Diagnosis of Alzheimer's disease using Effective Connectivity of rs-fMRI

Brain functional networks based on resting-state functional MRI (rs-fMRI) have numerous potentials about brain networks and has been utilized to discriminate healthy control (HC) form individual with AD and MCI. More importantly, the default mode network (DMN) in rs-fMRI has been progressively utilized in disease identification since it was created. Previous research has mainly focused on analyzing single DMN with different techniques. In this study we carry out a classification measure to analyze the effective connectivity obtained by the DMN along with frontal-parietal network (FPN), dorsal attention network (DAN) and silence network (SN). Effective connectivity features set were generated by the spectral Dynamic Causal Modeling (spDCM) and later converted into a vectorial features space for individual subjects. SoftMax classifier and sequential feature selection (SFS) methods were utilized to diagnosis classification of disease. We acknowledge that, the four spDCM corresponding to DMN, DAN, FPN and SN seed exactly cover various functional region of individual subjects. Our classification results show the potency of using effective connectivity to analyze the AD identification from seed networks which can enhance the diagnosis accuracy.