Automated detection of human monkeypox using skin lesion images and transfer learning CNN

A virus that causes skin lesions is the cause of monkeypox disease, which has been detected in the African continent recently. The people are afraid and panicked because of the deadly outcomes of virus infections during the COVID-19 epidemic. Human monkeypox (MPX), the deadliest infection, is contracted by direct contact with animals and can also be transmitted via body fluids, coughing, sneezing, and infected objects such as bedding. The symptoms of MPX infections are similar to those of chickenpox and measles, manually screening for infections at an early stage takes time, is prone to community spread, and is prone to human error. Computer-assisted detection of MPX may improve clinical efficacy and reduce healthcare worker’s load. The main objective of this research is to provide an automated system for MPX infection self-diagnosis. In this work, an automatic method for identifying MPX-infected people is developed using a convolutional neural network (CNN) and a publicly accessible skin lesion database. To identify MPX infection, pre-trained CNNs—GoogLeNet, EfficientNet, and VGG19 are used. For training benchmarked models, a transfer-learning technique is applied. To create a more reliable network, a data augmentation technique is combined with a repeating 10-fold holdout cross-validation scheme. EfficientNet obtained the maximum classification accuracy of 98.68%, sensitivity of 98.96%, and specificity of 98.86%. The experimental findings reveal that CNN has promise for identifying MPX-positive individuals based on pictures of skin lesions. The framework that is being provided may serve as an interface component for cloud-based medical diagnostic systems. • Human monkeypox is the serious infection. • Automated detection of monkeypox using skin lesion images. • Explored GoogLeNet, EfficientNet, and VGG19 foe classification. • Transfer-learning is used to re-train the benchmark models.