Deep learning based detection of monkeypox virus using skin lesion images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Epidemiology

Background:

• The 2022 Monkeypox Virus (MPV) outbreak declared a Public Health Emergency of International Concern (PHEIC) highlights the need for rapid diagnostic tools.
• Monkeypox symptoms manifest on the skin, making imaging a viable method for detection.
• Existing diagnostic methods may be slow, necessitating advanced technological solutions.

Purpose of the Study:

• To investigate the efficacy of deep learning models for diagnosing monkeypox using skin lesion images.
• To evaluate the performance of various pre-trained deep neural networks for MPV detection.
• To assess the potential for deploying efficient deep learning models on mobile devices for widespread use.

Main Methods:

• Utilized a publicly available dataset of monkeypox skin lesion images.
• Tested five pre-trained deep neural networks: GoogLeNet, Places365-GoogLeNet, SqueezeNet, AlexNet, and ResNet-18.
• Performed hyperparameter tuning and evaluated models using accuracy, precision, recall, f1-score, and AUC.

Main Results:

• ResNet-18 achieved the highest accuracy at 99.49%.
• All tested modified models demonstrated validation accuracies exceeding 95%.
• Explainable AI techniques (LIME, GradCAM) were integrated for model interpretability.

Conclusions:

• Deep learning models, particularly ResNet-18, show high potential for accurate and efficient monkeypox diagnosis from skin images.
• The developed models can be deployed on resource-limited devices like smartphones, facilitating accessible screening.
• Explainable AI enhances clinical trust and utility of the diagnostic tool.