Pulmonary Diffuse Airspace Opacities Diagnosis from Chest X-Ray Images Using Deep Convolutional Neural Networks...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Accurate COVID-19 diagnosis from chest X-rays is challenging.
• Deep Learning (DL) models, particularly Deep Convolutional Neural Networks (DCNNs), show promise for automated detection.
• Conventional DCNN training methods (Gradient Descent-based Training) are sequential and require extensive parameter tuning, hindering parallel processing.

Purpose of the Study:

• To develop a real-time, parallelizable COVID-19 detection model for chest X-ray analysis.
• To improve the efficiency and accuracy of COVID-19 diagnosis using DL.
• To address the limitations of traditional Gradient Descent-based Training in DCNNs.

Main Methods:

• Proposed a novel DL model utilizing the Whale Optimization Algorithm (WOA) for training fully connected layers of DCNNs.
• Benchmarked the WOA-trained DCNN on the COVID-Xray-5k dataset.
• Conducted comparative analysis against classic DCNN, DUICM, and Matched Subspace classifier with Adaptive Dictionaries.

Main Results:

• The proposed WOA-based DCNN achieved an average accuracy of 99.06%.
• Demonstrated a 1.87% performance improvement over the best-performing comparison model.
• Class Activation Maps (CAM) were used to identify potentially infected regions, correlating with expert clinical findings.

Conclusions:

• The WOA-based DL model offers a highly accurate and efficient approach for COVID-19 detection from chest X-rays.
• The method's parallel implementation capability makes it suitable for real-time applications.
• Further validation on larger datasets is recommended for comprehensive evaluation.