Quantitative Chest Computed Tomography and Machine Learning for Subphenotyping Small Airways Disease in Long COVID

Area of Science:

• Pulmonary Medicine
• Radiology
• Artificial Intelligence

Background:

• Long COVID, a condition following SARS-CoV-2 infection, can lead to persistent respiratory issues.
• Quantitative CT (QCT) and machine learning (ML) offer advanced tools for analyzing lung imaging.
• Small airway disease (SAD) is a recognized complication, but its imaging phenotypes in post-COVID patients require further investigation.

Purpose of the Study:

• To integrate quantitative CT (QCT) and machine learning (ML) for identifying imaging phenotypes in post-hospitalized COVID-19 survivors.
• To specifically investigate small airway disease (SAD) phenotypes and their correlation with pulmonary function tests (plethysmography).

Main Methods:

• A retrospective analysis of 257 adult COVID-19 survivors who underwent volumetric inspiratory and expiratory chest CT and plethysmography 6-12 months post-hospitalization.
• Quantitative CT (QCT) parameters were extracted using AI-Rad Companion Chest CT.
• Hierarchical clustering of QCT parameters was employed to identify distinct patient phenotypes.

Main Results:

• Four distinct imaging phenotypes were identified: "SAD" (14%), "intermediate" (39%), "younger fibrotic" (31%), and "older fibrotic" (16%).
• The SAD cluster exhibited higher residual volume (RV)/total lung capacity (TLC) ratios and lower FEF25-75/forced vital capacity (FVC) on plethysmography.
• The older fibrotic cluster showed the lowest TLC and FVC, while the younger fibrotic cluster had lower RV/TLC ratios and higher FEF25-75.

Conclusions:

• Integrating QCT and ML successfully identified distinct imaging phenotypes in long COVID patients.
• A unique SAD phenotype was identified, strongly associated with specific pulmonary function abnormalities.
• This approach provides a novel method for characterizing lung abnormalities in the context of long COVID.