comparison of coronary ct angiography image quality with and without breast shields
DESCRIPTION
Comparison of Coronary CT Angiography Image Quality With and Without Breast ShieldsTRANSCRIPT
-
5/19/2018 Comparison of Coronary CT Angiography Image Quality With and Without Breast Shields
1/8
AJR:200, March 2013
all patients [57]. Therefore, by the as low
reasonably achievable [4] principle of usi
the lowest radiation possible to achieve a
agnosis, continued methods of radiation
duction should be sought to minimize pati
radiation exposure without significantly d
grading diagnostic imaging quality.
One potential practical strategy that ph
sicians could employ is the use of bismu
shields [8, 9] to reduce radiation exposu
to breast tissue among younger women d
ing coronary CTA (Fig. 1). Although dosi
etry data are available with radiation reduct
at the breast tissue level ranging from 30%
almost 60% [10], this technique has not gnered widespread use and is not currently r
ommended by guidelines because of conce
that unshielded techniques, such as tube c
rent modulation, may be superior to shie
ing [10], concerns that breast shields co
increase radiation dose through scatter [1
and a lack of data concerning their effect
image quality [1113]. The impact of bre
shields on image quality of nongated thora
and pulmonary embolus CT studies has be
Comparison of Coronary CTAngiography Image Quality Withand Without Breast Shields
Edward Hulten1,2
Patrick Devine1
Timothy Welch1
Irwin Feuerstein3
Allen Taylor4
Sara Petrillo5
Minnetta Luncheon1
Binh Nguyen1
Todd C. Villines1
Hulten E, Devine P, Welch T, et al.
1Cardiology Service, Walter Reed National Military
Medical Center, Bethesda, MD. Address correspondence
to E. Hulten (ehulten@partners .org).
2Present address: Noninvasive Cardiovascular Imaging,
Departments of Medicine and Radiology, Brigham and
Womens Hospital, Harvard Medical School, Boston,
MA 02115.
3Food and Drug Administration, Silver Spring, MD.
4Washington Hospital Center, Washington, DC.
5Mid-Atlantic Kaiser Permanente Group, Rockville, MD.
Cardiopulmonary Imaging Original Research
AJR2013; 200:5295 36
0361803X/13/2003529
American Roentgen Ray Society
Coronary CT angiography (CTA) is
a highly accurate method for the
noninvasive evaluation of coro-
nary artery and heart disease [1].
Increased clinical utilization of coronary CTA
has resulted in concerns over potential long-
term risks of cumulative ionizing radiation re-
lated to coronary CT and other medical imaging
modalities [2]. Recent estimates of the theoretic
imposed risk from coronary CTA and other
medical ionizing radiation imaging techniques
have varied widely, and although the risk is
generally thought to be low, it may be higher
among younger women [24]. Women young-
er than 50 years who are at significant lifetimerisk for breast cancer (the commonest malig-
nancy of women) may undergo a significant
chest radiation dose during coronary CTA. Al-
though current dose-reduction strategies, such
as prospective ECG-triggered scanning, lower
tube potential (kilovoltage), decreased scan
length, and novel scan techniques, have been
shown to markedly reduce the estimated radia-
tion dose in coronary CTA, these techniques are
not consistently applied and are not feasible in
Keywords:breast shield, cardiac CT, coronary CT
angiography, radiation
DOI:10.2214/AJR.11.8302
Received November 23, 2011; accepted after revision
February 29, 2012.
T. C. Villines reports moderate speaker honoraria from
Boehringer-Ingelheim Pharmaceuticals unrelated to the
topic of this manuscript. All other authors have no
financial disclosures or conflicts of interest to declare.All authors had full access to all of the data in the study
and take responsibility for the integrity of the data and
the accuracy of the manuscript.
The opinions and assertions contained herein are the
authors alone and do not represent the views of the
Walter Reed National Military Medical Center, the U.S.
Army, or the Department of Defense.
JO URNAL CLUB FOCUSON:
OBJECTIVE.The purpose of this study is to compare the image quality of coronary
angiography performed with and without breast shields.
MATERIALS AND METHODS.This study involved a retrospective cohort of 72 wom
with possible angina who underwent 64-MDCT retrospective ECG-gated coronary CT angio
raphy at a single academic tertiary medical center. Images of 36 women scanned while weari
bismuth-coated latex breast shields and 36 control subjects scanned without shields, match
by heart rate and body mass index, were graded on a standardized Likert scale for image quity, stenosis, and plaque by two independent board-certified readers blinded to breast shields
RESULTS.Seventy-two patients (mean [ SD] age, 53 9 years) were included. The presc
heart rate, body mass index, and Agatston score did not differ between groups. The median
timated radiation dose was 13.4 versus 16.1 mSv for those with and without breast shields (p
0.003). For shielded versus unshielded scans, 86% versus 83% of coronary segments were r
ed excellent or above average (p= 0.4), median image quality was 2.0 for both groups, me
signal was 474 75 and 452 91 HU (p= 0.27), mean noise was 33.9 8.5 and 29.8
HU (p= 0.04), and median signal-to-noise ratio was 14.4 and 14.7 (p= 0.56), respectively
CONCLUSION.Breast shields for women undergoing coronary CT angiography slig
ly increased noise but did not negatively affect signal, signal-to-noise ratio, quality, or int
pretability. Breast shield use warrants further study.
Hulten et al.Coronary CT Angiography With and Wi thout Breast Shields
Cardiopulmonary ImagingOriginal Research
Downloadedfro
mwww.ajronline.orgby114.79.32.222on07/09/14fromI
Paddress114.79.32.222.CopyrightARRS.Forpersonaluseonly;allrightsreser
ved
-
5/19/2018 Comparison of Coronary CT Angiography Image Quality With and Without Breast Shields
2/8
530 AJR:200, March 2013
Hulten et al.
reported, but data for use in coronary CTA are
not available [11, 1416]. This validation re-
mains important because coronary anatomy
by CT is more technically demanding and in-
cumbent on intricately detailed image qualitywhen compared with nongated CT of thoracic
anatomy. Because of the potential for breast
shields to offer simple but clinically impor-
tant radiation protection, the relative lack of
breast shield use for coronary CTA vis--vis
alternative radiation protection methods, and
the lack of available image quality studies, we
undertook a comparative image quality study
of coronary CTA with and without in-plane
bismuth breast shields. Thus, the purpose of
this study is to evaluate coronary CTA im-
age quality with and without in-plane bismuth
breast shield use.
Materials and Methods
Patients
We retrospectively identified a cohort of wom-
en older than 18 years without known coronary ar-
tery disease (CAD) referred for symptoms of pos-
sible angina who underwent coronary CTA scans
with bismuth-coated in-plane latex breast shields
(AttenuRad, Cone Instruments). Such shields con-
sist of a 1-mm-thick bismuth sheet impregnated in
rubber that mounts in a foam offset. The shields
were placed on the anterior chest exterior to the
patient garments, allowed repeat use, and did not
require any specia l hygiene considerations beyond
that of any other medical equipment exposed topatient contact. The cost for a medium-size breast
shield used in our study was $98.25 [17]. The
breast-shielded patients served as a convenience
sample, the use of shields having been mandated
by our institutional review boards radiation safe-
ty officer before approval of a research protocol
evaluating chest pain [18]. However, because the
risk-to-benefit ratio of breast shields remains un-
proven, beyond the requirements of this research
protocol, bismuth shields are not otherwise re-
quired in our hospital for coronary CTA. Thus,
these patients were compared with a control pop-
ulation of 36 clinically scanned patients random-
ly selected among all women who had undergone
coronary CTA using a comparable CT acquisition
protocol without breast shields during the study
period. Patients were matched by heart rate (HR)
and body mass index (BMI), because these covar-
iates have been consistently shown to be among
the strongest patient characteristics that influ-
ence image quality on coronary CTA. All coro-
nary CTA studies were performed at Walter Reed
Army Medical Center, a single-center university-
affiliated urban tertiary medical center, from Jan-
uary 2006 through February 2010. The research
protocol was approved by our hospital institu-
tional review board. Clinical information was ob-
tained from electronic health databases (inpatient,
outpatient, laboratory, and radiologic) of the De-
partment of Defense Military Healthcare System.
Coronary CTA
Per usual protocol for coronary CTA at our in-
stitution [19], all patients were prescribed variable
doses (typically 50100 mg) of oral metoprolol to
be taken 1 hour before the scheduled scan [20]. Ad-
ditional metoprolol was administered IV, if need-
ed, immediately before coronary CTA to obtain a
goal prescan HR of less than 60 beats/min. Nitro-
glycerin 0.40.8 mg sublingual was given 1 minute
before contrast-enhanced image acquisition.
All scans were performed using the same 64-
MDCT scanner (LightSpeed VCT, GE Health-
care). An initial unenhanced prospectively ECG-
triggered scan was acquired without breast shields
for calcium scoring and contrast-enhanced scan
planning. After a timing bolus series, a contrast-
enhanced scan was obtained with contrast agent
(Isovue, Bracco Diagnostics) injected IV at flow
rates of 4.56.0 mL/s through an antecubital vein,
followed by a 40-mL normal saline flush. Sixty-
four overlapping 0.625-mm slices were acquired
per rotation, with a rotation time of 350 ms. All
scans were performed using a tube potential of
120 kV. Additional acquisition variables were ad-
justed to individually optimize scans (pitch range,
0.160.25; range of tube current, 400750 mA).
When appropriate, ECG-dose modulation of tubecurrent was used, with maximal current output
generally occurring between 40% and 80% of the
R-R interval. Because prospectively ECG-trig-
gered scanning was not in widespread use at the
beginning of the study period, all coronary CTA
studies included in this study were retrospective-
ly ECG-gated. Scans for all patients were recon-
structed at 0.625-mm slice thickness using filtered
back projection and the GE Healthcare standard
soft-tissue kernel.
Existing scans were previously interpreted
jointly by a card iologist and a radiologist tra ined
in the performance and interpretation of coronary
CTA. For the assessment of image quality for our
study, noncardiac findings were not reassessed.
For the assessment of coronary findings for our
retrospective study, the scans were independent-
ly reread by two independent cardiac CT board-
certified readers using the Society of Cardiovas-
cular CT 18-segment coronary model [21] while
blinded to use of breast shields, clinical variables,
symptoms, and the original scan clinical interpre-
tation. To achieve blinding, one researcher loaded
Fig. 1Appearance of bismuth in-plane breast shields.A,Shield is shown in position on patients chest.B,Shield is shown with 10 -cm ruler.C andD,Breast shield can be seen at top of axial slice from64-MDCT cardiac CT angiography image (C) and in 3Dreconstruction of same image (D).
Downloadedfro
mwww.ajronline.orgby114.79.32.222on07/09/14fromI
Paddress114.79.32.222.CopyrightARRS.Forpersonaluseonly;allrightsreser
ved
-
5/19/2018 Comparison of Coronary CT Angiography Image Quality With and Without Breast Shields
3/8
AJR:200, March 2013
Coronary CT Angiography With and Without Breast Shields
all image series and panned the breast shield and
anterior chest tissue from the FOV on the work-
station before coronary evaluation by the two
blinded readers. After independent interpretation,
each segment was then read and scored by consen-
sus. Complete datasets from 36 women scanned
with bismuth-coated latex breast shields and 36
HR- and BMI-matched female control subjects
scanned without breast shields were randomly
graded on a standardized Likert scale for quality,
stenosis, and plaque presence and type (i.e., non-
calcified, partially calcified, or entirely calcified)
[21, 22]. Specifically, segments were graded us-
ing all available acquired phases for image qual-
ity and for the worst stenosis within each segment,
in accordance with Society of Cardiovascular CT
guidelines [21]. Quality ranged from 1 (excellent),
2 (good), 3 (below average), to 4 (uninterpretable).
Stenosis severity was graded as 0 (no stenosis), 1
(< 25% diameter worst stenosis), 2 (2549% ste-
nosis), 3 (5069% stenosis), 4 ( 70% stenosis),
and 5 (100% stenosis). Plaque within each seg-
ment was rated as noncalcified (no appreciable
calcium within the plaque or segment), partially
calcified, or entirely calcified by each rater. Seg-
ments smaller than 1.5 mm were not assessed, and
anatomically absent segments (e.g., ramus) were
coded as missing. Coronary calcium was quanti-
fied according to the Agatston method [23]. Im-
age signal and image noise were measured as the
mean ( SD) of Hounsfield units within a 1-cm di-
ameter region of interest in the aortic root and at
the level of the left main coronary artery origin,
respectively [21]. Signal-to-noise ratio was calcu-
lated as the signal divided by the noise. Estimated
radiation dose in millisieverts was calculated as
the dose-length product multiplied by the conv
sion factor for chest CT of 0.014 [6].
Statistical Analysis
Continuous variables with normal distributi
were expressed as mean ( SD) and were co
pared using Student ttest for independent gro
and one-way analysis of variance for betw
group comparisons. Categorical variables w
expressed as frequencies (percentages) and w
compared by the Pearson chi-square test. V
ables with skewed distributions were expressed
median (interquartile range [IQR]) and were co
pared by Mann-Whitney Utest. Interrater relia
ity was measured with the kappa statist ic. Beca
of a skewed distribution, calcium scores were n
ural logtransformed before analysis for tren
one was added to zero calcium scores. A tw
tailedpvalue less than or equal to 0.05 was c
sidered significant. All analyses were perform
using Stata (version 11.0, StataCorp).
Results
Seventy-two patients with a mean age
53 9 years were included. The prescan H
was 56 10 beats/min, BMI was 30 5 kg/m
median Agatston score was 0 (IQR, 014; 3
of scores were > 0), and the estimated radiat
dose was 14 5 mSv (Tables 1 and 2).
Of 1296 theoretically possible coron
segments for 72 patients using an 18-segm
coronary model, 17% were anatomically
sent and 14% were rated as too small (<
mm). Arteries that were depicted on the t
oretic 18-segment model but were absenttoo small were most commonly ramus, l
posterolateral, left posterior descending, a
second obtuse marginal segments. Amo
891 segments rated, 99% of graded coron
segments were evaluable (10 segments w
noninterpretable because of poor image qu
ity secondary to coronary motion). The m
dian quality rating for both shielded and u
shielded patients was good (2 [IQR, 220
There was no significant difference betwe
groups in the prevalence of normal, nono
structive (< 50% worst stenosis), and o
structive (50%) CAD (Table 3). There w
slightly higher rates of no coronary plaq(9% prevalence difference; p < 0.001) a
slightly lower rates of noncalcified plaq
(2% difference;p=0.012) and partially c
cified plaque (6% difference; p < 0.001)
the shield group, although rates of calcifi
and noncalcified plaque were similar (Table
Evaluation of segments located closer to
shield did not identify an effect of proximity
breast shield placement on image quality sco
TABLE 1: Demographics of Patients Who Underwent Coronary CT
Angiography With and Without Breast Shields
CharacteristicWithout Shield
(n= 36)With Shield
(n= 36) p
Age (y), mean SD 54.1 1.7 51.3 1.5 0.27a
Body mass index, mean SD 29.9 0.9 29.2 0.8 0.84a
Hypertension, no. (%) of patients 22 (61) 19 (53) 0.48a
Hyperlipidemia, no. (%) of patients 21 (58) 15 (42) 0.16a
Smoking, no. (%) of patients 2 (6) 4 (11) 0.39a
Family history, no. (%) of patients 15 (42) 5 (14) 0.01a
Diabetes mellitus, no. (%) of patients 4 (11) 2 (6) 0.39a
Heart rate (beats/min), median (IQR) 54 (4762) 55 (5061) 0.66b
Left ventricular ejection fraction (%), median (IQR) 72 (6277) 66 (5873) 0.29b
NoteIQR = interquartile range.aStudent ttest or chi-square test.bWilcoxon rank-sum test.
TABLE 2: Results of Coronary CT Angiography With and WithoutBreast Shields
Result Without Shield (n= 36) With Shield (n= 36) p
Agatston score
Median (IQR) 0 (029) 0 (02) 0.44a
Minimum 0 0 Not applicable
Maximum 563 784 Not applicable
Tube current (mA), median (IQR) 593 (573642) 606 (575643) 0.47a
Dose-length product, median (IQR) 1151 (8891405) 956 (8301060) 0.003a
Radiation dose (mSv), median (IQR) 16.1 (12.420.0) 13.4 (11.614.8) 0.003a
Signal (HU), mean SD 452 91 474 75 0.27b
Noise (HU), mean SD 29.8 8.3 33.9 8.5 0.04b
Signal-to-noise ratio, median (IQR) 14.7 (11.521.7) 14.4 (11.717.9) 0.56a
Image quality score, median (IQR) 2 (22) 2 (22) 0.50a
NoteIQR = interquartile range.aWilcoxon rank-sum test.bStudent ttest or chi-square.
Downloadedfro
mwww.ajronline.orgby114.79.32.222on07/09/14fromI
Paddress114.79.32.222.CopyrightARRS.Forpersonaluseonly;allrightsreser
ved
-
5/19/2018 Comparison of Coronary CT Angiography Image Quality With and Without Breast Shields
4/8
532 AJR:200, March 2013
Hulten et al.
(for shielded vs unshielded patients, distal left
anterior descending quality, 2 vs 2 [p= 0.4];
mid-left anterior descending quality, 2 vs 2
[p= 0.8]; and proximal right coronary artery,
2 vs 2 [p= 0.4]).
Table 4 shows predictors of image quality
by univariate and multivariate linear regres-
sion. BMI (= 0.018;p< 0.0001) was signif-
icant after multivariate analysis. As depicted
in Figure 2, comparing scans with and with-
out breast shields, 86% versus 83% of coro-
nary segments were rated as excellent or good
(p= 0.4). There was no statistically significant
difference in the percentage of below-average
(poor) quality or nonevaluable segments for
shielded patients (14%) compared to unshield-
ed patients (17%;p= 0.39). Patients without
shields had a small, but significantly higher,
number of uninterpretable segments (n = 9)
compared with shielded patients (n= 1; p=
0.01). There was a small but statistically sig-
nificant increase in image noise of 4 HU (14%
relative increase) between scans performed
with (33.9 1.4 HU) and without (29.8 1.4
HU) breast shields (p= 0.04). There was no
statistically significant difference in mean sig-
nal (474 75 vs 452 91 HU;p= 0.27) and
median signal-to-noise ratio (14.4 vs 14.7;p=
0.56) for scans performed with and without
shields. At the segment level, none of the 18
coronary segments was significantly differ-
ent between shielded and unshielded patients
for quality, stenosis severity, or plaque score.
Interrater agreement was excellent for quality
(= 0.82;p< 0.001), stenosis (= 0.82;p