Multimodal DL Model Predicts Breast Cancer Recurrence from Longitudinal Mammograms

A deep learning model combining pretreatment mammograms and clinical data predicted breast cancer recurrence with an AUC of 0.80 (95% CI 0.74-0.85) in a single-center cohort of 3923 patients, outperforming unimodal approaches.

• For recurrence subtype (local/regional/distant) among patients who recurred, the model achieved a balanced accuracy of 65.1% (95% CI 60.2%–71.1%) and an AUC of 0.72 (0.68–0.78).
• The model was trained and evaluated within a single institution without external validation.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.