NCFU mathematicians have developed a neural network-based system that allows identifying 10 types of skin issues and offers unmatched precision.

The system implies image processing and involves various skin information that incorporates data such as the patient’s sex, age, and the location of the tumor.

As Pavel Lyakhov (Associate Professor, Head of Dept for Mathematical Modelling) noted, diverse data, if employed to make decisions while trying to detect health issues, will enhance precision through establishing connections among visual and statistical data pieces.

The newly developed system can now recognize 10 diagnostic categories of pigmented skin lesions – from dermatofibroma, nevus, solar lentigo and various ketaroses to melanoma and other skin cancers. The highest recognition accuracy achieved is 83.6%, which is way above the accuracy offered by visual examination.

Dmitry Bespalov (Rector, NCFU) stressed that the AI-based automated systems can really boost diagnostic accuracy.

Now, the researchers’ further plans include designing a mobile app so that anyone can check and identify their skin issues and then seek medical assistance, if needed.