FMRI activation mapping using wavelet-based SPM (WSPM) integrated with gray-matter graphs

Abstract

In many fMRI task-evoked studies, localized brain activity can be detected by GLM fitting and statistical hypothesis testing. Statistical parametric mapping (SPM) is the classical method that requires Gaussian pre-smoothing of the data. Instead, the wavelet transform provides a compact representation of activation patterns. Wavelet based SPM (WSPM) is an extension of SPM that combines wavelet processing with voxel-wise statistical testing. However, classical wavelets used in WSPM are designed for regular Euclidean grids and thus not adapted to the convoluted nature of the cerebral cortex. We recently showed how WSPM using graph wavelets tailored to the gray-matter structure of the cortex can improve detection of brain activity in single-subject studies. Here we extend this approach to group-level analysis by modifying the design of the brain graph.