Quadruple Augmented Pyramid Network for Multi-class COVID-19 Segmentation via CT.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Coronavirus disease (COVID-19) is a highly infectious respiratory illness.
• Chest computed tomography (CT) is crucial for diagnosing, prognosing, and monitoring COVID-19 complications.
• Accurate segmentation of affected lung regions in CT is vital for quantitative assessment.

Purpose of the Study:

• To develop and evaluate a novel multi-class segmentation framework for COVID-19 related abnormalities on chest CT scans.
• To assist radiologists in precisely estimating the volume of lung tissue affected by COVID-19.
• To improve the efficiency and accuracy of COVID-19 diagnosis and management through automated image analysis.

Main Methods:

• Implementation of a Quadruple Augmented Pyramid Network (QAP-Net) within an encoder-decoder segmentation architecture.
• Utilizing augmented pyramid networks to capture features from varying CT image sizes and facilitate semantic segmentation.
• Leveraging CNNs for spatial inter-connections and down-sampling to ensure sufficient feature transfer.

Main Results:

• The proposed QAP-Net achieved a competitive Dice coefficient of 0.8163 for COVID-19 CT segmentation.
• The method demonstrated superior performance compared to existing state-of-the-art segmentation techniques.
• The framework effectively segmented consolidation and ground-glass opacities in COVID-19 chest CT images.

Conclusions:

• The QAP-Net framework provides an efficient and accurate solution for segmenting COVID-19 lung abnormalities on chest CT.
• This AI-driven approach can significantly aid radiologists in the clinical assessment of COVID-19 severity.
• The proposed method shows promise for enhancing the diagnostic workflow and patient management for COVID-19.