XMP-Net: An XAI-Based Modified Xception Model for Recognizing Monkeypox and Other Skin Diseases

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Dermatology

Background:

• Accurate and timely diagnosis of skin conditions is crucial for effective treatment and public health.
• Emerging infectious diseases like monkeypox require rapid and reliable diagnostic tools.
• Deep learning offers potential for automated skin condition classification from images.

Purpose of the Study:

• To develop and evaluate XMP-Net, a modified Xception-based deep learning model for classifying skin conditions.
• To specifically enhance the identification of monkeypox among normal skin, chickenpox, and measles.
• To improve model interpretability using Grad-CAM and LIME for clinical trust.

Main Methods:

• Utilized a modified Xception architecture (XMP-Net) for image classification.
• Trained the model on a dataset of four skin condition categories: normal, chickenpox, measles, and monkeypox.
• Employed Grad-CAM and LIME for visual explanation of model predictions.

Main Results:

• XMP-Net achieved high accuracy: 98.33% for normal, 98.25% for monkeypox, 84.21% for measles, and 77.27% for chickenpox.
• Monkeypox classification showed strong performance with 91.80% precision, 98.25% recall, and 94.92% F1-score.
• Explainable AI methods highlighted image regions crucial for accurate classification.

Conclusions:

• XMP-Net demonstrates significant potential for accurate skin condition diagnosis, especially for monkeypox.
• Explainable AI integration enhances the interpretability and clinical utility of deep learning models.
• The research provides a foundation for developing accessible diagnostic tools for resource-limited settings.