🧪 New in The Lancet: testing conversational AI in actual clinic workflows.

Beth Israel Decon. evaluated Google's AMIE across 98 urgent primary care visits.

Physician observers never had to halt a session for safety, and the AI's top 7 differentials captured 90% of final diagnoses. #MedSky #MedAI

Examining patient/industry trade-offs in medical AI:
📉 85% of health research funds wasted
🧑‍⚕️Need for patient, physician involvement
⚖️ Emphasis on balance in partnerships
twitter
#MedAI #Ethics #GDPR #AIChasm #RCTs #LongitudinalStudies

https://tnyp.me/PVuATQWG/s

🧪 In a new study, Google's open MedGemma 4B (4 billion parameters) outscored Gemini 2.5 Pro on zero-shot chest X-ray, pathology, and retina classification, with a 500-fold compute cost gap against the priciest model compared. Dermatology was the exception (71.8% vs. 81.0% accuracy). #MedSky #MedAI

Two Bridge2AI Investigators receive the IEEE Innovation in Societal Infrastructure Award 🏆 Read more about our Investigators awarded for their leadership in privacy-preserving clinical research: bit.ly/4doccGq

#ArtificialIntelligence #BiomedicalInformatics #MedSky #MedAI #AISky #MLSky

Zhang et al. tested when a local AI agent should hand a diagnosis to a clinician. It diagnosed 551 simulated cases (five runs each). If the runs agreed, it proceeded alone. If not, a clinician would review. That caught 49 of 52 errors but also flagged 230 correct answers.
#MedSky #MedAI

A scoping review examines how AI and digital technologies are reshaping…

#AIinHealthcare #NeonatalCare #DigitalHealth #MedAI
https://news.google.com/rss/articles/CBMiX0FVX3lxTE5rUm9RS3dPQ1NRY2QwcVpuU3RxMF9ob2Q4N0VKeFRSenRWaGVfZ3RfdFQ3NUVfTmltRXFEUEZNaFFVQmtlQVMwdnJlSElILWRsZGJ0dTh6dG9KVUtoU3Rr

Join the CM4AI-AnVIL DREAM Challenge: Help pioneer in silico microscopy, build "virtual cells", and develop #GenerativeAI models for protein localization 🚀 

🏆 Up to $25k cash + $40k GCP credits
🔗 bit.ly/4AK1uE9

#AISky #MLSky #ArtificialIntelligence #BiomedicalInformatics #MedSky #MedAI #Genomics

🧪 Up to 92% of residents already use genAI, but letting models draft assessments risks never-skilling early learners. A new framework maps cognitive displacement by training stage: defer high-displacement tasks until core clinical reasoning can stand on its own.
#MedSky #MedAI

Health systems are treating the AI "kill switch" as a core design requirement before tools ever reach patients.
#MedSky #MedAI

🧪 An RCT of 848 caregivers in Japan found an AI chatbot boosted HPV vaccine literacy over government leaflets, but it didn't budge vaccination decisions at two weeks (40.3% vs 39.6%).

LLM conversational tools can clarify facts, yet trust still relies on doctors to make an impact.
#MedSky #MedAI 🛟

"Prospective evidence for conversational medical AI is hard, but non-negotiable" highlights why real-world trials are essential to assess safety, trust, workflow integration and clinical adoption of conversational AI. #MedAI #DigitalHealth