Detection of Monkey pox Disease Using Computer Vision-Based Transfer Learning Models: A Comparative Study

Abstract

The world is now concern about monkey pox where the world is still infected by the deadly covid-19 virus. Monkey pox is not as infectious as covid-19 but it is a contagious disease. Monkey pox cases are being reported in many different countries in recent time . It is assumed that world will be facing another pandemic due to monkey pox if necessary precautions are not taken. Monkey pox transmitted through human to human or animal to human or animal to animal directly (physical contact: sex, skin to skin touch) or indirectly. It will be catastrophic if community transmission takes place. Artificial Intelligence(AI) has great improvement on image processing . It extracts unique features from the image. Machine learning has been used on the medical purpose by image diagnosis like cancer detection, covid detection. It is also used for classification. Monkeypox eruptive phase includes maculopapular rash on the skin. But it is difficult to identify because rash can be appeared for many reasons. Also monkeypox and smallpox lesions are quite similar so it makes more difficult to distinguish for doctors which one is it. Medical test such as PCR needed to identify monkeypox which is costly. Machine Learning provides some algorithms for image processing and classification. As skin lesion occurs due to monkey pox, monkey pox can be detected by image diagnosis in the early stage. People can then do tests to confirm monkey pox. Thus, Community transmission can be lessened and unnecessary cost for PCR or other diagnosis will be saved by adopting such machine learning application. We used MSLD dataset and deep learning algorithms: Mobile Net, InceptionResNet-V2 and DenseNet-201 to classify into monkey pox and Others. We developed a CNN model SkinNet-9 for monkey pox detection. Mobile Net and DenseNet-201 gave the best accuracy of 95.42% among these algorithms.