Scholarlink

<p>Optimizing surgical decisions in breast cancer is critical. Choosing between mastectomy and breast-conserving surgery (BCS) is complex due to heterogeneous pre-operative clinical factors. We developed BrCaM (<i>Br</i>east <i>Ca</i>ncer <i>M</i>odel), a machine learning–based Clinical Prediction Model designed to analyze pre-operative surgical decision patterns. A dataset of 5100 patients (age range: 18–96 years) treated at a Breast Unit with standardized protocols was used. Surgeon-guided feature selection and an end-to-end machine learning pipeline were implemented. Multiple algorithms were evaluated; AdaBoost performed best using 10-fold cross-validation. BrCaM achieved an overall accuracy of 95% in distinguishing BCS from mastectomy. Feature selection identified clinically meaningful predictors that reflect established criteria influencing surgical decisions. In this retrospective setting, BrCaM captures real-world surgical decision patterns based on clinical factors. These findings support the consistency of current clinical practice and provide a foundation for future prospective validation as a clinical decision-support adjunct.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

BrCaM an artificial intelligence model for surgical decision making in breast cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月16日

查看全文
  • Daniela Evangelista,
  • Vasuk Gautam,
  • Luca Silvestri,
  • Mario Zanfardino,
  • Monica Franzese,
  • Massimiliano D’Aiuto

摘要

Optimizing surgical decisions in breast cancer is critical. Choosing between mastectomy and breast-conserving surgery (BCS) is complex due to heterogeneous pre-operative clinical factors. We developed BrCaM (Breast Cancer Model), a machine learning–based Clinical Prediction Model designed to analyze pre-operative surgical decision patterns. A dataset of 5100 patients (age range: 18–96 years) treated at a Breast Unit with standardized protocols was used. Surgeon-guided feature selection and an end-to-end machine learning pipeline were implemented. Multiple algorithms were evaluated; AdaBoost performed best using 10-fold cross-validation. BrCaM achieved an overall accuracy of 95% in distinguishing BCS from mastectomy. Feature selection identified clinically meaningful predictors that reflect established criteria influencing surgical decisions. In this retrospective setting, BrCaM captures real-world surgical decision patterns based on clinical factors. These findings support the consistency of current clinical practice and provide a foundation for future prospective validation as a clinical decision-support adjunct.