Professional Documents
Culture Documents
4
4
DOI 10.1007/s10554-017-1160-9
ORIGINAL PAPER
Abstract Left atrial (LA) enlargement and dysfunction Intra- and inter-observer variability was assessed in ten
are markers of chronic diastolic dysfunction and an impor- patients. Significant differences were found for LA vol-
tant predictor of adverse cardiovascular and cerebrovascu- ume and phasic function. The Bi-ALM significantly under-
lar outcomes. Accordingly, accurate quantification of left estimated LA volume and overestimated LA function in
atrial volume (LAV) and function is needed. In routine comparison to Simpson’s method (Bi-ALM vs. Simpson’s
clinical cardiovascular magnetic resonance (CMR) imag- method: LAVmax: 80.18 vs. 98.80 ml; L AVpre−ac: 61.09 vs.
ing the biplane area-length method (Bi-ALM) is frequently 80.41 ml; LAVmin: 36.85 vs. 52.66 ml; L AEFTotal: 55.17 vs.
applied due to time-saving image acquisition and analysis. 47.85%; LAEFPassive: 23.96 vs. 19.15%; LAEFBooster: 40.87
However, given the varying anatomy of the LA we hypoth- vs. 35.64%). LA volumetric and functional parameters
esized that the diagnostic accuracy of the Bi-ALM is not were reproducible on an intra- and inter-observer levels for
sufficient and that results would be different from a pre- both methods. Intra-observer agreement for LA function
cise volumetric assessment of transversal multi-slice cine was better for Simpson’s method (Bi-ALM vs. Simpson’s
images using Simpson’s method. Thirty one patients of the method; ICC LAEFTotal: 0.84 vs. 0.96; ICC LAEFPassive:
FIND-AFRANDOMISED-study with status post acute cerebral 0.74 vs. 0.92; ICC L AEFBooster: 0.86 vs. 0.89). The Bi-
ischemia (mean age 70.5 ± 6.2 years) received CMR imag- ALM is based on geometric assumptions that do not reflect
ing at 3T. The study protocol included cine SSFP sequences the complex individual LA geometry. The assessment of
in standard 2- and 4 CV and a stack of contiguous slices in transversal slices covering the left atrium with Simpson’s
transversal orientation. Total, passive and active LA emp- method is feasible and might be more suitable for an accu-
tying fractions were calculated from LA maximal volume, rate quantification of LA volume and phasic function.
minimal volume and volume prior to atrial contraction.
Keywords Cardiovascular magnetic resonance imaging ·
Left atrial volume · Left atrial function · Comparison ·
* Laura Kristin Wandelt
Biplane area-length method · Simpson’s method
laura.wandelt@stud.uni‑goettingen.de
1
Institute for Diagnostic and Interventional Radiology,
University Medical Center Göttingen, Robert‑Koch Straße Introduction
40, 37075 Göttingen, Germany
2
German Center for Cardiovascular Research (DZHK), Left atrial (LA) enlargement and dysfunction repre-
Partner Site Göttingen, Robert‑Koch Straße 40,
sent important markers of chronic diastolic dysfunc-
37075 Göttingen, Germany
3
tion. Accordingly, quantitative assessment of LA vol-
Department of Cardiology and Pneumology, University
ume (LAV) and function plays an important role for risk
Medical Center Göttingen, Robert‑Koch Straße 40,
37075 Göttingen, Germany stratification and prognosis of several cardiovascular
4 and cerebrovascular diseases [1, 2]. Studies addressed
Department of Paediatric Cardiology and Intensive
Care Medicine, University Medical Center Göttingen, the association of LA enlargement and dysfunction with
Robert‑Koch Straße 40, 37075 Göttingen, Germany the development and maintenance of atrial fibrillation,
13
Vol.:(0123456789)
Int J Cardiovasc Imaging
the prediction of pulmonary vein isolation success, first CMR image acquisition
ischemic stroke, stroke recurrence and the outcome of
patients with acute myocardial infarction [2–9]. Cardio- All CMR scans were performed on a 3T MR scanner
vascular magnetic resonance (CMR) imaging allows a (Skyra, Siemens Healthineers, Erlangen, Germany) with
reliable and reproducible application in terms of quanti- a 32 channel cardiac coil in the supine position. On the
fication of LA volume and function [1]. The evaluation of basis of scout images in axial, coronal and sagittal ori-
standard steady-state free-precession (SSFP) multi-slice entations, retrospectively ECG-gated breath-hold steady-
short-axis images using Simpson’s method of disks is state free-precession (SSFP) cine images were acquired
regarded as the common standard for LA volume quan- in the 2- and 4-chamber view with the following prop-
tification using CMR [10]. However, the image acquisi- erties: repetition time 41.4 ms, echo time 1.51 ms, flip
tion as well as the analysis of multi-slices images and angle 53°, field of view 340 mm, matrix size 256 × 256,
by association the application of the Simpson’s method slice thickness 6 mm. A stack of contiguous breath-hold
is time-consuming and consequently may be of limited SSFP cine slices with 0 slice spacing was obtained in
value in daily clinical routine. Therefore, the biplane transversal orientation with full coverage of both atria
area-length method (Bi-ALM), based on the evaluation and both ventricles at end expiration. Typical imaging
of routinely performed 4- and 2-chamber views, is fre- parameters were: repetition time 48.58 ms, echo time
quently applied for LA volume and function quantifica- 1.51 ms, flip angle 59°, field of view 340 mm, matrix size
tion due to timesaving image acquisition and evaluation 256 × 256, number of slices 25, slice thickness 7 mm.
[11]. Recent studies focused on the comparability of the
results of the biplane area-length with the standard short-
axis method and basically demonstrated that the biplane- CMR image analysis
area length method provides a reliable and reproducible
but also less accurate alternative to the time-consuming Volumetric analysis was performed offline with commer-
reference standard [11–14]. An alternative approach for cially available software (QMass MR 7.6, Medis medi-
the determination of LA volume and phasic function is cal imaging systems, Leiden, The Netherlands). For both
the evaluation of transversal, instead of short-axis, multi- methods LA volumes were assessed at three time-points
slice images. Currently, there are no data available that of the cardiac cycle: LA maximal volume at left ventricu-
address the agreement of the biplane area-length method lar end-systole before mitral valve opening (LAVmax),
with the volumetric analysis of multi-slice transversal LA volume at left ventricular diastole immediately prior
images. Thus, this study aimed to analyse the correlation to LA contraction (LAVpre−ac) and LA minimal volume
of two different slice positioning methods and to compare at late left ventricular diastole after LA contraction
the results in stroke patients to findings of the literature. (LAVmin) [16, 17]. LAVmax, LAVpre−ac and L AVmin were
We hypothesized that this approach might generate dif- also normalized to body surface area (BSA) as calcu-
ferent results than the previous published comparisons to lated by the Mosteller formula [12]. For the biplane area-
the standard short-axis method. length method, semi-automated segmentation of the LA
was performed in the 2- and 4-chamber view. LA vol-
umes were automatically calculated by the software using
the following equation:
Methods
0.85 × 4CV Area × 2CV Area
LA volume (LAV) =
Lmin
Patients
Lmin corresponds to the shorter long-axis length of the LA
A total of 31 subjects (17 women, 14 men, mean age either from the 2-chamber or the 4-chamber view [16, 18].
70.5 ± 6.2 years, range 62–92 years) of the FIND- In the transversal stack, LA endocardial contours were man-
AFRANDOMISED-study with status post acute cerebral ually traced in each slice and Simpson’s method of disks was
ischemia and no history of atrial fibrillation underwent applied for volumetric analysis. This approach is based on
CMR imaging for the quantification of left atrial vol- the Simpson’s rule, where the sum of cross-sectional areas
ume and phasic function [15]. The study was performed of each slice is applied against the slice thickness and the
according the principles of the declaration of Helsinki interval between the slices [14]. It is not consistently defined
and was approved by the Institutional Review Board. All if the left atrial appendage (LAA) must be included in the
patients provided written informed consent before being measurements of left atrial volume. While some investiga-
included in the study. tors consider the LAA to be part of the LA volume [19, 20]
13
Int J Cardiovasc Imaging
other studies exclude the LAA for both methods [12, 14, ICC 0.6–0–74, fair when ICC 0.4–0.59 and weak when
21]. Because of an unreliable detection and differentiation ICC < 0.4 [24, 25].
of the left atrial appendage in the 2-chamber view as well as
in the transversal slices and therefore for a better compara-
bility, we decided to exclude the LAA from the volumetric
measurements in both methods. The pulmonary veins were Results
also excluded in both methods. The left atrial appendage and
the pulmonary veins were excluded in both methods. The CMR examination was feasible in all 31 participants
segmentation of the LA in the transversal stack took about and adequate image quality was obtained in all studies.
15 min whereas the evaluation of the 2- and 4-chamber view Baseline characteristics of the participants are shown in
was distinctly faster (approximately 5 min). Both observ- Table 1. No patient presented with chest wall deformaties,
ers performing the evaluation of the left atria had at least 2 that could have influenced the LA geometry. Exemplary
years of experience in CMR. Total, passive and active LA images of a 2- and 4-chamber view and a stack of transver-
emptying fractions (LAEFTotal LAEFPassive, LAEFBooster) sal slices of a participant are shown in Figs. 1 and 2. The
were calculated from the measured LA volumes based on values obtained for LA volumes and phasic functions are
the following equations [16, 22]: presented in Table 2. Statistically significant differences
(p < 0.05) between the two methods were found for LA vol-
ume and phasic function. The biplane area-length method
( )
LAVmax − LAVmin × 100
LAEFTotal =
LAVmax significantly underestimated LA volume and overestimated
( )
LAVmax − LAVpre−ac × 100 LA phasic function compared to the Simpson’s method in
LAEFPassive = transversal slices (Table 2). The highest deviation of the
LAVmax biplane area-length method for LA volume was observed
( )
LAVpre−ac − LAVmin × 100 for LAVmin (ml) with an underestimation of 28.2%, the
LAEFBooster =
LAVpre−ac lowest deviation was found for L AVmax (ml) with an under-
estimation of 17.9%. For LA phasic function percental
deviation of the biplane area-length method was highest for
Statistical analysis LAEFPassive (%) with an overestimation of 35% and lowest
for LAEFBooster (%) with an overestimation of 15.4% com-
Statistical analysis was performed using Microsoft Excel pared to the Simpson’s method in transversal slices. Gen-
(Microsoft Corporation, Redmond, WA, USA), Statistica der-specific results are shown in Table 3. There was a trend
version 12 (Dell Inc., Tusla, Oklahoma, USA) and IBM towards higher LAVmax, LAVpre−ac and LAVmin in men than
SPSS Statistics version 24 (IBM Corporation, Armonk, in women for both volumetric techniques.
New York, USA).
All continuous variables are presented as
mean ± standard deviation (SD) unless otherwise noted. Table 1 Patients characteristics (n = 31)
For metric variables, the assumption of a normal distribu-
Mean ± SD
tion was checked graphically by using histograms. In the
case of a skewed distribution, nonparametric tests were Age (years) 70.5 ± 6.2
used. The correspondence of the results for left atrial vol- Gender 17 Female/14 male
ume and function measurements between both methods Height (cm) 168.6 ± 9.5
was determined by Bland–Altman-analyses, Wilcoxon Weight (kg) 82.1 ± 17.1
signed-rank test and Student´s paired t-test [23]. The Body surface area (m²) 2 ± 0.2
main hypothesis was checked in a confirmatory way with Body mass indices (kg/m²) 28.9 ± 5.7
a two-sided type-one error of 5%. For the multiplicity- n = 31
correction the Bonferroni approach was used. In addition,
Hypertension 24
an exploratory evaluation of gender differences for left
Diabetes 3
atrial volume and function was performed using Student´s
Smoking 15
unpaired t-test and Mann–Whitney U test. Intra- and
Hyperlipidemia 10
inter-observer variability for both methods were assessed
Congestive heart failure 1
in ten randomly selected patients with Bland–Altman-
Coronary artery disease 1
analyses and intraclass correlation coefficients (ICC),
Myocardial infarction 2
using a model of absolute agreement. Reproducibility
was considered excellent when ICC > 0.74, good when SD standard deviation
13
Int J Cardiovasc Imaging
Fig. 1 SSFP 2- chamber
view and 4-chamber view of
a 75 year old female patient
showing a representative seg-
mentation of L AVmax, LAVpre−ac
and LAVmin
13
Int J Cardiovasc Imaging
Fig. 2 SSFP transversal stack covering the left atrium of a 75 year old female patient showing a representative segmentation of L
AVmax (Soft-
ware: QMass MR, Medis)
Within the last years, the evaluation of left atrial enlarge- fibrillation as well as patients with coronary artery disease
ment and dysfunction by CMR gained high attention. Pre- and systolic LV dysfunction [11–14]. The findings have
vious studies compared the faster biplane area-length been partly different. On the one hand, Nacif et al. showed
method with the short-axis method for volumetric analysis that in patients with atrial fibrillation the results of the
of the left atrium in healthy participants and patients with biplane area-length method for LA volume (LAVmax) did
cardiovascular diseases as for example patients with atrial not differ significantly from those obtained by Simpson’s
13
Int J Cardiovasc Imaging
LAV left atrial volume; LAEF left atrial ejection fraction, BSA body surface area
*Unadjusted descriptive p-values
LAVmax (ml) −0.66 (−2.5 to 1.18) 0.991 (0.967 to 0.998) 1.34 (−0.76 to 3.44) 0.994 (0.978 to 0.999)
LAVmax/BSA (ml/m²) −0.29 (−1.26 to 0.68) 0.994 (0.977 to 0.998) 0.73 (−0.42 to 1.89) 0.994 (0.976 to 0.999)
LAVpre−ac (ml) −0.43 (−2.14 to 1.28) 0.991 (0.968 to 0.998) 0.44 (−1.76 to 2.65) 0.990 (0.962 to 0.998)
LAVpre−ac/BSA (ml/m²) −0.17 (−1.06 to 0.71) 0.994 (0.977 to 0.999) 0.31 (−2.05 to 1.43) 0.989 (0.957 to 0.997)
LAVmin (ml) 0.31 (−1.08 to 1.7) 0.994 (0.974 to 0.998) −0.31 (−2.05 to 1.43) 0.988 (0.953 to 0.997)
LAVmin/BSA (ml/m²) 0.17 (−0.53 to 0.87) 0.994 (0.978 to 0.999) −0.06 (−0.96 to 0.84) 0.989 (0.958 to 0.997)
LAEFTotal (%) −0.73 (−2.44 to 0.99) 0.962 (0.858 to 0.99) 0.96 (−1.33 to 3.26) 0.837 (0.380 to 0.959)
LAEFPassive (%) −0.38 (−2.2 to 1.45) 0.916 (0.658 to 0.979) 0.98 (− 2.69 to 4.65) 0.741 (−0.066 to 0.936)
LAEFBooster (%) −0.57 (−3.09 to 1.94) 0.891 (0.564 to 0.973) 0.63 (−1.87 to 3.13) 0.856 (0.422 to 0.964)
(b) Inter-observer variability
Transversal slices Biplane area-length
Bias (95% CI) ICC Bias (95% CI) ICC
LAVmax (ml) −5.39 (−8.36 to − 2.42) 0.962 (0.390 to 0.993) 0.70 (− 2.00 to 3.41) 0.992 (0.971 to 0.998)
LAVmax/BSA (ml/m²) −2.71 (−4.22 to 1.20) 0.974 (0.528 to 0.995) 0.34 (− 1.06 to 1.75) 0.993 (0.971 to 0.998)
LAVpre−ac(ml) −3.52 (−6.13 to − 0.90) 0.972 (0.741 to 0.994) 0.40 (− 2.69 to 3.49) 0.981 (0.925 to 0.995)
LAVpre−ac/BSA (ml/m²) −1.72 (− 2.96 to − 0.49) 0.981 (0.800 to 0.996) 0.23 (−1.38 to 1.85) 0.98 (0.921 to 0.995)
LAVmin (ml) −0.64 (−3.51 to 2.24) 0.976 (0.905 to 0.994) 0.43 (−2.67 to 3.53) 0.966 (0.863 to 0.992)
LAVmin/BSA (ml/m²) −0.29 (−1.72 to 1.15) 0.98 (0.92 to 0.995) 0.32 (−1.24 to 1.89) 0.97 (0.883 to 0.993)
LAEFTotal (%) −2.34 (−5.46 to 0.79) 0.816 (0.317 to 0.953) −0.46 (−4.54 to 3.62) 0.667 (−0.485 to 0.919)
LAEFPassive (%) −1.08 (−3.79 to 1.64) 0.883 (0.554 to 0.971) 0.01 (−2.98 to 3.00) 0.865 (0.432 to 0.967)
LAEFBooster (%) −2.07 (−4.89 to 0.75) 0.824 (0.345 to 0.955) −0.40 (−4.37 to 3.58) 0.734 (−0.163 to 0.935)
CI confidence interval, ICC intraclass correlation coefficient, LAV left atrial volume, LAEF left atrial ejection fraction, BSA Body surface area
method in the short-axis [12]. In contrast, Sievers et al. short-axis whereas LA SV and LA EF did not differ signifi-
illustrated that the values obtained for LAVmax (LA ESV) cantly [11]. Another study in patients with coronary artery
and LAVmin (LA EDV) by the biplane area-length method disease and systolic LV dysfunction found no statistically
were significantly higher in healthy participants and significant difference between biplane area-length method
patients with atrial fibrillation than those quantified in the and short-axis concerning LAVmax (LA ESV). In contrast
13
Int J Cardiovasc Imaging
13
Int J Cardiovasc Imaging
Fig. 3 Bland–Altman-plots for the comparison of the biplane area-length method and Simpson’s method
Compliance with ethical standards Ethical approval The present study has been approved by the eth-
ics committee and has been performed in accordance with the ethical
standards laid down in the 1964 Declaration of Helsinki and its later
Conflict of interest None.
amendments. All study participants gave their informed consent prior
to their inclusion in the study.
13
Int J Cardiovasc Imaging
13