Preoperative MRI deep learning radiomics predicts breast cancer recurrence risk

DLRC model integrating deep learning, radiomics, and clinicopathological features from preoperative breast MRI predicted 3-year recurrence-free survival with an AUC of 0.79 (95% CI 0.67-0.91) in external validation (n=196).

• Retrospective study of 428 women (training/internal validation) and an additional 196 women for external validation.
• The combined DLRC model outperformed deep learning alone (AUC 0.75) and deep learning radiomics (AUC 0.82) with AUCs of 0.95, 0.89, and 0.79 in training, internal, and external validation, respectively.
• Performance was attenuated in external validation (AUC 0.79 vs 0.89 internal), highlighting need for further prospective testing.

Evidence grade: 3

Automated summary

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