Statistical parametric mapping of functional MRI data using wavelets adapted to the cerebral cortex

Wavelet approaches have been successfully applied to the detection of brain activity in fMRI data. Spatial activation patterns have a compact representation in the wavelet domain. However, classical wavelets designed for regular Euclidean spaces are not optimal for the topologically complicated gray-matter (GM) domain where activation is expected. We hypothesized that wavelet bases that are adapted to the structure of the GM, would be more powerful in detecting brain activity. We therefore combine (1) a GM-based graph wavelet transform as an advanced spatial transformation for fMRI data with (2) the wavelet-based statistical parametric mapping framework (WSPM). We introduce suitable design choices for the graph wavelet transform and evaluate the performance of the proposed approach both on simulated and real fMRI data. Compared to SPM and conventional WSPM, the graph-based WSPM shows improved detection of finely 3D-structured brain activity.