A new study reveals people trust human doctors more than AI, rating them higher on identical information. AI medical advice faces skepticism due to unfamiliarity, perceived lack of empathy, and fear of errors.

Read Full Article

#DoctorVsAI #TrustInDoctors #AIinMedicine #MedicalAI #HumanTouch

People accept euthanasia decisions made by AI less than those made by human doctors. As the role of AI in society and medical care is expanding, the researchers highlight that it’s important to understand the reactions of ordinary people so that future systems can be perceived as morally acceptable. 🤖⚖️💔

Read Full Article

#Euthanasia #AIethics #MedicalAI #HumanVersusAI #SocialAcceptance

Commercial LLMs (GPT-4o, Gemini 1.5 Pro-002, Llama 3.1 8B Instruct) 🧠💊, including after finetuning, showed poor average generalization accuracy (21-42%) 📉 when attempting to answer medical questions related to new drugs or updated recommendations.

Read Full Article

#CommercialLLMs #GPT4o #Gemini15Pro002 #Llama31 #MedicalAI

AI models struggle to distinguish between users’ beliefs and facts, which could be particularly harmful in medical settings ⚕️🤖💔

Read Full Article

#AIEthics #MedicalAI #Misinformation #UserBeliefs #HealthTech

AI model can read and diagnose a brain MRI in seconds

Read Full Article

#AI #BrainMRI #MedicalAI #FastDiagnosis #NeuroTech

AI chatbots give inaccurate medical advice says Oxford Uni study

Read Full Article

#AIMedicalSafety #OxfordStudy #ChatbotAccuracy #MedicalAI #HealthTech

AI cancer tools risk “shortcut learning” rather than detecting true biology

Read Full Article

#AICancerTools #ShortcutLearning #BiologyFirst #MedicalAI #DiagnosticsReliability

🩺 A new study by Mass General Brigham researchers found that generative AI models struggle with clinical reasoning, particularly in generating differential diagnoses.

🧑‍⚕️ While AI models achieved over 90% accuracy in final diagnoses when given all information, they failed to produce appropriate differential diagnoses more than 80% of the time in a study evaluating 21 large language models.

📊 Researchers developed a new metric, PrIME-LLM, to evaluate AI's competency across stages of clinical reasoning, not just final accuracy, revealing imbalances in performance.

📈 Newer AI models generally outperformed older ones, and Grok 4 and GPT-5 achieved the highest PrIME-LLM scores (78%), but the study emphasizes the continued need for human oversight in healthcare AI.

#AIinHealthcare #MedicalAI #HealthTech

AI chatbots misdiagnose in over 80% of early medical cases, study finds

Read Full Article

#AImisdiagnosis #medicalAI #healthtechnews

Substantial amount of medical information provided by 5 popular chatbots inaccurate and incomplete, with half of their answers to health questions “problematic”. Grok generated significantly more highly problematic responses. Gemini generated the fewest highly problematic responses.

Read Full Article

#MedicalAI #AIinHealthcare #DigitalHealth

AMA 2026 survey shows 80%+ of physicians now use AI, boosting efficiency and diagnostics but raising privacy and skill-loss concerns

Read Full Article

#AIinMedicine #MedicalAI #HealthcareInnovation

🏥 AI Outperforms Doctors in Emergency Triage
A Harvard study published in Science revealed AI systems diagnosed more accurately than human doctors in high-pressure emergency medicine triage. In one experiment with 76 patients, AI achieved 67% accuracy compared to 50%-55% for human doctors. OpenAI's o1 reasoning model reached 82% accuracy with more detail, surpassing expert humans.

🩺 AI Excels in Treatment Planning
AI also outperformed 46 doctors in creating longer-term treatment plans, scoring 89% compared to humans' 34% in case studies involving antibiotic regimes and end-of-life processes.

💡 AI as a "Second Opinion" Tool
Researchers emphasize AI is not replacing doctors but acting as a powerful "second opinion" tool. The study focused on text-based data, excluding patient appearance or distress levels. Experts see AI reshaping medicine through a "triadic care model" of doctor, patient, and AI.

#AIinHealthcare #MedicalAI #FutureofMedicine

Detection of multiple diseases via retinal imaging

Read Full Article

#RetinaImaging #DiseaseDetection #MedicalAI

✅ LLM-generated summaries for emergency department patients received higher ratings from physicians than human-authored summaries across accuracy, completeness, and clinical utility in a blinded study of 99 encounters.

✅ LLM summaries were rated 0.78 points higher in accuracy, 0.43 points higher in completeness, and 0.67 points higher in clinical utility compared to physician-authored summaries.

✅ The LLM utilized 53.2% of available patient notes, prioritizing History & Physical (H&P) notes (97.0%) and imaging reports (81.8%), demonstrating context-sensitive document selection that varied by chief complaint.

✅ Qualitative analysis showed LLM summaries were often more inclusive and neutrally phrased, while physician summaries offered more contextual nuance but sometimes omitted key details, suggesting potential for a hybrid approach.

#AIinHealthcare #MedicalAI #EmergencyMedicine