A deep learning based approach for automatic detection of COVID-19 cases using chest X-ray images

Area of Science:

• Medical Imaging and Artificial Intelligence
• Radiology and Diagnostic Tools
• Computational Pathology

Background:

• The COVID-19 pandemic necessitates rapid and efficient diagnostic methods to curb transmission of SARS-CoV-2.
• Radiological images, specifically chest X-rays, contain vital information for identifying lung infections.
• AI-assisted automated detection shows promise as a supplementary diagnostic tool for COVID-19.

Purpose of the Study:

• To propose and evaluate a novel AI-based method for detecting COVID-19 and pneumonia using chest X-ray images.
• To enhance diagnostic capabilities for infectious lung conditions through automated image analysis.
• To compare the performance of different AI architectures for accurate classification.

Main Methods:

• A three-step process involving image segmentation, feature extraction, and classification.
• Conditional Generative Adversarial Network (C-GAN) for lung image segmentation.
• Integration of key point extraction (BRISK) and deep neural networks (DNNs) for feature extraction, followed by machine learning (ML) classification.

Main Results:

• The proposed method successfully segmented lung images and extracted discriminatory features.
• A comparative analysis demonstrated the effectiveness of various AI and ML model combinations.
• The highest testing classification accuracy achieved was 96.6% using the VGG-19 model with the BRISK algorithm.

Conclusions:

• The developed AI-assisted method is efficient for screening COVID-19 infected patients.
• This approach can augment conventional diagnostic tests, improving the management of respiratory infections.
• Automated analysis of chest X-rays holds significant potential for rapid disease detection in pandemic scenarios.