"I’m Healing by Photo" - modern technologies help check a mole for safety

It is fair to note that the idea is not new; it has already been embodied by Stentfort researchers in the past, and it has also become the topic of a grantproject by our Belarusian colleagues. Moreover, to date there are already up to forty mobile applications that can help a user try to explore their mole's safety. The Ukrainian team, having analyzed the experience of predecessors and taking into account all drawbacks and advantages, created the resource Nevus, which will be useful not only for an ordinary user but also for medical professionals working in this field.

The main focus of the project is on the ability to distinguish a nevus (a mole) from melanoma and seborrheic keratosis. For this, you simply need to photograph and upload your mole, after which a neural network with 25 million parameters will compare it with 10 thousand other images. As a result, three numbers will be obtained, reflecting the probability that the uploaded photo is a mole, melanoma, or keratosis. Based on the magnitude of this number, a hypothesis about the diagnosis is made.

Is the system wrong? The answer is yes, as is the case with doctors. There is always a probability of error, but it is interesting to note that the specificity and sensitivity of the created method are quite comparable to those achieved by colleagues in the Stentfort project, where the number of images in the database was ten times larger.

Currently, development of a mobile application with the same name is underway.

And in the meantime, we can already test other existing apps for our moles, and in the very near future we will be able to use our Ukrainian service.