Reproducibility of automated and manual determination of the pulmonary artery to aorta ratio (PA:A) of ‘real-world’ CT scans

The PA:A as a marker of acute exacerbations of COPD (AECOPD) was validated using research CT scans. Automation and validation using ‘real-world’ community-acquired CTs would facilitate uptake into clinical practice. We aimed to validate the reproducibility of the PA:A in a community practice-based study with raters (R) of varying levels of expertise and a semi-automated software (SlicerCIP, chestimagingplatform.org) against a thoracic radiologist (gold standard). Randomly selected CTs from a community practice database were anonymized. Two medical students (R1, R2) and a gynecologist (R3) performed between 4 and 3 consecutive randomized, blinded measures of the PA:A (trials). A general and a thoracic radiologist each performed 1 trial. R2 performed a Slicer-assisted trial. 33 CTs (18 contrast, 15 non-contrast) from 6 subjects (4 COPD, 1 asthma, 1 PH, 1M:5F, age 67 mean years(±16SD), PA:A ratio 0.86(±0.11)). Results are shown in Table1. <fig><object-id>erj;64/suppl_68/PA1657/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig> The PA:A of ‘real-world’ CTs exhibit excellent intra- and inter-rater agreement and reproducibility between radiologists matching those of research CTs. With experience, clinicians obtain similar results. While Slicer-assisted performance is less than expert radiologist and trained clinician, it is better than untrained clinician. Whether the PA:A of ‘real-world’ CTs will be equally predictive for AECOPD as that of research CT scans remains to be determined.