Survival machine learning model of T1 colorectal postoperative recurrence after endoscopic resection and surgical operation：a retrospective cohort study

Abstract Objective To construct a postoperative recurrence prediction model for patients with T1 colorectal cancer after endoscopic resection and surgical operation used survival machine learning algorithms. Methods Based on Two tertiary first-class affiliated hospitals, the case data of 582 patients with T1 colorectal cancer after endoscopic resection and surgical operation were obtained, and the information included patient personal information, treatment modalities, pathologically relevant information were extracted. After Boruta algorithm feature selection, Predictors of significant contributions were determined. The patients were divided into training set and test set at a ratio of 7༚3, five survival machine learning models: Random Survival Forest (RSF), GradientBoosting (GB), Survival Tree (ST), CoxPH and Coxnet for develop models.To interpret results based on SHAP algorithm. Results People at high risk of lymph node metastasis have a poorer prognosis, Different treatment methods have no significant impact on the prognosis of recurrence, The C-index and IBS scores of Random survival forest model in the test data are 0.848, 0.098, Its time-dependent AUC is 0.918, The interpretability analysis of the model showed that submucosal invasion depth < 1000µm, tumor budding grade BD1, Lymphovascular invasion and Perineural invasion is absent, well differentiated cancer cells, and tumor size < 20mm have positive effects on the model, Feature with negative gain is a contributing factor to the absence of recurrence in patients. Conclusions The prognostic model constructed by survival machine learning for patients with colorectal cancer has good performance. It can provide accurate individualized prediction.