Daily Briefing

What Changed

Combined deployment of expert-guided vision and language tools now sets a new efficiency bar for diagnostic training, requiring integration of perceptual and documentation support to sustain throughput gains. [1]

Diagnostics & Pathology | Co-Annotator improved diagnostic efficiency for ophthalmology residents. arXiv posted a preprint study of Co-Annotator with ophthalmology residents across two academic institutions. Combined guidance increased correct diagnoses per minute by 40% and reduced comment editing time by 67% without compromising diagnostic accuracy [1].

Operations & Workflow | Parts consumption and cost per case can anchor medical-device service AI business cases. Ryan Makely said on the AI in Healthcare and Life Sciences Podcast that medical device field service leaders can build a hard-cost case for AI investment from parts consumption and cost per case [2].

Industry & Products | Bayesian Health, founded by Dr. Suchi Saria, partners with health systems. In the NEJM AI Grand Rounds podcast, Dr. Suchi Saria explained that deploying such real‑time clinical intelligence at the point of care augments frontline teams and addresses a significant shortfall in current health‑system AI approaches [3].

Research & Evidence | An LLM matched or exceeded clinician review for substance-use and self-harm detection. medRxiv posted a preprint using a two-phase diagnostic-accuracy design at a UK Type 1 Emergency Department with a conflict-adjudicated reference standard. The LLM's balanced accuracy matched or outperformed clinician review in all three domains (alcohol 0.942 v 0.930, p=0.635; drug 0.959 v 0.791, p<0.001; self-harm 0.982 v 0.908, p=0.004) [4].