Doha: Microsoft has developed a new medical AI tool, which it claims can "sequentially investigate and solve medicine's most complex diagnostic challenges" that even expert physicians struggle to answer.
According to Qatar News Agency, this innovative tool, named Microsoft AI Diagnostic Orchestrator (MAI-DxO), has been benchmarked against real-world case records published each week in the New England Journal of Medicine (NEJM). Reports suggest that MAI-DxO correctly diagnoses up to 85% of NEJM case proceedings, a rate more than four times higher than a group of experienced physicians. Notably, the AI tool achieves these results more cost-effectively than traditional methods employed by physicians.
Microsoft explains the methodology behind the AI tool's impressive performance. They created interactive case challenges from the NEJM case series, termed the Sequential Diagnosis Benchmark (SD Bench). This benchmark involves transforming 304 recent NEJM cases into stepwise diagnostic encounters. During these encounters, models or human physicians can iteratively ask questions and order tests. As new information becomes available, the model or clinician updates their reasoning, gradually narrowing toward a final diagnosis. These diagnoses are then compared to the gold-standard outcomes published in the NEJM.