NHS AI Mammography System Detects 25% More Missed Cancers While Cutting Radiologist Workload by Up to 44%
Summary
A groundbreaking NHS AI mammography system detects 25% more missed cancers while cutting radiologist workload by up to 44%, offering a powerful solution to the UK's radiologist shortage, though human panels incorrectly overruled the AI on 93 cancer cases, highlighting urgent need for improved human-AI collaboration.
Key Points
- A large-scale AI mammography study across multiple NHS screening services shows the system achieves significantly higher cancer detection sensitivity than the original first human reader, detecting 25% of interval cancers missed in the standard double-read workflow.
- In a major reader study comparing traditional double-read workflows to an AI-enabled workflow, the AI system delivers non-inferior sensitivity and specificity while reducing total human reading workload by an estimated 36–44%, offering a potential solution to the UK's growing radiologist shortage.
- Despite promising results, challenges remain, including human arbitration panels incorrectly overruling correct AI recall decisions on 93 cancer cases, underscoring the need for improved AI explainability, better human-AI trust, and continuous monitoring to manage data drift in live clinical environments.