Health of Things Melanoma Detection System (HTMDS) -Detection and segmentation of melanoma in dermoscopic images applied to edge computing using deep learning and fine-tuning models Provisionally Accepted

José Jerovane da Costa Nascimento^1* Adriell Gomes Marques² Yasmim Osório Adelino Rodrigues² Guilherme Freire Brilhante Severiano² Icaro de Sousa Rodrigues² Carlos Dourado Jr² LUÍS FABRÍCIO DE FREITAS SOUZA^3*

• ¹Federal University of Ceara, Brazil
• ²Federal Institute of Education, Science and Technology of Santa Catarina, Brazil
• ³Universidade Federal do Cariri, Brazil

Skin cancer is a pathology that has been alarmingly developing worldwide. According to the World Health Organization (WHO), melanoma is a type of cancer that affects people globally in different parts of the human body, leading to the death of thousands each year worldwide.Intelligent diagnostic tools through automatic detection in medical images are extremely effective in assisting medical diagnosis. Systems based on Computer Aided Diagnostics (CADs) are extremely important for pre-diagnosis by images, and the use of tools based on artificial intelligence for monitoring, detection, and segmentation of the pathological region are increasingly used in intelligent solutions integrated into smart city systems, through cloud data processing with the use of edge computing. This study proposes a new approach for smart cities capable of being integrated into computational systems for monitoring and assisting medical diagnosis, a fully automatic intelligent system based on the Internet of Things (IoT) for health systems in smart cities. Named Health of Things Melanoma Detection (HTMDS), the method presents a deep learning approach based on the Yolo V8x network for detecting melanomas in dermatoscopic images of melanoma, proposing new fine-tuning techniques for segmentation of the melanoma region. The new approach achieved satisfactory results above 98% accuracy for detection and above 99% accuracy for skin cancer segmentation, surpassing different works found in the state of the art in various methods, including manual, semi-automatic, and automatic approaches.