Researchers have created an algorithm capable of diagnosing 134 skin diseases. A tool that they hope to one day make available to the general public to increase the early management of certain cancerous lesions.
Vitiligo, dermatitis, psoriasis, eczema, melanoma, shingles, herpes, acne… In its White Book published in September 2018, the French Society of Dermatology estimates that 16 million people (33% women and 28% men) suffer from a skin disease in France. Almost 80,000 skin cancers are diagnosed each year, about 70% of them are carcinomas (the least serious form).
Artificial Intelligence at the service of dermatology
These skin conditions are common, but it is often difficult to get an appointment with a dermatologist quickly, or to distinguish between malignant and benign conditions. That’s why Korean researchers have been working on an Artificial Intelligence (AI) capable of accurately classifying skin disorders, predicting malignancy and suggesting treatment options. Their works were published in the Journal of Investigative Dermatology.
“Recently, there have been remarkable advances in the use of AI in medicinesays Dr. Jung-Im Na from the Department of Dermatology at Seoul University (South Korea), lead author of the study. For specific problems, such as distinguishing between melanoma and nevi (large mole, editor’s note), AI has shown results comparable to those of human dermatologists”. Nevertheless, no AI was previously able to differentiate between malignant and benign conditions and “distinguish skin cancer from many other skin disorders such as inflammatory and infectious conditions.”
134 skin diseases diagnosed by an algorithm
The researchers therefore developed an algorithm based on 220,000 images of Asians and Caucasians affected by 174 different skin diseases. The system was equipped with “neurons” capable of identifying and interpreting these images. In total, this technology would be able to diagnose 134 skin diseases and offer primary treatment.
The performance of the algorithm was compared to that of 21 dermatologists, 26 dermatology students and 23 people from the general public. Result: its performance would be similar to that of dermatology students but slightly lower than that of dermatologists. Never mind, because the researchers did not hope to replace humans with technology: “We expect AI to support humans as Augmented Intelligence to make diagnoses faster and more accurately”they explain.
This algorithm, if one day made available on smartphones as desired by researchers, could become a valuable tool for the general public, since it would encourage patients to consult a healthcare professional and contribute to early treatment. of certain otherwise neglected cancerous lesions. Nevertheless, it will still take work before this technology is as precise on a smartphone as it is in the clinic…
.
