Quantitative CT for detecting COVID‑19 pneumonia in suspected cases

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• COVID-19 pandemic significantly impacts global health and economies.
• Accurate diagnosis of COVID-19 pneumonia is crucial for patient management.
• Computed tomography (CT) is a key imaging modality for COVID-19 assessment.

Purpose of the Study:

• To evaluate the diagnostic performance of quantitative CT using various threshold segmentation techniques.
• To differentiate between COVID-19 pneumonia and other lung conditions based on CT imaging features.
• To identify optimal quantitative CT parameters for COVID-19 diagnosis.

Main Methods:

• Retrospective analysis of 47 suspected COVID-19 patients (9 confirmed, 38 excluded).
• Utilized an improved 3D convolutional neural network (VB-Net) for automated lung lesion segmentation.
• Applied eight threshold segmentation methods to quantify ground glass opacity (GGO) and consolidation.
• Employed receiver operating characteristic (ROC) curves to assess diagnostic accuracy.

Main Results:

• Volume of GGO (VOGGO) and GGO percentage in the whole lung (GGOPITWL) at -300 HU showed high diagnostic value (AUC=0.769).
• Significant differences in infection volume and consolidation metrics were observed between confirmed and excluded cases.
• Specific thresholds (-300 HU) proved effective for differentiating COVID-19 pneumonia based on GGO and consolidation.

Conclusions:

• Quantitative CT offers a valuable image quantification method for auxiliary COVID-19 diagnosis.
• This approach can assist in confirming COVID-19 pneumonia in suspected individuals.
• The study highlights the potential of AI-driven CT analysis for pandemic response.