Time-Domain Quantum Diffusion Graph Networks for fMRI in Alzheimer’s Disease Diagnosis

This study investigates fMRI functional network characterization for Alzheimer’s disease diagnosis and the identification of regional connectivity patterns associated with cognitive decline. We propose a Biomedical Temporal Quantum Walk Graph Network (BTQWGN), which integrates quantum walk dynamics with temporal graph convolution for multi-scale network representation learning. Functional brain networks are constructed using Pearson correlation, while quantum diffusion kernels are derived from a Hamiltonian operator. The kernels are subsequently combined with empirical connectivity to obtain quantum-enhanced adjacency matrices. The categorical classification loss and the group-level consistency objective jointly enhance predictive accuracy while reducing inter-subject variability in connectivity biomarkers. Comprehensive experiments demonstrate that BTQWGN achieves superior performance over machine learning baselines and graph neural architectures in Alzheimer’s disease diagnosis. Furthermore, the model identifies clinically meaningful brain regions involved in memory storage and emotional regulation, providing biologically interpretable insights into pathological mechanisms.