Computed Tomography Image Processing Analysis in COVID-19 Patient Follow-Up Assessment.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Pulmonology

Background:

• COVID-19 pandemic necessitates rapid diagnostic tools.
• Chest CT scans are crucial for COVID-19 diagnosis and monitoring.
• Automated analysis of CT scans can expedite diagnosis and treatment.

Purpose of the Study:

• To develop an automated system for identifying ground-glass opacity (GGO) and pulmonary infiltrates (PIs) in COVID-19 CT images.
• To assess disease progression using CT image analysis.
• To complement existing diagnostic methods like RT-PCR.

Main Methods:

• Utilized oversegmentation mean shift and superpixel-SLIC for pulmonary parenchyma segmentation.
• Employed Tree Random Forest (TRF) classification based on superpixel features (position, intensity, texture, saliency).
• Applied watershed segmentation to identify GGO and PIs within segmented pulmonary parenchyma using detailed cluster features.

Main Results:

• Achieved >92% precision and recall for pulmonary parenchyma identification in COVID-19 CT scans (twofold cross-validation).
• Demonstrated 96% precision and 96% recall for GGO and PI identification (twofold cross-validation).
• The proposed methodology effectively segments and classifies key indicators of COVID-19 lung involvement.

Conclusions:

• The developed automated system accurately identifies GGO and PIs in COVID-19 CT scans.
• This approach can significantly aid in early diagnosis, disease monitoring, and patient management.
• High accuracy suggests potential for integration into clinical workflows for COVID-19 assessment.