基于机器学习算法构建胃癌新辅助化疗患者术前营养不良的预测模型OA

Objective To investigate factors influencing preoperative malnutrition in gastric cancer patients receiving neoadjuvant chemotherapy,and to construct a predictive model using machine learning algorithms,providing an auxiliary tool for the early clinical identification of high-risk patients.Methods The clinical data of 480 gastric cancer patients who received neoadjuvant chemotherapy in Fujian Cancer Hospital from January 2021 to July 2025 were collected.They were randomly divided into modeling group(336 cases)and validation group(144 cases)at a ratio of 7∶3.The key predictive factors of malnutrition were screened through least absolute shrinkage and selection operator(LASSO)regression,seven machine learning models were constructed,and the predictive efficacy was compared.The Shapley additive explanations(SHAP)method was adopted to analyze the feature importance of the optimal model.Results The incidence of malnutrition in modeling group and validation group was 48.8%(164/336)and 45.8%(66/144),respectively.LASSO regression identified five related factors:age≥65 years old,female,hypoalbuminemia,body mass index(BMI)<18.5kg/m2,and 4 cycles of neoadjuvant chemotherapy.The extreme gradient boosting(XGBoost)model performed the best in both modeling group and validation group.SHAP analysis showed that the order of feature importance was BMI,neoadjuvant chemotherapy cycle,gender,age,and serum albumin level.Conclusion The XGBoost prediction model constructed in this study has good efficacy and interpretability,and can be used to screen individuals at high risk of preoperative malnutrition in patients with gastric cancer undergoing neoadjuvant chemotherapy.