Diagnosis of Alzheimer's Disease Based on Structural Graph Convolutional Neural Networks

In recent years, classification methods based on multi-modality images have been widely applied in the diagnosis of Alzheimer’s Disease(AD), and have achieved better performance than methods based on single-modality image. However, most multi-modality based classification methods only extract features from a single-modality image independently, and then concatenate them into a long vector for classification, without properly considering the correlation between different modalities. Based on this, this study proposes a structure graph convolutional neural network(GCN) method for AD classification. Firstly, using 90 regions of interest(ROIs) from different modal images as nodes in the graph, extract the positional relationships and specific features of ROIs to construct a structural diagram. Then, the GCN network is used to aggregate the adjacent node features of the structural graph, and pooling layer is used to obtain graph representations representing different modal image features. Finally, the final AD classification diagnosis result is output through the softmax layer. Experiments results show that this method has better classification performance than existing methods.