Diagnostic performance of FDA-cleared and CE-marked AI systems for mammography screening: a systematic review and meta-analysis.
Putu Irma Wulandari, Ziba Gandomkar, Mary Rickard et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the diagnostic performance of FDA/CE-cleared AI systems for mammography screening, analyzing 42 studies. The pooled standalone AI AUC for digital mammography was 0.89, with a sensitivity of 76.3% and specificity of 89.6%. The findings suggest that AI can effectively reduce radiologist workload in clinical workflows.
Key findings
- Pooled standalone AI AUC for digital mammography was 0.89 (95% CI: 0.85-0.92).
- AI sensitivity was 76.3% (95% CI: 66.2-84.2%) and specificity was 89.6% (95% CI: 84.5-93.2%).
- Prospective evaluations showed non-inferior cancer detection with significant workload reduction compared to standard double reading.
- AI as an additional reviewer detected more cancers with minimal increase in recalls.
Why it matters
The integration of AI systems in mammography screening could enhance diagnostic accuracy while alleviating the workload of radiologists, potentially leading to improved patient outcomes and more efficient clinical workflows.
What to keep in mind
Risk of bias was frequently high, particularly for flow and timing in the included studies.