Dermoscopic Skin Lesion Classification Using Deep Learning

Abstract

Skin cancer is one of the most common and serious diseases worldwide, and early diagnosis plays a critical role in improving patient survival rates. Among different skin cancer types, melanoma is considered the most aggressive due to its high metastatic potential. However, accurate diagnosis of skin lesions remains challenging because benign and malignant lesions often exhibit very similar visual characteristics. In recent years, deep learning-based approaches, particularly Convolutional Neural Networks (CNNs), have shown remarkable success in medical image analysis and automated disease classification. In this study, a deep learning-based system is proposed for the automatic classification of dermoscopic skin lesions using the International Skin Imaging Collaboration (ISIC) dataset. The main objective of the study is to develop a reliable and efficient model capable of classifying different types of skin lesions with high accuracy. To improve the generalization capability of the model and reduce overfitting, preprocessing and data augmentation techniques such as image resizing, normalization, rotation, flipping, and zooming were applied. The proposed method utilizes a transfer learning approach based on the pre-trained DenseNet121 model. DenseNet121 enhances feature propagation and improves classification performance through its densely connected architecture. The model was evaluated using standard performance metrics, including accuracy, precision, recall, F1-score, and ROC-AUC. The validation accuracy of the model was approximately 78%, while the ROC-AUC score reached 0.88, indicating a good ability to distinguish between benign and malignant skin lesions. According to the classification report, the model achieved a precision of 0.73, a recall of 0.94, and an F1-score of 0.82 for the benign class, whereas for the malignant class it achieved a precision of 0.91, a recall of 0.61, and an F1-score of 0.73. This research supports the development of an automated decision support system for the early diagnosis of skin cancer that can assist dermatologists in clinical practice. Furthermore, the study highlights the potential of transfer learning-based deep learning approaches in medical imaging applications. The findings may contribute to the advancement of computer-aided diagnostic systems, particularly in regions where access to dermatology specialists is limited.