Pneumonia Classification Using Deep Learning from Chest X-ray Images During COVID-19.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Reverse-transcription polymerase chain reaction (RT-PCR) has limitations including cost, time, and expertise requirements.
• Chest X-ray (CXR) imaging offers a faster, more accessible alternative for disease screening.

Purpose of the Study:

• To develop and evaluate a deep learning model for classifying CXR images.
• To differentiate between COVID-19, bacterial pneumonia, non-COVID-19 viral pneumonia, and normal scans.
• To assess the model's performance in two-way, three-way, and four-way classification tasks.

Main Methods:

• Utilized a pretrained AlexNet deep learning model.
• Trained the model on CXR images from various public databases.
• Evaluated performance across multiple classification scenarios (e.g., COVID-19 vs. normal, multi-class).

Main Results:

• Achieved high accuracy in two-way classifications, including 99.16% for COVID-19 vs. normal scans.
• Demonstrated strong performance in four-way classification with 93.42% accuracy, 89.18% sensitivity, and 98.92% specificity.
• The model showed promising results for differentiating various pneumonia types from normal CXR scans.

Conclusions:

• Deep learning, specifically the AlexNet model, is effective for classifying CXR images for COVID-19 and other respiratory conditions.
• The proposed model provides a rapid, accurate, and accessible tool for aiding in the diagnosis of respiratory diseases.
• This approach can supplement existing diagnostic methods, improving patient outcomes during outbreaks.