PoxDetect: Advancing Monkeypox Diagnosis with Machine Learning for Skin Lesion Classification

EasyChair Preprint 15298

Abstract

The monkeypox virus can infect nonhuman primates as well as humans. This dissertation examines the use of machine learning (ML) for early detection and classification of monkeypox disease caused by the varicella-zoster virus infects both humans and nonhuman primates. Emphasizes the importance of diagnosing monkeypox skin lesions in their early stages for effective treatment and disease prevention. Utilized images of monkeypox lesions obtained from Kaggle, then augmented to develop and test various test our own ML models. A basic mobile app was created to allow users to capture and send images for analysis by the models.Primarily focused on evaluating the effectiveness of different ML models,including ResNet50, InceptionV3, Xception, DenseNet121, and MobileNet. Results showed that the MobileNet and Xception models performed best with MobileNet achieving a mean accuracy 0.97 and an F-1 score of 0.968, while Xception achieved an accuracy of 0.986 and F-1 score of 0.98. Potential impact of ML in healthcare particularly in disease classification and identification.

Keyphrases: DenseNet, Inception, MobileNet, Monkeypox, ResNet, Skin illness, Xception, machine learning

@booklet{EasyChair:15298,
  author    = {Md.Saiful Islam and Mahede Hasan and Sheikh Fazle Rabbi},
  title     = {PoxDetect: Advancing Monkeypox Diagnosis with Machine Learning for Skin Lesion Classification},
  howpublished = {EasyChair Preprint 15298},
  year      = {EasyChair, 2024}}