AI-assisted mammography screening
Software that reads screening mammograms alongside or before radiologists, catching more cancers and cutting the reading workload in large trials.
Overview
Deep-learning readers for mammography have regulatory clearance in Europe and the United States and are being tested as a replacement for one of the two human readers used in European screening programmes. The Swedish MASAI randomised trial found AI-supported screening detected more cancers with a lower reading workload, and the Danish and German programmes have reported similar results in practice. Risk-prediction models from the same images are being studied to personalise screening intervals.
How it works
Convolutional networks trained on millions of mammograms produce a suspicion score per breast; low-scoring exams go to single reading and high-scoring ones are flagged for human review or recall.
- Higher cancer detection in randomised and real-world studies
- Large reduction in radiologist workload
- Consistent performance across sites
- Long-term effect on interval cancers and mortality still being measured
- Regulatory and liability frameworks vary
- Risk of over-detection needs monitoring
Latest papers
topQuery for this technology: (TITLE:"AI-assisted mammography screening" OR ABSTRACT:"AI-assisted mammography screening") AND (cancer OR tumor OR tumour OR oncology OR carcinoma OR lymphoma OR leukemia OR leukaemia OR myeloma OR sarcoma OR melanoma OR glioma). Results are unfiltered search hits about AI-assisted mammography screening, not a curated reading list.