Monkeypox Detection using Skin Lesion Images: A Comparative Analysis of Deep Learning Models

The current monkeypox virus outbreak has brought attention to the necessity of quick and accurate diagnostic techniques, particularly in settings with limited resources. This work investigates automated monkeypox identification using deep learning methods based on images of skin lesions. Three deep learning models are used and their classification performance are assessed: Vision Transformer (ViT), ResNeXt, and Convolutional Neural Network (CNN). The CNN model demonstrates great sensitivity but low precision, despite its efficiency. This is greatly enhanced by the ResNeXt model, which is achieved a high balance between sensitivity and specificity with improved test accuracy. Both models are surpassed by the ViT model, which achieves nearly flawless classification with an F1-score of 0.99 and both test accuracy, precision, and recall of 99.69%. These findings imply that transformer-based architectures, as opposed to traditional convolutional methods, can improve diagnostic accuracy. This experiment shows how computer-aided diagnostics can help medical personnel identify diseases quickly and accurately by using deep learning for automated monkeypox diagnosis. According to the results, deep learning may be crucial in enhancing the effectiveness of image-based analysis for infectious disease detection, especially in outbreak situations where prompt screening is crucial.