COVID-19: a new deep learning computer-aided model for classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Infectious Disease Diagnostics

Background:

• Chest X-ray (CXR) imaging is a crucial, feasible diagnostic tool for early detection of COVID-19.
• The COVID-19 pandemic, caused by a novel coronavirus, necessitates rapid and accurate diagnostic methods.
• Accurate classification of COVID-19 infection from CXR scans is vital for patient management and disease control.

Purpose of the Study:

• To propose a novel, lightweight model named Chest X-Ray COVID Network (CXRVN) for analyzing grayscale CXR images.
• To evaluate the CXRVN model's performance against pre-trained models using multiple COVID-19 datasets.
• To assess the model's efficiency in terms of memory usage, processing time, and classification accuracy.

Main Methods:

• Developed the CXRVN, a lightweight neural network architecture featuring a single fully connected layer.
• Trained and evaluated CXRVN on three distinct COVID-19 CXR datasets, utilizing mini-batch gradient descent and Adam optimizers.
• Employed fine-tuning and transfer learning techniques to compare CXRVN with pre-trained models (GoogleNet, ResNet, AlexNet).

Main Results:

• The CXRVN model demonstrated high accuracy, achieving 96.7% on Dataset-2 and 93.07% on Dataset-3 after GAN augmentation.
• The average accuracy of the CXRVN model across experiments was 94.5%.
• CXRVN exhibited reduced memory usage and processing time compared to pre-trained models, analyzing images in milliseconds.

Conclusions:

• The proposed CXRVN model offers an efficient and accurate solution for COVID-19 detection using grayscale CXR images.
• Its lightweight architecture makes it suitable for rapid analysis, contributing to early diagnosis and pandemic control.
• CXRVN's performance, validated by standard metrics, highlights its potential as a valuable tool in medical diagnostics.