Int J Cardiovasc Imaging (2014) 30:1603–1612
DOI 10.1007/s10554-014-0499-4
ORIGINAL PAPER
Impact of a vendor-specific motion-correction algorithm on image
quality, interpretability, and diagnostic performance of daily
routine coronary CT angiography: influence of heart rate
on the effect of motion-correction
Heon Lee • Jeong A. Kim • Ji Sung Lee •
Jon Suh • Sang Hyun Paik • Jai Soung Park
Received: 3 January 2014 / Accepted: 12 July 2014 / Published online: 20 July 2014
Ó Springer Science+Business Media Dordrecht 2014
Abstract To investigate the impact of a vendor-specific
motion-correction algorithm on morphological assessment
of coronary arteries using coronary CT angiography
(cCTA) and to evaluate the influence of heart rate (HR) on
the motion-correction effect of this algorithm. Eighty-four
patients (mean age 56.3 ± 11.4 years; 53 males) were
divided into two groups with a HR of C65 and \65 bpm
during cCTA, respectively. Images were assigned quality
scores (graded 1–4) on coronary segments. Interpretability
was defined as a grade [1. Catheter angiography was used
to determine the diagnostic accuracy of cCTA for detecting
significant stenosis (C50 %). We compared the image
quality, interpretability and diagnostic accuracy between
the standard and motion-correction reconstructions in both
groups. The motion-correction reconstructions showed
significantly (p \ 0.05) higher image quality in the proxi-
mal and middle right coronary artery (RCA) in the low HR
group (57.2 ± 5.0 bpm; n = 51) and proximal-to-distal
RCA, posterior descending artery, and proximal and distal
H. Lee (&)
S. H. Paik
J. S. Park
Department of Radiology, Soonchunhyang University Hospital
Bucheon, 1174 Jung-Dong, Wonmi-Gu, Bucheon,
Gyeonggi-Do 420-767, Korea
e-mail: acarad@naver.com
J. A. Kim
Department of Radiology, Ilsan Paik Hospital, Ilsan 411-706,
Republic of Korea
J. S. Lee
Department of Biostatistics, Soonchunhyang University College
of Medicine, Seoul 140-743, Republic of Korea
J. Suh
Department of Cardiology, Soonchunhyang University Hospital
Bucheon, Bucheon 420-767, Republic of Korea
left circumflex artery in the high HR group (71.1 ±
4.6 bpm; n = 33). The per-segment interpretability was
significantly higher using motion-correction algorithm in
the middle RCA in the low HR group and in the proximal
and middle RCA in high HR group. Overall, the image
quality and interpretability were improved using motion-
correction reconstructions in both groups (p \ 0.05).
Motion-correction reconstruction demonstrated higher
(p \ 0.05) diagnostic accuracy in 25 patients from both
groups. Use of the motion-correction algorithm improves
the overall image quality and interpretability of cCTA in
both groups. However, it may be more beneficial to the
patients with a higher HR.
Keywords Computed tomography
Coronary CT
angiography
Image reconstruction
Diagnostic accuracy

Technical advancements
Introduction
Although coronary CT angiography (cCTA) has become
accepted as a powerful noninvasive diagnostic tool for the
evaluation of coronary artery disease in selected patients,
the limited temporal resolution of most of the currently
used multi-detector row CT (MDCT) scanners may cause
motion artifacts, which limits coronary artery evaluation
[1–5]. Motion artifacts and image quality degradations are
more pronounced at higher heart rates (HRs) because of the
shortening of the end-systolic and mid-diastolic imaging
window [6–8]. This often requires the use of rate-control-
ling premedications with beta-blockers for optimization of
coronary CTA acquisition although some patients cannot
achieve the target HR [9, 10]. On the other hand, several
technological solutions have been introduced to increase
123
1604 Int J Cardiovasc Imaging (2014) 30:1603–1612

temporal resolution and reduce motion artifacts, including with heart rate over 65 beats/min (bpm). The patients were
software- or hardware-based solution such as multi-seg- divided into two groups to compare the effect of the
ment reconstruction and dual-source CT, respectively motion-correction algorithm according to the HR during
[6, 11–13]. the CT scan. One group included patients with average HR
Recently, a new vendor-specific, software-based solu- C65 bpm (high HR group) and the other group consisted of
tion for motion correction [SnapShot Freeze (SSF); GE patients with average HR \65 bpm (low HR group) during
Healthcare, Waukasha, WI, USA] was released. In contrast CT scan. The study was approved by the institutional
to multi-segment reconstruction, which combines scan data review board.
from the same period of subsequent heart cycles [14], this
algorithm uses scan data from adjacent cardiac phases
cCTA protocol
within a single cardiac cycle to correct for the beating
motion of the coronary arteries. Promisingly, this algorithm
cCTA was performed on a 64-slice MDCT scanner CT
has been reported to result in significant improvements in
system (Discovery CT 750 HD scanner; GE Healthcare,
image quality and diagnostic accuracy in patients with a
Milwaukee, WI, USA) with using 50 % adaptive statistical
high heart rate during cCTA performed without use of
iterative reconstruction. A standardized scanning protocol
beta-blockers [15]. In their study, however, the patient
with a collimation of 64 9 0.625 mm, gantry rotation time
population was relatively small, and the CT scans were
of 35 ms and a pitch of 0.2 was used. Scans were acquired
primarily aimed at evaluating non-coronary cardiac struc-
in a cranio-caudal direction with simultaneous recording of
ture. Accordingly, despite their positive results, the use of
the patient’s ECG signal to allow for retrospective regis-
rate-controlling agents may not be abandoned in daily
tration of image reconstruction to the desired cardiac
routine CT practice for coronary evaluation even using
phase. The scan range extended from the level of the carina
motion-correction algorithm because the single-source CT
to just below the dome of the diaphragm. Tube voltage was
scanner available for this algorithm has an inherent limi-
100–120 kVp, and the tube current ranged between 300
tation of temporal resolution, and the effectiveness of this
and 750 mA. ECG-dependent tube current modulation was
strategy has not yet been thoroughly documented. Fur-
used in the current study with tube current reduced to 20 %
thermore, we believed that motion correction may be
of maximum in the systolic phase. Pharmacological HR
useful even in patients with a lower HR because it has been
control was used when patient’s HR was over 65 bpm.
proven that still considerable motion artifacts are noted on
Immediately before the scan, with the patient positioned on
coronary segments in the patients with a lower HR [11],
the table, 0.4 mg sublingual nitroglycerin is administered
although reducing the average HR has been reported to be
to all the patients. The scan delay time was determined by
beneficial for reducing artifacts [3, 16]. In addition, we
test-bolus injection of 20 ml of a non-ionic contrast med-
expected motion-correction algorithm to be also useful in
ium (Xenetix; 350 mgI/ml, Guerbet, France) at 5 ml/s
the patients using beta-blockers because, in some patients,
through an 18G intravenous antecubital catheter, followed
the target HR cannot be achieved despite the use of rate-
by 30 ml of saline using a dual-syringe injector (Opti-
controlling agents.
VantageTM DH; Covidien, Mansfield, MA, USA). The
Thus, we performed the present study to assess the
peak time of test bolus enhancement as measured by
effect of this vendor-specific, motion-correction algorithm
repetitive scanning at the level of the aortic root was used
on image quality, interpretability, and diagnostic accuracy
as the delay time. The contrast medium volume in milli-
among patients undergoing daily routine cCTA that
liters for actual contrast medium enhancement was deter-
includes the use of rate-controlling agent if necessary, and
mined by multiplying the scan duration by the injection
then to compare the motion-correction effect of this algo-
rate (i.e. 5 ml/s). If the scan duration was \10 s, a mini-
rithm on morphological assessment of coronary arteries in
mum of 50 ml of contrast medium was used. Radiation
the patients with low and high heart rate.
dose for coronary CTA was determined by the dose length
product, which was converted to millisievert (mSv) by
multiplying by the conversion factor of 0.014 [17].
Materials and methods

Patients Image post processing and interpretation

We retrospectively enrolled 84 consecutive patients who Data sets were reconstructed at 50 % adaptive statistical
underwent cCTA as part of a routine clinical evaluation for iterative reconstruction (ASIR; GE Healthcare, Milwaukee,
possible coronary artery disease. Rate-controlling agent WI, USA) in the current study. For standard (STD)
(propranolol 50–100 mg) was administered to the patients reconstruction, reconstructions were generated at 20

123
Int J Cardiovasc Imaging (2014) 30:1603–1612 1605

different RR-positions in 5 % increments (0–95 %). Image
reconstruction parameters comprised an individually
adapted field of view encompassing the heart, a matrix size
of 512 9 512 pixels and a section thickness of 0.625 mm
with an increment of 0.625 mm. To generate motion-cor-
rection reconstruction, raw cardiac CT data of the single
best phase for entire coronary arteries were determined and
processed off-line with the use of an advanced coronary
motion-correction algorithm. The best phase was defined as
a phase where the coronary motion was at minimum
through the entire length of each vessel among 20 different
phases. All 168 (single-phase images with both STD or
motion-correction reconstructions) datasets were assigned
to be interpreted. All image sets were anonymized and
presented randomly. An experienced observer then visually
interpreted the two reconstructions (both STD and motion-
correction algorithm) of the study. A few weeks later,
second and third interpretations of the reconstructions were
performed by the same observer to determine intraobserver
variability. Additionally, a second experienced reader
independently interpreted the same studies to determine
interobserver variability. CT image analysis was performed
using semi-automated post-processing software (Aquarius;
TeraRecon, San Mateo, CA, USA). The use of axial
datasets, curved multiplanar reformats, and volume-ren-
dered reconstructions were at the discretion of each reader.
Considering the complexity in analysis due to coronary
variation, only ten segments of the major epicardial coro-
nary arteries were evaluated on coronary CTA regardless
of size: left main artery, left anterior descending artery
(LAD: proximal, middle, and distal), left circumflex artery
(LCx: proximal and distal), right coronary artery (RCA:
proximal, middle, and distal), and posterior descending
artery (PDA). Image quality of the coronary arteries was
assigned the following quality scores (graded 1–4) on a per
segment level of major coronary arteries: (1) nondiagnos-
tic, poor image quality with severe coronary motion-related
or other artifacts; (2) acceptable quality with moderate
motion-related or other artifacts; (3) good quality with mild
motion-related or other artifacts; and (4) excellent quality
with minimal or no motion artifacts. Interpretability was
defined as grade [1 on per-segment level. To compare the
diagnostic accuracy between the two reconstructions, the
segments were graded as normal/having ‘‘not significant’’
or having ‘‘significant’’ (C50 % maximal luminal diameter
stenosis) coronary artery disease with the most normal
proximal and distal lumen used as the reference diameters.
Invasive coronary angiography (iCAG) was used as the
reference standard to determine the presence or absence of
coronary stenosis. The larger ten coronary segments that
were evaluable by coronary CTA were interpreted on
iCAG.
Statistical analysis
Statistical analysis was performed with SPSS 19.0 for
Macintosh (IBM, Somers, NY, USA). A two-tailed p value
of 0.05 was deemed to be significant. In both groups with
high and lower HRs, image quality, interpretability, and
diagnostic accuracy were compared between the motion-
correction and STD reconstructions. Diagnostic accuracy
was determined for the ability of cCTA to identify signif-
icant coronary artery disease at the 50 % stenosis threshold
compared with iCAG. Comparisons of ordinal variables
were performed with Wilcoxon signed-rank tests, and
comparisons of categorical variables were performed using
McNemar’s or Fisher’s exact tests. Data are presented as
the mean ± standard deviation (SD). The Cohen’s weigh-
ted Kappa statistics was used to determine intraobserver
and interobserver agreements between the two CT inves-
tigators for image quality and interpretability. All cCTA
images generated from the STD and motion-correction
reconstruction were initially analyzed by an experienced
cardiac radiologist with 9 years of experience. The

Table 1 Patient characteristics Low HR group High HR group p Total

(N = 51) (N = 33) (N = 84)

Age (average ± SD, years) 54.7 ± 11.3 58.9 ± 11.3 0.0998 56.3 ± 11.4
Male sex (%) 66.7 % 57.6 % 0.3991 61.1 %
Body mass index (kg/m2) 25.0 ± 3.3 25.0 ± 4.0 0.9313 25.0 ± 3.6
Bold value indicates that DM (%) 23.5 % 24.2 % 0.9403 23.80 %
statistically significant Hypertension (%) 56.9 % 60.6 % 0.7340 58.3 %
p values were calculated by Smoking (%) 35.3 % 24.2 % 0.2846 31.0 %
Student’s t test, Pearson Chi
Dyslipidemia (%) 61.4 % 48.3 % 0.2701 56.2 %
square test, or Fisher’s exact test
as appropriate Mean HR during study (bpm) 57.2 ± 5.0 71.1 ± 4.6 <0.05 62.6 ± 8.4
HR heart rate, N number, VPC HR variability during study (bpm) 5.3 ± 11.8 6.0 ± 8.6 0.7647 5.6 ± 10.6
ventricular premature Radiation dose (mSv) 7.7 ± 3.2 8.2 ± 3.9 0.5404 7.9 ± 3.5
contraction, p values between Arrythmia (VPC) 3.9 % 3.0 % [0.9999 3.6 %
low and high HR group

123
1606 Int J Cardiovasc Imaging (2014) 30:1603–1612

intraobserver and interobserver reproducibility for deter-
mination of the image quality and interpretability in both
STD and MC reconstructions were evaluated in a subset of
40 patients who were randomly selected from the patient
population. To determine the intraobserver variability, the
image analysis was repeated twice 1 month later by the
same investigator in these 40 patients. Subsequently, a
second experienced investigator who was blinded to the
initial analysis reevaluated the images of 40 patients to
determine the interobserver variability.

Table 2 Image quality of the All patients HR \65 HR C65

standard versus motion- (N = 84) (N = 51) (N = 33)
correction (MC) algorithms
Standard MC p Standard MC p Standard MC p
Mean Mean Mean

LM 3.6 3.7 0.125 3.9 3.9 – 3.3 3.5 0.125

RCA
P 3.1 3.6 <0.05 3.3 3.7 <0.05 2.7 3.3 <0.05
M 2.9 3.5 <0.05 3.2 3.7 <0.05 2.4 3.2 <0.05
D 3.5 3.7 <0.05 3.7 3.8 0.063 3.1 3.5 <0.05
PDA 3.4 3.6 <0.05 3.7 3.8 0.125 3.0 3.4 <0.05
LAD
P 3.4 3.5 0.109 3.7 3.7 0.250 3.0 3.1 0.500
M 3.3 3.5 <0.05 3.6 3.7 0.500 2.8 3.1 <0.05
Bold values indicate that
statistically significant D 3.4 3.4 0.125 3.6 3.6 0.500 3.1 3.1 0.500
p value by Wilcoxon signed LCx
rank test P 3.2 3.3 <0.05 3.6 3.6 1.000 2.6 3.0 <0.05
N the number of patients, MC D 3.2 3.4 <0.05 3.5 3.6 0.063 2.7 3.0 <0.05
Motion-correction algorithm, All 3.3 3.5 <0.05 3.6 3.7 <0.05 2.9 3.2 <0.05
P proximal, M middle, D distal

Fig. 1 Medial axial reformat

(MAR) images of a coronary
tree with standard (a) and
motion-correction algorithm
(b) of a 60-year-old male with a
HR of 62 bpm (range
60–63 bpm). Note that the
image quality of the proximal
and mid-RCA was improved
from 2 with standard (arrow in
a) to 4 (arrow head) with
motion-correction
reconstruction (arrow head in b)

123
Int J Cardiovasc Imaging (2014) 30:1603–1612
Results
The study group consisted of 84 patients (53 males and 31
females) with a mean age (±SD) of 56.3 ± 11.4 years
(range 33–82). The mean HR (±SD) was 62.6 ± 8.4 bpm
(range 42–83 bpm) and the mean HR variability was
5.6 ± 10.6 bpm. Beta-blockers were administered to 18
patients. In the low HR group (n = 51), the mean HR was
57.2 ± 5.0 bpm (range 42–64 bpm), and the mean HR
variability was 5.3 ± 11.8 bpm. In the high HR group
(n = 33), the mean HR was 71.1 ± 4.6 (range 65–83 bpm),
and the mean HR variability was 6.0 ± 8.6 bpm.
The patient characteristics are provided in Table 1. Most
of motion-correction reconstructions were performed at
75 % (n = 49), and only two cases of reconstructions were
generated at 45 % (n = 1) and 30 % (n = 1) of the R–R
interval, respectively, in the lower HR group. In the high
HR group, however, motion-correction reconstructions
were conducted at 75 % (n = 16), 45 % (n = 12), 50 %
(n = 3), 70 % (n = 1), and 40 % (n = 1) of the R–R
interval. The mean effective radiation dose (SD) was
7.92 ± 3.49 mSv.
Fig. 2 Volume rendering and curved multiplanar reconstruction
Images of the RCA using the standard protocol (a) and motion-
correction algorithm (b). The image quality of the mid-RCA was
improved from 1 on standard (arrow in a) to 3 on motion-correction
reconstruction (arrow head in b) in a 60-year-old male with a HR of
74 bpm (range 70–79 bpm)
1607
Image quality and interpretability
The overall image quality was higher using motion-cor-
rection versus STD reconstruction on a per-segment level
of the coronary arteries in both HR groups and in all
patients (p \ 0.05) (Table 2). In the low HR group,
motion-correction reconstruction significantly (p \ 0.05)
improved the image quality in the proximal and middle
segment of RCA (Fig. 1). In the high HR group, the image
quality was significantly (p \ 0.05) higher on a per-seg-
ment level of all three segments of the RCA (Fig. 2), the
middle segment of the LAD (Fig. 3), the proximal and
distal segments of the LCx (Fig. 4), and the PDA. In all
patients, images with MC reconstruction demonstrated
significantly (p \ 0.05) higher image quality on a per-
segment level of the three segments of RCA, the middle
LAD, the proximal and distal segments of the LCx, and the
PDA.
Compared with STD reconstruction, motion-correction
reconstruction was associated with an improvement in per-
segment, per-artery, and per-patient interpretability, with a
significant reduction in the number of nondiagnostic seg-
ments. In the high HR group, interpretability was higher
using motion-correction versus STD reconstruction
[96.4 % (318/330) vs. 90.6 % (299/330), respectively;
Fig. 3 Curved multiplanar reconstruction images of the left anterior
descending artery (LAD) using the standard (a) protocol and motion-
correction algorithm (b). Using the motion-correction algorithm, the
image quality of mid-LAD was improved from 2 (arrow in a) to 3
(arrow head in b) in a 37-year-old male with a heart rate (HR) of
66 bpm (range 65–69 bpm) (arrow head)
123
1608
p \ 0.01] on a per-segment level of the coronary arteries.
In the low HR group, interpretability was also higher using
motion-correction than STD reconstruction [99.8 % (509/
510) vs. 98.0 % (500/510), respectively; (p \ 0.01)]. The
overall interpretability was higher with the use of MC
versus STD reconstruction [98.5 % (827/840) vs. 95.1 %
(799/840); p \ 0.01] on a per-segment level of coronary
arteries in all patients (Table 3).
‘‘Almost perfect’’ intraobserver and ‘‘substantial’’ inter-
observer agreement were observed for the determination of
image quality in STD reconstructions [Kappa = 0.83 (95 %
confidence interval [CI] 0.79–0.87), and Kappa = 0.65
(95 % CI 0.59–0.70), respectively] as well as in motion-
correction reconstructions [Kappa = 0.85 (95 % CI
0.81–0.89), and Kappa = 0.62 (95 % CI 0.55–0.69),
respectively].
Diagnostic accuracy
iCAG was performed in 25 (16 in low HR group) of 84
patients within 6 weeks after cCTA. Among them, 55
obstructive segments (C50 %) were observed by iCAG in
76.0 % of patients. Compared with STD reconstructions, the
use of MC reconstruction was associated with an significant
(p \ 0.05) improvement in per-segment and per-artery
Fig. 4 Curved multiplanar reformation images of the left circumflex
artery (LCx) with the standard and motion-correction protocol. a, b A
56-year-old male with a HR of 75 bpm (range 74–79 bpm). The
image quality of the proximal LCx was improved from 2 with
standard (arrow in a) to 3 with motion-correction algorithm (arrow
123
Int J Cardiovasc Imaging (2014) 30:1603–1612
diagnostic accuracy. No significant difference, however, was
observed at the per-patient level (Table 4).
The use of MC reconstruction was also associated with
an improvement in sensitivity, specificity, positive predic-
tive value (PPV), and negative predictive value (NPV) in
per-segment base, and specificity, PPV, and NPV in per-
artery base. However, no statistically significant differ-
ences in PPV and NPV in per-segment base and sensitivity,
PPV, and NPV in per-artery base between STD and
motion-correction reconstruction were observed (Table 4).
Discussion
A new vendor-specific, software-based solution for motion
correction has been introduced recently. This technique
characterizes the motion path and velocity of coronary
arteries from adjacent cardiac phases and uses this infor-
mation to calculate an optimal estimate of the vessel
position at the target phase [15]. In contrast to multi-seg-
ment reconstructions, which utilize scan data from the
same period of subsequent heart cycles to improve tem-
poral resolution [14], this technique combines information
from the adjacent cardiac phase within a single cardiac
cycle to compensate for beating coronary motions [14, 15].
head in b). c, d 55-year-old female with a HR of 82 bpm (range
80–84 bpm). The image quality was improved from 2 (arrow in c) to
3 (arrow head in d) in the distal segment on motion-correction
reconstruction
Int J Cardiovasc Imaging (2014) 30:1603–1612 1609

The aims of our study were first to investigate the impact
of this vendor-specific motion-correction algorithm on the
morphological assessments of coronary arteries and, second,
to evaluate the influence of HR on the motion-correction
effect of this algorithm. Overall, the present study demon-
strated that application of this motion-correction algorithm
to cCTA showed a benefit in image quality and interpret-
ability over a standard algorithm in the patient group with the
high HR. However, in the patient group with low heart rate,
motion-correction algorithm still showed an improvement in
image quality although it was limited to proximal and middle
segment of RCA (Tables 2, 3). A previous report demon-
strated that this motion-correction reconstruction algorithm
is effective in terms of image quality, interpretability, and
diagnostic accuracy for evaluation of coronary arteries and
diagnosis of coronary artery disease in patients with a higher
HR. Our study is in agreement with that report by Leipsic
et al. [15] in this regard.
However, limited experience with this technique and its
application using a single-source CT scanner with inher-
ently insufficient temporal resolution make us hesitate to
abandon the use of beta-blockers in real-world situations.
Furthermore, we believed that this motion-correction
algorithm may play a role in patients with lower HR
because it has been proven that still considerable motion
artifacts are noted on coronary segments in such patients
[11]. In the present study cohort, the average HR was
62.6 ± 8.4 bpm after the use of additional rate-control
agents in patients with a pre-scan HR over 65 bpm. Despite
this, we observed a 39.2 % (n = 33) prevalence of an
average heart rate over 65 bpm during the CT study and a
9.4 % (31/330 segments) prevalence of non-interpretable
coronary segments at the time of coronary CTA, which
improved to 3.6 % (12/330 segments) with the use of
motion-correction algorithm in this high HR group. As
expected, in patients with a high HR, motion-correction

Table 3 Interpretability of the standard versus motion-correction (MC) algorithms

All patients HR \65 HR C65

Standard MC p Standard MC p Standard MC p

N % N % N % N % N % N %

Per-segment
LM 3 (3.6) 1 (1.2) 0.157 0 (0.0) 0 (0.0) – 3 (9.1) 1 (3.0) 0.157
RCA
P 7 (8.3) 0 (0.0) <0.05 2 (3.9) 0 (0.0) 0.157 5 (15.2) 0 (0.0) <0.05
M 11 (13.1) 2 (2.4) <0.05 4 (7.8) 0 (0.0) <0.05 7 (21.2) 2 (6.1) <0.05
D 2 (2.4) 0 (0.0) 0.157 0 (0.0) 0 (0.0) – 2 (6.3) 0 (0.0) 0.157
PDA 4 (4.8) 0 (0.0) <0.05 1 (2.0) 0 (0.0) 0.317 3 (9.4) 0 (0.0) 0.083
LAD
P 2 (2.4) 2 (2.4) 1.000 0 (0.0) 0 (0.0) – 2 (6.1) 2 (6.1) 1.000
M 2 (2.4) 1 (1.2) 0.317 0 (0.0) 0 (0.0) – 2 (6.1) 1 (3.0) 0.317
D 1 (1.2) 0 (0.0) 0.317 1 (2.0) 0 (0.0) 0.317 0 (0.0) 0 (0.0) –
LCx
P 5 (6.0) 4 (4.8) 0.317 1 (2.0) 1 (2.0) – 4 (12.1) 3 (9.1) 0.317
D 4 (4.8) 3 (3.6) 0.317 1 (2.0) 0 (0.0) 0.317 3 (9.1) 3 (9.1) –
All 41 (4.9) 13 (1.6) <0.05 10 (2.0) 1 (0.2) <0.05 31 (9.4) 12 (3.6) <0.05
Per-artery
LM 3 (3.6) 1 (1.2) 0.157 0 (0.0) 0 (0.0) 1.0000 3 (9.1) 1 (3.0) 0.157
RCA 15 (18.3) 2 (2.4) <0.05 6 (12.0) 0 (0.0) <0.05 9 (28.1) 2 (6.3) <0.05
PDA 4 (4.8) 0 (0.0) <0.05 1 (2.0) 0 (0.0) 0.317 3 (9.4) 0 (0.0) 0.083
LAD 5 (6.0) 3 (3.6) 0.317 1 (2.0) 0 (0.0) 0.3173 4 (12.1) 3 (9.1) 0.564
LCx 8 (9.5) 6 (7.1) 0.157 2 (3.9) 1 (2.0) 0.3173 6 (18.2) 5 (15.2) 0.317
All 35 (8.4) 12 (2.9) <0.05 10 (3.9) 1 (0.4) <0.05 25 (15.3) 11 (6.8) <0.05
Per-patient
All 22 (26.2) 8 (9.5) <0.05 9 (17.7) 1 (2.0) <0.05 13 (39.4) 7 (21.2) <0.05

Bold values indicate that statistically significant

p value by McNemar’s test
N the number of nondiagnostic segments, % the percentage of nondiagnostic segments, MC motion-correction algorithm, P proximal, M middle, D distal

123
1610 Int J Cardiovasc Imaging (2014) 30:1603–1612

reconstruction resulted in significant improvement of the phase) to reconstruct images with the STD algorithm and
image quality and interpretability of multiple segments of applies the motion-correction algorithm to this best phase
the RCA, LCx, and LAD on per-segment levels. By con- to reconstruct motion-corrected images. We then compared
trast, this motion-correction algorithm has been proven to a pair of coronary images reconstructed using the STD and
exhibit image quality improvement limited to the proximal motion-correction techniques in each patient. Conse-
and middle segments of the RCA in the low HR group. quently, there would be less room for this motion-correc-
This finding coincides with our observations that relatively tion technique to improve the image quality of coronary
few artifacts in coronary arteries were to be corrected by arteries. Thus, the effect of motion-correction technique
the motion-correction algorithm in this group. According to may be underestimated. Second, we included 10 segments
previous reports [3, 16], motion artifacts are minimized of larger coronary arteries but excluded smaller branches
when the motion of the coronary arteries is minimal. from the comparison to avoid complexity in data analysis
Coronary arteries have been shown to undergo heteroge- due to multiple coronary variations [16]. Third, again,
neous movement throughout the cardiac cycle, possibly although the use of this motion-correction reconstruction
causing motion artifacts on CT images when the velocity resulted in significant improvement in diagnostic accuracy
exceeds the temporal resolution of the scanner. The fastest in per-segment and per-artery base, our study may be
velocity was measured in the RCA, and the velocities of underpowered to evaluate the diagnostic performance of
RCA were faster than those of the LAD and LCx. Fur- cCTA with coronary angiography as a reference standard.
thermore, LAD not only has the largest width of the mid- In this study, we performed cCTA in the patients with
diastolic velocity trough but also the lowest minimum atypical chest pain with low-to-intermediate pre-test
velocity in systole [18, 19]. Accordingly, motion artifacts probability in accordance with recently published reporting
and image quality degradation of LAD and LCx were less guidelines [21]. Furthermore, patients in this study were
frequent than those of the RCA. Therefore, the effect of referred for CAG usually when significant stenosis was
motion correction in improving image quality of LAD and detected on cCTA. Thus only a few patients underwent
LCx in the patients with a lower heart rate may be limited
as demonstrated in the present study. However, as shown in
Table 3, considering that most segments described as non- Table 4 Diagnostic performance of the standard versus motion-cor-
interpretable are those of a fast-moving RCA, and motion- rection (MC) algorithms
correction algorithm improved the image quality of those Diagnostic Standard MC p
segments significantly, we believe that this motion-cor- performances
rection algorithm still has a role in patients with a lower
Segments (N = 250)
heart rate for improving the image quality of RCA where
Sensitivity % (N) 87.3 (48/55) 94.6 (52/55) <0.05
the fastest motion occurred by reducing the motion artifacts
Specificity % (N) 96.9 (189/195) 99.0 (193/195) <0.05
in proximal and mid-segment.
PPV % (N) 88.9 (48/54) 96.3 (52/54) 0.2702
In contrast to a previous report [15], although the use of
NPV % (N) 96.4 (189/196) 98.5 (193/196) 0.2001
the motion-correction algorithm was associated with an
Accuracy % (N) 94.8 (237/250) 98.0 (96/100) <0.05
improvement in per-segment and per-artery sensitivity, Arteries (N = 100)
specificity, PPV, and NPV, and diagnostic accuracy, there Sensitivity % (N) 94.9 (37/39) 94.9 (37/39) [0.9999
were no statistically significant improvements in some of Specificity % (N) 90.2 (55/61) 96.7 (59/61) <0.05
these measures over the STD approach in our study cohort PPV % (N) 86.1 (37/43) 94.9 (37/39) 0.2690
(Table 4). We believe it may be attributed to less motion NPV % (N) 96.5 (55/57) 96.7 (59/61) [0.9999
artifacts and better image quality due to the relatively Accuracy % (N) 92.0 (92/100) 96.0 (96/100) <0.05
lower average HR (62.6 ± 8.4 bpm) during CT acquisition Patients (N = 25)
in our study group [16], compared with the higher average Sensitivity % (N) 94.7 (18/19) 89.5 (17/19) 0.3173
HR in the previous study (71.8 ± 12.7 bpm) [15, 16]. Specificity % (N) 66.7 (4/6) 83.3 (5/6) 0.3173
Nevertheless, application of this MC algorithm improved PPV % (N) 90.0 (18/20) 94.4 (17/18) [0.9999
per-segment and per-artery based diagnostic accuracy NPV % (N) 80.0 (4/5) 71.4 (5/7) [0.9999
(p \ 0.05) and may reduce the need for additional studies Accuracy % (N) 88.0 (22/25) 88.0 (22/25) 1.0000
because of non-evaluable segments. However, further
Bold values indicate that statistically significant
studies including larger population are warranted to
p value for sensitivity, specificity, accuracy by McNemar’s test
investigate the effect of MC algorithm on diagnostic
p value for positive predictive value (PPV) and negative predictive value
accuracy in both low and high HR group [20]. (NPV) by Fisher’s exact test
We acknowledge several limitations in this study. First, MC and N stand for motion-correction algorithm and number of segments,
we chose only the cardiac phase with the least motion (best arteries, and patients, respectively

123
Int J Cardiovasc Imaging (2014) 30:1603–1612
Fig. 5 Curved multiplanar reconstruction and invasive coronary
angiography images of the right coronary artery (RCA) in a
55-year-old female with a heart rate of 82 bpm (range 80–84 bpm).
Although focal significant stenosis of distal RCA (white arrow) was
clearly demonstrated, due to the focal motion artifact (black arrow)
superimposed, the stenosis of the middle RCA was not detected in
MPR images with standard algorithm (a). However, significant
luminal stenosis of the middle (black arrow head) and distal RCA
(white arrow head) were found in MPR image reconstructed using the
motion-correction algorithm (b). Cross-sectional views of stenosis of
the middle RCA in standard and motion-correction reconstruction
image are demonstrated in left lower in a and b, respectively. Note the
prominent motion artifact of the stenotic segment in the cross-
sectional image at the mid-RCA on standard reconstruction. The right
coronary angiogram clearly shows significant stenosis of the middle
and distal RCA (arrows) (c)
1611
coronary angiography and we used a small number of
patients to evaluate the diagnostic accuracy of cCTA.
Further study is needed to validate the effect of motion-
correction algorithm on diagnostic accuracy, particularly
when the stenosis is located in the RCA (Fig. 5) [20].
In conclusion, our results suggest that the motion-cor-
rection algorithm is more beneficial to the patients with a
higher HR in terms of improving the image quality and
interpretability of cCTA. However, in patients with a lower
HR, this new algorithm was also useful for evaluating the
RCA where the fastest motion of coronary artery occurred.
Thus, we believe this new motion-correction algorithm
may be used to improve the image quality and interpret-
ability of coronary arteries in patients undergoing cCTA
with a rate controlling agent, with no further costs such as
additional radiation or contrast input.
Acknowledgments This work was supported by Soonchunhyang
University Research Fund.
Conflict of interest The authors have no conflict of interest to declare.
References
1. Schoepf UJ, Zwerner PL, Savino G et al (2007) Coronary CT
angiography. Radiology 244(1):48–63
2. Achenbach S (2006) Computed tomography coronary angiogra-
phy. J Am Coll Cardiol 48(10):1919–1928
3. Leschka S, Wildermuth S, Boehm T et al (2006) Noninvasive
coronary angiography with 64-section CT: effect of average heart
rate and heart rate variability on image quality. Radiology
241(2):378–385
4. Bastarrika G, Lee YS, Huda W et al (2009) CT of coronary artery
disease. Radiology 253(2):317–338
5. Miller JM, Rochitte CE, Dewey M et al (2008) Diagnostic per-
formance of coronary angiography by 64-row CT. N Engl J Med
359(22):2324–2336
6. Dewey M, Teige F, Laule M et al (2007) Influence of heart rate
on diagnostic accuracy and image quality of 16-slice CT coronary
angiography: comparison of multisegment and halfscan recon-
struction approaches. Eur Radiol 17(11):2829–2837
7. Seifarth H, Wienbeck S, Pusken M et al (2007) Optimal systolic and
diastolic reconstruction windows for coronary CT angiography
using dual-source CT. AJR Am J Roentgenol 189(6):1317–1323
8. Chung CS, Karamanoglu M, Kovacs SJ (2004) Duration of
diastole and its phases as a function of heart rate during supine
bicycle exercise. Am J Physiol Heart Circ Physiol 287(5):H2003–
H2008
9. Abbara S, Arbab-Zadeh A, Callister TQ et al (2009) SCCT
guidelines for performance of coronary computed tomographic
angiography: a report of the Society of Cardiovascular Computed
Tomography Guidelines Committee. J Cardiovasc Comput To-
mogr 3(3):190–204
10. Scharf M, Bink R, May MS et al (2011) High-pitch thoracic CT
with simultaneous assessment of coronary arteries: effect of heart
rate and heart rate variability on image quality and diagnostic
accuracy. JACC Cardiovasc Imaging 4(6):602–609
123
1612
11. Achenbach S, Ropers U, Kuettner A et al (2008) Randomized
comparison of 64-slice single- and dual-source computed
tomography coronary angiography for the detection of coronary
artery disease. JACC Cardiovasc Imaging 1(2):177–186
12. Schnapauff D, Teige F, Hamm B et al (2009) Comparison
between the image quality of multisegment and halfscan recon-
structions of non-invasive CT coronary angiography. Br J Radiol
82(984):969–975
13. Matt D, Scheffel H, Leschka S et al (2007) Dual-source CT
coronary angiography: image quality, mean heart rate, and heart
rate variability. AJR Am J Roentgenol 189(3):567–573
14. Dewey M, Laule M, Krug L et al (2004) Multisegment and halfscan
reconstruction of 16-slice computed tomography for detection of
coronary artery stenoses. Invest Radiol 39(4):223–229
15. Leipsic J, Labounty TM, Hague CJ et al (2012) Effect of a novel
vendor-specific motion-correction algorithm on image quality
and diagnostic accuracy in persons undergoing coronary CT
angiography without rate-control medications. J Cardiovasc
Comput Tomogr 6(3):164–171
16. Delhaye D, Remy-Jardin M, Rozel C et al (2007) Coronary artery
imaging during preoperative CT staging: preliminary experience
with 64-slice multidetector CT in 99 consecutive patients. Eur
Radiol 17(3):591–602
17. Einstein AJ, Elliston CD, Arai AE et al (2010) Radiation dose
from single-heartbeat coronary CT angiography performed with a
320-detector row volume scanner. Radiology 254(3):698–706
123
Int J Cardiovasc Imaging (2014) 30:1603–1612
18. Ding Z, Friedman MH (2000) Quantification of 3-D coronary
arterial motion using clinical biplane cineangiograms. Int J Card
Imaging 16(5):331–346
19. Husmann L, Leschka S, Desbiolles L et al (2007) Coronary artery
motion and cardiac phases: dependency on heart rate—implica-
tions for CT image reconstruction. Radiology 245(2):567–
576
20. Min JK, Arsanjani R, Kurabayashi S et al (2013) Rationale and
design of the ViCTORY (Validation of an Intracycle CT Motion
CORrection Algorithm for Diagnostic AccuracY) trial. J Cardio-
vasc Comput Tomogr 7(3):200–206
21. Taylor AJ, Cerqueira M, Hodgson JM et al (2010) ACCF/SCCT/
ACR/AHA/ASE/ASNC/NASCI/SCAI/SCMR 2010 appropriate
use criteria for cardiac computed tomography. A report of the
American College of Cardiology Foundation Appropriate Use
Criteria Task Force, the Society of Cardiovascular Computed
Tomography, the American College of Radiology, the American
Heart Association, the American Society of Echocardiography,
the American Society of Nuclear Cardiology, the North American
Society for Cardiovascular Imaging, the Society for Cardiovas-
cular Angiography and Interventions, and the Society for Car-
diovascular Magnetic Resonance. J Cardiovasc Comput Tomogr
4(6):407.e1– 407.e33