Patient selection
This multicenter retrospective cohort study included all patients who underwent CMR
for evaluation of pediatric or congenital heart disease on an open scanner at Stony
Brook University or the University of Michigan C.S. Mott Children’s Hospital, between
2/8/2012 and 7/24/2014. This study was approved by the Institutional Review Boards
at both centers, and requirement for informed consent in this retrospective study
was waived. Patients of any age with congenital heart disease were included; patients
without congenital heart disease were included if they were 18 years or younger at
the time of CMR. Patients were excluded if digital images were not available for review.
Medical records were reviewed for patient demographics, indication for the study,
and reason for performing the study on the open scanner.
For quantitative comparison of signal-to-noise ratio (SNR) and contrast-to-noise ratio
(CNR), a subset of 25 CMR studies were matched 1:1 by patient body surface area and
diagnosis with CMR studies performed on a 1.5 Tesla scanner at the same institution.
It should be noted that SNR and CNR depend on a number of imaging parameters, including
voxel size, acceleration factor, readout bandwidth, and number of averages. Therefore,
imaging parameters on both scanners are summarized in Table 1.
Table 1. Typical imaging parameters at 1.0 Tesla and 1.5 Tesla
CMR imaging
CMR was performed with a Panorama High Field Open 1.0 Tesla scanner (Philips, Best,
The Netherlands), using a solenoid body coil (medium, large or extra large as appropriate
for patient size). Gradient amplitude is 28 mT/m; maximum slew rate is 120 mT/m/ms.
Sensitivity encoding (SENSE) was not used for image acquisition, i.e. SENSE acceleration
factor?=?1, to maintain adequate signal-to-noise ratio (SNR). Typical imaging parameters
are presented in Table 1, with some minor differences by center. Cine imaging was performed with a breathhold,
electrocardiographic gated, segmented k-space balanced steady-state with free precession
(SSFP) sequence. Black blood imaging was performed with a breathhold, electrocardiographic
gated, double inversion recovery turbo spin echo sequence. Imaging parameters varied
by weighting of the sequence. Phase contrast imaging was performed free-breathing.
Magnetic resonance angiogram (MRA) was performed with a T1-weighted fast field echo
sequence, after injection of 0.2 mmol/kg of gadoteridol (ProHance, Bracco, Monroe
Township, New Jersey) for studies performed at the University of Michigan, or gadopentetate
dimeglumine (Magnevist, Bayer, Leverkusen, Germany) for studies performed at Stony
Brook University. Two post-contrast dynamics were performed to highlight the anatomy
of interest (e.g. pulmonary arteries or aorta) as well as the venous phase. Coronary
artery imaging was performed with a respiratory navigator gated, electrocardiographic
gated, three-dimensional SSFP sequence. Late gadolinium enhancement (LGE) was performed
12–15 minutes after contrast injection, using a breathhold, electrocardiographic gated,
phase-sensitive inversion recovery sequence. In 12 patients at the University of Michigan,
LGE imaging was performed with a three-dimensional, respiratory navigator gated, electrocardiographic
gated, phase-sensitive inversion recovery sequence, to improve in-plane spatial resolution
while maintaining adequate SNR.
CMR at 1.5 Tesla was performed on a commercially available scanner (Ingenia, Philips,
Best, The Netherlands), using a phased-array body surface coil. Gradient amplitude
is 33 mT/m; maximum slew rate is 200 mT/m/ms. Typical imaging parameters are presented
in Table 1. It should be noted that TR was longer on the 1.0 Tesla scanner than the 1.5 Tesla
scanner due to the lower gradient performance characteristics (maximum gradient strength
and slew rate) of the 1.0 Tesla scanner. Phase contrast imaging was performed free-breathing.
MRA was performed with two post-contrast dynamics, which was sufficient to highlight
the anatomy of interest, e.g. pulmonary arteries or aorta, as well as the venous phase
of the contrast uptake. LGE was performed using a breathhold, electrocardiographic
gated, phase-sensitive inversion recovery sequence.
Scan time for both scanners was measured from the start of the first imaging sequence
through the final imaging sequence. Thus, this included patient rest time between
breath-holds, and potential need to pause during imaging if there was any patient
discomfort.
Image analysis
All images were evaluated by a single experienced observer. For each sequence type,
image quality was scored qualitatively, taking into consideration resolution, blurring
effects, low SNR, delineation of structures, and artifacts, according to a 4-point
scale (4 – excellent with no artifacts; 3 – good with minor artifacts; 2 – below average,
with significant artifacts affecting interpretation; 1 – poor, nondiagnostic) 8]. Imaging examples are presented in Fig. 1.
Fig. 1. Images were scored on a 4-point scale for image quality and diagnostic utility. (a) SSFP imaging in short-axis in a 26 year-old patient with tetralogy of Fallot and
complete atrioventricular septal defect status post repair: No artifacts and good
endocardial definition, scored 4 for both quality and diagnostic utility. (b) SSFP imaging in the four-chamber plane in a 32 year-old patient with dextrocardia
and congenitally corrected transposition of the great arteries, who previously failed
CMR on a 1.5 Tesla scanner: Flow-related artifact (arrows), which did not affect interpretation
of ventricular size or function, scored 3 for quality, but 4 for diagnostic utility.
(c) Volume-rendered reconstruction of gadolinium-enhanced MRA in a 57 year-old patient
with pulmonary hypertension and reported history of atrial septal defect repair in
a foreign country. The left upper pulmonary vein (white arrowhead) drains into the
left innominate vein; the right upper and middle veins drain into a baffle within
the superior vena cava (black arrowhead) to the left atrium. Scored 4 for both quality
and diagnostic utility. (d) SSFP imaging in short axis in a 16 year-old patient with unbalanced atrioventricular
septal defect status post Fontan: Significant coil artifact, obscuring portions of
the heart, scored 2 for both quality and diagnostic utility (for right ventricular
size and function)
Clinical questions were extracted from the medical record, and were categorized as:
ventricular size/function, pulmonary artery anatomy, regurgitant fraction, LGE, aortic
root dimensions, coronary artery anatomy, aortic arch anatomy, ratio of pulmonary
to systemic blood flow (Qp:Qs ratio), and pulmonary venous anatomy. A given study
could have multiple clinical questions. Studies were scored for diagnostic utility
on a 4-point scale, based on the interpreter’s confidence in their ability to answer
the specific clinical questions (4 – high confidence of diagnosis; 3 – answers question
adequately; 2 – low confidence; 1 – could not answer the clinical question). Ventricular
size/function was scored on the ability to define and contour the endocardial border.
Anatomy of the pulmonary arteries, coronary arteries, aortic arch, and pulmonary veins
was scored on definition of the respective structures, course, and presence/absence
of stenosis as appropriate. Flow measurements were scored by internal consistency
of data. LGE images were scored by ability to determine presence or absence of LGE,
and extent if present.
Due to the qualitative nature of grading image quality and diagnostic utility, a 20%
subset of the cohort was randomly selected to evaluate reproducibility. A second reader,
blinded to the scores of the first reader, re-evaluated the images to evaluate interobserver
agreement. The initial reader, blinded to initial scores, also re-evaluated the images
to evaluate intraobserver agreement.
SNR and CNR were calculated for cine SSFP, black blood, MRA, 3-dimensional SSFP, and
LGE images. A region of interest (ROI) of approximately 1.0 cm2 was drawn in two locations (Fig. 2). For cine SSFP and LGE images, a midventricular short-axis slice was used, and ROIs
were drawn in the blood pool and in the interventricular septum. For black blood and
3-dimensional SSFP images, ROIs were drawn in the blood pool and in any adjacent myocardium.
For the MRA, the first post-contrast dynamic was used, and ROIs were drawn in the
aorta or main pulmonary artery (depending on timing of the contrast) and in the lung
field. SNR was defined as the mean signal intensity in the blood pool (for cine SSFP,
MRA, 3-dimensional SSFP and LGE images) or myocardium (for black blood images) divided
by the standard deviation of signal intensity in that ROI 9]. CNR was defined as the difference in signal intensities of the two ROIs, divided
by the average of the standard deviations of the two ROIs 9].
Fig. 2. Regions of interest for contrast to noise measurement. For cine images, an ROI of
~1.0 cm2 was drawn in the blood pool and in the septum. Only ROI1 (the blood pool) was used
for signal to noise ratio measurement
Statistical analysis
Data are presented as median (interquartile range [IQR]), mean?±?standard deviation,
or number (percent), as appropriate. SNR, CNR, and scan time on 1.0 Tesla and 1.5
Tesla studies were compared for each sequence using t-test. Age and body surface area
among patients on 1.0 Tesla and 1.5 Tesla patients were not normally distributed,
and were compared using Wilcoxon matched-pairs signed rank test. A p-value??0.05
was considered statistically significant. Intraobserver and interobserver agreement
was calculated using weighted kappa, to take into consideration closeness of agreement.