Do you suddenly have a suspicious spot on your skin? With artificial intelligence, you can get an initial analysis of your skin.
Skin cancer is one of the most common forms of cancer in Belgium. There are nearly 46,000 new skin cancer patients in our country every year. Although the number of people with skin cancer is increasing, the number of dermatologists is not. To ensure that the right people get access to the right care as soon as possible, SkinVision developed an AI-powered application to detect skin cancer. “We want to provide an answer to the very first question: does this spot need additional monitoring or not?” says Henk van der Molen, Senior Machine Learning (ML) Architect at SkinVision.
To the doctor?
Do you suddenly have a suspicious spot on your skin? Then our first reflex is to schedule a visit to the dermatologist to have the spot checked for skin cancer, although chances are you will be put on a long waiting list. Skin cancer can be treated well in the early stages. So it is important that the right people get the right care as soon as possible.
SkinVision is a clinically proven application that detects skin cancer. “Based on a photo, the AI model performs an analysis of your skin to determine whether or not certain spots may indicate skin cancer,” Van der Molen said.
The app helped a total of more than 1.8 million users worldwide.
Henk van der Molen, Senior Machine Learning (ML) Architect at SkinVision
The company wants this to be the first stepping stone to checking whether a doctor’s visit is necessary, as the waiting list of dermatologists today is very long. Currently, 112,225 Belgians are active on the app and a total of 201,117 skin checks were performed. The app helped a total of more than 1.8 million users worldwide.
Own data
Over the past decade, SkinVision built up a rich set of data from people using the application. “We collected about six million photos over that period,” van der Molen states. “Over half of that data was reviewed by a panel of three dermatologists. Based on that information and the continuous influx of new data, we train and improve the AI model,” he continues.
Although AI is part of the foundation, “we don’t see ourselves as a pure AI company,” van der Molen states. From the start, SkinVision has focused on combining AI with people. “In high-risk situations, the expertise of a dermatologist is called in. AI mainly does the initial selection,” says van der Molen.
Quality control
“The SkinVision app is good at recognizing what is not skin cancer. After all, most users don’t have it,” van der Molen begins. In this way, we can already rule out a large group of people who do not need additional monitoring by a doctor.” If no direct ruling can be made by the AI model, the expertise of a dermatologist is called in.
“We have a team of dermatologists who review photos daily, both high and low risk. Users with a doubt are then given additional detailed results after two days,” van der Molen said.
SkinVision is powered by proprietary data but to assess it, external data is used. “We have our app tested by universities and hospitals that assess the accuracy of the model,” he says. A recent 2021 study by Erasmus University in Rotterdam tested the app’s accuracy on about 800 patients. The researchers concluded that the app is at least as accurate as most general practitioners and probably even more accurate.
Universal camera quality
The SkinVision app analyzes your skin based on photos you take with your smartphone. Depending on which device or operating system you have, the photo quality can vary greatly. To ensure that this does not affect the quality of the skin analysis or model, SkinVision designed its own camera module.
“We don’t use the phone’s camera app but an in-house developed camera module that specializes in taking skin photos,” van der Molen said. “That way we ensure that the quality of the photos remains under control, regardless of which device you use.”
When you want to take a picture with the app, three check marks appear on the screen. The goal is to get those check marks all to green. To do this, you need to hold the camera at different angles and perspectives so you can take the perfect skin photo. “Only once the three checkmarks jump to green is the quality of the photo enough to be judged,” van der Molen explains.
First selection
“We don’t want to replace the doctor’s role and don’t diagnose,” van der Molen emphasizes. The number of people diagnosed with skin cancer is increasing, but the number of dermatologists is not. “With our application, we want to already exclude the group of people who don’t need to go to the doctor, so that more help is available for patients who do need care.” That way the right people get expedited access to care.
Insurers
In Western countries, we are used to good health care. To use the SkinVision app, however, you have to pay. This stops a lot of people who, when in doubt, decide to go to the doctor anyway. “That’s just what we don’t want. That’s why we cooperate with insurers. For them it is of course important that both costs and waiting times remain as low as possible so that everyone gets the right care in time. So insurers certainly understand the added value of our application,” he says.
SkinVision has been cooperating with some insurers in the Netherlands since 2019, as well as in Belgium since a year. This gives the customers of those insurers free access to the application.
AI can help get the right people the right care on time.
Henk van der Molen, Senior Machine Learning (ML) Architect at SkinVision
Skin cancer is a common cancer and waiting times for dermatologists are enormous. Early detection of signs is important with this type of cancer. “AI can play an important role by keeping people with harmless spots away from dermatologists, thus reducing waiting times for those at higher risk of skin cancer,” van der Molen concludes.