A novel framework based on deep learning for COVID-19 diagnosis from X-ray images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Coronavirus disease (COVID-19) poses a significant global health threat, necessitating rapid and accurate diagnostic methods.
• X-ray imaging remains a vital tool for COVID-19 detection in resource-limited settings, despite challenges like human error and time constraints.
• Deep learning (DL) offers promising solutions for automated COVID-19 diagnosis, addressing the critical need for accuracy in disease control.

Purpose of the Study:

• To develop a novel deep neural network framework for highly accurate, online recognition of COVID-19 from medical X-ray images.
• To improve the accuracy and efficiency of COVID-19 diagnosis, particularly in regions with limited access to advanced diagnostic kits.

Main Methods:

• A modified DenseNet-121 architecture was employed for the neural network.
• Image data was processed using a dedicated loader, with a loss function to minimize prediction errors.
• A weighted random sampler balanced the training phase, and an optimizer refined neural network attributes.

Main Results:

• The proposed framework demonstrated high diagnostic performance, achieving an accuracy of 99.81% in experiments involving various pneumonia types.
• The system proved effective in distinguishing COVID-19 cases from other forms of pneumonia.

Conclusions:

• The developed deep learning framework provides a highly accurate method for online medical image recognition.
• This system can serve as a valuable auxiliary tool for radiologists, enhancing the accuracy of initial COVID-19 screening.