International Journal of Computer Science and Technology

International Journal of Computer Science and Technology | Vol. 1, No. 2, February 2020 | pp. 9–16

Review Article

Title: A Comprehensive Survey of Deep Learning Techniques for Automated Skin Lesion Classification in Dermatology

Names of Authors: Fatima Zahra¹, Tariq Al-Mansoor², and David Miller³

Authors’ Affiliations:
¹Department of Biomedical Engineering, Cairo University, Giza, Egypt
²Department of Artificial Intelligence, King Abdulaziz University, Jeddah, Saudi Arabia
³Department of Computer Science, University of Oxford, Oxford, United Kingdom

Abstract: Early detection of malignant melanoma and other pigmented skin lesions remains a critical factor in improving patient survival rates in modern dermatology. Manual visual inspection by clinicians is inherently subjective, time-consuming, and prone to diagnostic variability, necessitating reliable computer-aided diagnostic (CAD) systems. In recent years, deep learning (DL) architectures—particularly convolutional neural networks (CNNs)—have revolutionized medical image analysis by automating feature extraction and classification tasks with high precision. This paper delivers a comprehensive, structured survey of state-of-the-art deep learning methodologies applied to automated skin lesion classification using benchmark datasets such as ISIC and HAM10000. We examine the evolution of deep architectures ranging from traditional AlexNet and VGG models to advanced residual networks (ResNet), dense networks (DenseNet), and vision transformers (ViT). Special attention is dedicated to preprocessing challenges, including hair artifact removal, color normalization, class imbalance handling via data augmentation, and generative adversarial networks (GANs). Furthermore, we analyze quantitative performance metrics across studies, noting that ensemble CNN models achieve diagnostic accuracy exceeding 94.8% and area under the ROC curve (AUC) values up to 0.99. Open research gaps are highlighted, including model interpretability, cross-domain generalization under varying clinical imaging equipment, and the deployment constraints of lightweight models on mobile edge devices. Finally, future directions are outlined to bridge the translational gap between algorithmic development and routine clinical workflows.

Keywords: Deep learning, Convolutional neural networks, Skin lesion classification, Melanoma detection, Medical image analysis, Computer-aided diagnosis

Manuscript Timeline: Received: November 10, 2019; Revised: December 15, 2019; Accepted: January 12, 2020; Published: February 1, 2020

Citation: Zahra, F., Al-Mansoor, T., & Miller, D. (2020). A comprehensive survey of deep learning techniques for automated skin lesion classification in dermatology. International Journal of Computer Science and Technology, 1(2), 9–16. DOI: 10.46882/2020/IJCST/000002

Subscribe to read the full article: https://internationalscholarsjournals.org/subscribe-to-read