MRI deep learning for triple-negative breast cancer: pooled AUC 0.85 but low-certainty evidence

Meta-analysis of nine studies (2985 patients): MRI deep learning distinguishes triple-negative from other breast cancer subtypes with pooled AUC 0.85 (95% CI 0.81-0.88), sensitivity 0.83, specificity 0.87; certainty of evidence low.

• Pooled positive likelihood ratio was 6.47 (95% CI 3.98-10.54), negative likelihood ratio 0.24 (95% CI 0.18-0.33), and diagnostic odds ratio 26.53 (95% CI 13.86-50.81).
• Heterogeneity was moderate and no publication bias was detected.
• GRADE certainty of evidence was low; limited external validation and methodological variability warrant standardized multicenter validation before clinical implementation.

Evidence grade: low

Automated summary

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