Post on 06-Mar-2018
transcript
Informatics and Information inin
Radiation Oncology :
OncoSpace
John W. Wong, Ph.D.Todd McNutt, Ph.D.
Department of Radiation Oncology
Supported in Part by Elekta Impac
OCI_50th 2008, JWW
Supported in Part by Elekta-Impac
Status of health care informationStatus of health care information
“Today, most business ---- down to the smallest corner y,grocery store have better information about their sales and inventories than even affluent medical practices have about their patients ”practices have about their patients. ………….
--- Michael BloombergMichael Bloombergto the Academy National Health Policy Conference, 05/07
OCI_50th 2008, JWW
NIH Roadmap 3:Re-engineering the Clinical Research Enterprise
• R.A. Harrington, M.D., Duke Research Institute (2005) : g , , ( )– “One of the greatest inefficiencies of the current
model of clinical research in our country is the lack of t i i i f t t ( hi h i l d h da sustaining infrastructure (which includes shared
resources, common data standards, and effective use of information technology among researchers), as well as the lack of a convenient forum to share best practices and learn from one another’s mistakes and successes”successes .
OCI_50th 2008, JWW
Cancer Bio informatics Grid (caBIG)• NCI: cancer Bioinformatics Grid (caBIG) provides infra-
Cancer Bio-informatics Grid (caBIG)
structure support – clinical trials management systems
integrative cancer research– integrative cancer research– tissue banks and pathology– Image workspaceImage workspace
• Not directed to address specific research or clinical questionsq
OCI_50th 2008, JWW
Data “Loss” at the Institutional LevelData Loss at the Institutional Level
• Data we are capturing – Labs, Images, Treatment Plans
• Data we are sending away– Patients in protocolsPatients in protocols
• Data we are storing – Disparate databases
Data (experience) we are not capturing• Data (experience) we are not capturing– Discarded treatment plans (and
decision making process)
• Information and knowledge are Not captured systematically
OCI_50th 2008, JWW
g p y y• Not utilized efficiently to impact research and patient care
Challenges of data longevity and re-useChallenges of data longevity and re use
• RTOG– Formed 1968, funded since 1971– Activated 300 trials
• 40 on-going• 60,000 patients enrolled
Q f– Established QA, credentialing process for RTP and dosimetry
– Centralized date repository; lacks secondary researchCentralized date repository; lacks secondary research– No measure of impact on community practice
OCI_50th 2008, JWW
Multiple Informatics Initiatives at JHUMultiple Informatics Initiatives at JHU
• Johns Hopkins University Health SystemsC itt f H lth I f ti– Committee for Health Informatics
– Johns Hopkins Medical Image Archive (JHMIA)– I4M: Integration of Imaging, Information and Intervention in g g g,
Medicine– Clinical Trial Groups
Industrial collaborations– Industrial collaborations• Microsoft (Almaga -- Healthcare Informatics)• IBM (Computational Medicine)• Harris Corporation (Multi-disciplinary clinic)• …………
OCI_50th 2008, JWW
The JHMIA Program– (Radiology)e J og a ( ad o ogy)
A single archive where all medical images and other non textual data (and associated reports etc ) fromnon-textual data (and associated reports, etc.) from across a Healthcare Enterprise are stored
• Clinical (300 TB now --- 700 TB in 2 years)• Research• Waveform• Genomic (planned)
P t i ( l d)• Proteomic (planned)
• Medical Image Archive Medical Data Archive?
OCI_50th 2008, JWW
• Medical Image Archive Medical Data Archive?
JHMIAA E t i I A hiAn Enterprise Image Archive
Current Participants Committed• JHH Radiology• JHH Vascular Surgery• JHH Peds Cardiology
• BMC Vascular• BMC Adult Echo Cardio• Endoscopy (GI)
• JHH Rad Oncology• JHH Adult Echo Cardio• Surgery
Potential• OB/GYNg y
• Cardiology• BMC Radiology• Ophthalmology
• Pathology• Howard County General Hospital
p gy
OCI_50th 2008, JWW
JHMIA and I4MJHMIA and I M
Analytic Database(s):Analytic Database(s):Query and Security
Analytic and Change Tools:Extraction of Information
Decision Support:
Web-service
Data-miningStatistical Modeling
OCI_50th 2008, JWW
• Goal: To improve both medical research and patient care
The enterprise model: JHMIA and I4MThe enterprise model: JHMIA and I M
Radiation Oncology
Infra-structure: JHMIA Radiology TeraMedica
RoboticSurgery Ophthalmology
Infra-structure: JHMIA, Radiology, TeraMedica,
Data-miningShape and Change Tools Decision SupportAnalytic Database
• Challenges to implement across multi-disciplines:– Data Standards
W kfl P d d M t Diff– Workflow, Procedure, and Management Differences• Different intervention time-scale
OCI_50th 2008, JWW
OncoSpace: pClosed Loop Adaptive Radiation Therapy
Data Information Intervention ResponsePatient
Data – Information – Intervention – Response
Real Time Image Guided Intervention
T t t R ti i ti E l T t t A tTreatment Re-optimization; Early Treatment Assessment,…
Population : New protocol, New dose level, New standards
OncoSpace Infrastructure
OCI_50th 2008, JWW
OncoSpace Infrastructure
OncoSpace:OncoSpace:Radiation Oncology as the I4M test-bed
Radiation Oncology
RoboticSurgery
Ophthalmology
Infra-structure: JHMIA, Radiology, TeraMedica,
Decision Support
Shape and Change Tools
Analytic Database
OncoSpace
Decision SupportData-mining
OCI_50th 2008, JWW
Extending the OncoSpace Model: g pSharing Research and Clinical Care
I4M Infr
Institute n
ra-structu
Institute 1
JHU Genomics
I4M Infra-structure
JHU
ure
Ophthalmology
Radiation Oncology
OCI_50th 2008, JWW
OncoSpace
• OncoSpace is a new research infra-structure based on pRadiation Oncology as a “use-case” model
• Bioengineering Research Partnershiplti di i li R di ti O l R di l– multi-disciplinary: Radiation Oncology, Radiology,
Physics and Astronomy, Computer Science and Biostatistics
– multi-institutional: Hopkins, clinical partner sites– IMPAC
OCI_50th 2008, JWW
Distributed Research ModelDistributed Research ModelCurrent Trial Practice Hypothetical Future Practice
Patient Tx Follow up
Treatment T t t
Patient Tx Follow up
TreatmentProtocol
TreatmentProtocol Journal
Publications
Journal Publication Publication of
Data to DB’s
Increased potential for data reuseSTOP
START
OCI_50th 2008, JWW
STARTOVER
Data Delivery in Cooperative Research:Data Delivery in Cooperative Research:Hitting the Wall
FTP and GREP are not adequate• You can GREP 1 MB in a second• You can GREP 1 GB in a minute
You can GREP 1 TB in 2 days
• You can FTP 1 MB in 1 sec• You can FTP 1 GB / min (~1
$/GB)
q
• You can GREP 1 TB in 2 days• You can GREP 1 PB in 3 years
50 MB l l DICOM t f t k 1 i
$/GB)• 2 days;1K$ / 3 years and 1M$
• 50 MB local DICOM transfer takes 1 min• 100 patients x 10 (3D) scans = 5 - 10 TB• A factor of 10 improvement in access speed
cannot offset the growth in data andcannot offset the growth in data and complexity
R thi k d t b ’ f ti
OCI_50th 2008, JWW
• Rethink databases’ function– following the CS community
OncoSpace: Adapting the SkyServer Approachp p g y pp
• SDSS is a collaborative effort to map 25% of the
• Shared resources– Methodology
sky• SkyServer publishes data
from the SDSS
– Software– Expertise– Experiencefrom the SDSS
• >> 100’s of new discoveries in astrophysics
p• New opportunities
– AnalysisVisualization
• Increased scale and scope for research
– Visualization– User experience
• Skyserver.sdss.org
OCI_50th 2008, JWW
Alex Szalay PhD - JHUJim Gray PhD - Microsoft
OncoSpace: Adapting the SkyServer ApproachO coSpace dapt g t e S ySe e pp oac
• Active Databases
• There is too much data to move around,take the analysis to the data!
• Do all data manipulations at databaseB ild t d d f ti i th– Build custom procedures and functions in the database
• Established Web-service for broad accessEstablished Web service for broad access– Query across multiple databases
• Automatic parallelism guaranteed
OCI_50th 2008, JWW
p g
Database ConsiderationDatabase ConsiderationOperational vs Analytical
•Workflow management•Patient records and archival
•Decision Support•On-line Analytical Processing•Multidimensional analysis•Time variantilt
erin
g
•Non-volatile
How do I organize my data?Typically Hierarchical
How might I analyze my data?Star Schema
D i t t l ion &
Fi
Dimension
Dimension Dimension
Dimension
Dimension
FACTS
DICOM RTOO principles
Design to support analysis…Fast query
Ext
ract
iD
ata
E
OIS, OCIS, EPRTPS, PACS
OncoSpace
OCI_50th 2008, JWW
OncoSpace: Work in progress (McNutt)Membership
VocabularyOncoSpace
D tData Preparation
Analytical Database
• Technologies– SQL Server 2005
Ruby on Rails
OCI_50th 2008, JWW
– Ruby on Rails
lab_valuesid
test_name
test_value
test_unit
test_date
patient id
lesionsid
lesion_site
lesion_date
lesion_size
lesion_size_unit
lesion_type
current_t_stage
pathology_featuresid
organ_site
anatomic_site
name
value
unit
patient_representationsid
dicom_uid
representation_name
image_set_modality
primary_tumorsid
tumor_site
t_stage
n_stage
m_stage
overall_stage
tumor_histology
transformationsid
is_rigid
x_translate
y_translate
z_translate
x_rotate
y rotate
Transformation
p _
created_on
updated_on
user_id
version
current_n_stage
current_m_stage
current_overall_stage
current_grade
created_on
updated_on
primary_tumor_id
patient_id
medical_historiesid
organ_system
disease
initial_diagnosis
patient_id
feature_grade
collection_date
patient_id
created_on
updated_on
user_id
version
representation_date
x_start
y_start
z_start
x_pixel_dimension
y_pixel_dimension
z_pixel_dimension
created_on
tumor_grade
presentation_date
created_on
updated_on
patient_id
user_id
version
y_rotate
z_rotate
deformable_transformation...
deformable_transformation
from_patient_rep_id
to_patient_rep_id
created_on
updated_on
user id
Dose_dose_grids
id
dose_per_fraction
x_start
y_start
z_start
family_historiesid
disease_type
disease name
user_id
version
created_on
updated_on
user_id
version
updated_on
patient_id
user_id
version
patientsid
last_name
first_name
radiation_summariesid
target
radiotherapy_sessionsid
number_of_fractions
user_id
version
PatientRadiation
grids
x_pixel_dimension
y_pixel_dimension
z_pixel_dimension
dose_grid_blob
patient_representation_id
created_on
updated_on
user_id
disease_name
relation
age_at_diagnosis
collection_date
patient_id
created_on
updated_on
user_id
version
medical_record_number
birth_date
gender
race
first_contact
postal_code
clinical_site
protocol
target
radiation_technique
radiation_protocol
dose_per_fraction
nominal_total_dose
simulation_date
rt_completion_date
treatment_interruption_rea...
treatment termination reason
prescribed_dose_per_fraction
start_date
completion_date
technique
modality
beam_energy
dicomrt_plan_uid
radiation_summary_id
Summaries
version version
created_on
updated_on
user_id
version
treatment_termination_reason
primary_tumor_id
lesion_id
created_on
updated_on
patient_id
user_id
version
patient_representation_id
dose_grid_id
created_on
updated_on
user_id
version
social_historiesid
social_history_type
social_history_name
social_history_value
social_history_unit
collection_date
clinical_eventsid
event_type
event_name
event_date
patient_id
created_on
updated on
outcomesid
days_since_enrollment
outcome_type
outcome_name
outcome_grading_scale
outcome_grade
patient id
prescribed_drugsid
drug_name
drug_type
dose
dose_unit
frequency
delivery route
roi_dose_per_sessionsid
roi_name
volume
mean_dose
max_dose
min_dose
stddev_dose
roi_dose_summariesid
roi_name
number_of_fractions
volume
mean_dose
max_dose
min dose
roi_geometriesid
roi_name
volume
x_center_of_mass
y_center_of_mass
z_center_of_mass
surface_mesh
patient_id
created_on
updated_on
user_id
version
surgical_proceduresid
procedure_name
attending_physician
procedure_date
patient_id
created_on
updated on
OCI_50th 2008, JWW
updated_on
user_id
version
patient_id
created_on
updated_on
user_id
version
delivery_route
start_date
stop_date
patient_id
created_on
updated_on
user_id
version
dose_volume_histogram
roi_geometry_id
radiotherapy_session_id
created_on
updated_on
user_id
version
min_dose
stddev_dose
dose_volume_histogram
created_on
updated_on
patient_id
user_id
version
binary_mask
patient_representation_id
created_on
updated_on
user_id
version
updated_on
user_id
version
Hopkins OncoSpace
Clinicians Researchers Bio-Statisticians
View/Analyze
Data
View/Analyze
Data
View/Analyze
Data
MS WebServicesTools Security
Project 3
Active Data base
Services y
Project 2
IMPAC/
Project 1
OCI_50th 2008, JWW
PACs MIS IMPAC/RTP Labs
OtherUsers
OtherUsers
Hopkins Institution 2Institution 2 Institution ncaBIG caBIG
Globus Globus Globus Web
ServicesWeb
Services
Globus Web Services
Globus Web Services
Active Data Base Active Data Base Active Data Base
OCI_50th 2008, JWW
OncoSpace: 4 Projects
1 Integration of clinical workflow with data collection to
OncoSpace: 4 Projects
1. Integration of clinical workflow with data collection to populate OncoSpace.
2. Optimize database architecture for secured distributed web-access
3. Tools for query, analysis, and navigation of OncoSpace to derive information from various classes of questionsto derive information from various classes of questions
4. Bio-statistic research and development to support data mining and ensure valid decision making from the OncoSpace Systems and Nested Experiments.
OCI_50th 2008, JWW
Influence of Shape
OCI_50th 2008, JWW
Overlap Volume Histogram (OVH)
1mmV l
Cumulative Overlap Volume Histogram (COVH)
OVH maps the shape of OAR to a volume-distance plane throughtarget expanding and shrinking.
1mm
1mm
Volume
1mm
Distance0 1mm 2mm-1mm-2mm
ExpansionShrinkage
OCI_50th 2008, JWW
Sphere: radius 7cmRectangle: 3.7*3.7*12.1cmSpatial resolution: 0 1*0 1*0 1cm
Target
OAR2
Spatial resolution: 0.1 0.1 0.1cmImage size: 291*291*291pix
OAR1
0.8
0.9
1
0.3
0.35
0.5
0.6
0.7OAR1OAR2
0.2
0.25OAR1OAR2
0.2
0.3
0.4
0.1
0.15
OCI_50th 2008, JWWOncoSpace 2008, JWWDistance in cm unit Distance in cm unit-6 -4 -2 0 2 4 6 80
0.1
-6 -4 -2 0 2 4 6 80
0.05
Statistic Mapping T
0.8
1
0.8
1DVH of parotidLOVH MOVH
HOVH
0 2
0.40.50.6
0 2
0.40.50.6
[ , , ]L M HDVH T OVH OVH OVH=
-2 0 2 40
0.2
Distance (cm)0 2 4 6 8
Distance (cm)2 4 6 8
Distance (cm)0 20 40 60 80
0
0.2
Dose (Gy)[ , , ]L M H
Parotid: V(30Gy)<50% of volume
⇓Dose corresponding 50% of volume:
32 5 [0 4685 4 411 5 189 ]Gy T cm cm cm=⇓
Distance corresponding 50% of volume:
OCI_50th 2008, JWWOncoSpace 2008, JWW
32.5 [0.4685 , 4.411 , 5.189 ]Gy T cm cm cm=
Treatment plan evaluation (parotids)Re-plan 22: R parotid (31Gy)
8
10
4
6
8
of P
TVH
0
2
d3:O
VH
23
44
6
8-2
-2-1
01
2
-2
0
2
d1:OVH of PTVLd2:OVH of PTVM
OCI_50th 2008, JWWOncoSpace 2008, JWW: 0—25Gy: 25—30Gy
: 30—35Gy: 35—40Gy
: >60GyDose corresponding 50% of volume:
•
× ∗+ :50—55Gy
:55—60Gy:40—45Gy:45—50Gy
Re-plan 22: right parotid
0 9
1
0 9
1
0.7
0.8
0.9
0.7
0.8
0.9
Left parotid
0.4
0.5
0.6
0.4
0.5
0.6
0.1
0.2
0.3
0.1
0.2
0.3
58.1 63 70, ,PTV PTV PTVDose in Gy Dose in Gy
Brainstem Cord
0 10 20 30 40 50 60 70 800
0 10 20 30 40 50 60 70 800
Right parotid: 50% volume 31Gy 24.4Gy:Re-plan
:Original plan
OCI_50th 2008, JWWOncoSpace 2008, JWW
OncoSpace:pPhysician’s Tool for Personalized Medicine
• Enable natural recall of past experience’s with patientsp p p– display of data that match physician’s way of thinking
• Allow other physicians to share the experience• Allow other centers to contribute and use OncoSpace to
broaden the stored experience caBIG compliance to insure data reuse and sharing• caBIG compliance to insure data reuse and sharing
OCI_50th 2008, JWW
OncoSpace:OncoSpace:as a Collaborative Research Model
All queries should be IRB approved• All queries should be IRB approved• RTOG as a service to legitimize query
– Data is live for re-useData is live for re use– Demonstrate Patient Care Improvement
OCI_50th 2008, JWW
ChallengesChallenges
• OncoSpace Designp g• Statistical Research• Quality Assurance• HIPPA and Security
OCI_50th 2008, JWW
OncoSpace:Ph i i ’ T l f P li d M di ias Physician’s Tool for Personalized Medicine
• Enable natural recall of past experience’s with patientsp p p– display of data that match physician’s way of thinking
• Allow other physicians to share the experience• Allow other centers to contribute and use OncoSpace to
broaden the stored experience caBIG compliance to insure data reuse and sharing• caBIG compliance to insure data reuse and sharing
OCI_50th 2008, JWW
OncoSpace:as a Collaborati e Research Modelas a Collaborative Research Model
• All queries should be IRB approvedq pp• RTOG as a service to legitimize query
– Data is live for re-use– Demonstrate Patient Care Improvement
OCI_50th 2008, JWW
ChallengesChallenges
• OncoSpace Designp g• Statistical Research• Quality Assurance• HIPPA and Security
OCI_50th 2008, JWW
Future Role of Physics in Radiation Therapyutu e o e o ys cs ad at o e apy
• Technological focus– Dose escalation; proton?– How high can we deliver?
H hi h d d?– How high do we need? • Bridging discoveries to RT
– Room to de-escalate– Room to de-escalate• Information and Informatics
– Improve effectiveness and pefficiency of research
– Disseminate knowledge for care
OCI_50th 2008, JWW
Future Role of Physics in Radiation Therapyutu e o e o ys cs ad at o e apy
• Technological focus– Dose escalation; proton?– How high can we deliver?
H hi h d d?– How high do we need? • Bridging discoveries to RT
– Room to de-escalate– Room to de-escalate• Information and Informatics
– Improve effectiveness and pefficiency of research
– Disseminate knowledge for care
OCI_50th 2008, JWW