A Novel Health Care System for Improving Dermatological Diagnosis: A Model for Skin Lesion Classification

Melanoma, nevus, and basal cell carcinoma (BCC) lesions are three types of skin cancer that pose a substantial threat to people's health all over the world. Early identification makes a significant contribution to the improvement of patient outcomes and the reduction of mortality rates. Within the scope of this research article, we investigate various deep-learning approaches for categorizing three lesions. The primary objective is to strengthen dermatological diagnosis. Furthermore, we overcome the issue of less annotated data directly affecting the model performance. We proposed the multi-output convolutional neural network (MO-CNN) model with two transfer learning approaches, VGG-19 with frozen layers and AlexN et with fine-tuning. The performance of our suggested MO-CNN model is compared to that of state-of-the-art methods like CNN, Transfer learning, and DSCC. According to the findings, the MO-CNN model achieves high accuracy, precision, recall, and Fl-Score levels. The model's accuracy reaches 97.81%, the precision of the model reaches 99.19%, the recall of the model reaches 97.04 %, and the Fl-Score reaches 97.01 %. The results of this study shed light on the potential of deep learning models, in particular the D-CNN architecture, to provide accurate and efficient classification of skin cancer. The findings of this study contribute to the development of enhanced diagnostic tools for melanoma and nevus lesions, which will ultimately result in improvements to patient treatment and prognosis in the field of dermatology.