A robust semantic lung segmentation study for CNN-based COVID-19 diagnosis.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Accurate and timely diagnosis of COVID-19 is crucial for patient management and public health.
• Chest X-ray (CXR) imaging is a widely available tool for respiratory illness assessment.
• Deep learning offers potential for automated analysis of medical images.

Purpose of the Study:

• To develop and evaluate a deep learning system for diagnosing COVID-19 using CXR images.
• To segment lung regions accurately from CXR scans for further analysis.
• To classify CXR images into three categories: Normal, Viral Pneumonia, and COVID-19.

Main Methods:

• Utilized the COVID-19 Chest X-Ray Dataset for semantic lung segmentation with DeepLabV3+.
• Applied image preprocessing techniques to enhance CXR images from the COVID-19 Radiography Database.
• Employed a modified AlexNet (mAlexNet) for feature extraction and Support Vector Machine (SVM) for classification.

Main Results:

• Successfully segmented lung regions from CXR images.
• Achieved a 99.8% classification success rate for distinguishing between Normal, Viral Pneumonia, and COVID-19 cases.
• Demonstrated superior performance compared to existing state-of-the-art methods.

Conclusions:

• The proposed deep learning approach provides a highly accurate method for COVID-19 diagnosis using CXR.
• Automated analysis of CXR images can significantly aid in the rapid identification of COVID-19.
• The system shows promise for integration into clinical diagnostic workflows.