Early detection of human Mpox: A comparative study by using machine learning and deep learning models with ensemble...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Dermatology

Background:

• Mpox diagnosis relies on clinical expertise, which can be challenging with visually similar conditions.
• Early and accurate detection of Mpox is crucial for effective public health management and patient outcomes.

Purpose of the Study:

• To enhance early Mpox diagnosis using machine learning (ML) and deep learning (DL) models.
• To develop an ensemble model integrating ML and DL for improved classification accuracy and robustness.
• To evaluate the performance of various ML and DL models in distinguishing Mpox from other skin lesions.

Main Methods:

• Utilized the Mpox Skin Lesion Dataset v2.0 with six categories: chickenpox, cowpox, Mpox, measles, hand-foot-mouth disease, and healthy skin.
• Evaluated four models: Logistic Regression, K-Nearest Neighbors, Vision Transformer (ViT), and ConvMixer.
• Developed an ensemble model combining ViT and ConvMixer predictions for enhanced diagnostic performance.

Main Results:

• The Vision Transformer (ViT) model achieved 93.03% accuracy in Mpox lesion detection, outperforming traditional ML models.
• The ensemble model further improved diagnostic performance, demonstrating balanced precision and recall across all lesion categories.
• The proposed AI approach showed superior classification accuracy compared to existing studies in differentiating Mpox from similar conditions.

Conclusions:

• Integrating ML and DL models in an ensemble framework significantly improves Mpox detection capabilities.
• This AI-driven diagnostic approach provides a scalable, accurate, and efficient solution for Mpox diagnosis, especially in resource-limited settings.
• Future research directions include enhancing model interpretability, incorporating federated learning, and validating with real-world clinical data.