A deep learning model for mass screening of COVID-19

Area of Science:

• Artificial Intelligence
• Medical Imaging Analysis
• Deep Learning

Background:

• Accurate and rapid diagnosis of respiratory illnesses like COVID-19 is crucial during pandemics.
• Chest X-rays are a common diagnostic tool, but interpretation can be challenging and time-consuming.
• Automated analysis of medical images can potentially improve diagnostic speed and accuracy.

Purpose of the Study:

• To develop a convolutional neural network (CNN) model named 'COVID-Screen-Net' for multi-class classification of chest X-ray images.
• To distinguish between COVID-19, bacterial pneumonia, and normal cases using automated feature extraction.
• To optimize the CNN model for efficiency and accuracy in classifying respiratory conditions.

Main Methods:

• Development of a custom CNN architecture ('COVID-Screen-Net') with optimized convolution and activation layers.
• Automatic feature extraction from chest X-ray images, with visualization using GradCam.
• Hyperparameter tuning to minimize computation time and enhance model efficiency.
• Model evaluation on diverse datasets, including hospital-acquired and web-available images.

Main Results:

• Achieved an average accuracy of 97.71% and a maximum recall of 100% in classifying chest X-ray images.
• Demonstrated superior performance compared to existing systems for COVID-19 screening.
• Validated effectiveness through real-time dataset analysis by radiology experts.

Conclusions:

• 'COVID-Screen-Net' shows significant potential as a tool for rapid, low-cost mass screening of COVID-19.
• The model can assist healthcare professionals by reducing the burden of interpretation during global health crises.
• The developed AI tool offers a promising approach to enhance diagnostic capabilities in medical imaging.