Development of an accurate AI-based dermatology assistant for skin disease recognition using YOLOv8 models

• Main Authors: Huzaini, Muhammad Irfan Darwish, Mansor, Hasmah, Gunawan, Teddy Surya, Ahmad, Izanoordina
• Format: Proceeding Paper
• Language: English
• Published: IEEE 2024
• Subjects: T Technology (General)
• Online Access: http://irep.iium.edu.my/114525/7/114525_Development%20of%20an%20accurate%20AI-based.pdf; http://irep.iium.edu.my/114525/; https://ieeexplore.ieee.org/document/10675537; https://doi.org/10.1109/ICSIMA62563.2024.10675537

The AI-based dermatology assistant for skin disease recognition is a state-of-the-art technological solution designed to address the disparity in access to professional dermatological services between urban and remote areas. This research employs the YOLOv8 model, which is a deep learning algorithm, to determine the most effective methods for detecting skin diseases. The research compares various algorithms, such as SVM, YOLOv3, YOLOv4, and Dual-Architecture CNN, through a comprehensive review of existing AI applications in dermatology. After 500 training epochs, the YOLOv8 Small Model was the most accurate, achieving a precision of 84%, a recall of 77.1%, and a mean average precision (mAP) of 84%. The potential of the proposed AI-based assistant to significantly improve healthcare accessibility and diagnostic accuracy in underserved areas is demonstrated through rigorous testing, which validates its effectiveness. This groundbreaking utilization of artificial intelligence in dermatology signifies a substantial stride forward in providing equitable healthcare solutions.