Machine Learning Models for Breast Cancer Risk Prediction Using Clinical, Imaging, and Genomic Predictors. A Comprehensive Study

Breast cancer is the most commonly diagnosed cancer among women worldwide, and survival outcomes are strongly dependent on stage at diagnosis. Early detection is therefore critical for reducing mortality and disease burden. However, traditional breast cancer risk prediction models show limited accuracy in certain populations, including younger women and those with dense breast tissue. In recent years, machine learning (ML)–based models have been developed to improve risk prediction by integrating clinical, imaging, and genomic data, but evidence regarding their performance and clinical utility remains fragmented and heterogeneous. The purpose of this research is to systematically evaluate ML-based breast cancer risk prediction models and quantitatively synthesize their predictive performance. A systematic review will be conducted in accordance with PRISMA 2020 guidelines. Data extraction will follow the CHARMS checklist, and methodological quality and risk of bias will be assessed using PROBAST-AI. Where appropriate, random-effects meta-analyses will be performed to pool discrimination metrics, primarily the area under the receiver operating characteristic curve (AUC).