Dual Attention Graph Convolutional Network Fusing Imaging and Genetic Data for Early Alzheimer's Disease Diagnosis

Alzheimer's Disease (AD) poses a significant global neurodegenerative challenge, underscoring the urgency of early clinical intervention. Our paper presents a novel approach for early AD diagnosis, focusing on a dual attention graph convolutional network that integrates multi-modal data. This methodology involves constructing image and gene graphs based on the image and genetic information of the subject. Graph convolution networks are then employed to extract embedded information from each graph. Enhanced diagnostic precision is achieved by utilizing self-attention and cross-attention mechanisms, facilitating the fusion of multi-modal state information crucial for early AD identification. Rigorous validation of the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset underscores the model's efficacy. Our method has demonstrated remarkable proficiency in diagnosing early-stage AD through experimental verification, assisting doctors in making accurate diagnoses of AD.