Nature-Inspired Multi-Level Thresholding Integrated with CNN for Accurate COVID-19 and Lung Disease Classification in...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Diagnostic Technologies

Background:

• Accurate COVID-19 classification from chest X-rays is challenging due to overlapping features with other lung diseases.
• Current diagnostic methods for lung conditions exhibit suboptimal performance.
• A significant diagnostic gap exists in precise segmentation and classification of lung pathologies.

Purpose of the Study:

• To develop a novel hybrid framework for precise segmentation and classification of lung conditions, specifically differentiating COVID-19.
• To address the limitations of existing methods in accurately diagnosing COVID-19 from chest X-rays.
• To improve the diagnostic accuracy and efficiency of identifying lung diseases.

Main Methods:

• A hybrid approach combining multi-level thresholding with metaheuristic optimization algorithms (Animal Migration Optimization, Electromagnetism-like Optimization, Harmony Search Algorithm) for enhanced image segmentation.
• Utilizing a Convolutional Neural Network (CNN) for the classification of segmented chest X-ray images.
• Categorizing images into COVID-19, viral pneumonia, or normal based on segmented features.

Main Results:

• The proposed framework achieved exceptional diagnostic performance: 99% accuracy, 99% sensitivity, and 99.5% specificity.
• The method demonstrated robustness and effectiveness in clinical image classification tasks.
• High diagnostic metrics confirm the reliability of the hybrid approach.

Conclusions:

• This study presents a novel, technically integrated solution for automated COVID-19 diagnosis using chest X-rays.
• The framework's high accuracy and computational efficiency suggest potential for real-world clinical deployment.
• The research offers a promising advancement in medical diagnostics for lung conditions.