A new artificial intelligence (AI)-based model can predict the risk of skin cancer from a photo. The AI model does this with greater accuracy than is possible with existing methods.
This is an algorithm developed by researchers at Erasmus MC. The model was trained using over 2,800 facial photos taken under the same conditions. It is known to the researchers which people developed skin cancer after taking the photo. Based on this dataset, the AI model managed to identify 200 features.
Based on this, the algorithm predicts the risk of developing skin cancer. Although the AI model relies on a photo of a patient's face, the assessment is not limited to the face.
Explainable AI
The researchers use 'explainable AI', a method that provides insight into what an algorithm bases its predictions on. This is in contrast to many other methods, where this is far from clear. This lack of clarity is why AI models are sometimes described as a 'black box', where it is unclear how the model arrives at its result.
At the same time, the researchers stress that explainable is not the same as interpretable. With such an algorithm, researchers can explain exactly what it does. However, the meaning of these actions is not always clear.
Identify risk factors
The model is able to identify risk factors for skin cancer that are invisible to the human eye. Dermatologist Dr Marlies Wakkee explains: "As a dermatologist, I pay attention to things like pigment spots, wrinkles or the fact that someone has lived in a sunny country. The algorithm picks up more features than just those classic risk factors for skin cancer."
"We visualise the 200 features on which the model bases the prediction. If we try to translate those back to what is known about skin cancer risk, we see that the algorithm fishes out features from the faces that we already know are associated with skin cancer. Think wrinkles, for example."
Raising awareness
The researchers see prevention and awareness opportunities for their AI model. Among other things, they hope that by alerting people early on to a high risk of skin cancer, they will be more aware of the sun and keep a closer eye on their skin.
In the near future, researchers plan to conduct more research on the algorithm. For example, by testing the algorithm on a larger dataset. And including photos of people of more ages and ethnic backgrounds. The researchers will also make the AI model public in the research community so that other scientists can build on it.
By: Wouter Hoeffnagel
Image: 3534679 via Pixabay
