Summary of the discussion on fMRI analysis at AHM06
- Short-term needs/questions:
- What are the capabilities available in Slicer for display of fMRI data and its integration with other types of data (structural and DTI)? Wendy gave an overview of the functionalities, but this seems like an opportunity for the educational Core to get involved.
- Need for a browser of fMRI data for quality control of the raw data. One of the tools mentioned was the matlab scripts by Tom Nichols at U. Michigan. They are freely available and should probably be investigated by Core 2.
- fMRI regularization using MRFs and anatomical priors seemed to be promising and is being integrated into Slicer. It largely addresses the problem of over-smoothing that Core 3 groups encounter when using standard Gaussian smoothing methods to boost SNR.
- Registration of Anatomical/DTI/fMRI: this is a research topic that is not solved yet. Difficulties include
- Non-linear warps in the EPI data
- It is unclear what is the proper way to treat the tensor data when warping the images
- In group registration, the mapping is not necessarily one-to-one, since the anatomy might be substantially different in different subjects. However, Core 3 groups feel that having an atlas and being able to transform from an individual to the atlas and back is crucial.
- Validation of registration is a challenge.
- On characterization of activation pattern, one question is what defines a shift of the same activation area vs. a separate activation area. Challenges here include
- Misalignment of fMRI with anatomy (see registration)
- Developing tools for estimating variance of the activation pattern.