Attention-based VGG-16 model for COVID-19 chest X-ray image classification

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• Computer-aided diagnosis (CAD) using Chest X-rays (CXR) offers a cost-effective approach for early COVID-19 detection compared to PCR or CT scans.
• Existing CXR-based CAD methods often neglect spatial relationships within lung regions, limiting diagnostic accuracy.
• Identifying COVID-19's impact on lung ROIs requires models that understand inter-region correlations.

Purpose of the Study:

• To develop a novel attention-based deep learning model for improved COVID-19 diagnosis from CXR images.
• To enhance the capture of spatial relationships between Regions of Interest (ROIs) in CXR scans.
• To fine-tune the classification process using VGG-16's features for accurate COVID-19 detection.

Main Methods:

• A novel deep learning model integrating an attention module with the VGG-16 architecture was developed.
• The attention module was employed to capture spatial dependencies between ROIs in CXR images.
• Fine-tuning was achieved by incorporating the 4th pooling layer of VGG-16 alongside the attention mechanism.

Main Results:

• Extensive experiments were conducted on three distinct COVID-19 CXR datasets.
• The proposed attention-based model demonstrated stable and promising performance.
• Results indicated superior classification accuracy compared to current state-of-the-art methods.

Conclusions:

• The developed attention-based deep learning model shows significant potential for CXR image classification in COVID-19 diagnosis.
• The model's ability to analyze spatial relationships enhances its suitability for early disease detection.
• This approach offers a promising, cost-effective alternative for widespread COVID-19 screening.