FIGURE 1.

Study workflow of this study: in Stage I, machine learning (ML) and natural language processing (NLP) models were trained respectively; in Stage II, two fusion approaches were used to combine and integrate the prediction information of each model; then, ROC, Shapley additive explanations (SHAP), and nomograms were used as evaluation and explanation tools; finally, we propose the utilization of the proposed model by quantifying the liver metastasis (LM) risk of colorectal cancer (CRC) postoperative patients.