Fig. 1.

Flowchart of the study. Step 1: the development of DL-based CV models on esophageal variceal images (from center #1 Jintan Hospital) by 12-month EV bleeding; Step 2: training and internally validating MMML models, integrating clinical structured data and outputs by deep learning models (from center #1 Jintan Hospital), to predict 12-month EV bleeding; Step 3: externally test the MMML models (at Center #2 Soochow University). AutoML, automated machine learning; CV, computer vision; DL, deep learning; GBM, gradient boost machine; GLM, general linear model; MMML, multimodal machine learning; LIME, local interpretable model-agnostic explanation; PDP, partial dependence plot; RF, random forest; SHAP, SHapley additive explanations; XGBoost, eXtreme gradient boosting