Abstract:

A key predictor of survival in the treatment of skin diseases is early detection and accurate diagnosis based on access to relevant data. There is a need to educate society on this matter and to provide more accessible self-assessment solutions for the early detection of dermatological conditions. In the treatment process, early diagnosis is a critical factor for disease remission, effective mitigation of its impact, and improved survival chance. The study focuses on predicting dermatological changes using a dedicated system that enables image capture and editing that then utilizes machine learning for analysis and identification. The application allows for multi-class prediction of pathological skin lesions, in contrast to current solutions based on binary classification. Based on the obtained results and conducted tests, it can be stated that the presented system demonstrated the ability to accurately identify seven skin disease units, achieving an overall accuracy of 80%.