Efficient Technique for Monkeypox Skin Disease Classification with Clinical Data using Pre-Trained Models

Abstract

Monkeypox is an infectious zoonotic disease with clinical features similar to those actually observed in victims with smallpox, however being medically less severe. With the control of infectious smallpox diseases in 1980 as well as the termination of smallpox by immunization, monkeypox has become the most significant orthopoxvirus affecting global health. It is very important to prevent and diagnose this disease immediately and efficiently before its spread worldwide. Currently, the traditional system is used for the diagnosis of this infectious disease, in which a medical practitioner identifies monkeypox disease with swabs of fluid from skin rash. This approach has a lot of limitations such as it requires medical expertise, is costly and slow, and its result is not satisfactory. AI-based technologies may assist prevent and identify this infectious disorder. Because of the limitations, this proposed work suggests an AI-based diagnosis system which can detect monkeypox virus efficiently and immediately. Five transfer learning models are applied on image -based dataset with some pre-processing and optimization techniques for monkeypox virus detection. The Inception-Resnet outperformed by achieving 97% accuracy, VGG16 achieved 94% accuracy, Inception achieved 96% accuracy, VGG19 achieved 91% accuracy, and Resnet50 achieved 71% accuracy. The positive results of this investigation suggest that this strategy outperforms the current approaches. The dataset used in this proposed work is obtained from Kaggle online repository and some new patients’ data are added from various sources. This suggested strategy can be used by health professionals for screening.