Construction and validation of machine learning-based predictive model for colorectal polyp recurrence one year after endoscopic mucosal resection

Yi-Heng Shi ^1,2, Jun-Liang Liu ¹, Cong-Cong Cheng ^1,2

⁰Department of Gastroenterology, The Affiliated Hospital of Xuzhou Medical University, Xuzhou 221002, Jiangsu Province, China.

World Journal of Gastroenterology +

|

March 24, 2025

View abstract on PubMed

Summary

A new machine learning model predicts colorectal polyp recurrence after endoscopic mucosal resection (EMR). This tool aids in personalized follow-up, improving patient care and reducing cancer risk.

Area Of Science

Gastroenterology and Hepatology
Oncology
Medical Informatics

Background

Colorectal polyps are precancerous lesions for colorectal cancer.
Early detection and removal of polyps reduce cancer mortality.
Endoscopic mucosal resection (EMR) has a high recurrence rate, lacking predictive models.

Purpose Of The Study

To develop and validate a machine learning (ML) model for predicting colorectal polyp recurrence one year post-EMR.
To identify key risk factors for polyp recurrence.
To improve post-EMR patient management strategies.

Main Methods

Retrospective and prospective data collection from 1694 and 166 patients, respectively.
Feature selection via logistic regression, with five ML algorithms employed.
Model performance evaluated using AUC, DCA, and SHAP analysis.

Main Results

Identified 8 independent risk factors for recurrence.
The eXtreme Gradient Boosting (XGBoost) model achieved high AUCs (0.909-0.963) across datasets.
SHAP analysis highlighted smoking history, family history, and age as key predictors.

Conclusions

The XGBoost model demonstrates excellent predictive performance for colorectal polyp recurrence.
This ML model can guide individualized colonoscopy follow-up schedules.
The findings support enhanced patient surveillance after EMR.

Keywords:

Colorectal polyps Machine learning Predictive model Risk factors SHapley Additive exPlanation