COVID-19 pulmonary consolidations detection in chest X-ray using progressive resizing and transfer learning...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonary Medicine

Background:

• COVID-19, a viral respiratory illness, presents with pulmonary consolidations crucial for diagnosis and prognosis.
• Deep learning, particularly Transfer Learning, shows promise in analyzing X-ray images for disease detection.
• Limitations in Transfer Learning can be addressed by techniques like Progressive Resizing.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for classifying pulmonary consolidation in X-ray images.
• To enhance the accuracy of COVID-19 detection by integrating Transfer Learning with Progressive Resizing.
• To improve the diagnostic capabilities for identifying normal, pneumonia, and SARS-CoV-2 related consolidations.

Main Methods:

• Utilized Convolutional Neural Networks (CNNs) with Transfer Learning and Progressive Resizing techniques.
• Employed EfficientNet and customized VGG-19 architectures for image classification.
• Incorporated GradCam for feature interpretation and visual analysis of X-ray images.

Main Results:

• The combined approach of Transfer Learning and Progressive Resizing on EfficientNet demonstrated substantial classification improvements.
• The customized VGG-19 model achieved benchmark scores across all evaluation metrics compared to the baseline.
• GradCam analysis provided insights into model predictions, aiding in score assimilation.

Conclusions:

• The proposed model effectively assists medical experts in COVID-19 identification and diagnosis through X-ray analysis.
• This approach holds significant clinical implications for remote and peripheral health centers lacking expert radiologists.
• Enhanced AI-driven analysis of pulmonary consolidations can improve diagnostic accuracy and patient outcomes.