In a way it’s not surprising, diagnosis is all about a series of judgements based on appearance or test results.
The diagnosis algorithm was created from a database by processing 130 images of known skin lesions representing 2000 diseases and creating a “nice taxonomy” out of the data, which was tested against the judgement of 21 dermatologists on 370 images of known diagnosis. The team from Stamford have published their study in Nature.
It is important because it means that in future skin lesions, which often look very similar to each other, can be more accurately diagnosed at the primary care level and where necessary treatment will be started more quickly.
It’s also important because it is the first one of many conditions which will benefit from AI. That’s a shame if like me you find diagnosis the most interesting part of medical life, but hey! you can still compare your own decisions with the Robot. The point is that you would no longer need a doctor to do the diagnosis, it could be at the pharmacy or maybe you could do it yourself via your mobile phone; or of course in the podiatry or osteopathy surgery.
All the same, treatment is still going to be in the hands of trained professional people. I’m not sure that I can see how a robot could enucleate a corn, or provide trigger point therapy, for a very long time yet, but….watch this space.