Enhanced Simple Linear Iterative Clustering (ESLIC) for Precise Skin Lesion Segmentation in Dermoscopic Images
The main challenge in dermoscopy image analysis lies in the lesion segmentation process, which is often affected by variations in color, texture, lighting, and irregular lesion boundaries. The Simple Linear Iterative Clustering (SLIC) method is widely used for superpixel-based segmentation, but it still faces limitations in preserving the morphological details of lesions in complex areas. This study proposes Enhanced Simple Linear Iterative Clustering (ESLIC) as a new approach to improve the quality of skin lesion segmentation. ESLIC combines a dynamic compactness adjustment mechanism, local texture analysis, and gradient-based boundary refinement to generate superpixels that are more adaptive to image characteristics. The study was conducted using 277 dermoscopy images consisting of benign lesions and melanoma lesions. Before segmentation, the images underwent image enhancement through filtering, normalization, contrast enhancement, and sharpening. ESLIC’s performance was evaluated based on segmentation quality and its impact on the classification process using a Convolutional Neural Network (CNN). The experimental results show that ESLIC is capable of producing a more accurate representation of lesion areas compared to conventional superpixel methods, particularly in areas with heterogeneous textures and complex lesion boundaries. The findings of this study indicate that ESLIC can improve the effectiveness of feature extraction and has the potential to become a key component in the development of artificial intelligence-based melanoma diagnosis support systems.