fmri in namic
DESCRIPTION
fMRI in NAMIC. Facilitator: Polina Golland Presenters: Jim Fallon and Andy Saykin. fMRI and NAMIC. NAMIC Core 1 projects focus on structure Anatomical DTI Many of us are interested in fMRI Core 1: analysis Core 3: tool for study of the disease Potential for new collaborations. - PowerPoint PPT PresentationTRANSCRIPT
NA-MICNational Alliance for Medical Image Computing http://na-mic.org
fMRI in NAMIC
Facilitator: Polina Golland
Presenters: Jim Fallon and Andy Saykin
National Alliance for Medical Image Computing http://na-mic.org
fMRI and NAMIC
• NAMIC Core 1 projects focus on structure– Anatomical– DTI
• Many of us are interested in fMRI– Core 1: analysis– Core 3: tool for study of the disease
• Potential for new collaborations
National Alliance for Medical Image Computing http://na-mic.org
fMRI Status Update
• Basic analysis functions in ITK (GE/Kitware)• User Interface in Slicer (BWH)
• Advanced detection/analysis– MIT/BWH – anatomically guided fMRI detection– UC Irvine – localization of activation peaks– Other Core 1 groups
• Integrated visualization of anatomy & function
National Alliance for Medical Image Computing http://na-mic.org
Our goals
• Not to replicate existing analysis tools
• To identify problems that are – important to Core 3– interesting to Core 1
• Use NAMIC to create new collaborations
National Alliance for Medical Image Computing http://na-mic.org
Our findings
• Some of the “problems” have already been “solved”
• Many items on the “wish list” are in reach– Especially with help of Core 2
• There are some really hard and interesting problems
National Alliance for Medical Image Computing http://na-mic.org
No Smoothing Gaussian MRF
Anatomically guided fMRI detection
With anatomy
Wanmei Ou, MIT
National Alliance for Medical Image Computing http://na-mic.org
Quality control for fMRI• Spatiotemporal browser designed for quality control
during preprocessing of single subject time series data or contrasts– Easy loading of raw scan formats– Easy navigation through time & space– Quantify signal to noise– Identify temporal spikes optional smoothing– Identify spatial distortion
• B0 field map and phantom optional adjustment
• Also feature to identify outliers in group data
• Tom Nichols at U. Michigan has something like this tool in Matlab. Core 2?
National Alliance for Medical Image Computing http://na-mic.org
Managing fMRI findings• fMRI activation cluster utility
– Need to create functional ROI (fROI) label maps for use in subsequent analyses
– Assuming user has created a thresholded activation map
• Ability to choose activation clusters to include in the label map
• User should be able to choose label values and provide a name in a text field for each cluster
– Extract data from these clusters• Individual time series or for group data
• Core 2?
National Alliance for Medical Image Computing http://na-mic.org
Outline of this discussion
• Presentations (15-20min)– Jim Fallon – Andy Saykin
• Questions (15-20min)– Ask the speakers more detailed questions
• Discussion/brainstorming on how we might solve these problems
National Alliance for Medical Image Computing http://na-mic.org
Major Themes
• Integration with anatomical and DTI:– Anatomically accurate and precise integration of all
modalities, including fMRI, DTI, into a single analysis framework
• Characterizing fMRI activation areas:– Invent new ways to describe active areas and how
they change from an experiment to an experiment. This ties into population analysis of activation.
National Alliance for Medical Image Computing http://na-mic.org
Jim Fallon
National Alliance for Medical Image Computing http://na-mic.org Fallon
Occip
Heschl’s
Frontal pole
7
ITG
STG
CB
DMPFC
DLPFC
VMPFC
LOF
IFG
Critical samples in BOLD
National Alliance for Medical Image Computing http://na-mic.org
Anterior ViewAnterior-Inferior View
Variability in population
NA-MICNational Alliance for Medical Image Computing http://na-mic.org
National Alliance for Medical Image Computing http://na-mic.org
DLPFCBA 46
BA 7
SLF-2
NA-MICNational Alliance for Medical Image Computing http://na-mic.org
National Alliance for Medical Image Computing http://na-mic.org
McCarthy, 2004
NA-MICNational Alliance for Medical Image Computing http://na-mic.org
beta map fBIRN phantom sensorimotor task
Activation patterns mixture model (Kim, et al, 2005)
Thresholded voxels (p<0.05)
Add 20% “gutter region” around each strictly defined area (e.g., DLPFC) to capture “rogue” functional activations in different subject and patientpopulations…”DLPC PLUS”
National Alliance for Medical Image Computing http://na-mic.org
Andy Saykin
National Alliance for Medical Image Computing http://na-mic.org
fMRI Specific Applications• Tool for assessment of test-retest reproducibility
of fMRI experiments– Simple approach would be calculating intraclass correlation
coefficients for voxels and ROIs• Useful but limited value because of fluctuations in exact peak and
distribution of activation foci
– A more sophisticated approach would include identification and extraction of key spatiotemporal features
• Prior knowledge could be used to inform regarding importance• Reproducible features could be quantified
• A related tool would provide an analysis of longitudinal stability and change– Consider reliable change index approach applied to activation
maps
National Alliance for Medical Image Computing http://na-mic.org
Multimodality Integration
• Registration of fMRI, DTI and anatomic MR– individual and group data
• Easy mapping between atlas space and native scan space– Permit warping from native space to atlas space or
vice versa
• Automated parcellation of cortical surface and subcortical gray matter structures– Generate label maps– Extract quantitative data from labeled ROIs or fROIs
• e.g. examine atrophy within functionally derived ROI
National Alliance for Medical Image Computing http://na-mic.org
Multimodality Integration - II
• Integrate measures of connectivity– Voxel by voxel and labeled ROI measures of
connectivity within single subject time series• Resting & Task-induced connectivity
• Changes in strength of connectivity over time
– important for learning and habituation experiments
– Relation to existing work • PLS, SEM, DCM, POI, other?
– Visualization tool to display strength of connectivity including functional and neuroanatomic (tractography)
National Alliance for Medical Image Computing http://na-mic.org
Questions and Discussion
National Alliance for Medical Image Computing http://na-mic.org
Major Themes
• Integration with anatomical and DTI:– Anatomically accurate and precise integration of all
modalities, including fMRI, DTI, into a single analysis framework
• Characterizing fMRI activation areas:– Invent new ways to describe active areas and how
they change from an experiment to an experiment. This ties into population analysis of activation.