EXPLORING CONSISTENT FUNCTIONAL BRAIN NETWORKS DURING FREE VIEWING OF VIDEOS VIA SPARSE REPRESENTATION

Functional brain mapping under naturalistic stimuli such as video watching has been receiving greater interest in recent years. We presented a sparse representation based data-driven strategy to explore consistent functional brain networks during free viewing of continuous video streams. Compared with the traditional independent component analysis (ICA) based method, the novelty of our method is taking the intrinsic sparsity of whole-brain fMRI data into consideration and identify those highly descriptive dictionary atoms for sparse representation of fMRI signals. Our experimental results demonstrate that meaningful consistent functional brain networks can be mapped during free viewing of video stream by our method. We also compared the proposed method with ICA-based method.