Dr Art Papier Discusses Future Efforts to Reduce Disparities in Dermatology

We need more educational or clinical projects that have a real-world impact on reducing disparities in dermatology, said Art Papier, MD, dermatologist, CEO, VisualDx.

Art Papier, MD, dermatologist, CEO, VisualDx, talks about disparity prevention in dermatology, including the second year of Project IMPACT, and considerations in artificial intelligence (AI) and machine learning.

Transcript

What are your goals for the second year of Project IMPACT?

I think we really want to see specific projects that are either educational or clinical that have real-world impact. Project IMPACT, the name, we chose that really because we want to be substantial. We don't want to just be performative and be just discussing an issue. We want to see some results. In the second year, we want to see more people involved and we want to see specific projects, and we're hoping to do specific projects around medical education.

How can we ensure future projects are including underserved populations in their research?

I think there's tremendous interest in primary care. So, in rural areas or inner city areas where you have federally qualified health clinics serving the underserved, there's a real need to bring the kind of information that those health care professionals can use in real time. In my work, I've been thinking about this problem, "What's the right information for the right patient population?" So we're thinking about, if the clinic is urban, what do they need? If the clinic's rural, what's the patient population? How do we really bring customized information to each population?

Since dermatological conditions can appear differently depending on skin type, what steps should be taken in AI development?

People are beginning to understand that AI and machine learning are totally dependent on good quality data. It's garbage in, garbage out. You have to have good data, and if you train your algorithms just on white skin, it's not going to work on brown skin.

So in our work at VisualDx, as an example, we've been collecting imagery for over 20 years, and since the start of our efforts, it's always been equitable. We have something like 32% of the imagery in the system is in brown skin, and we have tags on that imagery so that when we train our AI, it's equitable. This is key to having equity in AI and machine learning—it's the data you train on and how precise the data is.