Post by Else Kröner Fresenius Center for Digital Health

New Nature Portfolio publication: AI agent MIRA supports clinical workflows in electronic health records as a co-pilot. A research team led by Jakob Nikolas Kather presents the next stage of medical AI. They developed MIRA, an AI agent that autonomously evaluated medical information, ordered tests, and prepared diagnostic and treatment decisions within electronic health records in a simulated hospital information system. In retrospective simulations using real patient cases, the system achieved higher diagnostic accuracy than the physician comparison groups while also making guideline-concordant and safe treatment decisions. Dyke Ferber (first author): “MIRA was able to carry out clinical workflow steps autonomously within the test environment. It identified missing information, ordered tests, interpreted findings in line with clinical guidelines, and prepared treatment decisions. AI tools should support medical professionals, creating more time for patient care, while meeting the highest standards of safety, transparency, and reliability. With MIRA, we were able to demonstrate that this is possible.” ◻️ Publication: Towards Autonomous Medical Artificial Intelligence Agents, Nature 2026. doi: 10.1038/s41586-026-10675-5 ◻️ Authors: Dyke Ferber, Lars Hilgers, Dr. Christiane Höper, Benedict Kinny-Köster, Jan Eckardt, Katharina Egger-Heidrich, Marius Bill, Martin Schneider, Jan Clusmann, Lejla Kadric, Marcel Oehme, Dr. med. Maximilian Mayrhofer-Schmid, Dr. Alexander Oeser, Georg Wölflein, Isabella Wiest, Moritz Middeke, John Iafrate, Daniel Truhn, Dirk Jäger, Jakob Nikolas Kather ◻️ Link to publication: https://lnkd.in/dPhBReJt Technische Universität Dresden Universitätsklinikum Carl Gustav Carus Dresden TUD | Faculty of Medicine DKFZ Deutsches Krebsforschungszentrum Universität Heidelberg Heidelberg University Hospital NCT Heidelberg - Nationales Centrum für Tumorerkrankungen (NCT) Heidelberg