Multi-Class Classification of Skin Diseases Using ResNet50

Abstract

The World Health Organization (WHO) reports that cancer is one of the leading causes of death globally and is responsible for approximately 10 million deaths yearly. Globally, about 1 in 6 deaths are attributed to cancer. As the body’s outermost organ, the skin is vulnerable to various diseases, and accurate diagnosis is crucial for effective treatment. However, limited access to dermatologists and expensive skin biopsies pose challenges to efficient diagnosis. In this research, we utilized the ResNet50 model to classify seven types of skin diseases. We employed the HAM10000 dataset, which consists of images representing disease classes such as akiec, bbc, bkl, df, mel, vasc, and nv. The ResNet50 model was trained and evaluated using accuracy, precision, recall, and F1score performance metrics. The research findings reveal that the ResNet50 model achieved a high level of accuracy, with an accuracy rate of 91.71% on the test data. The precision, recall, and F1-Score metrics also demonstrated excellent performance, with 91.89%, 92.04%, and 91.87% respectively. These results confirm the model’s effectiveness in classifying various skin diseases, demonstrating high precision for accurate identification and high recall for locating all actual instances. The balanced F1-Score emphasizes the model’s proficiency in diagnosis, advancing skin disease classification with ResNet50. Overall, these findings contribute to advancing skin disease diagnosis using the ResNet50 model and highlight its efficacy in achieving accurate and reliable classification results.

Topics & Concepts