Advancing Melanoma Detection with MelDetect System

This article presents a systematic review and development of MelDetect, a multimodal deep learning system aimed at improving melanoma detection. The study highlights the rising incidence of melanoma, a highly aggressive skin cancer, and emphasizes the importance of early detection for enhancing survival rates. MelDetect integrates dermoscopic images with clinical metadata to enhance diagnostic accuracy, achieving a test accuracy of 81.83% and a macro-average AUC of 0.95 using the HAM10000 dataset. The system demonstrates high sensitivity for critical lesion classes, with recall rates of 89.63% for melanocytic nevi and 92.86% for vascular lesions. The article discusses traditional melanoma detection methods, including the limitations of visual inspection and the need for automated systems. It also outlines the experimental design, dataset selection, and the multimodal integration framework utilized in the study. The findings indicate that MelDetect could serve as a reliable, non-invasive tool for early melanoma detection and clinical decision support.