Data science and artificial intelligence are dramatic in their effects and have resulted from a data explosion (more in the last few years than in the history of humanity)…and with the willingness of healthcare professionals and engineers to work together…and also in Dermatology. Here are some learning points from the EADV 2020 Annual meeting taken from a presentation by Professor A. Navarini (Basel, Switzerland):

 

 

In this publication (June 2020) in Nature Medicine of the Deep Learning system developed by Google:

-the study concerned 26 dermatoses making up 80% of what presents to the primary care physician.

No alt text provided for this image

 

 

-This is how the system works. Note that the pictures are taken with a smartphone…that is what is done in reality and what is so much missing in the benchmark studies done until now:

No alt text provided for this image
It is not yet accurate if only one diagnosis is allowed. However differential diagnosis is what is done in the clinical setting and if 3 diagnoses are allowed, accuracy increases dramatically (from 0.58 to 0.93)
No alt text provided for this image
-Clinical information can be added to the physical examination (images). The more the pictures, the more accurate it becomes. The history is important, but if supplied with the images and not evaluated, the accuracy reduces, which could suggest a diagnostic doubt ?….after all dermatology is visual !
No alt text provided for this image
Source of information: Oral presentation at EADV 2020 by Professor Alexander A. Navarini