First Derm's new AI app helps users identify skin conditions

By Laura Lovett
Share

A quick Google image search is often the go-to for people when they notice a rash cropping up on their skin. This starts the process of trying to match up their skin condition to the one in the photos. 

But a new app called Skin Image Search by First Derm has just launched in beta today and aims to use artificial intelligence to help people figure out exactly what their skin condition is. 

“In many of these [Google searches] these images are wrongly labeled,” Dr. Alexander Börve, CEO and founder of First Derm, told MobiHealthNews. “We hope that our search will give you a much better understanding of this disease and what you have.”

Users can upload a photo of their skin to the platform. Then the AI will match up the image to possible conditions. Börve describes the technology as a triage device. 

“The AI isn’t as good as replacing a doctor. It is more of a skin image search. It is very good for triaging for example,” Börve said.  “You can also click a link and can read more about [the condition].”

Currently the technology is able to narrow down the condition for users. Right now the platform can identify the condition with 40 percent accuracy. It can narrow down the issue to the top five conditions that it could be with 80 percent accuracy, according to the company. 

But Börve said the accuracy is increasing over time. 

“This is a first result and over time we will get better. Every month the statics I gave you will be better,” Börve said. “Hopefully next month or in two months we will have 60 percent accuracy on one disease and 90 percent of the top five or maybe 80 percent the top three. Our goal is to be 80 percent accurate with one disease and 95 percent accurate with one disease and two differential diagnoses. That is the ultimate goal, and then obviously that we would add more diseases.”

Right now the technology has 250 skin diseases in its data base. But the technology is continuously training using a deep Convolutional Neural Network (CNN) with skin disease images.

The company started out as an on-demand online dermatology app and has collected hundreds of thousand of anonymous, amateur skin disease images that have been used to improve the service and were used to train the CNN, according to a statement. 

In addition to the new app, the company will continue to offer its platform that lets users anonymously take photos of an external skin issue and send them to licensed dermatologist who will respond to inquires within 24 hours of receiving the picture with an assessment of the problem. 

Up next, Börve said the company is embarking on clinical trials for Skin Image Search.