2004 All Hands Meeting Morphometry BIRN Bruce Rosen, MD PhD Jorge Jovicich PhD Steve Pieper, PhD...
-
Upload
jordan-cameron -
Category
Documents
-
view
220 -
download
1
Transcript of 2004 All Hands Meeting Morphometry BIRN Bruce Rosen, MD PhD Jorge Jovicich PhD Steve Pieper, PhD...
2004 All Hands Meeting
Morphometry BIRNBruce Rosen, MD PhD
Jorge Jovicich PhD
Steve Pieper, PhD
David Kennedy, PhD
Morphometry BIRN: The PI
“ The Principal Investigator, Dr. Rosen, does not have training or experience in computer science, programming, database design or digital networking…”
mBIRN Summary Statement – 8/04
Morphometry BIRN
“Strengths of the application include the substantial public health significance of the proposed mBIRN, a well qualified team of site investigators…, strong institutional support enhanced by the existing BIRN-CC, development of novel shape-modeling approaches, segmentation, and DTI methods, and the innovation of a number of project goals to link visual imaging with machine learning methods”
“Jorge Jovicich is to be commended for the skills he brings to this work. Leaders such as Randy Buckner, David Kennedy, Michael Miller, and Arthur Toga are invaluable assets for such an endeavor”
Morphometry BIRN Goals
Progress Highlights
Future Work
Morphometry BIRN: Outline
Overall Goal:
Develop capability to analyze and mine data
acquired at multiple sites using processing and
visualization tools developed at multiple sites
Morphometry BIRN
Overall Goal:
Develop capability to analyze and mine data acquired at multiple sites
using processing and visualization tools developed at multiple sites
Context: Human Brain MR Based Morphometry
Initial Application: Alzheimer’s, Depression, Ageing Brain
Participants: MGH, BWH, Duke, UC Los Angeles,
UC San Diego, Johns Hopkins, UC
Irvine,
Washington University, MIT
Morphometry BIRN
Human Data Protection
Multi-site data
acquisition
Data Upload
Integration and Application of
Processing Tools
Human Imaging
Database
Morphometry BIRN: Progress
Simplified diagram
of building blocks
SRB
Human Data Protection
Multi-site data
acquisition
Data Upload
Integration and Application of
Processing Tools
Human Imaging
Database
Morphometry BIRN: Progress
Simplified diagram
of building blocks
SRB
Raw data De-faced data
• De-facing: automated de-facing without brain removal• Pipeline: image formats, BIRN ID generation, defacing, QA, upload
Accomplishment: Developed a robust automated methods for
bulk MRI de-identification and upload to database(diverse inputs, sharable outputs, common
package)
De-identification and Upload Pipeline
• UCSD (fMRI): A. Bischoff, C.Notestine, B.
Ozyurt , S. Morris, G.G. Brown
• MGH (NMR): B. Fischl
• BWH (SPL): S. Pieper
• UCI: D. Wei
• Duke: B. Boyd
Morphometry BIRN: Progress
Human Data Protection
Multi-site data
acquisition
Data Upload
Integration and Application of
Processing Tools
Human Imaging
Database
Morphometry BIRN: Progress
Simplified diagram
of building blocks
SRB
Multi-site Structural MRI Data Acquisition & Calibration
Methods: common acquisition protocol, distortion correction, evaluation by scanning human phantoms multiple times at all sites
•MGH (NMR): J. Jovicich, A. Dale, D. Greve, E. Haley
•BWH (SPL): S. Pieper•UCI: D. Keator•UCSD (fMRI): G. Brown •Duke University (NIRL): J. MacFall CorrectedUncorrected
Image intensity variability onsame subject scanned at 4 sites
Accomplishment: develop acquisition & calibration protocols that improve reproducibility, within- and across-sites
Morphometry BIRN: Progress
Human Data Protection
Multi-site data
acquisition
Data Upload
Integration and Application of
Processing Tools
Human Imaging
Database
Morphometry BIRN: Progress
Simplified diagram
of building blocks
SRB
Shared Tools for Data Analysis
Tool Site
• Freesurfer MGH
• Slicer BWH
• LONI Pipeline UCLA
• LDDMM Johns Hopkins
• Query Interface UCSD
Morphometry BIRN: Progress
Shared Data
Data Site
AD BWH/MGH
AD UCSD
AD WashU
Depression Duke
MCI UCI
Integration and Application of Processing Tools
Various projects driving developments:
• Multi-site Imaging Research in Analysis of Depression• Data from one site processed with tools of multiple sites
• Multi-site Morphometry in Analysis of Alzheimer’s Disease• Data from multiple sites processed with tools of one site
• Semi-Automated Shape Analysis Project • Data from BIRN sites processed with tools of various sites
Morphometry BIRN: Progress
Integration and Application of Processing Tools
Projects driving developments:
• Multi-site Imaging Research in Analysis of Depression (MIRIAD)• Data from one site processed with tools of multiple sites
• Multi-site Morphometry in Analysis of Alzheimer’s Disease• Data from multiple sites processed with tools of one site
• Semi-Automated Shape Analysis Project• Data from BIRN sites processed with tools of various sites
Morphometry BIRN: Progress
DukeArchives
UCLAAIR Registration
and Lobar Analysis
BWHIntensity Normalizationand EM Segmentation
DukeClinical Analysis
1
2
3
4
BWH Probabilistic Atlas
(one time transfer)
UCSDSupercomputing
Goal: analyze legacy data using automated lobar
segmentation (UCLA) and cortical/subcortical
segmentations (BWH)
MIRIAD: Overview
N=200
MIRIAD Project: Accomplishments
Segmentation Duke BIRN-MIRIAD
Item (semi-automated) (fully-automated)
# of tissue classes 3 (Fig1) 23 (Fig2)
Time for 200 brains 400 hours 1 hour
Time for 200 lobe & 250 hours all lobes (Fig3) and 27 regional analysis regions included above
Improved computational capabilities
1 2 3
Integration and Application of Processing Tools
Projects driving developments:
• Multi-site Imaging Research in Analysis of Depression • Data from one site processed with tools of multiple sites
• Multi-site Morphometry in Analysis of Alzheimer’s Disease • Data from multiple sites processed with tools of one site
• Semi-Automated Shape Analysis Project • Data from BIRN sites processed with tools of various sites
Morphometry BIRN: Progress
AD Project: Overview
MGH Segmentation
Multi-Site Data Acquisition
De-identification and upload
SRB UCSDN=125
BWH/MGHN=118
Multi-site DataQueries and
Statistics
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0CVLT Discriminability Score
1000
2000
3000
4000
5000
6000
Le
ft H
ipp
oca
mp
al V
ol u
me
BWH/MGH and UCSD Data
HID
HID
Visualization and Scientific Search with
3DSlicer & Query Atlas
1
2
3
4
5
AD Project: Accomplishments
• Data sharing:• Successfully tested De-identification and Upload Pipeline (DUP)
• Integration of data
• Common database schemas for clinical and derived morphometry data at different sites: Human Imaging Database (HID)
• Mediated queries that interrogate databases at two sites
• Integration of processing tools
• MGH subcortical segmentation completed on UCSD data
• Statistical tools through the BIRN Portal and HID Query Interface
• Data visualization and interpretation using 3DSlicer and Query Atlas
Integration and Application of Processing Tools
Projects driving developments:
• Multi-site Imaging Research in Analysis of Depression • Data from one site processed with tools of multiple sites
• Multi-site Morphometry in Analysis of Alzheimer’s Disease • Data from multiple sites processed with tools of one site
• Semi-Automated Shape Analysis Project • Data from BIRN sites processed with tools of various sites
Morphometry BIRN: Progress
SASHA Project: Overview
MGH Segmentation
Data DonorSite (WashU)
De-identificationAnd upload
JHUShape Analysis
of Segmented Structures
SRB
BWHVisualization
Goal: comparison and quantification of structures’
shape and volumetric differences across patient
populations
1
2
3
4
5
Teragrid
N=45
SASHA Project: Accomplishments
Data: 46 hippocampus data sets (2070 comparisons) Each LDDMM comparison takes about 3 to 8 hours
Large Deformation Diffeomorphic Metric Mapping (LDDMM) using the TeraGrid
Improved computational capabilities
Single PC TeraGrid 1 comparison ~431 days
60 comparisons simultaneously ~7 days
Morphometry BIRN: Future
MRI Calibration (J. Jovicich)
Analysis, Visualization, Tools (S.Pieper)
Computational Informatics (D. Kennedy)
Outline of new extensions:
MRI Calibration: Extensions
More MRI Systems• Siemens 1.5T, 3T• GE 1.5T, 3T, 4T• Philips 1.5T, 3T• Picker 1.5T
Sources of variability corrections• Gradient unwarping • B0 inhomogeneities• B1 inhomogeneities • On-line motion correction
More Imaging modalities:• T1-based multi-spectral morphometry• T2-based multi-spectral morphometry• Diffusion MRI
Goal Remains:
Develop acquisition protocols and correction methods that minimize multi-site image variability
EYE MOTION FLOW
me FLASH
FLASH
me FLASH
FLASH
Correcting for Sources of Variability
B0 inhomogeneitiesOn-line motion correction
B1 inhomogeneities
T2-based multi-spectral morphometry
• Goal: quantify reproducibility of WM, GM, CSF, lesion segmentations
• Subjects with known dementia vascular lesions [Duke]
• Test-retest, 1.5T (Siemens, GE) + 3T (GE) + 4T (GE)
PD T2 T1 FLAIRRaw Data
Segmentations
CSF
WM
GMVascular lesion
Reproducibility of diffusion MRI
• Goal: quantify reproducibility of fractional anisotropy and apparent diffusion coefficient as function of:
• MR Signal-to-noise ratio (number of averages) [JHU]
• Diffusion weighting (b-value) [JHU]
• Echo time [JHU]
• Number/orientation of diffusion gradient-encoding directions [UCSD]
• B0 inhomogeneity corrections (field maps, time domain reconstruction) [Duke,UCSD]
EPI distortion correction Spiral blurring correction
MRI Calibration Deliveries Plan
By the end of 2006 we will provide: Protocol recommendations for multi-site, multi-fields
• T1-based structural multi-spectral protocol
• T2-based structural multi-spectral protocol
• Diffusion protocol
Correction recommendations that minimize variability
Software tools that perform the recommended corrections
De-identified human calibration data
Morphometry BIRN: Future
MRI Calibration (J. Jovicich)
Analysis, Visualization, Tools (S.Pieper)
Computational Informatics (D. Kennedy)
Outline of new extensions:
Analysis & Visualization Work Plan
Segmentation [MGH, UCSD, BWH]• Extension to 3T/4T; QA; Defacing
Shape Analysis [JHU, WashU, BIRN-CC]• Portal / TeraGrid Integration; DTI Mapping; Statistics
DTI Aims [BWH, UNC, UCI]• Tractography Tools; White Matter Atlases
Visualization [BWH]• Interactive: Structure+Shape+Tracts+Statistics
Query Atlas [BWH, UCSD]• Ontology / Informatics Integration with Analysis
Machine Learning [MIT, BIRN-CC]• Hypothesis Generation and Hypothesis Visualization
Query Atlas Prototype Cortical
Parcellation by Freesurfer
User Selected Features of Interest
Literature and Other Database Queries• Medline, BrainInfo,
SMART Atlas, CCDB, etc…
Diffusion MRI Acquisition
Six Gradient-encoding directionsBaseline scans
Diffusion MRI Examples
mBIRN White Matter Atlas• Under Development using
Slicer (BWH) by James Fallon (UCI)
Segmentation and Tractography
Parcellation • Freesurfer (MGH)
Tractography• DoDTI (H.J. Park)
Visualization• Slicer (BWH)
Full Integration with Slicer underway
Machine Learning Plan Apply Existing Machine Learning Code to mBIRN
Datasets• Mining MIRIAD Data for Provable Hypotheses• Expert Review and/or Further Testing of Proposed
Hypotheses
New Machine Learning Directions• Incorporate a priori Domain Knowledge and Constraints to
Strengthen Clinical Hypotheses
With BIRN-CC, Integrate Machine Learning Tools in Portal for Distribution
Morphometry BIRN: Future
MRI Calibration (J. Jovicich)
Analysis, Visualization, Tools (S.Pieper)
Computational Informatics (D. Kennedy)
Outline of new extensions:
Computational Informatics Work Plan
Aim 1: Where’s the Data?• Local/Global• Upload• Raw/Derived
Aim 2: More types of Data• Diffusion, T2, Genetics
Aim 3: Uses of Data• Quality assurance (acquisition, processing)• Querying• Statistics• mBIRN Information Services
• Knowledge Management
Database Tools:• HID• XNAT• LONI DB
OntologyHaystack / Semantic Web
RPDR
Clinical Measures
Genotype
Local Storage
BIRN Rack
SRBMCAT
HID
DU
P
Calibration & Analysis
Tools
GRID
PortalMediator
Institution A
BIRN Rack
SRBMCAT
Institution B
HID
… Workflow Control: - Queries (identify subject populations, extract data, etc.) - Statistical Analysis - Download Data for: > Visualization > More Statistics > More Processing
- Interoperable Queries (literature, homology, other databases, etc.)
Human Data Protection
StandardizedAcquisition
Protocol
Institution C
Informatics Architecture
Local DB
Human Imaging Database
• Goal: develop the image repository and relational database for clinical and derived morphometric data
Cortical Summary Data by Region
Subcortical Summary Data by Region
• BWH (SPL): J. Sacks• Duke University: S. Gadde, S. Anastasiadis• UCI: D. Wei• JHU: A. Kolasny, R. Yashinski• MGH (NMR): K. Song• UCSD (fMRI): B. Ozyurt• UCLA (LONI): K. Crawford• BIRN CC: J. Grethe
Single or Multi-site
(mediated) Query
Integrated Query Functions
Query Results
Integrated Query Functions
Single subject data: view, browse, download
Integrated Query Functions
Integrated Query Functions
Integrated Query Functions
mBIRN Ontologies
Brain
Cerebellum Cerebrum
Cerebral white matter …
Frontal cortex Temporal cortex
Superior temporal Mesial temporal
Amygdala Hippocampus
…
Cerebral cortex
…
…
…
Neuroanatomical ontology
Brain
Cerebellum Cerebrum
Cerebral white matter …
Frontal cortex Temporal cortex
Superior temporal Mesial temporal
Amygdala Hippocampus
…
Cerebral cortex
…
…
…
Brain
Cerebellum Cerebrum
Cerebral white matter …
Frontal cortex Temporal cortex
Superior temporal Mesial temporal
Amygdala Hippocampus
…
Cerebral cortex
…
…
…
Neuroanatomical ontology
MMSE
Cognitiveimpairment
Dementiaseverity
Cognition
Assessment
Neuropsychology
Alzheimer’s
Task and score description
MMSE
Cognitiveimpairment
Dementiaseverity
Cognition
Assessment
Neuropsychology
Alzheimer’s
Task and score description
Cognitive Assessment ontology
?
BIRN Portal
Web Based• Single Login to BIRN
Resources• Intuitive Interface• Flexible to Add Tools• Launch Local Visualization
Tools on Downloaded Data