The embodiment of the invention provides a method and a
system for predicting serious postpartum hemorrhage. The method comprises the following steps of: firstly, acquiring multi-
modal medical image data, and preprocessing and aligning the multi-
modal medical image data; performing segmentation
processing on the image through a sliding window strategy, and extracting structural features, texture features, hemodynamic features and
blood vessel distribution features from a segmentation result; then the features serve as nodes to construct a graph structure, the relation between the nodes is coded through a graph convolutional network, and the relevance of the features on the topology level is calculated; and finally, fusing the features according to the relevance, inputting a prediction model, and outputting a prediction result. According to the method, through multi-
feature fusion and multi-
modal data integration, the prediction capability and the prediction accuracy are improved, the robustness of the model in different data environments is enhanced, the bleeding risk can be recognized in time in the middle and later periods of
pregnancy of the pregnant woman, early intervention and treatment suggestions are provided, and the risk of bleeding is reduced. The occurrence rate and fatality of serious postpartum hemorrhage are obviously reduced.