Unveiling the Unique Dermatological Signatures of Human Pox Diseases Through Deep Transfer Learning Model Based on DenseNet and Validation with Explainable AI

Differentiating between human monkeypox, chickenpox, cowpox, and measles diseases based on skin symptoms alone pose a challenge due to their similarities, potentially leading to misdiagnosis and delayed treatment. Currently, doctors look at samples by hand or rely on confirmation tests that are not always easy to obtain, such as polymerase chain reaction (PCR) tests, which take a long time. A few studies have focused on individual disease classification using traditional deep learning techniques, but there is a lack of research addressing the specific problem of distinguishing between different diseases based on skin symptoms using deep transfer learning with greater accuracy. We propose utilizing transfer learning techniques, specifically based on DenseNet201, trained on a large dataset of annotated skin images, to develop an automated system capable of accurately differentiating between human monkeypox, chickenpox, cowpox, measles, normal and hand-mouth face disease. Finally, our proposed model resulted in a test accuracy of 0.90, a precision of 0.89, a recall of 0.91, and an F1 score of 0.90, which significantly outperformed all other models, avoided common skin problems and explainable AI used to detect which portion of the image our model categorized the sample as any disease.