =post-AAW metric value - pre-AAW metric Value

pre-AAW metric Value (5.1)

This was done to represent relative changes between 0 and 1.

Unit-less difference value data sets for each metric were then compared to detect any

handwriting changes after the AAW exercises. Based on clinical convention, increases in

global metrics (i.e., positive difference value), or lower Density Variation values (i.e.,

negative difference value) were regarded as improvements. In the specific case of

sentences, a positive difference value in Word Width was only considered an

improvement if it were greater than or equal to the difference value for the corresponding

Space Width. Comparisons between groups (pre-AAW group vs. post-AAW group) were

conducted with paired t-tests.

5.4 Test Results from PD Subjects

Table 5.2 shows resulting difference values for all subjects and metrics, averaged over

returned sample days: 14 days for Subjects 1-4 and 6-12; 7 days for Subject 5; and 10

days for Subject 13 respectively. Note that in general, these values are, for the most part,

positive and negative for global and local metrics, respectively. Moreover, on average,

change in Word Width is positive, and greater than the change in Space Width. These

Page 117: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 105

indicate, generally, that AAW reduces micrographic symptoms and enlarges handwriting

samples in this study.

Page 118: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 106

Table 5.2 Resulting difference values for all subjects from 1 to 13 and metrics results

evaluation: over 14 days, the underline marked results do not show improvement

assessed by metrics.

Page 119: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 107

Figure 5.4 Non-improvement percentage of tasks summary for Subjects 1-6: over 14 days.

(A) Top: mild UPDRS group results and (B) Bottom: moderate UPDRS group results.

To go beyond this overall assessment, we next examined the percentage of tasks that

were not improved with AAW, as measured by the metrics (and underlined on the Table).

For Subjects 1-8, mild UPDRS group, the non-improvement percentages are 4.5%, 13.6%,

13.6%, 13.6%, 22.7%, 13.6%, 4.5% and 9.1%; for Subject 9-13, moderate UPDRS group,

the non-improvement percentages are 18.2%, 54.5%, 45.5%, 31.8% and 40.9%;

respectively in Figure 5.4.

Page 120: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 108

Page 121: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 109

Table 5.3 Resulting difference values for all subjects from 1 to 13 and metrics results

calculation: day-to-day at least half the time, the underline marked results do not show

improvement assessed by metrics.

Page 122: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 110

Figure 5.5 Non-improvement percentage of tasks summary for Subjects 1-13: day-to-day

at least half the time. (A) Top: mild UPDRS group results and (B) Bottom: moderate

UPDRS group results.

However, since several variance values are quite large, another analysis was performed

which simply counted the number of days that a metric improved (Table 5.3); if this

number was at least half of total number of days for each subject, the metric was counted

as non-improved. Looking at the data in this way, the non-improved percentages for

Subjects 1-8, mild UPDRS group, are 13.6%, 4.5%, 22.7%, 18.2%, 18.2%, 13.6%, 13.6%

Page 123: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 111

and 13.6%; for Subjects 9-13, moderate UPDRS group, are 50%, 72.7%, 59.1%, 45.5%

and 50%; respectively showed in Figure 5.5.

Table 5.4 Statistical Analysis: p values for metrics used to quantify improvement. Top:

mildly affected group, middle: moderately affected group, and bottom: all subjects group.

Page 124: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 112

Results of paired t-tests (p < 0.05) for each metric are shown on Table 5.4. Examining

across moderately affected subjects, 15 of 22 metrics do not show statistical significance,

while for Subjects 1-8 (mildly affected) in the same functional grouping, only 5 of these

22 metrics fail.

5.5 Discussion and Conclusion

After a short-term self-administrated AAW exercise protocol, metric-based results show

significant improvement in handwriting performance for PD subjects with mild

micrographia (slightly slow and small handwriting) through statistical analysis. In

addition, comparing the relationship between different tasks and corresponding statistical

analysis results from each metric, we found the effect of AAW is less apparent in the

printed name than the other three tasks, see details in Table 5.5.

Table 5.5 Relationship between different tasks and corresponding statistical analysis

results from each metric.

Page 125: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 5.Quantitative Assessment of a Therapeutic Exercise in Mitigating

Micrographia Asssciated with Parkinson’s Disease 113

In this chapter, AAW exercise in mildly affected PD subjects’ handwriting shows the best

results while moderately affected subjects show marginal improvements. These were

consistent with the feedback of subjects who clearly indicated that they thought their

handwriting improved after AAW by their returned Voice and Customer Surveys.

However, some improvements were subtle, which subjects did not notice, yet they were

still detectable via the metrics, suggesting increased sensitivity. Fatigue reported by

subjects during AAW exercises, especially for moderately affected subjects set, is an

issue should be considered in future study.

We also note that examination of specific metrics with different geometric and

physiologic implications in expanded populations can aid in description of PD symptom

severity, and can also help identify which metrics are more useful in capturing various

neurological disorders. If this knowledge can be mastered, it could be programmed on a

digital tablet (i) to gather handwriting information from subjects, and (ii) to process this

information with statistical understanding in order to provide decision support and

guidance to clinicians for intervention planning.

Page 126: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 114

Chapter 6

Design of Graphic User Interface (GUI) for

Monitoring Micrographic and/or Tremulous

Handwriting

6.1 Introduction

Accurate and timely condition monitoring are essential but not always straightforward for

people with PD or ET. Studying how micrographia and tremor affect handwriting can

offer insight into detection of the associated symptom progressions. In this direction, we

developed and introduced quantifiable metrics with sensitivity to detect both feature

patterns from the images of handwriting samples in chapter 3 and chapter 4. These

metrics have been validated statistically on realistic samples from 12 PD subjects

corresponding to healthy and post-diagnosis states and 12 healthy subjects corresponding

to normal and AIT-affected states.

In this chapter, we present an infrastructure comprising a graphical user interface (GUI)

within which metrics described in Chapters 3 & 4 can be integrated for efficient and

effective delivery of healthcare in Section 6.2.1 and Section 6.3.1 respectively.

Page 127: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 115

Specifically, we design the GUI applications utilizing quantifiable metrics to characterize

micrographia associated with PD and tremor associated with ET through static

handwriting sample images; see example case studies in Section 6.2.2 and Section 6.3.2.

They provide new patient-clinician evaluation frameworks with the potential to support

decision-making in clinical practice, rapid screening of large populations, and/or early

diagnosis.

6.2 Infrastructure and Application of GUI for Monitoring

Micrographia Associated Parkinson’s Disease

6.2.1 Proposed Micrographia Monitoring Framework

Quantitative measurements associated with samples have been developed and validated

on 12 PD subjects’ historical signature samples in Chapter 3. They comprise five key

features linking to micrographia, namely, 1) sizing changes, 2) pixel density variation

from left to right in a string sample, 3) pixel distribution changes along horizontal and

vertical directions and the variation of peak number of HPP curve for a string sample, and

measure of morphology changes for single character by 4) skeletal points as well as 5)

radial distribution of pixels; see metric details in chapter 3. While investigation of these

metrics on expanded populations is needed, we also realize the potential of their use

within a GUI-based infrastructure for monitoring micrographia.

The proposed infrastructure (Figure 6.1) is aligned with rapidly growing tele-monitoring

technology aimed at efficient delivery of healthcare beyond existing practice. It

comprises a user-friendly GUI designed in MATLAB with three intuitive panels (Figure.

6.2): demographic information of the testing subject, pre-processing of images of writing

Page 128: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 116

samples (e.g., cropping, filtering samples), and analysis (using the previous described

metrics in chapter 3).

Figure 6.1 New home-based micrographia/tremor characterization, monitoring and

evaluation mode using the GUI infrastructure.

The idea is that the subject would produce handwriting samples either in the home setting,

or at the clinic. Next, a clinician would use the GUI for data analysis and tracking metric

variations across samples. The analysis can assist in making more informed decisions for

motor symptom variations. For instance, when the metric values indicate possible onset

of micrographia, the subject can be referred to a specialist for further examination.

The GUI application would also be useful for remotely tracking large at-risk and

symptomatic populations, as the built metrics can interpret static images for micrographia,

without the need of any additional infrastructure, subjects would only need to send their

Page 129: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 117

handwriting samples securely for analyzing. Such simple and replicable process would

eventually make the GUI infrastructure widely adoptable, at the same time, enabling the

collection of large data pool from which more reliable statistical inferences can be

developed.

Figure 6.2 Snapshot of GUI applications for micrographic handwriting on the computer

for one of sample image from Subject 06 (see details in chapter 3).

Page 130: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 118

From the patients’ perspective, the GUI infrastructure could be useful as it would help

plan focused and fewer clinical visits, which are challenging especially for the elderly.

Use of this infrastructure would also expedite the evaluation of subjects’ symptomatic

states, thereby lowering healthcare costs. Last but not least, the GUI framework can be

further supplemented with digitizing tablets, to perform analysis of dynamic hand

movements similar to those in [63].

6.2.2 Application of the GUI for Monitoring Micrographia for a Case Study

In this section, we illustrate how to use the GUI for a case study by clinicians or

researchers. Here, we choose to analyze the fifth siangture sample from Subject 06 shown

in Figure 3.8 (a).

At first, the user should rename the test sample image as a standard upload image name,

here the default name is “s.png”, and then can type all demographic information of the

tested subject in “patient information” panel, including name/ID #, sex, disease history

and current medication (if applicable and/or available), as well as specific notes for this

subject and sample. Here, we record the diagnosed symptom onsite time; see examples in

Figure 6.2.

Next, the user can provide a specific image name or ID number for the test sample in

“sample upload & preparation” panel, and by click the “pre-processed image” button to

visualize the original upload sample image and prepare new image used for further

analysis at the same time, see case example result in Figure 6.3.

Page 131: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 119

Figure 6.3 Resulting images shown for original uploaded sample and pre-processed

sample. The original sample image is from [59].

Then, the user can not only read the numerical metric results for tested sample in the

MATLAB command window, which contains the size descriptors (sample area, ink

deposite), string measurements (pixel density variation, area under HPP/VPP, number of

peaks in HPP), and/or results for metrics used for single character analysis; but also

present metric results in the pop-up plot/figure windows. An example analysis numerical

report of the studied signature sample is showed as following (Figure 6.4) with related

metric result visualization (Figure 6.5).

Page 132: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 120

Figure 6.4 Snapshot of numerical results of GUI applications (micrographic handwriting)

for the signature sample case on the computer.

Page 133: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 121

Figure 6.5 Snapshot of visualized results of GUI applications (micrographic handwriting)

for the signature sample case on the computer: (A-left column) HPP and VPP curves for

the test sample, and (B-right column) red triangle marked peaks in HPP/VPP.

At the end, the user can save all related information to form the database for each subject

(PD affected individuals), and further use these large amounts of historical and current

data to detect and track micrographia variation and corresponding treatment efficacy.

Page 134: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 122

6.3 Infrastructure and Application of GUI for Identifying Tremulous

Writing Associated Essential Tremor

6.3.1 Proposed Tremor Identifying Framework

Three metrics for describing three key features of single character affected by tremor

developed. They include 1) critical point metric presenting types and total ratio of critical

points exist within a character sample, 2) segmental curvature metric calculating the

curvature changes caused by tremor along ink trajectory, and 3) measure of the number

of peak change along both horizontal and vertical axes by HPP/VPP peak metric; find

metric details in chapter 4. While investigation of these metrics on actual handwriting

samples from ET affected people is needed for all these quantitative metrics, further

screening tests on large sets of high-risk senior populations is also a great challenge for

the current clinical evaluation approach. Therefore, we realize the potential use of these

metrics within a GUI-based infrastructure for tremor detection in the ET community.

In this dissertation, we also propose an infrastructure to detect tremor during handwriting

is aligned with the current healthcare trends in the US---that is, remote diagnosis and

precision medicine which is aimed at efficient and accurate healthcare beyond current

practices [104]. Similar to the GUI application we introduced in Section 6.2, this

approach comprises a user-friendly GUI designed in MATLAB with three intuitive

panels (Figure. 6.6): information of the testing subject, pre-processing of images of

writing samples, and analysis using the previous described metrics in chapter 4.

Page 135: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 123

Figure 6.6 Snapshot of GUI applications for tremulous handwriting on the computer for

one of sample image from Subject 05 (see details in chapter 4).

Page 136: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 124

6.3.2 Application of the GUI for Identifying Tremulous Writing for a Case Study

In this section, we present how to use the GUI application on a case study for clinicians

or researchers. Here, we choose to analyze a single character “c” affected by AIT selected

from one subject who participated our EMS experiement.

First, the user should rename the test sample image with a standard upload image name,

here the default name is “s.png”, and then can type all demographic informations of the

tested subject in “patient information” panel, including name/ID #, sex, disease history

and current medication (if available), as well as specific notes for this subject and sample

(e.g., AIT affected with specific EMS settings); see examples in Figure 6.6.

Figure 6.7 Resulting images shown for original uploaded sample and pre-processed

character sample.

Page 137: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 125

Next, the user can give a specific name or ID number for the test sample in the “sample

upload & preparation” panel, and by clicking the “pre-processed image” button. The

original upload sample image and prepared new image used for further analysis are

displayed at the same time; see case example result in Figure 6.7.

Figure 6.8 Snapshot of numerical results of GUI applications (tremulous handwriting) for

the character sample case on the computer.

Page 138: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 126

Then, the user can read the numerical metric results for the tested sample in the

MATLAB command window, which contains the size descriptors (sample area, ink

deposite), and results from tremulous metric set. An example analysis numerical report of

the studied character sample is displayed as seen in Figure 6.8 with the related metric

result visualization shown in Figure 6.9.

Figure 6.9 Snapshot of visualized results of GUI applications (tremulous handwriting) for

the character sample case on the computer.

Page 139: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 127

In the final step, the user can saved all related information to form the database for each

subject (ET affected individuals and/or healthy subjects with atificially induced tremor),

and the large amounts of historical and current data could be used as further comparison

for tracking tremor variations and corresponding treatment efficacy.

6.4 Discussion and Conclusion

Based on the success of research Aim 1 presented in chapter 3 and chapter 4, we built

two support tools by using a GUI programmed in MATLAB to obtain the symptomatic

indication results for static handwriting sample, which can finally be used for clinical to

preliminary detect and quantify the symptom types and variations through handwriting

for movement disorders (ET and PD). Long-term motor symptom progression tracking or

treatment efficacy monitoring can also be achieved through evaluating the “metric scores”

obtained from subjects’ handwriting samples by GUI applications.

These two GUI applications introduced here have the potential to further serve as a

Computer Aided Diagnostic (CADiag) tool with expanded population research. Here, we

introduce one of many possible directions for researchers. First, we need to select the

optimal weight value for each metric, which will be ultimately used for the symptomatic

indication index calculation. For the micrographia (PD) indication index, it combines

results from basic size metric set and mircographia metric sets for both string and

character analyses; for the tremor (ET) indication index, it integrates results from a basic

size metric set and a tremulous metric set. According to the significance and sensitivity,

many different weight combinations will be tested until we achieve satisfactory results:

Page 140: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 128

the sensitivity and the specificity will not be lower than a specified threshold (e.g. 85%)

for each symptom indication index. Then, a user-friendly interface for the final APP will

be built based on the feedback from our subjects. It should be easily used on a PC by

expert, and non-expert clinicians or even patients with little training on their own PCs.

The final tool will be able to test all of handwriting samples collected from our subjects;

see the flowchart in Figure 6.10. It may be a benefit in clinical settings by assisting

clinicians’ diagnoses of two major movement disorders, and may also provide objective

information for long-term tracking and treatment efficiency monitoring. We also expect

this CADiag tool can be used by ET or PD affected people themselves. It will inform

them of their motor symptom progression associated with treatment through a long-term

report, and also help them to identify the symptom fluctuations from morning to night.

After completing validation of the CADiag tool on a large target population, the further

possible version may be an IOS or Android system-based app. Then it can be easily

downloaded online and used on end-users’ smart devices.

Page 141: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 6. Design of Graphic User Interface (GUI) for Monitoring Micrographic

and/or Tremulous Handwriting 129

Figure 6.10 Computer aided diagnostic support system for the most common movement

disorders characterization

Page 142: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 7. Conclusions and Future Work 130

Chapter 7

Conclusions and Future Work

7.1 Concluding Remarks

Neurological disorders including PD and ET may cause deterioration in the handwriting

and signature of individuals afflicted with these conditions. Under therapy, with

medication, and after surgery, these individuals may recover functionality. It is then of

question how one measures the improvement of such interventions. If representative and

reliable measurements could be made, then it would enable healthcare providers to assess

improvement levels, use these measurements to compare different individuals’ recovery

regimes, and even use such measurements over time to track improvement and/or

worsening of symptoms. To our best knowledge, currently there exists no clinically

accepted means of quantitatively assessing the quality of handwriting and signature of

individuals with neurological disorders.

The proof-of-concept methodology presented in this dissertation serves to establish the

developed mathematical algorithms by which one can score handwriting and signature

samples of individuals and present opportunities for meaningfully applying the associated

metrics. These scores, hereafter called metric results, are able to capture and quantify

handwriting of individuals based on various features of the writing, such as geometry,

Page 143: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 7. Conclusions and Future Work 131

orientation, and oscillation characteristics.With regard to how the presented approach

compares with the existing literature, as emphasized above, those that rely on hand

kinematics through dysgraphia analysis were successful in identifying PD and/or ET, and

those based on static image analysis mainly focused on size effects by analyzing certain

letters, abstract shapes, specific words, or sentences. One exception is the work of

Helsper, et al. [18] in which the authors have advocated that certain morphological and

geometric patterns in static handwriting samples should be investigated to study

micrographia.

The approach described here, however, does not rely on any additional data collection

infrastructure; it requires only the data collected in the subjects' natural settings, and has

the advantage of utilizing large pools of historical handwriting samples thus allowing

retroactive studies. Needless to say, the approach proposed here could very well be

integrated into a dysgraphia analysis framework, to further strengthen the

characterization of movement disorders with static historical data.

One application of implementing the presented metrics for detecting micrographic

handwriting to assess the efficacy of treatments is introduced in this dissertation, a short

term therapeutic handwriting exercise (AAW) has been studied, and successful results

were obtained and analyzed for pre-exercise and post-exercise comparison over 14 days.

The metrics presented in this dissertation are effectively executed by importing sample

images via two GUI tools built with MATLAB for micrographia and tremor detection. In

light of all the results obtained through human subjects testing, such utility has the

potential to be used to support clinical practice, supplement digital tablet data, provide a

data-based archiving mechanism, be applied in controlled experiments to characterize PD

Page 144: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Chapter 7. Conclusions and Future Work 132

and ET, and assess relevant changes that may result from treatments and interventions

intended to mitigate symptoms.

Finally, there is supplemental work not reported here that writing samples for such tests

are not significantly affected by standard data transferal methods. Thus with simple

instructions, symptomatic people could use any or a variety of accessible ways to send

samples to an appropriate clinician for analysis.

7.2 Future Work

In the future, the following research directions are recommended:

Investigation on an expanded PD population to find out the relationship between

the metric values and UPDRS scores and represent the metric results in this

widely used clinical rating scale.

Study on metrics for tremulous writing with real ET affected people’ writing

samples to generate more absolute scales based on more extensive datasets.

Research on different AAW exercise routines with a large number of PD subjects

with varied symptom severities to identify most efficient routine for such severity

levels to could show better output results.

Programing current GUI tools into other open source formats which can be

adapted by medical staff with minimal training.

Further development of an IOS or Android-system-based app, integrated with

both metric sets for micrographia and tremor detection, which can be easily

downloaded online and used on smart devices by testing subjects themselves for

self-monitoring.

Page 145: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

References 133

REFERENCES

1. Samii, A., Nutt, J. G., & Ransom, B. R. (2004). Parkinson’s disease. Lancet, 363:

1783-1793.

2. Lyons, K. E., Pahwa, R., Comella, C. L., Eisa, M. S., Elble, R. J., Fahn, S., ... &

Watts, R. L. (2003). Benefits and risks of pharmacological treatments for essential

tremor. Drug safety, 26(7): 461-481.

3. Muangpaisan, W., Mathews, A., Hori, H., Seidel, D.( 2011). A systematic review of

the worldwide prevalence and incidence of Parkinson's disease, Journal of the

Medical Association of Thailand, 94(6): 749-755.

4. Zesiewicz, T. A., Shaw, J. D., Allison, K. G., Staffetti, J. S., Okun, M. S., & Sullivan,

K. L. (2013). Update on Treatment of Essential Tremor. Current treatment options in

neurology, 15(4): 410-423.

5. Panicker, J. N., & Pal, P. K. (2003). Clinical features, assessment and treatment of

essential tremor. JOURNAL-ASSOCIATION OF PHYSICIANS OF INDIA, 51: 276-

285.

6. Jankovic, J. (2008). Parkinson’s disease: Clinical features and diagnosis. Journal of

Neurological Neurosurgery and Psychiatry, 79: 368-376.

7. Jankovic, J. (2003). Pathophysiology and clinical assessment of parkinsonian

symptoms and signs. Neurological Disease and Therapy, 59: 71-108.

Page 146: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 134

8. Gibb, W. R., & Lees, A. J. (1988). The relevance of the Lewy body to the

pathogenesis of idiopathic Parkinson's disease. Journal of Neurology, Neurosurgery

& Psychiatry, 51(6): 745-752.

9. Barbosa, M. T., Caramelli, P., Cunningham, M. C. Q., Maia, D. P., Lima‐Costa, M. F.

F., & Cardoso, F. (2013). Prevalence and clinical classification of tremor in elderly—

A community‐based survey in Brazil. Movement Disorders, 28(5), 640-646.

10. Weintraub, D., Comella, C. L., & Horn, S. (2008). Parkinson’s disease—part 1:

pathophysiology, symptoms, burden, diagnosis, and assessment. American Journal of

Managed Care, 14 (2): S40-S48.

11. Tolosa, E., Wenning, G., & Poewe, W. (2006). The diagnosis of Parkinson’s disease.

Lancet Neurology 5: 75-86.

12. Pahwa, R., Lyons, K.E. (2003). Essential tremor: differential diagnosis and current

therapy, The American Journal of Medicine, 115(2), pp. 134-142.

13. Elble, R., Comella, C., Fahn, S., Hallett, M., Jankovic, J., Juncos, J.L., Lewitt, P.,

Lyon, K., Ondo, W., Pahwa, R., Sethi, K., Stover, N., Tarsy, D., Testa, C., Tintner,

R., Watts, R., Zesiewicz, T. (2012). Reliability of a new scale for essential

tremor, Movement Disorders, 27(12): pp. 1567-1569.

14. Belton, S.E., Cavalcante, L.H., Sipahi, R., Gouldstone, A., Jaeger, B.K. (2015, Nov)

Assistive Writing Device for Tremor Patients, NIH-IEEE 2015 Strategic Conference

on Healthcare Innovations and Point-of-Care Technologies, Bethesda, MD.

Page 147: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 135

15. Sapir S., Spielman J.L., Ramiq L.O., et al. (2007). Effect of intensive voice treatment

(the Lee Silverman Voice Treatment [LSVT]) on vowel articulation in dysarthric

individuals with idiopathic Parkinson disease: acoustic and perceptual findings.

Journal of Speech Language and Hearing Research, 50 (4): 899-912.

16. The National Collaborating Center for Chronic Conditions, editor. (2006).

Parkinson’s Disease. London: Royal College of Physicians, 29-47.

17. Goetz, C. G., Fahn, S., Martinez-Martin, P., Poewe, W., Sampaio, C., Stebbins, G. T.,

Stern, M. B., Tilley, B. C., et al. (2007). Movement Disorder Society-sponsored

revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS): Process,

format, and clinometric testing plan, Movement Disorders, 22(1), 41-47.

18. Helsper, E., Teulings, H.L., Karamat, E., Stelmach, G.E. (1996). Preclinical

Parkinson features in optically scanned handwriting, Handwriting and Drawing

Research: Basic and Applied Issues. IOS Press, Amsterdam, 241-250.

19. Smits, E. J., Tolonen, A. J., Cluitmans, L., et al. (2014). Standardized Handwriting to

Assess Bradykinesia, Micrographia and Tremor in Parkinson's Disease. PloS one, 9

(5): e97614.

20. Shukla, A. W., Ounpraseuth, S., Okun, M. S., et al. (2012). Micrographia and related

deficits in Parkinson’s disease: A cross-sectional study. BMJ Open, 2(3): e000628.

21. Bryant, M. S., Rintala, D. H., Lai, E. C., et al. (2010). An investigation of two

interventions for micrographia in individuals with Parkinson’s disease. Clinical

Rehabilitation, 24:1021-1026.

Page 148: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 136

22. Tsanas, A., Little, M. A., McSharry, P. E., & Ramig, L. O. (2011). Nonlinear speech

analysis algorithms mapped to a standard metric achieve clinically useful

quantification of average Parkinson's disease symptom severity. Journal of the Royal

Society Interface, 8(59): 842-855.

23. Wang, R., Medioni, G., Winstein, C. J., & Blanco, C. (2013). Home Monitoring

Musculo-skeletal Disorders with a Single 3D Sensor. In Computer Vision and Pattern

Recognition Workshops (CVPRW), 2013 IEEE Conference on (pp. 521-528). IEEE.

24. Objective Parkinson’s disease measurement. http://kineticsfoundation.org/. Accessed

Mar 20, 2016.

25. Giuffrida, J. P., Riley, D. E., Maddux, B. N., & Heldman, D. A. (2009). Clinically

deployable Kinesia™ technology for automated tremor assessment. Movement

Disorders, 24(5): 723-730.

26. Burkhard, P. R., Shale, H., Langston, J. W., & Tetrud, J. W. (1999). Quantification of

dyskinesia in Parkinson's disease: validation of a novel instrumental

method. Movement disorders, 14(5), 754-763.

27. Zietsma, R. C. (2013, May). Apparatus for use in diagnosing and/or treating

neurological disorder. United States Patent. US2013/0060124 A1.

28. Ünlü, A., Brause, R., & Krakow, K. (2006). Handwriting Analysis for Diagnosis and

Prognosis of Parkinson’s Disease. International Symposium on Biological and

Medical Data Analysis, LNCS, 4345: 441-450.

29. Walker, R. W., Zietsma, R., & Gray, W. K. (2014). Could a new sensory pen assist in

the early diagnosis of Parkinson's? Expert review of medical devices, 11(3): 243-

245.

Page 149: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 137

30. Sesa-Nogueras, E., Faundez-Zanuy, M., & Mekyska, J. (2012). An information

analysis of in-air and on-surface trajectories in online handwriting. Cognitive

Computation, 4(2): 195-205.

31. Hipkiss, A. R. (2014). Aging risk factors and Parkinson's disease: contrasting roles of

common dietary constituents. Neurobiology of Aging, 35 (6): 1469-1472.

32. Chillag-Talmor, O., Giladi, N., Linn, S., Gurevich, T., El-Ad, B., Silverman, B., ... &

Peretz, C. (2001). Use of a refined drug tracer algorithm to estimate prevalence and

incidence of Parkinson's disease in a large Israeli population.Journal of Parkinson's

disease, 1(1): 35-47.

33. Lerche, S., Hobert, M., Brockmann, K., et al. (2014). Mild Parkinsonian Signs in the

Elderly-Is There an Association with PD? Crosssectional Finding in 992 Individuals.

PLOS ONE, 9 (3): e92878.

34. Marek, K., Jennings, D., & Lasch, S. (2011). The Parkinson progression marker

Initiative (PPMI). Prog Neurobiol, 95: 629–663.

35. Wu, Y., Le, W., & Jankovic, J. (2011). Preclinical biomarkers of Parkinson disease.

Arch Neurol, 24: 309-317.

36. Benabid, A. L. (2003). Deep brain stimulation for Parkinson’s disease, Current

opinion in neurobiology, 13: 696-706.

37. Fahn, S. (2005). Does levodopa slow or hasten the rate of progression of Parkinson’s

disease? Journal of neurology, 252: iv37-iv42.

Page 150: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 138

38. Siebner, H. R., Ceballos‐Baumann, A., Standhardt, H., Auer, C., Conrad, B., Alesch,

F. (1999). Changes in handwriting resulting from bilateral high‐frequency stimulation

of the subthalamic nucleus in Parkinson's disease. Movement Disorders, 14(6), 964-

971.

39. Péran, P., Nemmi, F., Méligne, D., Cardebat, D., Peppe, A., Rasco, O., Caltagirone,

C., Demonet, J. F., Sabatini, U. (2013). Effect of levodopa on both verbal and motor

representations of action in Parkinson’s disease: A fMRI study. Brain and

Language, 125(3), 324-329.

40. Lemke, M. R., Brecht, H. M., Koester, J., Kraus, P. H., Reichmann, H.. (2005).

Anhedonia, depression, and motor functioning in Parkinson’s disease during

treatment with pramipexole. The Journal of Neuropsychiatry and Clinical

Neurosciences, 17(2), 214-220.

41. "Parkinson's Disease Foundation (PDF) - Hope through Research ..." Parkinson's

Disease Foundation. Web. 8 July 2014. <www.pdf.org>.

42. Medical Neurosciences.

http://www.neuroanatomy.wisc.edu/virtualbrain/BrainStem/20Substantia.html,

Accessed Mar 20, 2016.

43. Bilodeau, M., Keen, D. A., Sweeney, P. J., Shields, R. W., & Enoka, R. M. (2000).

Strength training can improve steadiness in persons with essential tremor. Muscle &

nerve, 23(5): 771-778.

44. Kwakkel, G., De Goede, C. J. T. & Van Wegen, E. E. H. (2007). Impact of physical

therapy for Parkinson's disease: a critical review of the literature, Parkinsonism and

Related Disorders, 13: S478-S487.

Page 151: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 139

45. Keus, S. H., Munneke, M., Nijkrake, M. J., Kwakkel, G., & Bloem, B. R. (2009).

Physical therapy in Parkinson's disease: evolution and future challenges, Movement

Disorders, 24(1): 1-14.

46. Hackney, M. E. and Earhart, G. M. (2009). Effects of dance on movement control in

Parkinson’s disease: a comparison of Argentine tango and American ballroom,

Journal of Rehabilitation Medicine: Official Journal of the UEMS European Board of

Physical and Rehabilitation Medicine, 41(6): 475.

47. Harmer, F., Li, P., Fitzgerald, K., Eckstrom, E., Stock, R., Galver, J., Maddalozzo, G.,

and Batya, S. S. (2012). Tai chi and postural stability in patients with Parkinson's

disease, New England Journal of Medicine, 366(6): 511-519.

48. Goodwin, V. A., Richards, S. H., Taylor, R. S., Taylor, A. H. and Campbell, J. L.

(2008). The effectiveness of exercise interventions for people with Parkinson's

disease: A systematic review and meta‐analysis, Movement disorders, 23(5): 631-640.

49. "IETF | Your Voice for Essential Tremor." International Essential Tremor

Foundation. www.essentialtremor.org,

50. Mostile, G., Fekete, R., Giuffrida, J. P., Yaltho, T., Davidson, A., Nicoletti, A., et al.

(2012). Amplitude fluctuations in essential tremor. Parkinsonism & related disorders,

18(7): 859-863.

51. Kinesia HomeView product for objective Parkinson’s Assessment.

http://glneurotech.com/kinesia/products/homeview/, Accessed Mar 20, 2016.

52. Giuffrida, J. P., Riley, D. E., Maddux, B. N., & Heldman, D. A. (2009). Clinically

deployable Kinesia™ technology for automated tremor assessment. Movement

Disorders, 24(5): 723-730.

Page 152: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 140

53. Comparison of handwriting in patients with PD and ET, https://s-media-cache-

ak0.pinimg.com/736x/23/14/3f/23143fa2c340fe9ccfcd90d110020a5b.jpg, Accessed

Mar 20, 2016.

54. The handwriting obtained from patients with essential tremor and Parkinson’s disease,

http://www.jkma.org/ArticleImage/0119JKMA/jkma-55-987-g002-l.jpg, Accessed

Mar 20, 2016.

55. Jankovic, J., Rajput, A. H., McDermott, M. P., and Perl, D. P. (2000). The evolution

of diagnosis in early Parkinson disease, Archives of neurology, 57(3): 369-372.

56. Oertel, W. H. (1996). Computational analysis of open loop handwriting movements

in Parkinson’s disease: a rapid method to detect dopamimetic effects. Mov Disord,

11: 289-297.

57. Sandyk, R., & Iacono, R. P. (1994). Reversal of micrographia in Parkinson's disease

by application of picotesla range magnetic fields. International Journal of

Neuroscience, 77(1-2): 77-84.

58. Van Gemmert, A. W., Teulings, H. L., Stelmach, G. E. (1998). The influence of

mental and motor load on handwriting movements in Parkinsonian patients. Acta

Psychol (Amst), 100: 161-175.

59. Mclennan, J. E., Nakano, K., Tyler, H. R., et al. (1971). Micrographia in Parkinson’s

disease. Journal of the Neurological Sciences, 15:141-152.

60. Tetrud, J. W. (1991). Preclinical Parkinson’s disease: Detection of motor and

nonmotor manifestations. Neurology, 41(5 suppl 2): 69-71.

Page 153: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 141

61. Van Drempt, N., McCluskey, A., & Lannin, N. A. (2011). A review of factors that

influence adult handwriting performance. Australian Occupational Therapy Journal,

58(5): 321-328.

62. Van Gemmert, W. A., Adler, C. H. and Stelmach, G. E. (2003). Parkinson’s disease

patients undershoot target size in handwriting and similar tasks, Journal of

Neurology, Neurosurgery & Psychiatry, 74: 1502-1508.

63. Letanneux, A., Danna, J., Velay, J.L., Viallet, F., and Pinto, S. (2014). From

micrographia to Parkinson's disease dysgraphia, Movement Disorders, 29(12):1467-

1475.

64. Drotar, P., Mekyska, J., Rektorova, I., Masarova, L., Smekal, Z. and Faundez-Zanuy,

M. (2015). Decision support framework for Parkinson's disease based on novel

handwriting markers, Neural Systems & Rehabilitation Engineering, IEEE Trans on,

23(3): 508-516.

65. Ho, K., Iansek, R., Marigliani, C., Bradshaw, J. L. and Gates, S. (1999). Speech

impairment in a large sample of patients with Parkinson’s disease, Behavioural

Neurology, 11: 131-137.

66. Scott, S., Caird, F. I. and Williams, B. O. (1984). Evidence for an apparent sensory

speech disorder in Parkinson's disease, Journal of Neurology, Neurosurgery &

Psychiatry, 47: 840-843.

67. McArdle, W. D., Katch, F. I. and Katch, V. L. (2010). Exercise physiology: Nutrition,

energy, and human performance. Lippincott: Williams & Wilkins.

Page 154: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 142

68. Poluha, P., Teulings, H. L., & Brookshire, R. (1998). Handwriting and speech

changes across the levodopa cycle in Parkinson’s disease. Acta psychological,

100(1): 71-84.

69. Michaelsen, S. M. and Levin, M. F. (2004). Short-Term Effects of Practice With

Trunk Restraint on Reaching Movements in Patients With Chronic Stroke A

Controlled Trial, Stroke, 35: 1914-1919.

70. Ramig, L. O., Countryman, S., O'Brien, C., Hoehn, M. and Thompson L. (1996).

Intensive speech treatment for patients with Parkinson's disease Short-and long-term

comparison of two techniques", Neurology, 47:1496-1504.

71. Ribaudo, "The ABC’s of Parkinson’s Disease Handwriting", in Parkinson’s Humor

Blog. http://parkinsonshumor.blogspot.com/2012/06/abcs-of-parkinsons-disease-

handwriting.html, Accessed Mar 20, 2016.

72. Zid. A and J. Russell, "Delay the Disease/Handwriting Challenge", in OhioHealth.

http://www.delaythedisease.com/tips-and-education/handwriting-challenge/,

Accessed Mar 20, 2016.

73. Summer, A. "Handwriting and Parkinson’s disease", in PD Warrior from Advance

Rehab Center. http://pdwarrior.com/handwriting-and-parkinsons-disease/, Accessed

Mar 20, 2016.

74. Nutt, J. G., Wooten, G. F. (2005). Diagnosis and initial management of Parkinson's

disease. New England Journal of Medicine, 353(10): 1021-1027.

75. Gelb, D. J., Oliver, E., Gilman, S. (1999). Diagnostic criteria for Parkinson

disease. Archives of Neurology, 56(1): 33-39.

Page 155: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 143

76. Zhi, N. Q., Jaeger, B. K., Gouldstone, A., Frank, S., Sipahi, R. (2016). Toward

monitoring Parkinson’s through analysis of static handwriting samples: a quantitative

analytical framework, the IEEE Journal of Biomedical and Health Informatics.

(Available online)

77. Zhi, N. Q., Jaeger, B. K., Gouldstone, A., Frank, S., Sipahi, R. (2015, Oct.). Objective

Quantitative Assessment of Movement Disorders through Analysis of Static

Handwritten Characters, Proceedings of ASME2015 Dynamic Systems and Control

Conference, Columbus, Ohio.

78. Zhi, N. Q., Jaeger, B. K., Gouldstone, A., Frank, S., Sipahi, R. (2015). A novel

quantitative assessment method to detect effects of essential tremor on static

handwriting, In Biomedical Engineering Conference (NEBEC), 2015 41st Annual

Northeast, IEEE.

79. Zhi, N. Q., Jaeger, B. K., Gouldstone, A., Frank S., Sipahi, R. (2014, Oct).

Quantitative assessment of a therapeutic exercise in mitigating micrographia

associated with Parkinson’s disease, IEEE Special Topic Conference on Healthcare

Innovation and Point-of-Care Technologies.

80. Zhi, N. Q., Jaeger, B. K., Gouldstone, A., Sipahi, R. (2015, Nov). A graphical user

interface for monitoring micrographia, NIH-IEEE 2015 Strategic Conference on

Healthcare Innovation and Point-of-Care Technologies.

81. Essential Tremor, http://medicalassessmentonline.com/terms.php?R=125&L=E,

Accessed Mar 20, 2016.

Page 156: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 144

82. Kim, E. J., Lee, B. H., Park, K. C., Lee, W. Y., & Na, D. L. (2005). Micrographia on

free writing versus copying tasks in idiopathic Parkinson's disease, Parkinsonism and

Related Disorders, 11(1): 57-63.

83. Dimauro, G., Impedovo, S., Pirlo, G., et al. (1997). A multi-expert signature

verification system for bankcheck processing. International Journal of Pattern

Recognition and Artificial Intelligence, 11(5): 827-844.

84. Smyth, M. S., Martin, J. H. J., (2000). X Ray crystallography. Molecular Pathology,

53(1): 8–14.

85. Oliveira, R. M., Gurd, J. M., Nixon, P., Marshall, J. C., Passingham, R. E. (1997).

Micrographia in Parkinson’s disease: the effect of providing external cues. Journal of

Neurology, Neurosurgery & Psychiatry, 63(4): 429-433.

86. Margolin, D. I., Wing, A. M., (1983). Agraphia and micrographia: Clinical

manifestations of motor programming and performance disorders. Acta

Psychologica, 54(1): 263-283.

87. Baltzakis, H. & Papamarkos, N. (2001). A new signature verification technique based

on a two-stage neural network classifier. Engineering Applications of Artificial

Intelligence,14: 95-103.

88. Bhattachatyya, D., Das, P., Bandyopadhyay, S. K., et al. (2009). Analysis of

handwritten signature images. Security Technology, 58: 43-50.

89. Impedovo, D. & Pirlo, G. (2008). Automatic signature verification: The state of the

art. IEEE Transactions on Systems, Man and Cybernetics-Part C: Application and

Reviews, 38(5): 609-635.

Page 157: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 145

90. Abu-Rezq, A. N. & Tolba, A.S. (1999). Cooperative self-organizing maps for

consistency checking and signature verification. Digital Signal Processing, 9: 107-

119.

91. Ferrer, M. A., Alonso, J. B., & Travieso, C. M. (2005, June). Offline geometric

parameters for automatic signature verification using fixed-point arithmetic. IEEE

Transactions on Pattern Analysis and Machine Intelligence, 27(6): 993-997.

92. Hanmandlu, M., Yusof, M. H. M., & Madasu, V. K. (2005). Off-line signature

verification and forgery detection using fuzzy modeling. Pattern Recognition, 38:

341-356.

93. Srinivasa Chakravarthy, V., & Kompella, B. (2003). The shape of handwritten

characters. Pattern recognition letters, 24(12): 1901-1913.

94. Howell, D. (2013). Fundamental statistics for the behavioral sciences, Cengage

Learning.

95. Teixeira, A., Rosa, Á. and Calapez, T. (2009). Statistical power analysis with

Microsoft excel: normal tests for one or two means as a prelude to using non-central

distributions to calculate power, Journal of Statistics Education, 17(1), n1.

96. Shulman, L. M. (2007). Gender differences in Parkinson's disease, Gender

medicine, 4(1): 8-18.

97. Van Den Eeden, S. K., Tanner, C. M., Bernstein, A. L., Fross, R. D., Leimpeter, A,

Bloch, D. A. and Nelson, L. M. (2003). Incidence of Parkinson’s disease: variation by

age, gender, and race/ethnicity,” American Journal of Epidemiology, 157(11): 1015-

1022.

Page 158: Quantitative assessment of micrographia and tremor in static …cj... · 2019-02-12 · Quantitative Assessment of Micrographia and Tremor in Static Handwriting Samples Analysis,

Resume 146

98. Djaldetti, R., Ziv, I. and Melamed, E. (2006). The mystery of motor asymmetry in

Parkinson's disease, The Lancet Neurology, 5(9): 796-802.

99. New America Cursive Alphabet, upper and lower case.

http://www.newamericancursive.com/files/3413/9652/7462/tn_sample.gif, Accessed

20 March 2016.

100. ProM 555 3 Mode Digital Muscle Stimulator Machine.

http://images10.newegg.com/NeweggImage/ProductImageCompressAll300/A1DW_

1_20121104_84843420.jpg, Accessed 20 March 2016.

101. The Anterior Forearm Muscle Image. http://media-3.web.britannica.com/eb-

media/35/113035-004-4ED08BE7.jpg, Accessed 20 March 2016.

102. Wacom Tablet Intuos 4, http://www.wacom.com/en-us/products/pen-tablets,

Accessed 20 March 2016.

103. NeuroGlyphics Software package has downloaded from website:

http://www.neuroglyphics.org/, Accessed 22 January 2013.

104. Collins, F.S. & Varmus, H. (2015). A new initiative on precision medicine. New

England Journal of Medicine, 372(9), 793-795.


Recommended