Episode Details

This episode unpacks a comprehensive review spanning ten years of AI in screening mammography. Listeners will learn why pooled accuracy figures mask deep implementation challenges, how AI is effectively reducing reading workloads by up to 44.2%, and why local calibration is far more critical than an algorithm's out-of-the-box claims.

Key points

• Standalone AI accuracy pooled an AUC of 0.890 across over 1.2 million examinations, but the prediction interval was so wide (0.731–0.960) that out-of-the-box performance cannot be assumed.
• A bivariate model showed a summary sensitivity of 73.3% and a specificity of 92.4%, highlighting that algorithms miss roughly one in four screen-relevant cancers and should not be used as autonomous replacements.
• The landmark MASAI randomized trial demonstrated a 44.2% reduction in screen-reading workload alongside a cancer detection rate ratio of 1.29.
• Non-randomized implementations showed higher detection gains (pooled ratio 1.22), but these are highly vulnerable to confounding factors like self-selection bias.
• System updates and hardware changes severely impact AI performance; one UK program saw recall rates jump threefold simply due to a mammography equipment software upgrade.

Source: Ten Years of Artificial Intelligence in Screening Mammography: A Systematic Review and Meta-Analysis of Diagnostic Accuracy and Clinical Implementation (Literature Published 2015-2025) - Diagnostics (Basel, Switzerland), 2026 (CC BY)

This spot is available. Reach clinicians, health-system leaders and medtech and pharma teams following AI in medicine. Sponsor the show

This episode is an AI-generated conversation summarising a public document; the hosts' voices are synthetic. It is for information only and is not medical advice. Always refer to the original source.

Full transcript: https://ai-in-medicine-podcast.vercel.app/episodes/mammography-s-ai-decade-from-retrospective-illusions-to-randomized-rea-1566ad