Artificial intelligence and screening for melanoma

Melanoma detection is a significant challenge for public health. According to the International Agency for Research on Cancer, 232,000 cases of malignant melanoma are detected every year.

This type of cancer can easily be treated if detected early. Unfortunately, many cases are only diagnosed at advanced stages, hence the need for effective screening.

Last May, a team of French, German and American researchers developed an artificial vision algorithm capable of distinguishing malignant cutaneous lesions from harmless moles. The machine processed over 100,000 images of benign or suspicious lesions before being tested.

Its efficacy was compared to that of 58 dermatologists from 17 different countries. Within the limits of this study, the AI-based machine was found to be more accurate than the human intelligence of the doctors. The dermatologists correctly identified 87% of melanomas compared to 95% correct diagnoses by the algorithm.