Enhanced prediction of preoperative mesenteric lymph node metastasis in colorectal cancer using machine learning with CT‐based data

The detection of lymph node (LN) involvement is fundamental for staging colorectal cancer (CRC) and aids in clinical decision‐making. Traditionally, determining LN status predominantly relies heavily on histological examination of LN specimens, which can occasionally lead to overtreatment. This study aims to develop a clinical prediction model using machine learning algorithms to assess the risk of mesenteric LN metastasis preoperatively, based on computed tomography images and clinicopathological data from CRC patients. Our findings demonstrate that the predictive model based on XGBoost algorithms exhibited the optimal performance, with area under the curve values consistently stable across training (0.836, 95% confidence interval [CI]: 0.750–0.902) and validation (0.831, 95% CI: 0.688–0.927) cohorts. The model was further elucidated using SHapley Additive Explanation values, which ranked predictors in the XGBoost model by their importance, providing insights into the model's decision‐making process. Additionally, the force plot visualizes the contribution of each variable to the prediction for individual samples. The as‐obtained model may have the potential to aid in clinical treatment planning, optimize the selection of surgical methods, and guide the decision‐making process for adjuvant therapy before surgery. This study developed a machine learning model to predict mesenteric lymph node metastasis (MLNM) in colorectal cancer patients using preoperative CT images and clinicopathological data. The LASSO algorithm was employed for feature selection, while XGBoost provided optimal prediction performance. SHAP values were utilized for model interpretation, highlighting key predictors. The model offers potential to enhance preoperative decision‐making and treatment planning.