Deep Residual Networks for Pigmented Skin Lesions Diagnosis

The skin is a barrier, protecting the body from external factors such as bacteria, viruses, and harmful environmental agents. Healthy skin is a crucial factor in creating an attractive appearance and a source of motivation for personal confidence. Skin conditions can also affect our emotions and mental well-being, with skin problems that can cause stress and self-consciousness. Maintaining skin health helps maintain an appealing outward appearance and plays a vital role in the early detection of health issues such as skin cancer and other dermatological conditions. Using advancements in deep learning has facilitated the diagnosis of various pigmented skin lesions, including actinic keratosis, basal cell carcinoma, seborrheic keratosis, dermatofibroma, malignant melanoma, nevus, and vascular lesions with Deep Residual Networks on images of various sizes. However, traditional methods have limitations in handling large image areas, especially in the presence of noise or complex lesion regions. In this study, we employed ResNet50, VGG16, and InceptionV3 for skin disease diagnosis, showing promising results compared to previous studies. The experimental results demonstrate high performance across accuracy, Area Under the curve, Precision, Recall, and F1-score.