Automated Monkeypox Skin Lesion Detection Using Deep Learning and Transfer Learning Techniques.

Area of Science:

• Medical Informatics
• Computer Science
• Virology

Background:

• The global spread of monkeypox (mpox) necessitates rapid and accurate diagnostic tools.
• Early detection is critical for effective mpox treatment and public health management.

Purpose of the Study:

• To evaluate and validate deep learning models for detecting mpox.
• To identify the best-performing model for mpox image classification.

Main Methods:

• Five pretrained deep learning models (VGG19, VGG16, ResNet50, MobileNetV2, EfficientNetB3) were assessed.
• Model performance was quantified using accuracy, recall, precision, and F1-score.
• The top model was validated on diverse datasets.

Main Results:

• MobileNetV2 demonstrated superior performance with 98.16% accuracy, 0.96 recall, 0.99 precision, and 0.98 F1-score.
• External validation confirmed MobileNetV2's efficacy, achieving 0.94% accuracy.
• The MobileNetV2 model outperformed existing literature methods for mpox image classification.

Conclusions:

• Deep learning, specifically the MobileNetV2 model, offers a promising approach for the early and accurate detection of mpox.
• This AI-driven method can serve as a valuable tool for clinical diagnosis, improving mpox management strategies.