Discriminative Group Sparse Representation for Mild Cognitive Impairment Classification

Abstract

Witnessed by recent studies, functional connectivity is a useful tool in extracting brain network features and finding biomarkers for brain disease diagnosis. It still remains, however, challenging for the estimation of a functional connectivity from fMRI due to the high dimensional nature. In order to tackle this problem, we utilize a group sparse representation along with a structural equation model. Unlike the conventional group sparse representation, we devise a novel supervised discriminative group sparse representation by penalizing a large within-class variance and a small between-class variance of features. Thanks to the devised penalization term, we can learn connectivity coefficients that are similar within the same class and distinct between classes, thus helping enhance the diagnostic accuracy. In our experiments on the resting-state fMRI data of 37 subjects (12 mild cognitive impairment patients; 25 healthy normal controls) with a cross-validation technique, we demonstrated the validity and effectiveness of the proposed method, showing the best diagnostic accuracy of 89.19% and the sensitivity of 0.9167.