Autism Disease Prediction with Self-Supervised Graph Contrastive Learning

The method based on graph learning is widely used in the field of disease prediction, and the graph convolutional neural network model has made great achievements in the field of disease prediction. However, there are still some limitations in the field of disease prediction based on the existing multi-modal medical data. To this end, we apply graph contrastive learning to the field of disease prediction, and propose a data enhancement method of self-supervised graph contrastive learning method. Graph contrastive learning focuses the model on important nodes and edges by introducing perturbations, so as to force the model to recognize the underlying semantic information, so as to better help it learn the graph representation. Experiments on two challenging medical datasets show that our approach is superior to the baseline approach, verifying the robustness of our proposed model and the effectiveness of graph contrastive learning in disease prediction.