Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 26 min ago
Peer-reviewedCC BYNature

Towards autonomous medical artificial intelligence agents

  • Dyke Ferber1,2
  • Lars Hilgers3,2
  • Christiane Höper2
  • Benedict Kinny‐Köster4,5
  • Jan‐Niklas Eckardt3,2
  • Katharina Egger‐Heidrich3
  • Marius Bill3
  • Martin M. K. Schneider3
  • Jan Clusmann3,6,2,7
  • Lejla Kadric8
  • Marcel Oehme9
  • Maximilian Mayrhofer‐Schmid10,11
8 more authors
  • Alexander Oeser12
  • Georg Wölflein2
  • Isabella C. Wiest2,7
  • Jan Moritz Middeke3,2
  • A. John Iafrate13
  • Daniel Truhn14
  • Dirk Jaeger1
  • Jakob Nikolas Kather15,16,17
17 affiliations
  1. 1Department of Medical Oncology, National Center for Tumor Diseases (NCT), Heidelberg University Hospital, Heidelberg, Germany
  2. 2Else Kroener Fresenius Center for Digital Health, Technical University Dresden, Dresden, Germany
  3. 3Department of Medicine I, Faculty of Medicine and University Hospital Carl Gustav Carus, TUD Dresden University of Technology, Dresden, Germany
  4. 4Department of General, Visceral and Transplantation Surgery, Heidelberg University Hospital, Heidelberg, Germany
  5. 5Department of Surgery, New York University Langone Health, New York, NY, USA
  6. 6Department of Medicine III, University Hospital RWTH Aachen, Aachen, Germany
  7. 7German Cancer Consortium (DKTK) partner site Dresden and German Cancer Research Center (DKFZ), Heidelberg, Germany
  8. 8Department of Nephrology, Heidelberg University Hospital, Heidelberg, Germany
  9. 9Department of Oral and Cranio-Maxillofacial Surgery, Heidelberg University Hospital, Heidelberg, Germany
  10. 10Department of Hand, Plastic and Reconstructive Surgery, Burn Center, BG Trauma Center Ludwigshafen, University of Heidelberg, Heidelberg, Germany
  11. 11Department of Hand-, Replantation-, and Microsurgery, BG Klinikum Unfallkrankenhaus Berlin and Chair of Hand-, Replantation-, and Microsurgery at the Charité Universitätsmedizin Berlin, Berlin, Germany
  12. 12Innovation Center Computer Assisted Surgery (ICCAS), University of Leipzig, Leipzig, Germany
  13. 13Department of Pathology, Massachusetts General Hospital Cancer Center and Harvard Medical School, Boston, MA, USA
  14. 14Department of Diagnostic and Interventional Radiology, University Hospital Aachen, Aachen, Germany
  15. 15Department of Medical Oncology, National Center for Tumor Diseases (NCT), Heidelberg University Hospital, Heidelberg, Germany. [email protected]
  16. 16Department of Medicine I, Faculty of Medicine and University Hospital Carl Gustav Carus, TUD Dresden University of Technology, Dresden, Germany. [email protected]
  17. 17Else Kroener Fresenius Center for Digital Health, Technical University Dresden, Dresden, Germany. [email protected]
Open at sourceDOICite

Abstract

. However, building physician copilots will require models that operate within the electronic health record (EHR), with governed access to patient data and the ability to initiate permitted EHR actions within defined safety constraints. Yet it remains unproven whether such a system can manage patient cases with physician-level performance. Here we show that MIRA (Medical Intelligence for Reasoning and Action), an autonomous artificial intelligence agent operating in a sandboxed EHR environment, can navigate a large clinical action space to obtain patient histories; order and interpret laboratory, imaging and microbiology tests; generate differential diagnoses; and formulate treatment plans such as prescribing medications, scheduling surgical procedures and planning admissions. In simulations on real patient cases spanning multiple diagnoses, MIRA outperformed physicians in diagnostic accuracy and made guideline-concordant, medication-safe and appropriate admission decisions. Compared with previous LLM applications that addressed isolated subtasks or provided free-text advice, these results suggest that an EHR-integrated artificial intelligence agent can turn clinical intent into structured, actionable EHR operations, possibly making it a more effective decision-support partner for physicians. Further work is needed to establish generalization, safety and governance through prospective, real-world studies.

Related

  1. Autonomous biomedical research with an artificial intelligence agent
    Kexin Huang, Serena Zhang, Hanchen Wang +25
    SciencePeer-reviewedLicence not statedFull text257 citedMolecular biology & gene regulationEpidemiology & public healthOpenAlex
  2. Human–artificial intelligence collaboration in emergency medical services: from diagnostic algorithms to shared decision support
    Hang A Park
    Clinical and experimental emergency medicinePeer-reviewedCC BY-NCFull textGeneral medicineEurope PMC
  3. Predictors of Intentions to Learn and Adopt Medical Artificial Intelligence Among Healthcare Professionals in Ghana: A Descriptive Cross‐Sectional Study
    Godwin Adzakpah, Daniel Kwame Kwansah Quansah, Mavis Puopelle Dakorah +10
    Health science reportsPeer-reviewedCC BY-NCFull textGeneral medicineEurope PMC
  4. A Tale of Two Countries: Challenges in the Adoption of Artificial Intelligence in Medical Imaging in India and Singapore.
    Charlene Jin Yee Liew, Swati Goyal
    The Indian journal of radiology & imagingPeer-reviewedCC BY-NC-NDFull textEpidemiology & public healthEurope PMC
  5. Theranostic Imaging Agents: Integrating Molecular Diagnostics, Precision Therapeutics, and Artificial Intelligence
    Shrestha Sharma, Syed Arman Rabbani, Mohamed El-Tanani +3
    BiomoleculesPeer-reviewedCC BY 4.0Full textMedical imaging & radiologyCrossrefEurope PMC

Cite

BibTeXRIS
BibTeX
@article{ferber2026towards,
  title = {{Towards autonomous medical artificial intelligence agents}},
  author = {Dyke Ferber and Lars Hilgers and Christiane Höper and Benedict Kinny‐Köster and Jan‐Niklas Eckardt and Katharina Egger‐Heidrich and Marius Bill and Martin M. K. Schneider and Jan Clusmann and Lejla Kadric and Marcel Oehme and Maximilian Mayrhofer‐Schmid and Alexander Oeser and Georg Wölflein and Isabella C. Wiest and Jan Moritz Middeke and A. John Iafrate and Daniel Truhn and Dirk Jaeger and Jakob Nikolas Kather},
  journal = {Nature},
  year = {2026},
  month = jun,
  doi = {10.1038/s41586-026-10675-5},
  url = {https://doi.org/10.1038/s41586-026-10675-5},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Towards autonomous medical artificial intelligence agents
AU  - Dyke Ferber
AU  - Lars Hilgers
AU  - Christiane Höper
AU  - Benedict Kinny‐Köster
AU  - Jan‐Niklas Eckardt
AU  - Katharina Egger‐Heidrich
AU  - Marius Bill
AU  - Martin M. K. Schneider
AU  - Jan Clusmann
AU  - Lejla Kadric
AU  - Marcel Oehme
AU  - Maximilian Mayrhofer‐Schmid
AU  - Alexander Oeser
AU  - Georg Wölflein
AU  - Isabella C. Wiest
AU  - Jan Moritz Middeke
AU  - A. John Iafrate
AU  - Daniel Truhn
AU  - Dirk Jaeger
AU  - Jakob Nikolas Kather
PY  - 2026
DA  - 2026/06/17
JO  - Nature
DO  - 10.1038/s41586-026-10675-5
UR  - https://doi.org/10.1038/s41586-026-10675-5
AB  - . However, building physician copilots will require models that operate within the electronic health record (EHR), with governed access to patient data and the ability to initiate permitted EHR actions within defined safety constraints. Yet it remains unproven whether such a system can manage patient cases with physician-level performance. Here we show that MIRA (Medical Intelligence for Reasoning and Action), an autonomous artificial intelligence agent operating in a sandboxed EHR environment, can navigate a large clinical action space to obtain patient histories; order and interpret laboratory, imaging and microbiology tests; generate differential diagnoses; and formulate treatment plans such as prescribing medications, scheduling surgical procedures and planning admissions. In simulations on real patient cases spanning multiple diagnoses, MIRA outperformed physicians in diagnostic accuracy and made guideline-concordant, medication-safe and appropriate admission decisions. Compared with previous LLM applications that addressed isolated subtasks or provided free-text advice, these results suggest that an EHR-integrated artificial intelligence agent can turn clinical intent into structured, actionable EHR operations, possibly making it a more effective decision-support partner for physicians. Further work is needed to establish generalization, safety and governance through prospective, real-world studies.
SN  - 0028-0836
SN  - 1476-4687
ER  -