NON-LINEAR ANALYSIS OF SURFACE

ELECTROMYOGRAPHY IN PARKINSON’S DISEASE

Matthew Flood1,2*

1School of Electrical & Electronic Engineering,University College Dublin,Belfield,Dublin 4.

Prof. Madeleine Lowery1,2

2Insight Centre for Data Analytics,

O'Brien Centre for Science,Belfield,Dublin 4.

Parkinson’s Disease

Bradykinetic gait

Muscle tremorMuscle rigidity

Akinesia

Parkinson’s Disease

No biomarkerNeed to quantify disease severity

Levodopa medication & deep brain stimulation

Limited research on nonlinear analysis of EMG

Aim

To distinguish Parkinsonian EMG from healthy controls using advanced signal analysis,

enabling early diagnosis and biomarkers to assess therapeutic

interventions.

Experimental Methods

Time (s)

EMG of Rectus Femoris (PD)

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

8.8 8.9 9 9.1 9.2

EM

G

Time (s)

EM

G

EMG of Rectus Femoris (Control)

-0.3

-0.2

-0.1

0

0.1

0.2

10.65 10.75 10.85 10.95

B. Upper leg muscles

A. Experimental

setup

C. Surface EMG

Recurrence Quantification Analysis

Embedding Dimension * Time Delay *

Trajectory Radius * Line length

0.20.1

0-0.1

-0.2-0.2

-0.1

0

0.1

-0.05

0

0.05

0.1

0.15

-0.15

-0.1

0.2

PD

Controlshttps://code.google.com/p/girs/downloads/list

Rössler System

EMG

Recurrence

Plot

EMGEM

G

(mV

)Tim

e

(sec)

PD Patient

Time - (seconds)

Tim

e - (

seco

nds)

Recurrence Plot of Right Rectus Femoris of PD Patient (Subject 2 - Trial 1)

0.5 1

0.5

1

0 0.5 1-1

0

1

Time- (Seconds)

Ampl

itude

(mV)

Plot of Right Rectus Femoris sEMG of Parkinson's Patient

%DET = 38.76%

Time - (Seconds)

Tim

e - (

Sec

onds

)

Recurrence Plot of Purely Stochastic Signal

0.5 1

0.5

1

0 0.5 1-1

0

1

Time - (Seconds)

Ampl

itude

(mV)

Plot of Purely Stochastic Signal

%DET = 0%Noise

Time (seconds)

Tim

e (s

econ

ds)

Recurrence Plot of Right Rectus Femoris of Healthy Control (Subject 3- Trial 6)

0.5 1

0.5

1

0 0.5 1-1

0

1

Time - (Seconds)

Ampl

itude

(mV)

Plot of Right Rectus Femoris EMG of Healthy Control

%DET = 3.545%

Healthy Control

Time (sec)

Time (sec)

Time (sec)

Recurrence Quantification Analysis

Agonist-antagonist EMG Coherence

Intermuscular Coherence

Frequency (Hz)

Intermuscular Coherence of PD and Control SubjectsControls

PD

Inte

rmus

cular C

oher

ence

0

0.1

0.2

0.3

0.4

0.5

0 10 20 30 40 50 60

Significance Threshold

θδ α β γ

EMG Kurtosis and skew Kurto

sis

Skewness

Negative Skew

Positive & Negative Kurtosis

Positive Skew

Results: Recurrence Quantification Analysis

Controls Parkinson's05

10152025303540

Left LeftRight Right

Determinism of EMG Signal%

DE

T

*

* p = 0.0475

δ (0-4 Hz) θ (4-8 Hz) α (8-15 Hz) β (15-30 Hz) γ (30-60 Hz)0

0.5

1

1.5

2

2.5

3

EMG Coherence Between Agonist & Antagonist

Muscles

Frequency Bands

Inte

rmusc

ula

r C

ohere

nce

Results: Agonist-antagonist EMG coherence

* * *

**Parkinson 's Disease Controls

* p < 0.001** p = 0.014

Results: EMG Kurtosis and Skew

Series10

0.05

0.1

0.15

0.2

0.25

SkewnessControls

Abso

lute

Skew

Series10

0.51

1.52

2.53

3.54

4.5

KurtosisControls Parkinson's Disease

Kurt

osi

s

*

* p = 0.0457

Conclusions

Determinism of EMG signal increases in Parkinson’s Disease, greater motor unit synchrony.

Coherence higher in patients in θ, α, β, & γ frequency bands, greater shared common input across muscles.

Kurtosis higher in Parkinson’s disease, greater signal structure.

Pathological synchrony of motoneuron firing patterns in isometric muscle contractions in Parkinson’s disease.

Future Work

Sensitivity analysis of recurrence quantification parameters to optimize for parkinsonian EMG.

Identify relationship between coherence and other nonlinear measures (e.g. Sample Entropy, LZ Complexity).

Use simulation modelling to identify mechanisms underlying difference in parkinsonian EMG.

Apply analysis methods to quantify the efficacy of exercise intervention in PD patients.

The researchers would like to thank Prof. Bente Jensen and colleagues at the Dept. of Exercise and Sports Sciences at the

University of Copenhagen, Denmark, for contributing their data.

This project was funded under the SFI Insight Centre for Data Analytics.

Acknowledgements