Diagnosis of Monkeypox Disease Using Transfer Learning and Binary Advanced Dipper Throated Optimization Algorithm

Area of Science:

• Medical Informatics
• Computational Biology
• Artificial Intelligence in Healthcare

Background:

• Monkeypox virus (MPXV) detection is critical, especially following the COVID-19 pandemic, due to potential public health crises.
• Skin lesions are a key indicator of MPXV infection, necessitating rapid and accurate diagnostic tools.
• Existing diagnostic methods may require enhancement for pandemic-level detection capabilities.

Purpose of the Study:

• To develop and evaluate a novel metaheuristic optimization approach for enhanced monkeypox detection.
• To improve the performance of feature selection and classification for identifying monkeypox-indicative skin lesions.
• To establish a robust system for rapid monkeypox case identification during potential pandemics.

Main Methods:

• Utilized deep learning (GoogLeNet) and transfer learning for feature extraction from skin lesion indicators.
• Employed a binary Dipper Throated Optimization (DTO) algorithm for efficient feature selection.
• Applied a Decision Tree classifier, optimized with a continuous DTO algorithm, for accurate classification.

Main Results:

• Achieved high performance metrics: F1-score of 0.92, sensitivity of 0.95, specificity of 0.61, p-Value of 0.89, and N-Value of 0.79.
• The optimized Decision Tree classifier resulted in an overall accuracy of 94.35% for monkeypox detection.
• Statistical tests (ANOVA, Wilcoxon signed rank test) confirmed the proposed method's superiority over alternatives.

Conclusions:

• The proposed metaheuristic optimization approach significantly enhances the accuracy of monkeypox detection using skin lesions.
• This AI-driven methodology offers a unique and valuable tool for rapid identification of monkeypox cases.
• The findings support the potential of this approach for early detection and management of MPXV outbreaks.