COVID-19 classification using chest X-ray images based on fusion-assisted deep Bayesian optimization and Grad-CAM...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• COVID-19 has severe global health and economic impacts.
• Chest X-rays (CXRs) are vital for COVID-19 detection, but manual analysis is challenging.
• Deep convolutional neural networks (DCNNs) show promise for automated medical image analysis.

Purpose of the Study:

• To propose a Bayesian optimized DCNN and explainable AI framework for COVID-19 classification from CXRs.
• To enhance the accuracy and efficiency of COVID-19 detection using automated image analysis.
• To provide visual explanations for AI-driven diagnoses.

Main Methods:

• A multi-filter contrast enhancement technique was applied to CXRs.
• Two pre-trained DCNN models (EfficientNet-B0, MobileNet-V2) were fine-tuned and trained using Bayesian optimization (BO).
• Features were extracted, fused using a slicing-based serial approach, and classified with machine learning classifiers. Grad-CAM was used for visualization.

Main Results:

• The proposed framework achieved high accuracies of 98.8%, 97.9%, and 99.4% on three public COVID-19 datasets.
• Bayesian optimization improved hyperparameter selection for DCNNs.
• Grad-CAM effectively highlighted infected regions in CXRs.

Conclusions:

• The developed AI framework demonstrates high efficacy in classifying COVID-19 from chest X-rays.
• The integration of Bayesian optimization and explainable AI enhances diagnostic capabilities.
• This approach offers a valuable tool for medical professionals in combating the COVID-19 pandemic.