A Task Performance-guided Model of Functional Networks Identification

摘要

Understanding the organization of brain cortical functions has long been an intriguing research domain. Since the popularity of whole-brain in vivo imaging techniques, such as functional magnetic resonance imaging (fMRI), researchers have developed various brain network analysis methods for functional network identification, including principal component analysis (PCA), independent component analysis (ICA), and the methods based on sparse representation. However, all these aforementioned methods were either data-driven or hypothesis-driven, while the individual behavioral or task performance interpretation of the identified networks remains to be examined. To this end, we proposed a framework that incorporates the behavioral measures of in-scanner task performance to a hybrid temporo-spatial dictionary learning and sparse representation pipeline to identify group-wise basic networks from task fMRI data. The identified holistic functional networks were intrinsically guided by behavioral measures that encode across-individual functional variations. This framework was applied to working memory task fMRI data and the results demonstrate the effectiveness of the proposed framework.