Developing and Testing Deep Learning Models for Monkeypox and Herpes Zoster Rash Differentiation

Abstract

Monkeypox (mpox) is a viral infection known for its pimple-like skin rashes found across the body. Since the 2022 outbreak, monkeypox has been mistaken for similar conditions, such as herpes zoster (shingles). In the past, deep learning, a form of machine learning in which software models learn to form conclusions using artificial neural networks (ANNs) similar to that of the human brain, have been utilized to identify various dermatological conditions. Utilizing the engineering method, the following research question was inquired: How can deep learning be applied to a mobile application to distinguish monkeypox viral lesions from herpes zoster viral lesions? ResNet50 and MobileNetV3, two deep-learning models, were trained utilizing transfer learning for the task of differentiating between monkeypox and shingles. ResNet50 and MobileNetV3 were then tested utilizing a separate testing set of images. Both models achieved an accuracy of 93.10 percent, with ResNet50 achieving a loss of 17.03 percent and MobileNetV3 achieving a loss of 29.30 percent. It was concluded that ResNet50 may be more precise due to its lower loss percentage, indicating fewer instances of errors. An end-user prototype of a mobile application was created to simulate these deep-learning models. These findings suggest that deep learning is an efficient and precise solution to prevent the misdiagnosis of monkeypox, potentially hindering the rising cases of this virus while aiding in proactive measures to identify and treat monkeypox victims.