DeepCOVNet Model for COVID-19 Detection Using Chest X-Ray Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• The COVID-19 pandemic necessitates accurate diagnostic tools.
• Differentiating COVID-19 from other respiratory conditions like pneumonia using chest X-rays (CXR) is critical.
• Technology-enabled solutions are vital for efficient disease detection.

Purpose of the Study:

• To propose and evaluate a deep learning model for COVID-19 detection using CXR images.
• To provide a methodical approach for preparing data to train robust deep learning models.
• To compare a custom-built model against pre-trained models for COVID-19 classification.

Main Methods:

• Development of a 3-convolutional layer Deep Neural Network named "DeepCOVNet".
• Training the model on refactored datasets combining images from multiple sources.
• Classifying CXR images into three categories: COVID-19, Normal, and Pneumonia.

Main Results:

• The DeepCOVNet model achieved a classification accuracy of 96.77%.
• The model demonstrated a F1-score of 0.96 in classifying COVID-19, Normal, and Pneumonia cases.
• Effective classification of COVID-19 patients was achieved using CXR images.

Conclusions:

• Deep learning models, such as DeepCOVNet, are effective for classifying COVID-19 from CXR images.
• Data preparation is a crucial step in building robust deep learning models for medical image analysis.
• The proposed approach offers a viable technological solution for COVID-19 screening.