Predictive model for minimal hepatic encephalopathy based on cerebral functional connectivity

Minimal hepatic encephalopathy (MHE) is a common neurocognitive complication of liver cirrhosis, which have few recognizable clinical symptoms. Previous functional magnetic resonance imaging (fMRI) studies have found that widespread cortical and subcortical functional connectivity (FC) changes were significantly in patients with MHE. The goals of this study were twofold: 1) to construct predictive models for MHE, based on brain regional functional connectivity, 2) and to test feature selection method on p-value ranker based kernel principle component analysis (kPCA). Our study included thirty-two cirrhotic patients with MHE and twenty age-, gender-, and eduction-matched healthy controls. Using 1.5T MR, we obtained resting-state fMRI for each subject. Functional connectivities between 116 pairs of brain regions in patients with MHE were compared with those in control participants. Then, p-value ranker based kPCA was applied in feature selection step to reduce the dimension of input data. The best parameters of feature selection were chose based on 10-fold cross-validation of vector machines (SVMs). Finally, We found FC-based diagnostic model was accurate in differing MHE from normal controls with 86.5% accuracy, 88% specifity and 85% sensitivity.