na-mic national alliance for medical image computing fmri within namic sandy wells, polina golland...
DESCRIPTION
National Alliance for Medical Image Computing fMRI Detection/Regularization Smarter strategies for smoothing –MRF priors (MIT/BWH) Wanmei Ou, Polina Golland, Sandy Wells –Surface-based vs. volumetric smoothing (MGH) Anastasia Yendiki, Doug Greve, Bruce Fischl Example: MIND fMRI reliability study –Sensorimotor paradigm –10 subjects on 2 visits at each of 4 sites –We thank Randy Gollub for providing the MIND dataTRANSCRIPT
NA-MICNational Alliance for Medical Image Computing http://na-mic.org
fMRI within NAMIC
Sandy Wells, Polina Golland
Discussion moderator: Andy Saykin
National Alliance for Medical Image Computing http://na-mic.org
fMRI Update• Algorithms for time-series analysis
– Regularization/smoothing– Segmentation/clustering
• Enabling methodologies– Joint analysis with other modalitites– Group analysis
• Core 1 / Core 3 projects to apply to clinical data• Core 1 / Core 2 projects to integrate into NAMIC-kit
National Alliance for Medical Image Computing http://na-mic.org
fMRI Detection/Regularization• Smarter strategies for smoothing
– MRF priors (MIT/BWH)• Wanmei Ou, Polina Golland, Sandy Wells
– Surface-based vs. volumetric smoothing (MGH)• Anastasia Yendiki, Doug Greve, Bruce Fischl
• Example: MIND fMRI reliability study – Sensorimotor paradigm– 10 subjects on 2 visits at each of 4 sites– We thank Randy Gollub for providing the MIND data
National Alliance for Medical Image Computing http://na-mic.org
Surface vs. Volume Smoothing
Surface
Volume
• Four subjects (fixed-effects, single visit), 15mm FWHM:
• Demonstrated better detection power
National Alliance for Medical Image Computing http://na-mic.org
Functional Hierarchy/Segmentation
• Hierarchical clustering of time series data (MIT)– Polina Golland, Bryce Kim, Danial Lashkari,
• Simultaneously estimate – Representative “signatures”– Which signature best describes each voxel
• Example: diverse set of visual and mental tasks– localizer, rest, movie, etc.; ~1 hour of fMRI data– 7 subjects
National Alliance for Medical Image Computing http://na-mic.org
Hierarchy in Single Subject
AuditoryMotor
High Visual
?
STS+
? ?
Visual Motor+Aud
Motor+AudRetinotopic
High Visual
Intrinsic Stimulus Dependent
STS?
National Alliance for Medical Image Computing http://na-mic.org
Group Analysis of 2 systemsIndividual Maps Group Average
NA-MICNational Alliance for Medical Image Computing http://na-mic.org
Enabling Methodologies
Core 1 / Core 2 / Core 3
National Alliance for Medical Image Computing http://na-mic.org
fMRI/DTI Connectivity• DTI-based Connectivity Analysis
– Path of interest analysis (MGH)– Probabilistic tractography (MT/BWH/Harvard)
• Strength of connection between ROIs• Tri Ngo, C-F Westin, Marek Kubicki, Polina Golland
• ROIs from fMRI– Color Stroop in Schizophrenia– 15 subjects in each group
• Implementation in NAMIC-kit in progress
National Alliance for Medical Image Computing http://na-mic.org
Anatomical Analysis• Cortical segmentation and flattening (MGH)
– Freesurfer tools, now compatible with Slicer– Doug Greeve, Bruce Fischl, Steve Pieper
• Conformal mapping of the cortex (Georgia Tech)– Yi Gao, John Melonakos, Allen Tannebaum– Filters in ITK
National Alliance for Medical Image Computing http://na-mic.org
Population Registration• Information-theoretic group-wise alignment (MIT/MGH/BWH)
– Integration into NAMIC-kit in progress– In the fututre: non-rigid deformations using B-splines
Unaligned input
Aligned output
–Serdar Balci, Lilla Zollei, Mert Sabuncu, Sandy Wells, Polina Golland
National Alliance for Medical Image Computing http://na-mic.org
EPI Registration/De-Warping• Combine segmentation and registration with Physics-
based modeling of susceptibility (MIT/BWH/fMRIB)– Accurate registration of fMRI to anatomical MR– Retrospective correction of EPI distortions– Clare Poynton, Sandy Wells, Mark Jenkinson
Acquired
Estimated