Artificial intelligence and dermatology: concrete help for quicker and more targeted diagnoses of dermatological lesions and in the correct use of drugs according to the medicine of the 4 ps, i.e. personalized, predictive, preventive and participatory. But also strategic support for the preparation of dermatologists through new training methods linked to AI. These are just some of the great opportunities that project dermatology into the future, provided, however, that a multidisciplinary task force dedicated to the training of doctors and all interested categories on the conscious and informed use of Ai which, if not understood and used well, can become extremely risky. This is the message launched by the experts present at the 98th National Congress of the Italian Society of Dermatology and Sexually Transmitted Diseases (Sidmast), which ends today in Giardini Naxos (Me).
“To be able to benefit from or even better use artificial intelligence in an active way it is necessary first of all to have adequate training – explains Pietro Rubegni, professor of Dermatology at the University of Siena – to obtain it it is important to organize and create multidisciplinary teams where young dermatologists interact daily with biostatisticians, bioengineers and biologists. This is what we are carrying forward in dermatology in Siena which has already traditionally collaborated closely with bioengineering for over 30 years, first with Gabriele Cevenini and for about five. years also with Alessandra Cartocci, biostatistician and fellow at Sidemast. Thanks to this great collaborative work we have internally developed what is called the ‘Health technology assessment group’.
There are currently three fields in which dermatology uses AI. First of all, it is used for the early diagnosis of skin tumors, accompanying the professional in the evaluation of skin lesions during the diagnostic process up to the final recognition of the neoformation. “It is as if the dermatologist and the AI went ‘hand in hand’ along this path – underlines Cartocci – but it is always the man who guides and lays the foundations. A position paper by the Eadv, European Academy of Dermatology and Venereology on the argument has in fact shown that most of the Apps developed and sold for automatic image recognition have failed miserably, however if the dermatologist selects the ‘right’ lesions to show to the AI the latter ‘wins’ compared to the doctors you select something that the artificial intelligence doesn’t know, the latter will make mistakes.”
The second field of application is that which allows an objective evaluation of the severity of the disease “Daniel Kahneman, Nobel Prize winner for economics – continues Cartocci – talks about ‘noise’, that is to say the states of mind which are influenced by what we happens around us and which therefore conditions our daily choices, even in medicine. AI eliminates external influences and allows us not to make mistakes, and therefore to be balanced in the evaluation objective and comparable assessments of the drug’s activity. AI manages to eliminate subjective influences and will tell us how much the drug will work, not how much it seems to me to work.”
Third, perhaps the most relevant and futuristic, adds Rubegni “is the possibility of predicting for that type of patient which will be the best therapy and with the least adverse effects”. And on this front, sharing data among experts will be decisive. A further developing field that sees protagonist is the Gan (generative adversarial network) method: a technique which, using real images of pathological manifestations, allows the creation of plausible images, even if ‘fake/synthetic’. “In medicine there is often little data – continues Professor Rubegni – the GANs will be able to increase them dramatically, allowing, for example from 50 images of melanoma, to produce hundreds, completely plausible and indistinguishable. These in turn can be used to teach to young people or train, through further AI methods, other models for automatic recognition”.
But, concludes the expert, “since most health systems today do not have the regulatory capacity to supervise and manage this rapidly evolving technology, we must ensure that we accompany the growth of AI with regulations that contain it”.
#Dermatologists #task #force #train #doctors