Detecting POX Disease in Skin Images Using Explainable Artificial Intelligence

Abstract

The intricate domain of pox diseases, including chickenpox, cowpox, monkeypox, hand, foot, and mouth disease (HFMD), and measles, and their profound influence on human health is the focal point of extensive research. In response to the compelling demand for precise disease classification, an exploration is undertaken in this study, delving into the intersection of machine learning (ML) methodologies and the pursuit of interpretability through the application of explainable artificial intelligence (XAI). In this study, we use ML methods for pre-processing, then train the data and apply the XAI approach to those trained models. First, ML methods scikit-image were used to segregate the 15,000 images into train (70%), test (20%) and valid (10%). Then, we used 8 CNN models, namely AlexNet, LeNet, SeNet, GoogleNet, SpinalNet, MobileNetV1, VGG, and ZFNet, to train the model. The accuracy of GoogleNet is 82%, which is much better than that of other CNN models.