Development and validation of a gradient boosting machine-based model for predicting tumor-infiltrating lymphocyte proportions in breast cancer.

objectiveTo construct and validate a multidimensional model for evaluating tumor-infiltrating lymphocyte (TIL) levels in breast cancer (BC) patients.

methodsThis retrospective study included 318 BC patients with 318 lesions confirmed by MRI and surgical pathology in the First Affiliated Hospital of Guangxi Medical University from January 2021 to December 2024. The patients were randomly split into a training set (n=228) and a validation (n=90) set, and further divided into low and high TIL groups based on immunophenotype assessment. Multivariate Logistic regression was used to identify independent predictors of TILs levels. A gradient boosting machine (GBM) model and a Logistic regression model were built. Model performance was assessed using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). An external validation cohort of 120 BC patients admitted between January 2025 and May 2025 was used to verify the predictive accuracy of the GBM model. Results Ki-67 level, internal enhancement pattern, multifocality, apparent diffusion coefficient (ADC) value, and neutrophil-to-lymphocyte ratio (NLR) were identified as independent predictors of high TIL levels. The GBM model demonstrated superior performance compared to the Logistic regression in the training set (AUC: 0.859 vs 0.724; P=0.014). Calibration curves indicated good agreement between predicted and observed probabilities in both models. DCA showed that the GBM model provided higher clinical utility. External validation yielded an AUC of 0.784 for the GBM model, with the calibration curve and DCA further confirming the model's good calibration and clinical applicability.

conclusionThe GBM-based multidimensional model reliably predicts TIL levels in BC patients, supporting prognosis evaluation and guiding personalized treatment strategies.