Deep Learning–Based Skin Cancer Diagnosis in the United States: Advances, Challenges, and Clinical Translation

Abstract

Skin cancer represents one of the most common and potentially fatal malignancies in the United States. Where timely and accurate diagnosis is critical for reducing mortality and healthcare burden. Current diagnostic practices, including visual examination and dermoscopy. This rely heavily on clinician expertise and are subject to inter-observer variability, particularly in early-stage disease. In recent years, deep learning has emerged as a promising tool for automating skin cancer detection from dermoscopic images. With offering improved diagnostic accuracy and scalability. This article provides a comprehensive analysis of deep learning-based approaches for skin cancer classification, with a focus on convolutional neural network architectures, commonly used datasets, and performance evaluation metrics relevant to U.S. clinical settings. Key challenges, including dataset imbalance, limited representation of diverse skin tones, overfitting on small cohorts, and high computational demands, are critically examined. Additionally, the study discusses emerging trends toward lightweight and deployable models suited for real-time clinical workflows and mobile health applications within the United States healthcare system. The findings aim to support the development of robust, generalizable, and clinically translatable deep learning solutions for skin cancer diagnosis.