A Calibrated Stacking Ensemble Framework for High-Fidelity Breast Cancer Mortality Prediction Ensemble for Cancer Prognosis

Abstract

Background: Accurate breast cancer mortality prediction is constrained by severe class imbalance and the intrinsic sensitivity-specificity trade-off, which frequently induces high false-positive rates in clinical settings. To address this limitation, we propose a rigorously calibrated Static Stacking Ensemble framework.

Methods: Model derivation utilized the METABRIC cohort () via 5-fold stratified cross-validation, while the SEER dataset () was strictly reserved for independent external validation. The architecture employs out-of-fold (OOF) prediction stacking of Deep Neural Network, XGBoost, and Random Forest base learners. To replace standard static decision boundaries (default ), we integrated a data-driven threshold optimization strategy derived from Youden’s J-statistic.

Results: Baseline architectures exhibited severe clinical limitations, including a degraded specificity of in standalone deep learning and catastrophic weight collapse in evolutionary ensembles. By establishing an optimized clinical decision boundary (), the proposed framework successfully balanced the prognostic trade-off. Evaluated on the independent cohort, the model yielded a specificity of , a sensitivity of , an overall accuracy of , and an -score of . Furthermore, the framework achieved an area under the receiver operating characteristic curve (ROC-AUC) of , a robust precision-recall curve area (PR-AUC) of , and a highly calibrated Brier score of .

Conclusion: The integration of OOF representation learning with mathematically optimized thresholding delivers a calibrated Clinical Decision Support System (CDSS). This framework minimizes prognostic false alarms and robustly stratifies high-risk patients across heterogeneous cohorts, significantly outperforming isolated baseline architectures.