Machine learning techniques for personalized breast cancer risk prediction
This Collaborative research project develops methods, validates tools, and disseminates machine learning techniques for personalized breast cancer risk prediction (ML_BCP). The techniques are compared against state-of-the-art model-based approaches like Breast Cancer Risk Assessment Tool (BCRAT) and Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA).
-
Background: Breast cancer risk prediction modeling allows researchers to identify high-risk patients and reduce unnecessary interventions. Currently, breast cancer risk prediction models tend to exhibit low discriminatory accuracy (0.53-0.64). We are employing machine learning (ML) approaches to address current limitations and improve accuracy of outcome forecasts. This study compares the discriminatory accuracy of ML-based estimates against a pair of established methods: Breast Cancer Risk Assessment Tool (BCRAT) and Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA).
-
Methods: Eight simulated datasets and two retrospective samples were used to compare the performance of the proposed ML methods against state-of-the-art BCRAT and BOADICEA models.
-
Results : Predictive accuracy reached 88.28% (AU-ROC curve) using ML-Adaptive Boosting, 88.89% using ML-Random Forest, 62.40% for BCRAT, and 59.31% for BOADICEA.
-
Conclusions: The significant prediction improvement of the accuracy of classification of women with and without breast cancer achieved with ML algorithms is important in personalized medicine and may suggest prevention strategies and individualized clinical management.
...TBD...
- TBD ...