According to a study released on Thursday in Science, Microsoft scientists claim that a diagnostic AI tool, which mimics a doctor’s step-by-step approach, was more than double as precise as standard doctors at resolving detailed medical cases. The system utilizes an ‘orchestrator’ to manage the process, simulating a group of clinical specialists working together to determine the necessary questions, exams, and tests for a patient, as described in the report.
Unlike existing medical AI models that examine all symptoms, lab results and imaging at once, this tool works sequentially. It first decides what questions to ask about patient history and what matters in the physical exam. Based on responses, it determines which tests to order, evaluates results and orders further testing as needed, updating hypotheses throughout the process. All the while, it is updating its working hypotheses and plans, the research leaders said in an interview.
Read Also: Filipinos have low trust in stablecoins for remittances
When given over 300 obscure and perplexing medical puzzles, the orchestrator-based system surpassed general physicians who relied on textbooks or the web. It also outperformed human doctors on emergency room scenarios taken from a major database. The AI was more than twice as effective at ‘working up’ the patient and reaching a conclusion compared to human practitioners using the same resources.
Traditional AI assessments often rely on multiple-choice questions where models achieve high accuracy, but real-world cases are messier. “AI has become very good at getting a 100% pass rate on those,” said Dr. Dominic King, Microsoft AI’s corporate vice president for health. “But of course, that’s not real-world. No patient ever came to see me with a nice summary.” The researchers also found that diagnostic processes cost less when the orchestrator designs the work-ups than when human doctors do.
Leave a Reply