It’s not a question of whether artificial intelligence (AI) will disrupt dermatology. Changes to the specialty — thanks to accumulating data and images — have already begun.
Researchers studying if machine-learning algorithms could diagnose the spectrum of pigmented skin lesions as accurately as human experts found machine-learning outperformed human experts but had limitations and didn’t take the place of humans, according to the study published June 11, 2019 in The Lancet Oncology.1
Questions remain whether dermatologists will embrace AI and be part of the specialty’s transformation, according to Allan C. Halpern, M.D., chief dermatology service, Memorial Sloan Kettering Cancer Center.
Before answering those questions, dermatologists need to understand AI and how it might interact with or disrupt the specialty, he says.
“Artificial intelligence is not new,” Dr. Halpern says. “In the 1950s, people used computers to develop what we call expert systems, which frankly worked, but not well. By the 1980s, there were new models for using computers, including machine learning with neural networks, decision trees and more. Since 2012, we’ve been in the third phase: deep learning.”
It’s one thing to describe a set of rules for something like a cat to a computer to recognize (e,g., triangular ears, long tail); it’s another to feed the computer’s memory with thousands of images of all kinds of animals in categories and allow the algorithms to teach themselves. Computers need large image libraries along with accurate associated data to identify animals, or skin lesions, Dr. Halpern explains.
Dr. Halpern is a consultant with Canfield Scientific, Lucid and SciBase.
1. Tschandl P, Codella N, Akay BN, et al. Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: an open, web-based, international, diagnostic study. Lancet Oncol.2019;20(7):938-947.
2. Esteva A, Kuprel B, Novoa RA, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115-118.