Medical artificial intelligence has demonstrated performance surpassing human doctors in certain areas of disease diagnosis and treatment decisions, according to research findings.
The international journal Nature recently published research on a medical AI agent called "Mira," developed by researchers at the Technical University of Dresden and Heidelberg University Hospital in Germany, as well as Google's medical AI "AMIE," according to Britain's Financial Times (FT) and the scientific community Wednesday.
Mira recorded strong performance in evaluations based on more than 500 emergency room clinical cases. After conveying various symptoms and medical histories through conversations with an AI playing the role of patients, researchers comprehensively measured its ability to make diagnoses, test plans, drug prescriptions, and surgery scheduling decisions.
As a result, Mira recorded a diagnostic accuracy of 87.1% across eight major disease categories, including appendicitis and pulmonary embolism. This exceeded the 78.1% accuracy of an evaluation group consisting of six doctors, including specialists and residents.
In particular, Mira showed higher accuracy than doctors in conditions such as pancreatic cancer and pneumonia, and also received favorable evaluations on drug safety aspects such as adherence to clinical practice guidelines and allergy verification.
"More Precise Prescriptions"… Google AI Proves Patient Management Ability
Google's medical AI, AMIE, also produced competitive results compared with human doctors. Researchers compared the performance of the AI with that of 21 primary care physicians through simulated consultations with actors playing the role of patients.
As a result, AMIE recorded a level similar to that of doctors in overall patient management capabilities. In particular, it received even higher evaluations in establishing test plans and adhering to clinical guidelines. In the drug prescription process, analysts found that it did not simply recommend medications but offered specific details such as dosage, duration, and method of administration, demonstrating excellent precision.
Google evaluated the research findings as "showing that AI can contribute to reducing the workload of medical staff and increasing communication time with patients." The company explained that if AI takes on repetitive and standardized tasks in medical settings, doctors can focus on more complex judgments and patient management.
"Cannot Replace Doctors Yet"… Challenges Remain for Real Clinical Application
However, researchers and experts stressed that this should not be interpreted to mean that AI can immediately replace doctors. In some cases, Mira made judgments that deviated from optimal treatment plans, and AMIE was also assessed as requiring additional verification before being applied to actual clinical environments.
Experts also point out that the cases used in the research were conducted in relatively structured environments. In actual hospitals, patients' symptoms are often unclear or multiple diseases appear simultaneously, so it remains uncertain whether AI can maintain research-level performance.
Julie Jacko, a professor of medical informatics at the University of Edinburgh in Britain, said, "This research is an important milestone in the development of medical AI," but added, "It is difficult to view the results as fully reflecting the complexity and uncertainty of actual clinical settings."
Jakob Kather, a professor at the Technical University of Dresden who participated in developing Mira, also said, "AI will play a role similar to an aircraft's autopilot system. It can support medical staff by performing routine tasks on their behalf, but the final responsibility and judgment will ultimately remain with the doctor."
The tragedy of a mother who wandered the roads for four hours—why are hospitals trapped in the prison of "defensive medicine"?
