Monkeypox and Chickenpox Skin Lesions Classification Using Hybrid Deep Learning Features

Abstract

In light of the global COVID-19 pandemic, the timely identification of infections has become more critical than ever before. Since a new monkeypox outbreak emerged in May 2022, there has been a pressing need to differentiate it from similar skin lesions like chickenpox and measles. Contracting monkeypox necessitates isolation for affected individuals. While Polymerase Chain Reaction (PCR) tests offer accurate diagnosis, their availability remains limited. Harnessing machine learning techniques presents an opportunity to facilitate early detection and mitigate transmission risks. To address this challenge, a series of four experiments were conducted using Convolutional Neural Network (CNN) models, using the Monkeypox Skin Lesion Dataset (MSLD). This research aims to propose a hybrid model for monkeypox classification. The feature was extracted using different deep learning CNN models. Standard machine learning algorithms were used for the final classification. The experiments achieved an excellent accuracy 96% when ResNet101 and InceptionNet (V2) features are combined. By leveraging machine learning, this approach demonstrates the potential for early detection, thereby playing a crucial role in reducing the transmission of the monkeypox virus.