Analysis on Skin Disease Segmentation and Classification using Machine Learning and Deep Learning in Dermoscopic Images

Skin diseases (SDs), particularly skin cancer, are leading causes of morbidity and mortality worldwide, with early detection being crucial for improving patient outcomes. Dermoscopy, a non-invasive imaging technique, is commonly used for evaluating skin lesions, but its analysis requires expertise and is time-consuming. Machine learning (ML) and deep learning (DL) methods have shown promise in automating the Segmentation process and Classification process of SDs (SCSD) using dermoscopic images. This review examines traditional ML algorithms and advanced DL architectures like Convolutional Neural Network model (CNNs), U-Net, and transfer learning (TL) approaches. Key challenges include dataset imbalances, the need for large labeled datasets, and the generalization of models across diverse skin types and lesion variations. The review also discusses common evaluation metrics and the strengths and limitations of current methods. By providing an overview of existing research, the paper aims to guide the development of more efficient and robust ML and DL models, advancing automated SD diagnosis and contributing to telemedicine and dermatology’s early diagnostic tools.