MRI Parameters

All examinations were performed on 1.5-T MRI systems (Magnetom Aera; Siemens Healthineers). Four-dimensional flow MRI examinations were performed at a sagittal three-dimensional volume covering the right-sided heart, caval veins, main pulmonary arteries, and pulmonary veins (Fig 2, A and B). Data were acquired with prospective electrocardiography gating and during free breathing by using respiratory navigator gating. Pulse sequence parameters were as follows: spatial resolution, 2.1–2.4 × 2.1–2.6 × 3.1–4.3 mm; temporal resolution, 40.0–41.6 msec; field of view, 340–420 × 238–341 × 108–151 mm; section thickness, 108–151 mm; echo time msec/repetition time msec, 2.6–2.8/5.0–5.2; flip angle, 15°; total scan time, 12–16 minutes; and velocity sensitivity, 80 cm/sec in all three velocity-encoding directions. Analysis of global left ventricular and right ventricular function was on the basis of two-dimensional cine steady-state free precession images that were part of the MESA MRI protocol.

Whole-heart four-dimensional (4D) flow data analysis workflow. Example of, A, magnitude and three-directional velocity-encoded images, B, three-dimensional (3D) segmentation with phase-contrast MR angiography used as a guide, and retrospective plane analysis in the, C, arterial and, D, venous system. IVC = inferior vena cava, LPA = left pulmonary artery, MPA = main pulmonary artery, RA = right artery, RPA = right pulmonary artery, RV = right ventricle, SVC = superior vena cava, Vz = velocity encoding on the z plane, Vx = velocity encoding on the x plane, Vy = velocity encoding on the y plane.

Deidentified 4D flow MRI data were centrally analyzed at the Northwestern University MRI Readings Center (Chicago, Ill). Four-dimensional flow MRI preprocessing included corrections for noise, eddy currents, Maxwell terms, and velocity aliasing with in-house software programmed by using software (Matlab version R2017A; MathWorks, Natick, Mass) (21). A three-dimensional phase-contrast angiogram was derived from the 4D flow MRI data (Figs 2, A, B) and served as an anatomic guide for manual segmentation (Mimics; Materialise, Leuven, Belgium) of the superior vena cava, inferior vena cava, right atrium, right ventricle, pulmonary arteries, and pulmonary veins (Fig 2) (22).

Three-dimensional blood flow imaging and regional flow quantification were performed by using commercial software (EnSight; CEI, Apex, NC; Fig 2) (23). Time-resolved path lines were generated by using emitter planes placed in the inferior vena cava and superior vena cava, as illustrated in Figure 2. The resulting traces were color coded according to blood flow velocity. Net flow, peak velocity, retrograde flow, and retrograde fraction were calculated by using a total of 14 two-dimensional analysis planes, which were manually positioned on anatomic landmarks. Three analysis planes were placed in the superior vena cava: at the right brachiocephalic vein, below the bifurcation of the superior vena cava into right brachiocephalic and left brachiocephalic branches, and at the emptying of the superior vena cava into the right atrium. Likewise, three analysis planes were also placed on the inferior vena cava: distal to the hepatic vein branches, below the diaphragm, and above the diaphragm. Additional analysis planes were positioned on the tricuspid valve, the main pulmonary artery, and the right and left pulmonary artery branches. Four pulmonary venous analysis planes were placed proximally on their emptying points into the left atrium.

To assess interobserver and intraobserver variability, flow quantitation of the 14 analysis planes was performed by two independent readers (O.R., a radiologist with 3 years of experience, and K.S., a consultant cardiologist with 10 years of experience) and by the same reader (O.R.) on two occasions. Analysis was performed in a subgroup composed of 13 participants with COPD and eight control participants randomly selected from our study cohort, yielding a total of 294 comparisons. Readers were blinded to the clinical status of the participants.