What will AI mean for dermatology?
A hot-button topic gets a closer look.
P028 – AI/Technology in Dermatology
Saturday, August 12 | 1-3 p.m.
Ballroom D
Everywhere you look these days, there’s a headline about artificial intelligence. Many are concerned about where it’s headed and its future impact.
Although AI has its liabilities and pitfalls in society, it could become a useful tool in dermatology, particularly with predicting and treating multiple conditions, including skin cancer.
Meenal Kheterpal, MD, FAAD, associate professor of dermatology at Duke University in Durham, North Carolina, is one of several speakers who will address that and other issues in the session “P028 – AI/Technology in Dermatology,” led by Trilokraj Tejasvi, MD, MBBS.
On Saturday, Aug. 12, Dr. Kheterpal argues that AI use in dermatology is still in early research stages, but it shows tremendous promise.
“AI application use cases in dermatology are immense,” she said. “Thus far, most published models are proof-of-concept projects. The applications include triage, diagnosis, quality analysis, and disease quantification. However, large-scale image datasets with disease annotations will be needed to create reproducible models. But we see AI playing an important role in medical and cosmetic dermatology in the next five years.”
Diagnosis and treatment
For diagnosing and treating skin cancer, AI opens the door to several potential uses, according to Dr. Kheterpal.
“There is a curated dataset for skin cancers housed by International Skin Imaging Collaboration, among others,” she said. “There are a number of research and commercial groups that have developed successful models that predict the likelihood of skin cancer. There are tools available for tracking lesions at high risk of skin cancer for clinical decision support. There are also groups detecting changes in lesions based on wide field images. All these developments will automate skin cancer detection when deployed with prospective validation into dermatology clinic settings.”
For now, Dr. Kheterpal said she believes these uses should stay limited to clinical settings.
“Personally, I feel wary of at-home detection apps and would caution patients from using them at this time,” she said.
The downside of AI considered
Even with all these advances and potential applications, Dr. Kheterpal said AI is not without its drawbacks when it comes to dermatology, not the least of which is a lack of regulatory guidance on how it should be used.
“Currently we are in the infancy stage of AI in dermatology with a lot of proof-of-concept studies and limited trials,” she said. “We need to ensure that the regulatory landscape is firmed up, starting with the FDA giving out clear guidance for AI software as a medical device.”
But that guidance won’t just be coming from the federal government. The AAD’s Augmented Intelligence Committee is working to develop guidelines for publishing AI model manuscripts and around the commercial deployment of models. This, Dr. Kheterpal said, will help create a resource that institutions can use to ensure they are buying a legitimate product.
“We also need to ensure that image data collected for analysis by these models is well-preserved per HIPAA compliance standards,” she added. “We have a lot of work ahead of us.”
Other scheduled speakers for the new AI session are Lilia M. Correa, MD, FAAD, George Han, MD, PhD, FAAD, Babar K. Rao, MD, FAAD, and Veronica Rotemberg, MD, FAAD.
