Detection of Monkeypox Disease from Human Skin Images with a Hybrid Deep Learning Model

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Virology

Background:

• Monkeypox is a zoonotic DNA virus with distinct lineages, transmissible through animal and human contact.
• Transmission occurs via direct contact with bodily fluids, blood, or respiratory secretions, manifesting as skin lesions.
• Accurate and timely detection of monkeypox is crucial for public health management.

Purpose of the Study:

• To develop and evaluate a hybrid artificial intelligence system for detecting monkeypox from skin images.
• To address data imbalance in monkeypox datasets using augmentation and preprocessing techniques.
• To compare the performance of various deep learning models for monkeypox classification.

Main Methods:

• Utilized an open-source, multi-class dataset including chickenpox, measles, monkeypox, and normal skin images.
• Applied data augmentation and preprocessing to mitigate class imbalance issues.
• Developed a novel hybrid deep learning model combining top-performing models with Long Short-Term Memory (LSTM) for enhanced detection.

Main Results:

• The hybrid AI system achieved a test accuracy of 87% for monkeypox detection.
• Cohen's kappa score for the model reached 0.8222, indicating substantial agreement.
• The study demonstrated the effectiveness of deep learning and hybrid models in classifying monkeypox lesions.

Conclusions:

• The proposed hybrid AI system shows significant potential for accurate monkeypox detection from skin images.
• The methodology effectively handles imbalanced datasets, crucial for medical image analysis.
• This AI-driven approach can support clinical diagnosis and epidemiological surveillance of monkeypox.