Automated Skin Lesion Border Detection Techniques – A Key Component in Melanoma Screening

Abstract

Accurate identification of lesion boundaries in dermoscopic images plays a critical role in automated melanoma screening, as the effectiveness of subsequent feature analysis and diagnostic decisions depends heavily on segmentation performance. This review presents a comprehensive overview of skin lesion border-detection methodologies, highlighting their evolution from classical image-processing techniques and hybrid approaches to advanced deep-learning–based segmentation models. In addition, it discusses commonly used public datasets, standard performance metrics, and typical segmentation challenges encountered in real-world clinical images. The paper further identifies existing limitations in current research and outlines promising future directions aimed at improving robustness, generalization, and clinical applicability. Special attention is given to dataset availability, fair benchmarking practices, and practical recommendations for developing reliable and clinically viable lesion segmentation systems.