The invention discloses a
radiation proctitis causal prediction method and
system fused with multi-
modal information. The method comprises the following steps: acquiring and preprocessing multi-
modal medical data of a patient; multi-
modal high-dimensional joint features are extracted through a deep network; a depth generation decoupling model is constructed, and the joint features are decoupled into independent intrinsic pathogenic mechanism characterization and exogenous
hybrid interference characterization; and based on a
causal inference theory, blocking the influence of the
hybrid characterization on prediction by applying intervention, and calculating a
radiation proctitis causal prediction value only by using the intrinsic pathogenic mechanism characterization. According to the method, through causal characterization decoupling and causal intervention prediction,
hybrid bias caused by heterogeneous data is effectively eliminated, shortcut learning of the model is avoided, and the generalization ability, robustness and technical reliability of the model in complex clinical environments such as cross-center and multi-device environments are improved.