Deep Learning for Breast Cancer Risk Prediction: Application to a Large Representative UK Screening Cohort.

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Artificial Intelligence in Healthcare

Background:

• Early breast cancer detection is crucial for improving patient outcomes.
• Current screening mammography has limitations in predicting future cancer risk.
• Artificial intelligence offers potential for enhanced risk assessment in breast cancer screening.

Purpose of the Study:

• To develop and evaluate a deep learning tool for predicting breast cancer risk using screening mammograms.
• To assess the model's performance on a large, multi-site UK dataset.
• To explore the interpretability of the AI model for risk prediction.

Main Methods:

• Utilized the OPTIMAM Mammography Imaging Database (>300,000 patients).
• Trained a deep learning model on screening mammograms and patient age to predict cancer occurrence within 3 years.
• Evaluated model performance using the area under the receiver operating characteristic curve (AUC) on a hold-out test set.

Main Results:

• The AI model achieved an overall AUC of 0.70 (95% CI: 0.69, 0.72) on the test set.
• Performance was consistent across different screening sites, ethnicities, and age groups.
• Saliency maps provided insights into mammographic features influencing AI-predicted risk.

Conclusions:

• The developed AI tool demonstrates good predictive performance for breast cancer risk.
• The model is effective on a large, multi-site UK population dataset.
• AI holds promise for augmenting breast cancer screening and risk stratification.