Sparse Group Bases for Multisubject fMRI Data

Considering that functional magnetic resonance imaging (fMRI) signals from multiple subjects (MS) can be represented together as a sum of common and distinct rank-1 matrices, new MS dictionary learning (DL) algorithm named sparse group (common + distinct) bases (sgBACES) is proposed. Unlike existing MS-DL algorithms ignore fMRI data’s prior information, it formulated penalized plus constrained matrix approximation, where l1 norm-based adaptive penalty, l0 regularization, lag-1 based autocorrelation maximization have been introduced in the minimization problem. Besides, spatial dependence among voxels has exploited for fine-tuning sparsity parameters. To my best knowledge, sgBACES first to effectively take both temporal information into account an MS-fMRI-DL framework. It also advantage not requiring separate coding stage. Studies on synthetic experimental datasets are used compare performance with state-of-the-art terms correlation strength computation time. emerged proposed enhanced signal-to-noise ratio (SNR) recovered time courses (TCs) precision maps (SMs). A 9.2% increase value over ShSSDL observed motor-task data.