Detecting monkeypox disease from skin lesion images using deep learning methods

Monkeypox disease virus, while trying to recover from COVID-19, poses a new outbreak threat. Although Monkeypox disease is not as deadly and contagious as COVID-19, it has the potential to evolve into a global pandemic with new cases being reported daily. Deep Learning techniques in medical imaging offer promising prospects for identifying which disease a person has. Images of skin lesions infected with Monkeypox virus can be used for early detection of the disease. However, there is currently no approved database by the World Health Organization. It is crucial to create a proper image dataset to train deep learning models accurately. This study consists of two parts. The first part presents a deep learning model trained with the Monkeypox Skin Image Dataset (MSID). The second part presents a deep learning model trained with a combined dataset created from Monkeypox Skin Image Dataset (MSID), Monkeypox Master (MM), and Monkeypox Original Images (MOI) datasets. These images are collected from various open-source and online sources and suitable for research purposes. In this study, five different deep learning models, namely DenseNet201, InceptionResNetV2, InceptionV3, NASNetLarge, and Xception, are tested and compared on the two created databases. By utilizing augmented datasets, the DenseNet201 model is evaluated and recommended, achieving identification of monkeypox disease with an accuracy of 99.33% and 98.52% on the MSID and HYBRID datasets, respectively. The proposed model contributes to the preservation of human health by preventing a potential pandemic.