A Hybrid Model of Deep Learning and Machine Learning with Image Feature Extraction Techniques for Melanoma: Supporting Clinical Diagnostic Decision Making

Abstract

Skin cancer is among the most common malignancies, with melanoma being the most dangerous form due to its high metastatic potential and poor prognosis if not detected early. Despite significant advancements in personalized medicine and image-based skin cancer classification algorithms, most approaches have not fully explored the unique visual characteristics of melanoma. This study focuses on analyzing the ABC features (Asymmetry, Border, Color) of melanoma lesions to extract information related to shape, boundary, and color patterns. These features play a crucial role in distinguishing malignant tumors from benign lesions. By developing a multidimensional analysis model, the research aims to provide deeper insights into the histological characteristics of individual lesions, thereby supporting more accurate and timely clinical diagnosis and treatment decisions.