Classification of COVID-19 and Pneumonia Using Deep Transfer Learning

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• COVID-19, caused by SARS-CoV-2, became a global pandemic in 2019, posing a significant public health challenge.
• Pneumonia, a similar viral respiratory illness, presents a range of severity and poses risks to vulnerable populations.
• Accurate and rapid detection of these respiratory diseases is crucial for effective management and containment.

Purpose of the Study:

• To develop and evaluate a deep transfer learning model for classifying COVID-19 and pneumonia from chest X-ray (CXR) images.
• To enhance classification precision by employing customized, pre-trained deep convolutional neural network (CNN) models.
• To compare the performance of various pre-trained CNN models for disease detection accuracy.

Main Methods:

• Utilized deep transfer learning techniques with customized, pre-trained deep CNN models.
• Extracted deep features from CXR images using models including MobileNetV2, InceptionV3, EffNet, and VGG19.
• Evaluated model performance primarily based on classification accuracy.

Main Results:

• Deep transfer learning effectively detected COVID-19 and pneumonia from CXR images.
• MobileNetV2 achieved the highest classification accuracy at 98%.
• Other models showed strong performance: InceptionV3 (96.92%), EffNet (94.95%), and VGG19 (92.82%).

Conclusions:

• Deep transfer learning, particularly with customized models like MobileNetV2, offers a highly accurate method for diagnosing COVID-19 and pneumonia from CXR scans.
• The study highlights the potential of AI in improving the speed and accuracy of respiratory disease detection.
• Further research can leverage these findings for clinical decision support systems.