Radiol Phys Technol (2017) 10:155160
DOI 10.1007/s12194-016-0378-x
Radiation dose optimization for the bolus tracking technique
in abdominal computed tomography: usefulness of real-time
iterative reconstruction for monitoring scan
Yuya Ishikawa1 Atsushi Urikura1 Tsukasa Yoshida1 Keisuke Takiguchi1
Yoshihiro Nakaya1
Received: 12 February 2016 / Revised: 14 September 2016 / Accepted: 14 September 2016 / Published online: 30 September 2016
 Japanese Society of Radiological Technology and Japan Society of Medical Physics 2016
Abstract The purpose of this study was to optimize the
monitoring dose, obtained using the conventional filtered
back projection (FBP) method and iterative reconstruction
algorithms, for the bolus tracking technique. A phantom
study was performed to assess the effect of the scan start
time in patients grouped according to different body
weights. An oval torso phantom was used for simulating
the time enhancement curve of the bolus tracking tech-
nique. To reproduce image noise levels in the two body
weight groups, the phantom diameter was adjusted with a
water-equivalent material. The tube currents were 10, 20,
30, and 50 mA. The monitoring scan was performed with
the conventional FBP method and real-time adaptive iter-
ative dose reduction by three-dimensional processing
(AIDR 3D). The results at different doses were compared
with those at 50 mA. The volume computed tomography
dose index was 1.31, 2.65, 3.93, and 6.56 mGy at tube
currents of 10, 20, 30, and 50 mA, respectively. The scan
start time, reconstructed using FBP, was significantly faster
at 10 and 20 mA in group A (5059 kg) and at 20 mA in
group B (C80 kg). The CT values in the region of interest
could not be measured at 10 mA because of artifacts. With
real-time AIDR 3D, both groups showed no significant
differences between the measurements obtained at 30 or
20 mA and those obtained at 50 mA. Our study demon-
strated that the real-time AIDR 3D algorithm improved the
accuracy of the CT measurements with the bolus tracking
technique.
& Yuya Ishikawa
y.ishikawa@scchr.jp
1 examinations [911]. Iterative reconstruction (IR) is an
Department of Diagnostic Radiology, Shizuoka Cancer
Center, 1007 Shimonagakubo, Nagaizumi, Sunto,
Shizuoka 411-8777, Japan
Keywords Computed tomography
Iterative
reconstruction
Radiation dose
Bolus tracking
Contrast
media
1 Introduction
Technical advances in multi-detector row computed
tomography (CT) have made it possible to obtain detailed
volume data with faster scan times [1, 2]. In multiphase
dynamic scanning for the detection of hepatic masses,
faster scanning contributes to improvement in diagnostic
imaging by reflecting the blood flow state of the liver [3].
Dynamic CT requires that the scan timing be optimized to
the patient. Test bolus injection and bolus tracking (BT) are
the two methods most commonly used to determine the
optimum scan timing for dynamic CT [35].
BT detects the arrival time of the contrast medium by
scanning arbitrary slices continuously to monitor the image
in real time. Scans are performed either intermittently or
continuously, and automatically detect the arrival time of
the contrast medium either visually or by using a region of
interest (ROI) in blood vessels to measure CT values.
Mehnert et al. [5] reported that BT that automatically
detects the arrival time of contrast medium is effective in
determining the timing of the hepatic artery phase in
multiphase imaging. In general, monitoring scans using BT
are performed at low doses of B50 mAs [58]. However,
as continuous scanning results in increased local doses, the
dose should be set as low as possible.
Recently, various technologies have been installed on
the latest CT scanners to reduce radiation doses during
image reconstruction technique that uses an iterative
algorithm. It first attracted attention for its ability to
156
maintain nearly the same image quality as that of images
obtained by using the conventional filtered back projection
(FBP) method while reducing both the dose and image
noise; thus, it is widely used in clinical settings [12]. In
some advanced CT scanners, real-time IR can be used to
obtain BT monitoring scan images, which is expected to
reduce BT monitoring scan doses. Almost no study has
been conducted to determine the optimum dose to improve
the accuracy of BT, and no study has been performed on
dose reductions made possible by IR.
The purpose of this study was to assess and optimize the
effects of patient body weight and an IR algorithm on the
monitoring dose used to obtain valid trigger timing when
using BT.
2 Materials and methods
2.1 Measurement of the time enhancement curve
This study was approved by our institutional review board
(No. 28-J70-28-1-3). To reproduce the time enhancement
curve (TEC) for BT during abdominal dynamic CT using a
phantom, we retrospectively measured the TECs of
dynamic liver CT cases where BT was used. Between April
and May 2014, 30 consecutive patients (male-to-female
ratio: 21:9, mean age: 68.8 9.8 years) underwent
dynamic liver CT at our institution. The patients with body
weights of up to 75 kg received intravenous non-ionic
iodinated contrast media at a dosage of 630 mgI/kg body
weight. Those with body weights of [75 kg received the
fixed dosage of 47.25 gI. The infusion time was stan-
dardized at 30 s. A circular ROI was set in the abdominal
aorta at the origin of the celiac artery for the monitoring
scan, which commenced 10 s after contrast agent infusion
was started. Real-time monitoring scans were performed
intermittently until the CT value of the ROI reached 80
Hounsfield units (HU) and then switched automatically to
continuous scan.
All data were obtained using a 320-row area detector CT
(Aquilion ONE, Toshiba Medical Systems, Tokyo, Japan).
The BT software used was Real Prep (Toshiba Medical
Systems). The scan parameters were as follows: tube
voltage, 120 kVp; tube current, 30 mA; rotation time,
0.5 s; detector configuration, 0.5 9 4 mm; slice thickness,
2.0 mm; and reconstruction kernel, FC11 (FBP).
Retrospective image reconstruction was performed for
the monitoring scans for all the patients. The ROI was set
in the abdominal aorta, and CT values were measured with
an interval of 0.5 s using Digital Imaging and Communi-
cations in Medicine (DICOM) information. The measured
CT values were plotted to calculate the enhancement unit
(EU) by subtracting the precontrast CT value [13]. In
Y. Ishikawa et al.
addition, because the contrast agent arrival times in the
ROI differed between the patients, the contrast media
detection time was normalized to 0 s. In this study, we
defined the arrival times as the time immediately before the
CT values increased to more than two times the variations
in the precontrast CT values. The contrast agent arrival
time was obtained from the DICOM header data recorded
on the images using an injector synchronization system.
2.2 Relationship between subject body weight
and image noise
To reproduce the relationship between subject body weight
and image noise when using a phantom for dynamic liver
CT, we divided the subjects described in Sect. 2.1 into five
body weight groups (4049 kg, n = 8; 5059 kg, n = 12;
6069 kg, n = 4; 7079 kg, n = 3; and C80 kg, n = 3).
Image noise was obtained for all body weight groups by
measuring the standard deviation (SD) of the CT values
measured prior to contrast agent detection in the ROI
described in Sect. 2.1.
2.3 TEC phantom
An overview of the phantom is shown in Fig. 1. We
enclosed a water-equivalent material (agar) inside an
elliptical phantom (Kyoto Kagaku Co., Ltd., Kyoto,
Japan) that mimics the human trunk. Cylindrical holes
were positioned on the dorsal side of the phantom to
reproduce the TEC in the monitoring scans by allowing a
simulated blood vessel to pass through the scan slices.
The simulated blood vessel rod had a concentration gra-
dient of continuous CT values (from 0 to 250 HU).
Movement of the rod was controlled by a power injector,
and it was passed through the scan slice at a constant
speed. In addition, to reproduce two types of noise levels
(group A: 5059 kg and group B: C80 kg), we adjusted
image noise by adding a uniform layer of water-equiva-
lent material to the outside perimeter of the phantom. The
circular ROI with an area of 50 mm2 was placed at the
simulated blood vessel slices.
The ROI values were measured five times under each
condition, and the TEC and image noise were compared
with the results described in Sect. 2.1. The data acquisition
parameters used were the same as those described in
Sect. 2.1.
2.4 Relationship between monitoring dose
and trigger timing
To assess the relationship between the monitoring dose
and trigger timing in BT, we measured the main scan
start time at different monitoring doses. Phantom slices
Radiation dose optimization for the bolus tracking technique in abdominal computed tomography 157

Fig. 1 A schematic diagram of the phantom used in the analysis. An oval torso phantom was filled with a water-equivalent material (agar).
A TEC was reproduced by scanning a simulated blood vessel passing through the scan cross section in the z-direction

were continuously scanned and monitored. The ROI
threshold was set at an EU of 100 HU. Differences in
trigger start times were compared by measuring the start
times of the main scans. Accurate scan start times were
obtained from the DICOM header data recorded on the
images by an injector synchronization system. Four tube
currents were used as follows: 10, 20, 30, and 50 mA.
The monitoring dose was defined as the CTDIvol dis-
played on the console.
The other scan parameters were the same as those
described in Sect. 2.1. Each condition was measured five
times, and the mean values of the results were compared
with the values measured at 50 mA. The data were
obtained using the same parameters as described in
Sect. 2.1, except for tube current.
2.5 Effect of adaptive iterative dose reduction
by three-dimensional processing on noise level
and trigger timing
To assess the effect of real-time adaptive iterative dose
reduction by three-dimensional processing (AIDR 3D,
Toshiba Medical Systems) on trigger timing under the
conditions described in Sect. 2.4, we measured the main
scan start time for each dose. All the results were compared
with the values measured at 50 mA.
2.6 Statistical analyses
We used the MannWhitney U test for the statistical
analysis of the standard deviation of the CT values in
Sect. 2.2. We used the paired t test for the statistical
analysis of the scan start times mentioned in Sects. 2.4 and
2.5. The significance level for all the statistical analyses
was \5 %. All statistical analyses were performed using
the EZR 1.31 software [14]. EZR is a graphical user
interface for R (The R Foundation for Statistical Comput-
ing, Vienna, Austria).
3 Results
3.1 Measurement of time enhancement curve
A graph showing the median TEC values obtained via
monitoring scans using BT on dynamic liver CT is shown
in Fig. 2. The time from contrast agent detection to an EU
of 100 HU was *4 s.
3.2 Relationship between body weight and image SD
The relationship between the subjects body weights and
the image SDs is shown in Fig. 3. The median image SD
158 Y. Ishikawa et al.

Fig. 2 A graph showing the TEC plot acquired from the BT scan Fig. 4 A graph showing simulated TEC plots acquired from the
images of abdominal dynamic CT phantom measurements (square plots). The dotted line with diamond
plots indicates the results presented in Sect. 2.1

Table 1 The trigger times of the monitoring scans for the tube cur-
rents obtained using FBP
Trigger time (s)
Tube current (mA) Group A p value Group B p value

50 16.7 0.05 16.6 0.20

30 16.5 0.19 n.s. 16.1 0.40 n.s.
20 16.0 0.15 \0.01 14.6 0.48 \0.01
10 15.6 0.43 \0.01 u.m.
n.s. no significant difference u.m. unmeasurable
* p values were calculated for 50 vs. 10, 20, and 30mA in the two
groups

3.4 Relationship between the monitoring scan doses

and scan start times

The volume CT dose indexes (CTDIvol) for the monitoring

Fig. 3 Comparison between patient body weight and image noise
(the standard deviation of the CT values from the abdominal aorta)
values were 39.0, 49.1, 63.3, 79.5, and 129.8 HU for the
40- to 49-, 50- to 59-, 60- to 69-, 70- to 79-, and C80-kg
body weight groups, respectively.
3.3 Measurement of the TEC and SD
with phantoms
The TECs obtained from the TEC phantoms are shown in
Fig. 4. The phantom TECs were nearly the same as the
results in Sect. 3.1. All of the EU plots up to 100 HU were
within two SD. The image SDs for the phantoms simulat-
ing the two weight groups (groups A and B) were
48.1 5.3 and 127.6 15.4 HU, respectively, and
showed values that were similar to the results in Sect. 3.2.
scan tube current settings of 10, 20, 30, and 50 mA were
1.31, 2.65, 3.93, and 6.56 mGy, respectively. The scan start
time in group A was slower at 30 mA but did not represent
a significant difference. The scan start times for 10 and
20 mA were significantly faster than that at 50 mA
(Table 1). The scan start time in group B did not signifi-
cantly differ at 30 mA, but the scan start time at 20 mA
was faster. CT values could not be measured at 10 mA.
3.5 Effect of AIDR 3D processing on scan start
times
The scan start times when real-time AIDR 3D was used are
shown in Table 2. In group A, the scan start times at 20 and
30 mA did not significantly differ from those at 50 mA, but
a significant difference was observed at 10 mA. In group B,
no significant difference was observed at any dose.
Radiation dose optimization for the bolus tracking technique in abdominal computed tomography 159

Table 2 The trigger times of the monitoring scans acquired using the B20 mA, which cause marked changes in CT values in the
IR algorithm ROI, and the threshold may be reached faster. In low-dose
Trigger time (s) monitoring (10 mA) in the high body weight group, streaks
and dark band artifacts caused by photon starvation and
Tube current (mA) Group A p value Group B p value
beam hardening were observed, which made it difficult to
50 16.7 0.05 16.6 0.20 measure CT values [15].
30 16.7 0.21 n.s. 16.6 0.18 n.s. The accuracies of the scan start time measurements
20 16.4 0.28 n.s. 16.4 0.41 n.s. under all dose conditions were improved through the use of
10 16.2 0.30 \0.05 16.3 0.75 n.s. real-time AIDR 3D in performing monitoring scans. At
C20 mA, the scan start times were the same as those at
n.s no significant difference
50 mA, regardless of patient body weight. Figure 5 shows
* p values were calculated for 50 vs. 10, 20, and 30 mA in the two
groups examples of clinical images of a monitoring scan obtained
from a 62-year-old man (body weight, 80 kg; height,
161 cm; and body mass index, 30.9 kg/m2), which were
4 Discussion
reconstructed with FBP (a) and AIDR 3D (b). The mean
CT (SD) values in the abdominal aorta (arrowhead) were
The TEC phantom that we used was effective at repro-
162 (104.3) and 153 (32.8) HU for FBP and AIDR 3D,
ducing CT value changes caused by BT on dynamic CT. It
respectively. The CT values and SD showed that AIDR 3D
also faithfully reproduced image noise changes resulting
was extremely effective when used in monitoring scans of
from different patient body weights.
clinical images and in suppressing apparent elevations in
FBP obtained equivalent scan start times regardless of
CT values caused by noise amplitude. Based on the pre-
body weight at all tube currents of C30 mA, under the
vious studies on physical image quality, IR changes fre-
imaging conditions used in this study. However, the scan
quency characteristics, unlike FBP [1618]. However, as
start times were shortened at B20 mA. The CT values in
IR maintains CT values [19, 20], the use of IR in moni-
group B were impossible to measure when the tube current
toring scans is effective in reducing noise. The monitoring
was set at 10 mA. Many previous clinical studies that used
scan doses used in the BT method used in this study could
BT set the tube current to B50 mA [58]. However, the
be reduced by 40 or 60 % by changing the tube current
tube current settings for BT monitoring scans often do not
setting from 50 to 30 or 20 mA, respectively.
account for patient body weight. To obtain optimum scan
This study has several limitations. First, the image noise
timing when using BT, stable CT values should be mea-
level was not necessarily identical between the real-time
sured. Noise increases at low tube current settings of
monitoring images and the retrospective reconstructed

Fig. 5 Clinical images of a monitoring scan obtained from a 62-year- rotation time of 0.5 s, and tube current of 30 mA. The mean (SD) CT
old man (body weight, 80 kg; height, 161 cm; and body mass index, values in the abdominal aorta (arrowhead) were 162 (104.3) and 153
30.9 kg/m2), which were reconstructed with FBP (a) and AIDR 3D (32.8) HU for FBP and AIDR 3D, respectively
(b). Both images were obtained with a tube voltage of 120 kVp,
160
images acquired by FBP and AIDR 3D. The accurate
image noise level of the real-time images might be dif-
ferent from our results. Second, the phantom was made of a
uniform material that did not include bone or intestinal
tract structures. Therefore, noise and artifacts originating
from these peripheral structures may not have been
reproduced. Third, as the phantom did not take patient
respiration and body movements into account, we could not
know the effect of these factors on the clinical images.
Fourth, this study did not assume differing thresholds and
contrast conditions. Finally, this study investigated a single
device supplied by one vendor and did not investigate CT
scanners and IR algorithms from other vendors.
5 Conclusion
Applicable real-time IR algorithms for dynamic liver CT
BT monitoring increased the accuracy of CT measurements
by reducing high noise levels. This indicates that the use of
IR with monitoring scans allowed for reducing doses to
below those used with conventional FBP.
Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of
interest.
References
1. Hein PA, Romano VC, Lembcke A, May J, Rogalla P. Initial
experience with a chest pain protocol using 320-slice volume
MDCT. Eur Radiol. 2009;19:114855.
2. Rogalla P, Kloeters C, Hein PA. CT technology overview:
64-slice and beyond. Radiol Clin North Am. 2009;47:111.
3. Hammerstingl RM, Vogl TJ. Abdominal MDCT: protocols and
contrast considerations. Eur Radiol. 2005;15(Suppl 5):E7890.
4. Sandstede JJ, Tschammler A, Beer M, Vogelsang C, Wittenberg
G, Hahn D. Optimization of automatic bolus tracking for timing
of the arterial phase of helical liver CT. Eur Radiol.
2001;11:1396400.
5. Mehnert F, Pereira PL, Trubenbach J, Kopp AF, Claussen CD.
Biphasic spiral CT of the liver: automatic bolus tracking or time
delay? Eur Radiol. 2001;11:42731.
6. Dinkel HP, Fieger M, Knupffer J, Moll R, Schindler G. Opti-
mizing liver contrast in helical liver CT: value of a real-time
bolus-triggering technique. Eur Radiol. 1998;8:160812.
Y. Ishikawa et al.
7. Sultana S, Awai K, Nakayama Y, Nakaura T, Liu D, Hatemura
M, Funama Y, Morishita S, Yamashita Y. Hypervascular hepa-
tocellular carcinomas: bolus tracking with a 40-detector CT
scanner to time arterial phase imaging. Radiology.
2007;243:1407.
8. Tatsugami F, Awai K, Takada H, Yoshikawa S, Takeda Y,
Morita H, Narumi Y. Reduction of interpatient variability of
arterial enhancement using a new bolus tracking system in
320-detector computed tomographic coronary angiography.
J Comput Assist Tomogr. 2013;37:7983.
9. Catalano C, Francone M, Ascarelli A, Mangia M, Iacucci I,
Passariello R. Optimizing radiation dose and image quality. Eur
Radiol. 2007;17(Suppl 6):F2632.
10. Matsubara K, Sugai M, Toyoda A, Koshida H, Sakuta K, Takata
T, Koshida K, Iida H, Matsui O. Assessment of an organ-based
tube current modulation in thoracic computed tomography.
J Appl Clin Med Phys. 2012;13:3731.
11. Singh S, Kalra MK, Hsieh J, Licato PE, Do S, Pien HH, Blake
MA. Abdominal CT: comparison of adaptive statistical iterative
and filtered back projection reconstruction techniques. Radiology.
2010;257:37383.
12. Willemink MJ, de Jong PA, Leiner T, de Heer LM, Nievelstein
RA, Budde RP, Schilham AM. Iterative reconstruction techniques
for computed tomography Part 1: technical principles. Eur
Radiol. 2013;23:162331.
13. Yamaguchi I, Kidoya E, Suzuki M, Kimura H. Evaluation of
required saline volume in dynamic contrast-enhanced computed
tomography using saline flush technique. Comput Med Imaging
Graph. 2009;33(1):238.
14. Kanda Y. Investigation of the freely available easy-to-use soft-
ware EZR for medical statistics. Bone Marrow Transplant.
2013;48:4528.
15. Barrett JF, Keat N. Artifacts in CT: recognition and avoidance.
Radiographics. 2004;24:167991.
16. Richard S, Husarik DB, Yadava G, Murphy SN, Samei E.
Towards task-based assessment of CT performance: system and
object MTF across different reconstruction algorithms. Med
Phys. 2012;39:411522.
17. Urikura A, Ichikawa K, Hara T, Nishimaru E, Nakaya Y. Spatial
resolution measurement for iterative reconstruction by use of
image-averaging techniques in computed tomography. Radiol
Phys Technol. 2014;7:35866.
18. Smith EA, Dillman JR, Goodsitt MM, Christodoulou EG,
Keshavarzi N, Strouse PJ. Model-based iterative reconstruction:
effect on patient radiation dose and image quality in pediatric
body CT. Radiology. 2014;270:52634.
19. Love A, Olsson ML, Siemund R, Stalhammar F, Bjorkman-
Burtscher IM, Soderberg M. Six iterative reconstruction algo-
rithms in brain CT: a phantom study on image quality at different
radiation dose levels. Br J Radiol. 2013;86:20130388.
20. Schindera ST, Odedra D, Raza SA, Kim TK, Jang HJ, Szucs-
Farkas Z, Rogalla P. Iterative reconstruction algorithm for CT:
can radiation dose be decreased while low-contrast detectability
is preserved? Radiology. 2013;269:5118.