Enhanced COVID-19 Detection from X-ray Images with Convolutional Neural Network and Transfer Learning

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Disease Diagnostics

Background:

• The global Coronavirus (COVID-19) pandemic necessitates rapid and scalable diagnostic tools.
• Early detection of COVID-19 is crucial for effective disease containment and patient management.
• Automated analysis of Chest X-ray (CXR) images offers a promising avenue for large-scale screening.

Purpose of the Study:

• To introduce a novel approach for automated COVID-19 detection using advanced Convolutional Neural Network (CNN) models.
• To evaluate the performance of six distinct CNN architectures on both binary and multi-class classification tasks using CXR images.
• To identify the most effective CNN model for accurate COVID-19 diagnosis from radiographic data.

Main Methods:

• Utilized a dataset of 15,000 Chest X-ray (CXR) images for training and validation.
• Employed six state-of-the-art CNN models: Xception, Inception-V3, ResNet50, VGG19, DenseNet201, and InceptionResNet-V2.
• Performed both binary (Normal vs. Abnormal) and multi-class (Normal, COVID-19, Pneumonia) classification.

Main Results:

• The Xception model achieved superior performance, with 98.13% accuracy in binary classification.
• In multi-class classification, the Xception model reached 87.73% accuracy.
• Other models like ResNet50 also demonstrated competitive results, validating the CNN approach.

Conclusions:

• Convolutional Neural Network models, particularly Xception, show high efficacy in automated COVID-19 detection from CXR images.
• The developed automated system has the potential to significantly aid in large-scale screening and epidemic control efforts.
• This AI-driven approach offers a scalable and efficient solution for early diagnosis of respiratory diseases like COVID-19.