DL-CRC: Deep Learning-Based Chest Radiograph Classification for COVID-19 Detection: A Novel Approach

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Chest radiographs (X-rays, CT scans) are accessible and cost-effective for lung infection detection.
• Automating radiograph analysis can expedite COVID-19 diagnosis.

Purpose of the Study:

• To propose a deep learning-based chest radiograph classification (DL-CRC) framework for automated COVID-19 detection.
• To accurately distinguish COVID-19 cases from pneumonia and normal cases using chest X-rays.
• To enhance model robustness by employing advanced data augmentation techniques.

Main Methods:

• A unique dataset was compiled from four public sources, including posteroanterior (PA) chest X-ray views.
• A data augmentation of radiograph images (DARI) algorithm, utilizing generative adversarial networks (GANs), was developed to create synthetic COVID-19 X-ray images.
• A customized convolutional neural network (CNN) model was trained using both actual and synthetic X-ray images.

Main Results:

• The DL-CRC framework achieved a COVID-19 detection accuracy of 93.94% with data augmentation.
• Without data augmentation, the accuracy was significantly lower at 54.55%.
• The customized CNN model outperformed widely adopted architectures like ResNet, Inception-ResNet v2, and DenseNet.

Conclusions:

• The proposed DL-CRC framework demonstrates high accuracy in automating COVID-19 detection from chest X-rays.
• Data augmentation is crucial for training robust models when limited COVID-19 data is available.
• This framework offers a fast, reliable complementary tool for existing COVID-19 diagnostic methods.