Voxel selection framework with signal decomposition for fMRI based brain activity classification

This paper presents an fMRI signal analysis methodology using Empirical mean curve decomposition (EMCD) and mutual information (MI) based voxel selection framework. Previously, the fMRI signal analysis has been carried out either using empirical mean curve decomposition (EMCD) model or voxel selection on raw fMRI signal. The first methodology does signal decomposition that makes voxel selection process easy while the latter methodology does selection of relevant voxels (or features). Both these advantages are added by our methodology in which the frequency component is considered by decomposing the raw fMRI signal using Empirical mean and the voxels are selected from EMCD signal. The proposed methodologies are adopted for predicting the neural response. Experimentations are carried out in the openly available fMRI data of six subjects and comparisons are made with existing decomposition model and voxel selection framework. The comparative results demonstrate the superiority of the proposed methodology.