Improvement of flexible design matrix in sparse Bayesian learning for multi task fMRI data analysis

Abstract

Detecting the active regions of the brain during cognitive functions is one of the important problems in cognitive neuroscience and disorder diagnosis. One of the promising approaches to solve this problem is to use General Linear Model (GLM) in functional Magnetic Resonance Imaging (fMRI) data. The main difficulty of the GLM method is to determine a flexible design matrix to model mentioned problem appropriately. In this paper, an approach to the critical construction of a flexible design matrix for precise detection of active regions of the brain, according to response in synthetic fMRI data based on GLM is presented. Should the design matrix is accurate, the next detection algorithm can extract a correct response from a very low signal to noise ratio (SNR); therefore, the presented design matrix is flexible to eschew over fitting and capture unfamiliar slow drifts. Using a sparse Bayesian learning method, some specific regressors are selected for flexible design matrix. Results show clearly prominent performance of suggested algorithm rather than conventional t-test methods and other conventional Bayesian analysis of fMRI data.