Study design and participants

All participants were older than 70 years and presented with at least one of the following
diagnoses: Diabetes, hypertension, prior stroke and/or heart failure. Patients with
arrhythmias during CMR were excluded from further analyses since arrhythmias make
flow measurements unreliable 4]. Patients with aortic stenosis (AS) were excluded from the main analyses due to the
difficulties in acquiring reliable flow results from the complex turbulent flow surrounding
the high velocity jet caused by the stenosis 2], 10], 11]. We performed a post-hoc exploratory flow analysis in the subgroup of patients with
AS.

To assess whether patients had mitral regurgitations (MR) all accessible SSFP images
were studied visually for a signal loss in the left atrium in front of the mitral
valve. This assessment was performed without regarding flow measurements.

Data acquisition

All scans were performed on a 1.5 Tesla Siemens Espree, Erlangen, Germany. The volumetric
measurements of the left ventricle (LV) were assessed using steady-state free precession
cine sequences (SSFP) (8 mm; no gap; 25 phases; field of view (300-360) x 360 mm adjusted
for each patient; matrix size (174-192) x 192 voxels) obtained at 7 to 10 s end-expiratory
breath-holds. Short axis images covering the entire LV were obtained. Long axis images,
including two-chamber, three-chamber and four-chamber views, were acquired for planning
of the short axis images and to aid in delineation of the ventricle. Furthermore,
for positioning of flow planes, two orthogonal views of the left ventricular outflow
tract were acquired.

Flow sequences (free-breathing through-plane phase-contrast non-navigator-gated sequences)
were obtained at the tip of the valve cusps in systole (judged from SSFP images) and
at the ST-junction. To assure perpendicularity to the aorta, the imaging plane of
the flow sequence was simultaneously viewed in two orthogonal SSFP images of the left
ventricular outflow tract. Flow measurements were done in the isocenter of the magnet.
Images were checked for aliasing and velocity encoding adjusted if needed. The following
parameters were used: 50 phases; TE?=?2.8 ms; TR?=?34.9; K-space segmentation factor
4; field of view (240-320) x 320 mm adjusted for each patient; matrix size (192-256)
x 256 voxels; pixel spacing 1.25 x 1.25 mm; slice thickness?=?5 mm; temporal resolution
was 12-26 ms. flip angle 30°; 3 averages; velocity encoding of 200 cm/s and increased
if aliasing present; acquisition time 1.14-2.26 min.

Data analysis

Dedicated software (CVI42 v. 5.1.0, Circle Cardiovascular Imaging Inc., Calgary, Canada)
was used for post-processing.

LV end-systolic and end-diastolic phases were identified and traced manually at the
endocardial border according to the LV blood pool area, excluding the papillary muscles
from the LV cavity and including the left ventricular outflow tract as part of the
LV cavity 12]. Endocardial borders were delineated using windowing, and the trabeculation was excluded
from the blood pool. Left ventricular epicardium was segmented to compare myocardial
mass in end-diastole and end-systole to avoid overestimation of end-systolic volume.

Aorta was traced semi-automated with manual correction. Background correction, consisting
of a ROI in the muscle and fat in the anterior thoracic wall, was used on all images.
For the valve level measurements the entire aortic area was included (the three leaf
clover see Fig. 2).

The stroke volumes measured with flow at valve level (SV
V
), at the ST-junction (SV
ST
) and the volumetrically measured SV from SSFP sequences (SV
ref
) were registered.

Statistics

Paired students t-tests were used for comparisons. Two comparisons were made:

The two flow measurements (SV
V
and SV
ST
) were compared to each other.

The two flow measurements (SV
V
and SV
ST
) were, in turn, compared to the volumetric measurements (SV
ref
). For this second analysis, patients with mitral regurgitations were excluded to
ensure that stroke volumes measured volumetrically would theoretically equal flow
measurements, i.e. only reflecting forward aortic flow.

Bland-Altman plots were used to visualize the difference and variability between two
measurements.

Interobserver variability was studied by separate blinded analyses of 10 randomly
selected scans by two investigators. The results were assessed with regression analyses
and Bland-Altman analyses and plots.