This is great to see. A key part of medicine is generating an appropriate differential rather than identifying one true diagnosis (as they explain). A ranked li

This is great to see. A key part of medicine is generating an appropriate differential rather than identifying one true diagnosis (as they explain). A ranked list of diagnoses lets you appropriately incorporate new information (history, lab data, imaging, etc) and gives you a framework for exploring further (thinking about value of information and what additional data you'd need to narrow your list or increase your confidence). I think in dermatology, radiology, and pathology (and other visual diagnostic fields) this work is a huge boon, but the conceptual framework is the same for diagnostics based on laboratory, physical exam, genetic, or any other findings. I don't think ML is necessary in creating excellent decision support tools but is increasingly valuable as you include more and more disparate sources of information.