Microsoft system tested on simulated diagnostic work-ups

A Microsoft team says its diagnostic system was more than twice as accurate as general physicians in working through more than 300 unusual and difficult medical cases. Cyprus Mail reported that the study was published in Science on Thursday and compared the system with doctors who could consult textbooks and the internet.

The system uses an “orchestrator” to coordinate a simulated panel of clinical experts. It selects questions about a patient’s history, identifies relevant elements of an examination and orders tests in stages, revising its possible diagnoses as results arrive. Researchers also reported that it produced less costly diagnostic work-ups than the physicians, although the account did not provide cost figures.

The result remains a test of simulated cases rather than evidence that the system improves patient care in routine practice. The comparison involved general physicians, and the report does not describe performance against relevant specialists. The system also outperformed doctors on scenarios drawn from an emergency-room admissions database. A planned collaboration with Beth Israel Deaconess Medical Center in Boston will separately test its ability to triage patients in emergency-department waiting rooms.

Companies claim lower melanoma recurrence or death risk

Moderna and Merck say an injection used with immunotherapy reduced the risk of melanoma recurring or a patient dying by 42% compared with immunotherapy alone. NOS reported that more than 1,100 patients participated and that the analysis used information from the first two years after treatment. The companies also reported a 38% reduction in the risk of the cancer spreading.

The injection is intended as part of treatment, not as a vaccine to prevent cancer. NOS said the researchers would present the full results at a European Society for Medical Oncology congress in Madrid. The available account does not provide absolute event rates, safety findings or details about randomisation and blinding, limiting assessment of the reported relative risk reductions. The two-year data also do not establish longer-term durability, and the findings are company claims rather than evidence that the combination has become standard clinical practice.