CoroNet

CoroNet identifies COVID-19 infections from chest X-ray images using a deep convolutional neural network based on the Xception architecture.

Key Features:

• Architecture: Built on the Xception deep convolutional neural network pre-trained on ImageNet for robust feature extraction from medical images.
• Training and testing data: Trained and tested on a curated dataset comprising COVID-19 and other chest pneumonia X-ray images sourced from two publicly available databases, enabling differentiation of bacterial, viral, and normal cases.
• Classification performance: For 4-class classification (COVID vs. Pneumonia Bacterial vs. Pneumonia Viral vs. Normal) overall accuracy is 89.6%, COVID precision is 93%, and COVID recall is 98.2%; for 3-class classification (COVID vs. Pneumonia vs. Normal) accuracy is 95%.
• Pre-processing: Minimizes the need for extensive pre-processing of input chest X-ray images.
• Scalability: Performance can be further improved with additional training data.

Scientific Applications:

• Clinical screening: Assists radiologists and clinicians in rapid identification of COVID-19 cases from chest X-rays.
• Differential diagnosis: Supports distinguishing COVID-19 from bacterial and viral pneumonia and from normal chest X-rays for diagnosis and follow-up.

Methodology:

Leverages a pre-trained Xception deep convolutional neural network within a deep learning framework and minimizes extensive pre-processing of chest X-ray images.