Graph Diffusion Reconstruction Network for Addictive Brain-Networks Identification.

Functional Magnetic Resonance Imaging(fMRI) can reveal complex patterns of brain functional changes. The exploration of addiction-related brain connectivity can be more precise with fMRI data. However, it is still difficult to obtain addiction-related brain connectivity effectively from fMRI data due to the complexity and non-linear characteristics of brain connections. Therefore, this paper proposed a Graph Diffusion Reconstruction Network (GDRN), which could capture addiction-related brain connectivity from fMRI data of addicted rats. The diffusion reconstruction module effectively maintained the unity of data distribution by reconstructing the training samples. This module enhanced the ability to reconstruct nicotine addiction-related brain networks. Experiments on the nicotine addiction rat dataset show that the proposed model can effectively explore nicotine addiction-related brain connectivity.