Predicting pathological upgrading of colorectal polyps after endoscopic submucosal dissection: development and validation of a machine learning model
Year: 2026
Type: article
Abstract: To develop and validate a machine learning–based model for predicting pathological upgrading of colorectal polyps after endoscopic submucosal dissection (ESD), providing quantitative support for clinical decision-making. We retrospectively analyzed 561 patients with colorectal polyps who underwent ESD between January 2020 and December 2025. Clinical characteristics, laboratory data, and endoscopic features were collected. Least absolute shrinkage and selection operator (LASSO) regression was used for feature selection. Ten machine learning models (SVM, LR, KNN, DT, RF, LightGBM, GNB, XGBoost, GBT, and MLP) were constructed. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis (DCA). The Shapley Additive exPlanations (SHAP) method was applied for model interpretation. Among 561 patients, 100 (17.8%) experienced pathological upgrading. The dataset was randomly and stratified into training set (n = 392) and test set (n = 169) at a 7:3 ratio. The upgrading group showed larger polyp size, positive fecal occult blood (FOB), solitary (single) lesions, villous components, reddish surface, uneven morphology, and erosion. Among ten candidate machine-learning algorithms compared via repeated stratified nested cross-validation within the training set, logistic regression showed the best discrimination (cross-validated AUC = 0.795, 95% CI 0.667–0.927) and was selected as the final model; on the internal test set, it achieved an AUC of 0.799 (95% CI 0.701–0.879). Calibration and DCA demonstrated reasonable agreement and potential clinical utility in exploratory analyses. SHAP analysis identified villous component, FOB status, and mucosal erosion as the top three predictors contributing most to the model’s predictions. A logistic regression-based model showed promising discrimination for predicting pathological upgrading after ESD in internal validation, with acceptable calibration and potential clinical benefit. External validation is needed before this model can support individualized clinical decision-making.
Source: Scientific Reports
Authors: Fuli Gao, Jian Chen, Luojie Liu, Xiaodan Xu
Institutions: Soochow University, Changshu No.1 People's Hospital
Language: English
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7227274371
