Analysis of Pulsatile Vessel Expansion in Healthy, COPD- and PH-Patients Using Dynamic Vessel Segmentation in...

The pulsatile expansion of pulmonary vessels carries dynamic cardiopulmonary information that may reveal disease earlier than structural changes alone.

To test (i) intra- and inter-scan repeatability of dynamic vessel metrics in healthy subjects, and (ii) whether chronic obstructive pulmonary disease (COPD) and postcapillary pulmonary hypertension (PH) exhibit disease-specific differences.

Retrospective, multi-cohort feasibility study with repeatability sub-study.

Healthy: 29 (11 female), COPD: 52 (15 female), PH: 25 (15 female) with isolated postcapillary PH.

1.5T, free-breathing spoiled gradient-echo sequence, TE = 0.82 ms, TR = 3 ms, flip angle = 5°.

2D U-Nets segmented pulmonary vessels throughout each series of 250 images. Dynamic metrics included the coefficient of variation (CV) of vessel area and CV of vessel signal. Delay between vessel signal and vessel expansion as %-of-RR-interval (SE-delay) together with the heart rate and lung area were calculated.

Repeatability: Friedman; intraclass correlation coefficient (ICC); standard error of measurement (SEM) and the minimal detectable change (MDC). Group comparisons: Kruskal-Wallis and multiple linear regression. p < 0.05 was considered significant.

Only the heart rate and the SE-delay presented significant changes across repetitive scans. Differences in Bland-Altman plots were evenly distributed across the range of measurements and symmetrically scattered around zero within 95% confidence intervals (mean bias 0.08%-5.12%). Across cohorts, CV vessel area and CV lung area differed significantly, showing lowest variability in COPD (2.67%; 7.03%) and highest in PH (4.58%; 16.74%). SE-delay was significantly prolonged in COPD (81.98%) compared to PH (67.20%) and healthy participants (62.97%). Cohort status (Healthy/COPD/PH) remained the strongest significant predictor for all parameters after adjustment (F-values 4.62-34.04), while age and gender had no significant influence (p > 0.391; p > 0.069).

Free-breathing lung MRI with automated vessel segmentation reveals distinct hemodynamic characteristics in COPD and PH. This method shows potential as a sensitive non-invasive tool for detection, phenotyping, and treatment monitoring in pulmonary pathology.

3.

2.