COVID-19 detection using chest X-ray images based on a developed deep neural network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• The COVID-19 pandemic presents a significant global health challenge, necessitating rapid and accurate diagnostic tools.
• Timely diagnosis of respiratory infections, including COVID-19, bacterial pneumonia, and viral pneumonia, is crucial for effective patient management and disease control.

Purpose of the Study:

• To develop and evaluate a deep learning-based method for classifying chest X-ray images into distinct categories: Healthy, Bacterial pneumonia, Viral pneumonia, and COVID-19.
• To assess the proposed model's performance and robustness in differentiating between various lung infection scenarios.

Main Methods:

• Utilized six diverse chest X-ray image databases, augmented with white Gaussian noise to simulate real-world conditions.
• Designed and implemented a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model for hierarchical feature extraction from chest X-ray data.
• Tested the model on the initial six databases and two additional independent datasets.

Main Results:

• The CNN-LSTM model demonstrated high accuracy, exceeding 90% for most classification scenarios, including differentiating COVID-19 from healthy and viral cases.
• Achieved a 99% accuracy rate specifically for distinguishing COVID-19 from healthy individuals.
• The model proved robust against noise up to 1 dB and maintained over 90% accuracy on additional test databases.

Conclusions:

• The proposed deep learning network is effective for the detection of COVID-19 and other infectious lung diseases via chest X-ray analysis.
• This approach shows promise in assisting radiologists by enabling faster and more accurate diagnoses, potentially improving patient outcomes.