Artificial intelligence for skin cancer detection

Skin cancer is the most common cancer in the world, with 2-3 million new cases per year, according to WHO. One in every five Americans will get skin cancer before the age of 70, according to skincancer.org. The survival rate of late stage skin cancer is less than 15%, but if detected early then skin cancer can be treated effectively.

AI is the key to the accurate and early detection and classification of skin cancer and other types of skin lesions, and we are currently building as a best-of-the-world SkinCancerAI system. After the first 2 months of work, our system won the First Runner-Up Prize at the largest international skin cancer challenge ISIC 2019 in 09/2019. Since then, we have been improving our system continuously, by adding new data and inventing more efficient algorithms.

While human dermatologists’ diagnosis of melanoma (the most dangerous type of skin cancer) has sensitivity of about 88% and specificity 76%, (see, e.g., https://academic.oup.com/annonc/article/29/8/1836/5004443), our system achieves sensitivity above 95\% for specificity above 80\%$ in melanoma detection.

We are currently in talks with the advanced dermatoscope maker Pixience (pixience.com) and other partners to bring our state of the art SkinCancer AI system to the dermatologists and the people.

ISIC2019
NT Zung (second from left) representing Torus Actions at ISIC2019
(MICCAI2019, Shenzhen, China, 10/2019)