The Protocol The Biometrics Weekly

Sign in Subscribe

AI/ML

AI/ML & Digital

Regulators Name the SaMD Statistical Gaps; The Journals Start Filling Them

MHRA's AI Airlock and FDA CDRH have now publicly enumerated the same regulatory-science gaps for AIaMD — and a wave of 2025–26 diagnostic-accuracy papers is racing to make those gaps citeable.

AI/ML & Digital

AI Sample-Size Reduction Has a Methodology; the Validation Infrastructure Is Missing

The arXiv sample-size preprint maps AI-derived prognostic covariates onto FDA's 7-step credibility framework — which makes the remaining gap between elegant method and trustworthy practice harder to ignore.

AI/ML & Digital

FDA's AI Pilot RFI and a Flurry of Methods Papers Force a Useful Question

A live FDA comment deadline, a serious AI sample-size preprint, and evidence that AI models overestimate their own generalizability by ~20% — not all pointing the same direction.

AI/ML & Digital

AI Diagnostic Submissions Finally Have a Principled F1 Toolkit

Two peer-reviewed methods papers give AI/ML SaMD analysts citable F1 inference tools — arriving precisely as FDA and MHRA name the metrics gap officially unresolved.