Can diagnosis be improved using machine learning? Elizabeth Tracey reports
Podcast: Download (Duration: 1:00 — 1.4MB)
Artificial intelligence or machine learning looks like an attractive way to help physicians make diagnoses, especially when the person has unusual symptoms or doesn’t fit the usual demographic. Such situations account for many missed diagnoses among those who visited emergency departments in a recent study led by David Newman-Toker at Johns Hopkins. Yet Newman-Toker says development of such tools may be a ways off.
Newman-Toker: It’s a much trickier problem when you’re talking about the average patient walking into the average clinical setting. Primary care, emergency department, specialty clinic or otherwise with an undifferentiated symptom, like headaches or dizziness or chest pain or abdominal pain. There, the types of data that would be required to train those systems to be truly expert don’t exist yet. That’s part of the problem. It’s going to take a while for us to mature those datasets and turning that into expert decision support tools. :29
Newman-Toker notes that data gathering tools are being increasingly utilized so progress may be quick. At Johns Hopkins, I’m Elizabeth Tracey.