中国现代手术学杂志2025，Vol.29Issue(5)：355-360,6.DOI:10.16260/j.cnki.1009-2188.2025.05.001
基于机器学习算法的腹腔镜结直肠癌根治术后吻合口漏风险预测
Machine learning-based prediction of anastomotic leakage following laparoscopic surgery in colorectal cancer patients
摘要
Abstract
Objective To analyze the factors influencing anastomotic leakage in colorectal cancer patients undergoing laparoscopic surgery,and to construct machine learning-based predictive models,thereby providing scientific evidence for clinical decision-making.Methods To analyze the factors influen-cing anastomotic leakage in colorectal cancer patients undergoing laparoscopic surgery,and to construct machine learning-based predictive models,thereby providing scientific evidence for clinical decision-making.Results Among the 400 patients,25(6.25%)developed anastomotic leakage.Analyses using the four machine learning models showed that gender had the highest feature importance in the logistic regression model,whereas preoper-ative albumin levels were identified as the most significant predictors in the decision tree,random forest,and BP neural network models.In terms of model performance,the random forest model achieved an accuracy of 0.950,specificity of 0.988,F1 score of 0.667,and AUC of 0.968 in the training set,outperforming the other three models.All models exhibited a slight decline in performance in the validation set.The random forest model still maintained an accuracy of 0.933,sensitivity of 0.500,specificity of 0.956,and AUC of 0.854,demonstrating favorable predictive capability.The logistic regression and decision tree models showed relatively low sensitivity and F1 scores,while the BP neural network model yielded acceptable accuracy but only moderate F1 score and AUC values.Conclusion The machine learning-based predictive models,especially the ran-dom forest model,can effectively predict the risk of anastomotic leakage following laparoscopic colorectal cancer surgery.These models are expected to assist clinicians in identifying high-risk patients and implementing timely,individualized preventive interventions.关键词
结直肠肿瘤/腹腔镜手术/吻合口漏/机器学习/预测模型Key words
colorectal neoplasms/laparoscopy/anastomotic leakage/machine learning/predictive model引用本文复制引用
禤锦峰,李炯先,黎峰,李嘉明,莫振昌,龚超..基于机器学习算法的腹腔镜结直肠癌根治术后吻合口漏风险预测[J].中国现代手术学杂志,2025,29(5):355-360,6.基金项目
广西壮族自治区卫生健康委自筹经费科研课题(Z-D20231667) （Z-D20231667）
