Chronic Kidney Disease Identification Based on Multi - Modal Graph Convolutional Network

Chronic kidney disease (CKD) is a widespread and significant chronic illness with a rising global prevalence. Due to the often subtle early symptoms, many patients are diagnosed only in the later stages of the disease, making early detection and intervention crucial to slowing its progression. Resting-state functional magnetic resonance imaging (rs-fMRI) has played a crucial role in advancing early CKD diagnosis. In this context, functional brain networks (FBNs) have attracted considerable attention as they provide insights into brain activity related to CKD. Graph convolutional networks (GCNs) have proven to be a promising method, enabling simultaneous FBN feature extraction and CKD identification while minimizing the reliance on manual feature engineering. To further enhance diagnostic accuracy, We present a multi-modal graph convolutional network (MMGCN) that incorporates various connectivity patterns. In this initial study, we combined multiple GCN modules to incorporate information from distinct connectivity patterns and evaluated the MMGCN approach using rs-fMRI scans from 100 subjects. The experimental findings demonstrate that our model achieved an 85.0 % accuracy, highlighting its potential in improving CKD diagnosis.