Wavelet based sparse source imaging technique

The present study proposed a novel multi-resolution wavelet to efficiently compress cortical current densities on the highly convoluted cortical surface. The basis function of the proposed wavelet is supported on triangular faces of the cortical mesh and it is thus named as the face-based wavelet to be distinguished from other vertex-based wavelets. The proposed face-based wavelet was used as a transform to gain the sparse representation of cortical sources and then was integrated into the framework of L1-norm regularizations with the purpose to improve the performance of sparse source imaging (SSI) in solving EEG/MEG inverse problems. Monte Carlo simulations were conducted with multiple extended sources (up to ten) at random locations. Experimental MEG data from an auditory induced language task was further adopted to evaluate the performance of the proposed wavelet based SSI technique. The present results indicated that the face-based wavelet can efficiently compress cortical current densities and has better performance than the vertex-based wavelet in helping inverse source reconstructions in terms of estimation accuracies in source localization and source extent. Experimental results further indicated improved detection performance of the face-based wavelet as compared with the vertex-based wavelet in the framework of SSI. It thus suggests the proposed wavelet based SSI can become a promising tool in studying brain functions and networks.