Classifying COVID-19 Patients From Chest X-ray Images Using Hybrid Machine Learning Techniques: Development and...

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Computational Biology

Background:

• COVID-19 pandemic caused global concern, with severe cases showing lung inflammation.
• Chest X-ray (CXR) is vital for diagnosis, typically interpreted by specialists.
• Machine learning (ML) shows promise in CXR analysis but often lacks computational efficiency.

Purpose of the Study:

• Introduce a novel hybrid ML model for accurate COVID-19 classification from CXR images.
• Reduce computational time for CXR interpretation.
• Evaluate and compare the model's performance against existing methods.

Main Methods:

• Retrospective analysis of 4200 CXR images from 5 public datasets.
• Application of ML techniques: decision trees, support vector machines, neural networks.
• Development of a two-layer hybrid classification model (MLHC-COVID-19) after preprocessing and feature extraction.

Main Results:

• The MLHC-COVID-19 model achieved high accuracy (0.962) and F-measure (0.962) on unseen COVID-19 CXR images.
• Demonstrated effectiveness in classifying COVID-19 CXR images with improved accuracy and reduced interpretation time.
• Integrated into a publicly available web-based computer-aided diagnosis system.

Conclusions:

• The MLHC-COVID-19 model effectively differentiates COVID-19, non-COVID-19, and healthy individuals from CXR images.
• Outperformed state-of-the-art deep learning techniques, suggesting its utility in early COVID-19 detection.
• The model's adaptability makes it valuable for future pandemic situations and is accessible via a web platform.