# AI development in medical diagnostics > Live situation record from CLSTR: https://clstr.news/situations/ai-development-in-medical-diagnostics > Updated: 2026-10-22T21:55:54.000Z. Sources: 49. Developments: 2. Recent developments in artificial intelligence show increasing capabilities in clinical diagnostic settings. A study published in The Lancet regarding Google’s AMIE chatbot indicated the feasibility of using conversational AI to help patients prepare for primary care appointments. While clinical evaluators rated conversation quality favorably in most cases, researchers noted instances of hallucination and one interaction rated as “somewhat harmful” by a physician due to patient anxiety caused by a suggested cancer diagnosis. In a separate advancement reported in the journal Science, Microsoft researchers introduced MAI-DxO, a system designed to mimic the step-by-step diagnostic process of physicians. This tool achieved an 83.2% accuracy rate in diagnosing over 300 complex and unusual cases from The New England Journal of Medicine, significantly exceeding the 36.1% accuracy rate of generalist physicians using medical references and the internet. The system utilizes an “orchestrator” to coordinate simulated clinical experts that sequentially determine necessary patient questions and tests. MAI-DxO also demonstrated superior performance in emergency department scenarios, reaching approximately 80.3% accuracy compared to 46.8% for human physicians. Beyond accuracy, research indicates the system may reduce diagnostic costs through more efficient testing paths. Microsoft officials, including AI lead Mustafa Suleyman and Health unit VP Dominic King, emphasized that the system has only been tested on historical data and requires further clinical validation. Microsoft plans to conduct real-world validation studies in collaboration with Beth Israel Deaconess Medical Center. Recent data suggests the system’s performance was notably high when integrated with OpenAI’s GPT-5 and GPT-5.5 models. ## Claims - MAI-DxO uses an orchestrator to coordinate multiple AI agents to simulate a medical panel. (corroborated by 29 sources) - The system performs diagnostic tasks sequentially by asking questions and ordering tests. (corroborated by 29 sources) - The MAI-DxO system achieved an 83% accuracy rate in diagnosing complex medical cases. (corroborated by 28 sources) - General physicians achieved a 36% accuracy rate in the same complex case simulation. (corroborated by 26 sources) - The system showed minimal improvement in accuracy for emergency room cases. (corroborated by 9 sources) - The system's performance was highest when using OpenAI's GPT-5 and GPT-5.5 models. (corroborated by 7 sources) - Microsoft plans to conduct clinical validation with Beth Israel Deaconess Medical Center. (corroborated by 4 sources) - Microsoft is collaborating with the Mayo Clinic to train a healthcare model. (corroborated by 2 sources) - The healthcare model developed with Mayo Clinic is targeted for release next year. (corroborated by 2 sources) ## Timeline ### 2026-10-22: Microsoft AI system MAI-DxO outperforms doctors in complex diagnoses Microsoft's MAI-DxO AI system achieved 83% accuracy in diagnosing complex medical cases, significantly outperforming general practitioners in a study published in Science. 42 sources. https://clstr.news/cluster/microsoft-ai-system-mai-dxo-outperforms-general-practitioners-in-complex-diagnoses ### 2026-10-08: Google AI chatbot AMIE shows feasibility in clinical diagnostic study A study in The Lancet shows Google’s AMIE chatbot can safely engage in clinical conversations to help patients prepare for doctor visits, matching confirmed diagnoses in over half of the cases studied. 7 sources. https://clstr.news/cluster/google-ai-chatbot-amie-shows-feasibility-in-clinical-diagnostic-study --- Cite as: AI development in medical diagnostics. CLSTR, https://clstr.news/situations/ai-development-in-medical-diagnostics