Artificial intelligence

AI model predicts risk of skin cancer better than existing methods

Researchers at Erasmus MC have developed an artificial intelligence (AI) model that predicts the risk of skin cancer based on a photograph of the face. The AI model outperforms the current methods doctors use to assess skin cancer risk.

4 likes
Reading time 2 min
AI skin cancer

What makes the AI model unique is its ability to identify risk factors for skin cancer that are invisible to the human eye. ‘As a dermatologist, I pay attention to things like pigment spots, wrinkles, or knowing that someone has lived in a sunny country. The algorithm picks up more features than just those classic risk factors for skin cancer’, explains dermatologist Dr. Marlies Wakkee, who led the research along with AI expert Dr. Gennady Roshchupkin.

The team trained the AI model with over 2800 facial photos, all taken under the same conditions. It was known to the researchers which people had developed skin cancer after the photo was taken. An algorithm automatically extracts 200 features from the photos, based on which it predicts the risk of skin cancer, either on the face or elsewhere on the body. This AI-generated prediction is more accurate than a prediction based on skin examination, questionnaires, and genetic research. The researchers published their findings in the scientific journal The Lancet’s eClinicalMedicine.

Explainable AI

The researchers used so-called explainable AI methods. This means it’s possible to understand what the algorithm bases its predictions on. This isn’t always the case with AI. Often it’s a kind of black box where it’s unclear how the model arrives at a particular outcome. ‘We visualize exactly what the 200 features are that the model bases the prediction on. When we try to translate them back into what is known about the risk of skin cancer, we see that the algorithm picks up features from the faces that we already know are associated with skin cancer. Think of wrinkles, for example’, Wakkee explains. In addition, the model picks up features that are less clear to humans. ‘Explainable isn’t the same as interpretable. We can explain what the algorithm does, but what that means in a human way isn’t so obvious’, says Roshchupkin.

In the future, Wakkee and Roshchupkin see two possible roles for their AI risk model for skin cancer: prevention and awareness. ‘If you know you have a high risk of skin cancer, you’ll be more conscious about sun exposure. We also hope that alerted individuals will better monitor their skin.’

Focus on research

The AI model isn’t yet usable in the clinic, but the focus is on further research. ‘For example, we want to test the algorithm on larger datasets with photos of people of different ages and ethnic backgrounds. By publicly sharing the model in the research world, fellow scientists can build on it. That could accelerate the entire AI community’, says Roshchupkin.

Also read