Eur Radiol

https://doi.org/10.1007/s00330-017-5159-3

COMPUTED TOMOGRAPHY

Can We Perform CT of the Appendix with Less Than 1 mSv? A

De-escalating Dose-simulation Study
Ji Hoon Park 1 & Jong-June Jeon 2 & Sung Soo Lee 1 & Amar C. Dhanantwari 3 & Ji Ye Sim 4 &
Hae Young Kim 1 & Kyoung Ho Lee 1,5

Received: 21 August 2017 / Revised: 13 October 2017 / Accepted: 27 October 2017

# European Society of Radiology 2017

Abstract the non-inferiorities of 1.5- and 1.0-mSv CT were accepted;

Objectives To systematically explore the lowest reasonably
achievable radiation dose for appendiceal CT using an itera-
tive reconstruction (IR) in young adults.
Methods We prospectively included 30 patients who
underwent 2.0-mSv CT for suspected appendicitis. From the
helical projection data, 1.5-, 1.0- and 0.5-mSv CTs were gen-
erated using a low-dose simulation tool and the knowledge-
based IR. We performed step-wise non-inferiority tests se-
quentially comparing 2.0-mSv CT with each of 1.5-, 1.0-
and 0.5-mSv CT, with a predetermined non-inferiority margin
of 0.06. The primary end point was the pooled area under the
receiver-operating-characteristic curve (AUC) for three ab-
dominal and three non-abdominal radiologists.
Results For the abdominal radiologists, the non-inferiorities
of 1.5-, 1.0- and 0.5-mSv CT to 2.0-mSv CT were sequentially
accepted [pooled AUC difference: 2.0 vs. 0.5 mSv, 0.017
(95% CI: -0.016, 0.050)]. For the non-abdominal radiologists,
Electronic supplementary material The online version of this article
(https://doi.org/10.1007/s00330-017-5159-3) contains supplementary
material, which is available to authorized users.
* Kyoung Ho Lee
kholeemail@gmail.com
however, the non-inferiority of 0.5-mSv CT could not be
proved [pooled AUC difference: 2.0 vs. 1.0 mSv, -0.017 (-
0.070, 0.035) and 2.0 vs. 0.5 mSv, 0.045 (-0.071, 0.161)].
Conclusion The 1.0-mSv appendiceal CT was non-inferior to
2.0-mSv CT in terms of diagnostic performance for both ab-
dominal and non-abdominal radiologists; 0.5-mSv
appendiceal CT was non-inferior only for abdominal
radiologists.
Key points
• For both abdominal and non-abdominal radiologists, 1.0-
mSv appendiceal CT could be feasible.
• The 0.5-mSv CT was non-inferior to 2.0-mSv CT only for
expert abdominal radiologists.
• Reader experience is an important factor affecting diagnos-
tic impairment by low-dose CT.
Keywords Appendicitis . Prospective studies . Tomography,
X-ray computed . ROC curve . Sensitivity and specificity
Abbreviations
IR Iterative reconstruction
AUC Area under the receiver-operating-characteristic curve

1
Department of Radiology, Seoul National University Bundang
Hospital, 82, Gumi-ro 173 Beon-gil, Bundang-gu, Introduction
Seongnam-si, Gyeonggi-do 463-707, Korea
2
Department of Statistics, University of Seoul, Seoul, Korea CT is widely used for diagnosing acute appendicitis [1, 2].
3
CT/AMI Clinical Science, Philips, Cleveland, OH, USA Many patients undergoing appendiceal CT are young and have
4
Department of Radiology, Hanil General Hospital, Seoul, Korea normal life expectancy, for whom the radiation dose should be
5
Program in Biomedical Radiation Sciences, Department of
kept "as low as reasonably achievable". Efforts have been made
Transdisciplinary Studies, Graduate School of Convergence Science to decrease the radiation dose of appendiceal CT [3–5], now
and Technology Seoul National University, Seoul, Korea particularly using iterative reconstruction (IR) techniques. The

use of IR has been advocated in various diagnostic indications
as IR improves image quality by reducing the image noise [6].
In many previous studies purposed to reduce the CT radia-
tion dose for various applications, the investigators have rarely
clarified how they chose the tested radiation dose. Arbitrarily
set low-dose levels for research purposes would be inevitably
insufficient or excessive in practice. This study limitation can
be potentially overcome using a radiation dose-simulation tool
that enables systematic exploration across multiple dose levels
[7–9].
This study was aimed to prospectively and systematically
explore the lowest reasonably achievable radiation dose for
appendiceal CT using an iterative reconstruction in young adults.
Materials and methods
Study overview
The institutional review board approved the protocol of this
prospective proof-of-concept study. Informed consent was ob-
tained from each participant. We adopted a Bdose de-
escalating^ study design, which is often used in phase I phar-
macologic clinical trials to minimise the sample size and the
number of comparisons in determining a drug dose [10].
We planned step-wise non-inferiority tests sequentially com-
paring original 2.0-mSv CT vs. each of simulated 1.5-, 1.0-, and
0.5-mSv CT in 30 patients for the diagnostic performance of six
radiologists with different levels of experience. The non-
inferiority for the next lower dose was tested only when the
previous higher dose passed the non-inferiority test (Fig. 1).
Fig. 1 Study schema. Grey boxes
are the four alternative study
conclusions for the lowest
achievable dose
Eur Radiol
This report was made in line with the guidelines of the
Standards for Reporting of Diagnostic Accuracy (STARD)
[11].
Participants
Eligible patients were young adults (18 to 44 years of age)
who presented to the emergency department in a tertiary hos-
pital and underwent CT for suspected appendicitis. Clinical
suspicion of appendicitis and the need for a CT examination
were determined at the discretion of the emergency physicians
or surgeons on service.
CT imaging
Intravenous contrast-enhanced CT images of the abdomen
and pelvis were acquired during the portal venous phase using
a 256-channel CT system (Philips Healthcare, Cleveland,
OH). The target median dose-length product was 130 mGy·
cm, which corresponded to an effective dose of 2.0 mSv with
a conversion factor of 0.015 mSv·mGy-1·cm-1 [12]. The radi-
ation output was automatically adjusted according to individ-
ual patient body size [13]. The 2.0-mSv protocol (Table 1)
assumed the standard of practice in our institution through a
series of clinical validation studies over the last 10 years [4,
14–18]. Two mSv was around the lowest dose level explored
in previous studies regarding the diagnosis of appendicitis [3,
5].
For each of the 2.0-mSv scans and simulated scans (de-
tailed below), three image data sets were acquired from each
patient: thick (4-mm thickness with 3-mm interval) transverse
Eur Radiol
Table 1 CT imaging parameters
Imaging parameter
Manufacturer
Model
Scan range
Detector configuration (mm)
Tube potential (kVp)
Effective tube current-time product for an average
adult* (mAs)
Actual tube current-time product (mAs)
Rotation speed (s)
Pitch
Automatic tube current modulation
Iterative reconstruction technique
Transverse and coronal images
Thickness (mm)
Interval (mm)
Additional thin-section transverse images
Thickness (mm)
Interval (mm)
*
The average adult was assumed to have a 29-cm water-equivalent diameter
images, thick (4-mm thickness with 3-mm interval) coronal
images, and thin (2-mm thickness with 1-mm interval) trans-
verse images. We have previously reported the technical ad-
vantages of obtaining both thick and thin transverse images
[17, 19]. All transverse images were reconstructed using a
knowledge-based IR technique (IMR, level 2 of 3, Philips
Healthcare). Other reconstruction parameters, such as the field
of view, were identical across the four dose levels and between
the two different section thicknesses.
Simulated low-dose CT
From the helical projection data of the original 2.0-mSv scans,
we simulated 1.5-, 1.0-, and 0.5-mSv scans using a dose sim-
ulation tool that takes into account both photonic and electron-
ic noise. Specifically, the tool makes use of the conditional
variance identity to properly account for the variance of the
input high-dose data and allows for the inclusion of real sam-
ples of detector noise [7]. The fidelity of the simulation tool
has been extensively validated by robust phantom [7] and
animal [8] studies, which demonstrated excellent agreement
between simulated and original images in both objective and
subjective image evaluations.
Radiologists
Considering that appendicitis is a very common disease that
can be encountered in many hospitals in different settings, we
opted to involve both abdominal and general radiologists.
Specification
Philips Healthcare
iCT 256
From 4 cm above the liver dome to 1 cm below the
ischial tuberosity
128 × 0.625, double sampling
100
49
Varied according to patient size
0.5
0.99
Angular and longitudinal tube current modulation
IMR, level 2 of 3
4
3
2
1
Specifically, we involved three abdominal radiologists
[readers 1–3 (K.H.L., H.J.K., and J.Y.S.), each with 17-, 14-
and 3-year clinical experience, respectively] and three general
radiologists (readers 4–6 each with 2-year clinical experi-
ence). Except for reader 1 who had clinical and research ex-
perience of low-dose appendiceal CT over 10 years, the re-
maining five radiologists were invited from other hospitals
with limited experience in low-dose abdominal CT.
The overall experiment of step-wise non-inferiority testing
was performed separately for the three abdominal radiologists
and for the three general radiologists. We intended the two
radiologist groups to reproduce the two extremes of different
practice settings, ranging from an academic centre to a com-
munity hospital, in terms of the availability of abdominal
radiologists.
Image review session
A study-coordinating radiologist (J.H.P.), who was not one of
the readers, and a statistician (J.J.J.) carefully planned and
arranged the image review sessions in line with the step-
wise non-inferiority testing scheme. First, the readers
reviewed the 1.5-mSv images and then the 2.0-mSv images.
If the non-inferiority of 1.5 mSv to 2.0 mSv was not accepted,
the study would be terminated with the conclusion that
2.0 mSv is the lowest achievable dose. If the non-inferiority
was accepted, the study would continue to the next steps, the
review of 1.0-mSv and then 0.5-mSv images, while the non-
inferiority was tested against the 2.0-mSv CT at each step. To

ensure memory extinction, we separated the four image re-
view sessions by an interval of at least 5 months.
Image review
The readers were informed of the patient eligibility criteria and
clinical findings of individual patients including blood test
results; however, they were blinded to the original CT reports
and final diagnosis. For each case, the readers initially
reviewed the thick transverse and coronal images. If the
readers were not fully confident in their assessment, they op-
tionally reviewed the thin transverse images using a multi-
planar sliding-slab averaging technique [19]. The readers rat-
ed the likelihood of appendicitis (using a 5-point Likert scale)
and appendix visualisation score (using a 3-point Likert scale)
and rendered a potential alternative diagnosis. The grading
criteria are detailed in Table S1 (Supplemental Materials) [4].
Reference standards
For patients who underwent abdominal surgery, the final di-
agnosis was determined based on surgical and pathological
findings. Histopathological diagnosis of acute appendicitis
was defined as transmural neutrophil infiltration in the
appendiceal wall [20]. In patients who did not undergo sur-
gery, an independent radiologist with 4-year clinical experi-
ence adjudicated the final diagnosis based on medical records
and standardised telephone interviews conducted at least 3
months after the CT examination.
Statistical analysis
The study coordinator and the statistician planned all analyses
before data collection. There was no major change in the study
protocol after the study registration at ClinicalTrials.gov
(registration No. NCT02556983).
The primary end point was the pooled area under the
receiver-operating characteristic curve (AUC) of three radiol-
ogists in diagnosing appendicitis. As we mentioned previous-
ly, all experiments and analyses were performed separately for
the three abdominal radiologists and three general radiolo-
gists. We used a fixed-sequence procedure that allowed
family-wise type I error to remain constant at 0.05 through
the step-wise non-inferiority tests [21]. Non-inferiority was
established when the upper bound of the two-sided 95% con-
fidence interval (CI) for the difference in the pooled AUC of
three radiologists lay below the non-inferiority margin [22,
23]. The non-inferiority margin was set as 0.06 by assuming
the pooled AUC for 2.0-mSv CT as 0.98 [24] and by consid-
ering a difference of 0.06 as clinically acceptable. A previous
study [25] involving 46 abdominal radiologists showed that
the lower quartile of the individual radiologists’ AUC with 2-
mSv CT was 0.92.
Eur Radiol
The sample size was determined as 30 patients, using the
method proposed by Chen et al [26]. Based on data from a
previous study [24], the estimated covariance parameters of the
AUC estimates of different modalities for the same readers, of
different readers for the same modality, and of different readers
and different modalities were set as 5.74 × 10-6, 1.87 × 10-5, and
2.27 × 10-6, respectively; the estimated variance for interaction
between reader and modality was set at 1.07 × 10-5. With these
assumptions, 16 to 29 patients were required, when the preva-
lence of appendicitis ranged from 25% to 45%, to prove the non-
inferiority in the pooled AUC of three radiologists with 80%
statistical power.
Secondary end points were the diagnostic sensitivity and
specificity, measures related to the diagnostic confidence, and
alternative diagnosis. The diagnostic sensitivity and specific-
ity were calculated by considering a likelihood score of ap-
pendicitis ≥ 3 as positive for appendicitis [14]. Outcome mea-
sures regarding diagnostic confidence included the 5-point
likelihood score of appendicitis in diagnosing or ruling out
appendicitis, the number of Bindeterminate^ diagnoses (grade
3), and the 3-point visualisation score in patients not having
appendicitis. For alternative diagnoses, we counted the cases
of correctly suggested diagnosis. For the secondary end
points, no formal statistical comparisons were made because
of low power and multiple comparison issues. Instead, we
drew clustered stacked bar charts to illustrate the outcome
measures regarding diagnostic confidence.
Inter-observer agreement among all six readers was mea-
sured using Krippendorff’s α statistic. An α value close to 1
indicated high agreement, a value close to 0 indicated that the
agreement was due to chance, and a value close to -1 indicated
that there was systematic disagreement. iMRMC (Center for
Devices and Radiological Health, Food and Drug
Administration, Silver Spring, MD) and Stata/SE version
14.0 (Stata, College Station, TX) were used.
Results
Patient characteristics
From August through October 2015, 57 consecutive patients
were eligible for the study, and 30 patients (mean age ± stan-
dard deviation, 27 ± 8 years) finally participated in the study.
They were 13 males (27 ± 7 years) and 17 females (28 ± 8
years). There were no dropouts or missing data.
Patient characteristics are summarised in Table 2 including
clinical risk scores for appendicitis [27, 28]. The median dose-
length product was 131 mGy·cm (interquartile range, 121 to
158). The median time interval between CT to appendectomy
was 4.5 h (interquartile range, 3.8–5.5). Appendicitis was path-
ologically confirmed in nine patients, including five without
Eur Radiol

Table 2 Patient characteristics (n = 30) 0.017 (-0.070, 0.035), respectively. However, the non-
Characteristic Study sample inferiority of 0.5 mSv to 2.0 mSv failed to be proven [pooled
AUC difference, 0.045 (-0.071, 0.161)]. Therefore, 1.0 mSv
Age (years) 27 ± 8 was concluded as the lowest achievable dose (Fig. 2).
Females 28 ± 8 For all the six radiologists combined, the pooled AUC dif-
Males 27 ± 7 ferences between 2.0 vs. 1.5 mSv, 2.0 vs. 1.0 mSv, and 2.0 vs.
Sex 0.5 mSv were -0.013 (95% CI: -0.053, 0.028), -0.007 (-0.042,
Females 17 (57%) 0.028), and 0.030 (95% CI: -0.033, 0.095), respectively
Males 13 (43%) (Table 3).
Body mass index (kg/m2) 21.7 (20.3–26.9)
< 18.5 (underweight) 1 (3%) Secondary end points
18.5–24.9 (normal) 19 (63%)
≥ 25.0 (overweight or obese) 10 (33%) With 2.0-, 1.5-, 1.0-, and 0.5-mSv CT, the individual readers’
Diameter of the abdomen (cm) diagnostic sensitivity ranged from 78%–100%, 89%–100%,
Anteroposterior 17.9 (16.1–20.9) 78%–100%, and 78%–100%, respectively; the specificity
Lateral 33.0 (30.4–35.3) ranged from 81%–100%, 76%–100%, 86%–95%, and 90%–
Effective 24.1 (22.0–27.8) 95%, respectively (Table 3). Reader 1’s specificity tended to
Blood test results decrease with lower radiation doses: otherwise, there was no
White blood cells (103/mm3) 9.0 (6.8–11.7) notable decrease in the abdominal radiologists’ sensitivities or
Segmented neutrophils (%) 70.8 (64.6–83.0) specificities with lower radiation dose. The sensitivities of
C-reactive protein (mg/dl) 1.4 (0.3–3.6) readers 4 and 5 tended to be compromised with lower radia-
Alvarado score tion doses: otherwise, there were no such trends for the gen-
Low risk (0–4) 22 (73%)
eral radiologists’ specificities.
Intermediate risk (5–8) 8 (26%)
The decrease in the diagnostic confidence in ruling in ap-
High risk (9–10) 0 (0%)
pendicitis with lower radiation doses was more pronounced
for the general radiologists than for the abdominal radiolo-
Appendicitis inflammatory response score
gists. In ruling out appendicitis, the diagnostic confidence of
Low risk (0–4) 22 (73%)
readers 1 and 2 tended to be compromised with lower radia-
Intermediate risk (5–8) 7 (23%)
tion doses, while the other readers did not show such trend.
High risk (9–12) 1 (3%)
There was no notable increase of indeterminate interpretations
Radiation dose
with lower radiation doses for any of the readers. The normal
Volume CT dose index (mGy) 2.3 (2.1–2.7)
appendix visualisation tended to be compromised with lower
Size-specific dose estimate (mGy) 3.5 (3.4–3.7)
radiation doses for readers 1, 2, 4, and 5: this was more pro-
Dose-length product (mGy·cm) 130 (121–158)
nounced for readers 4 and 5 than for readers 1 and 2. (Fig. S1
Data are mean ± standard deviation, n (%), or median (interquartile range) in Supplemental Materials). The suggestion of correct alterna-
tive diagnosis overall tended to be mildly compromised with
perforation or gangrene, three with perforation, and one with lower radiation doses (Table 4).
gangrene. No patient experienced negative appendectomy. The Krippendorff’s α statistic values regarding the likeli-
hood of appendicitis for 2.0-, 1.5-, 1.0-, and 0.5-mSv CT were
0.78 (95% CI: 0.65, 0.89), 0.77 (0.63, 0.89), 0.84 (0.73, 0.95),
Pooled AUC and 0.75 (0.61, 0.89), respectively.

For the three abdominal radiologists, the non-inferiority of each

of 1.5-, 1.0-, and 0.5-mSv CT to 2.0-mSv CT was sequentially
accepted. The pooled AUC differences between 2.0 vs. 1.5 mSv,
2.0 vs. 1.0 mSv, and 2.0 vs. 0.5 mSv were 0.007 (95% CI: -
0.016, 0.030), 0.003 (-0.037, 0.044), and 0.017 (-0.016, 0.050),
respectively. Accordingly, 0.5 mSv was concluded as the lowest
achievable dose.
For the three general radiologists, the non-inferiority of each
of 1.5- and 1.0-mSv CT to 2.0-mSv CT was sequentially ac-
cepted. The pooled AUC differences between 2.0 vs. 1.5 mSv
and 2.0 vs. 1.0 mSv were -0.031 (95% CI: -0.101, 0.038) and -
Discussion
In this study, we systematically explored the lowest achievable
radiation dose for appendiceal CT obtained using IR. The radi-
ologist’s experience level was an important factor affecting the
degree of diagnostic impairment with lower radiation dose.
This is an expected but important observation, as appendicitis
is a very common disease encountered in various hospital set-
tings with different availabilities of experienced radiologists.
Based on our results, we recommend 0.5 mSv as the lowest
 Eur Radiol

Fig. 2 Differences in the area under the receiver-operating characteristic lines represent points of no difference. Dotted vertical lines represent a
curve (AUC) values between (a) 2.0- vs. 1.5-mSv CT, (b) 2.0- vs.1.0- predetermined non-inferiority margin of 0.06. *Multi-reader multi-case
mSv CT, and (c) 2.0- vs. 0.5-mSv CT. Data points and error bars repre- receiver-operating characteristic analyses
sent the point estimates and 95% confidence intervals (CIs). Solid vertical

achievable dose for hospitals where experienced abdominal The carcinogenic risk from diagnostic radiation is a highly
radiologists are always available. For hospitals with limited controversial issue. Large epidemiologic studies [29, 30] have
availability of experienced abdominal radiologists, 1.0 mSv shown that conventional radiation doses from common CT
would be more acceptable as the lowest achievable dose. applications are associated with carcinogenesis at least in

Table 3 Diagnostic performance

Radiation dose

2.0 mSv 1.5 mSv 1.0 mSv 0.5 mSv

AUC
Reader 1 1.00 1.00 0.97 0.97
Reader 2 0.97 0.98 1.00 0.98
Reader 3 1.00 0.98 1.00 0.98
Reader 4 0.97 1.00 0.99 0.94
Reader 5 0.98 0.92 0.97 0.88
Reader 6 0.86 0.97 0.91 0.85
Reader 1-3* 0.99 0.98 0.99 0.97
Reader 4-6* 0.94 0.97 0.95 0.89
All readers* 0.96 0.98 0.97 0.93
Sensitivity
Reader 1 100% (9/9) 100% (9/9) 89% (8/9) 100% (9/9)
Reader 2 100% (9/9) 100% (9/9) 100% (9/9) 100% (9/9)
Reader 3 100% (9/9) 100% (9/9) 100% (9/9) 100% (9/9)
Reader 4 100% (9/9) 100% (9/9) 100% (9/9) 89% (8/9)
Reader 5 100% (9/9) 89% (8/9) 89% (8/9) 78% (7/9)
Reader 6 78% (7/9) 89% (8/9) 78% (7/9) 78% (7/9)
Specificity
Reader 1 100% (21/21) 100% (21/21) 95% (20/21) 90% (19/21)
Reader 2 95% (20/21) 90% (19/21) 95% (20/21) 95% (20/21)
Reader 3 90% (19/21) 95% (20/21) 95% (20/21) 95% (20/21)
Reader 4 81% (17/21) 76% (16/21) 86% (18/21) 95% (20/21)
Reader 5 90% (19/21) 90% (19/21) 95% (20/21) 90% (19/21)
Reader 6 95% (20/21) 100% (21/21) 95% (20/21) 95% (20/21)

Data are the area under the receiver-operating characteristic curve or % (n/N)
AUC = Area under the receiver-operating characteristic curve
*
Multiple-reader multiple-case receiver-operating characteristic analyses
Eur Radiol

Table 4 Number of correct alternative diagnoses carcinogenesis in adults [31], given that annual natural back-
Radiation dose ground radiation mostly ranges over 1 mSv [32].
The feasibility of sub-mSv CT has been reported for high-
2.0 mSv 1.5 mSv 1.0 mSv 0.5 mSv contrast diagnostic tasks such as screening chest CT [33] or CT
for urinary stone disease [34], particularly using IR techniques.
Right colonic diverticulitis (n = 5)
However, such a radical dose reduction has been regarded as
Reader 1 5 5 3 4
difficult to achieve for low-contrast diagnostic tasks such as he-
Reader 2 4 4 3 3
patic nodule detection [35]. The diagnosis of appendicitis can be
Reader 3 4 4 2 4
considered a high-contrast diagnostic task in many of patients
Reader 4 2 2 4 2
because the normal or inflamed appendix contrasts with peri-
Reader 5 2 1 2 2 appendiceal fat. In a small portion of patients with little peri-
Reader 6 1 0 2 2 appendiceal fat, who are usually slender, ultrasound is more de-
Gastroenterocolitis (n = 4) sirable than CT as the initial imaging modality.
Reader 1 4 4 3 3 This was a proof-of-concept study purposed to explore the
Reader 2 4 4 3 4 feasibility of sub-mSv appendiceal CT. For the routine use of
Reader 3 4 4 3 2 sub-mSv appendiceal CT in practice, further validation studies
Reader 4 4 4 3 3 are needed preferably with larger sample sizes, prospective CT
Reader 5 3 2 1 1 interpretations by multi-site radiologists, and clinical end points
Reader 6 2 2 3 2 such as negative appendectomy rate and appendiceal perforation
Complicated adnexal cyst (n = 2) rate [4, 14, 18]. Our study results can provide important infor-
Reader 1 2 2 1 1 mation for justifying and designing such future clinical studies.
Reader 2 2 2 2 2 We did not formally test images reconstructed using filtered
Reader 3 2 2 2 1 back projections in this study. Based on our clinical and re-
Reader 4 2 2 2 2 search experience in low-dose appendiceal CT over the last
Reader 5 1 1 1 2 10 years [4, 14–19, 24], we believe that sub-mSv appendiceal
Reader 6 1 2 1 1 CT using filtered back projections results in unacceptable image
Pelvic inflammatory disease (n = 2) quality (Fig. 3). The use of IR is likely essential to properly
Reader 1 2 2 0 0 achieve sub-mSv appendiceal CT.
Reader 2 2 2 1 1
Reader 3 1 1 0 1
Reader 4 2 1 0 1
Reader 5 1 1 0 1
Reader 6 1 1 0 1
Ureteral stone (n = 1)
Reader 1 1 1 1 0
Reader 2 1 1 1 1
Reader 3 1 1 1 1
Reader 4 0 0 0 1
Reader 5 0 0 0 0
Reader 6 0 0 0 0
Mesenteric lymphadenitis (n = 1)
Reader 1 1 0 0 0
Reader 2 1 1 1 1
Reader 3 1 1 1 1
Reader 4 1 1 0 1
Reader 5 0 0 0 0
Reader 6 1 0 1 0

Data are the number of patients

children. Most experts now agree that CT radiation below
1 mSv can be regarded as having a negligible effect on human
Fig. 3 Contrast-enhanced transverse CT images from a 41-year-old fe-
male with appendicitis (arrows). (a) Original 2.0-mSv image reconstruct-
ed with filtered back projection (FBP). (b) Simulated 0.5-mSv image with
FBP. (c) Original 2.0-mSv image with iterative reconstruction (IR). (d)
Simulated 0.5-mSv image with IR

The strengths of our study include the careful study design
by adopting the dose-de-escalation concept from pharmaco-
logic trials. This is distinctly different from the conventional
study design, which demands each of the involved radiolo-
gists to review all images at all the tested radiation dose levels
[36]. The advantages of the former study design over the latter
are as follows. First, the step-wise scheme allowed a chance
for the potential avoidance of unnecessary reading sessions,
although this advantage was not realised in our study. Second,
the pre-specified fixed-sequence procedure could increase the
statistical power, allowing a smaller sample size. Our study
design may be applicable to future studies exploring the low-
est achievable radiation doses for different applications.
Our study had limitations. The lower-dose images were
simulations, not real clinical images. Dose simulation is the
only method enabling head-to-head comparison across differ-
ent doses in the same patients while avoiding repeated radia-
tion exposure to the study patients. However, although the
dose simulation tool was based on sound CT physics and
has been validated extensively, the simulated images are not
perfectly identical to real clinical images. Second, the sample
size was too small to perform statistical analyses for the sec-
ondary end points. The diagnostic sensitivity and specificity,
as well as clinical outcomes, of sub-mSv appendiceal CT need
further validation, as we mentioned previously. Third, as only
a single CT machine was used in this study, our recommen-
dations regarding the lowest achievable dose may not be
generalisable to different CT machines. Fourth, like other
studies investigating the diagnosis of appendicitis, our study
was subject to verification bias.
In conclusion, 1.0-mSv appendiceal CT was non-inferior to
2.0-mSv CT in terms of diagnostic performance for both ab-
dominal and non-abdominal radiologists. The 0.5-mSv
appendiceal CT was non-inferior only for expert abdominal
radiologists. Further validation studies with a larger sample
size and clinical end points are needed for the wide adoption
of sub-mSv appendiceal CT.
Acknowledgements We thank Kyung Hwa Han in Yonsei University
College of Medicine for her advice on sample size calculation.
Funding This research was supported by grants from the Korea Health
Technology R&D Project through the Korea Health Industry
Development Institute (KHIDI), funded by the Ministry of Health &
Welfare, Republic of Korea (no. HI16C0451) and SNUBH Research
Fund (no. 02-2015-030).
Compliance with ethical standards
Guarantor The scientific guarantor of this publication is Kyoung Ho
Lee.
Conflict of interest One of the co-authors (Amar C. Dhanantwari) is an
employee of Philips. He contributed to imaging processing regarding
radiation dose simulation and manuscript editing, but did not interfere
Eur Radiol
with the medical interpretation proposed in this study. Otherwise, there
are no conflicts of interest to declare.
Statistics and biometry One of the authors has significant statistical
expertise.
Informed consent Written informed consent was obtained from all
subjects (patients) in this study.
Ethical approval Institutional Review Board approval was obtained.
Methodology
• prospective
• diagnostic study
• performed at one institution
References
1. Drake FT, Florence MG, Johnson MG et al (2012) Progress in the
diagnosis of appendicitis: a report from Washington State's Surgical
Care and Outcomes Assessment Program. Ann Surg 256:586–594
2. Park JH (2014) Diagnostic imaging utilization in cases of acute
appendicitis: multi-center experience. J Korean Med Sci 29:1308–
1316
3. Yun SJ, Ryu CW, Choi NY, Kim HC, Oh JY, Yang DM (2017)
Comparison of low- and standard-dose CT for the diagnosis of
acute appendicitis: a meta-analysis. Am J Roentgenol 208:W198–
W207
4. Kim K, Kim YH, Kim SY et al (2012) Low-dose abdominal CT for
evaluating suspected appendicitis. N Engl J Med 366:1596–1605
5. Aly NE, McAteer D, Aly EH (2016) Low vs. standard dose com-
puted tomography in suspected acute appendicitis: Is it time for a
change? Int J Surg 31:71–79
6. Willemink MJ, Leiner T, de Jong PA et al (2013) Iterative recon-
struction techniques for computed tomography part 2: initial results
in dose reduction and image quality. Eur Radiol 23:1632–1642
7. Zabic S, Wang Q, Morton T, Brown KM (2013) A low dose sim-
ulation tool for CT systems with energy integrating detectors. Med
Phys 40:031102
8. Muenzel D, Koehler T, Brown K et al (2014) Validation of a low
dose simulation technique for computed tomography images. PLoS
One 9:e107843
9. Mei K, Kopp FK, Bippus R et al (2017) Is multidetector CT-based
bone mineral density and quantitative bone microstructure assess-
ment at the spine still feasible using ultra-low tube current and
sparse sampling? Eur Radiol. https://doi.org/10.1007/s00330-017-
4904-y
10. Dowlati A, Robertson K, Radivoyevitch T et al (2005) Novel phase
I dose de-escalation design trial to determine the biological modu-
latory dose of the antiangiogenic agent SU5416. Clin Cancer Res
11:7938–7944
11. Bossuyt PM, Reitsma JB, Bruns DE et al (2015) STARD 2015: An
updated list of essential items for reporting diagnostic accuracy
studies. Radiology 277:826–832
12. Deak PD, Smal Y, Kalender WA (2010) Multisection CT protocols:
sex- and age-specific conversion factors used to determine effective
dose from dose-length product. Radiology 257:158–166
13. Martin CJ, Sookpeng S (2016) Setting up computed tomography
automatic tube current modulation systems. J Radiol Prot 36:R74–
R95
14. Ahn S (2014) LOCAT (Low-dOse Computed tomography for
Appendicitis Trial) comparing clinical outcomes following low-
Eur Radiol
vs standard-dose computed tomography as the first-line imaging
test in adolescents and young adults with suspected acute appendi-
citis: study protocol for a randomized controlled trial. Trials 15:28
15. Joo SM, Lee KH, Kim YH et al (2009) Detection of the normal
appendix with low-dose unenhanced CT: use of the sliding slab
averaging technique. Radiology 251:780–787
16. Kim SY, Lee KH, Kim K et al (2011) Acute appendicitis in young
adults: low- versus standard-radiation-dose contrast-enhanced ab-
dominal CT for diagnosis. Radiology 260:437–445
17. Seo H, Lee KH, Kim HJ et al (2009) Diagnosis of acute appendicitis
with sliding slab ray-sum interpretation of low-dose unenhanced
CT and standard-dose IV contrast-enhanced CT scans. Am J
Roentgenol 193:96–105
18. LOCAT group (2017) Low-dose CT for the diagnosis of appendi-
citis in adolescents and young adults (LOCAT): a pragmatic,
multicentre, randomised controlled non-inferiority trial. Lancet
Gastroenterol Hepatol 2:793–804
19. Lee YJ, Kim B, Ko Y et al (2015) Low-dose (2-mSv) CT in ado-
lescents and young adults with suspected appendicitis: advantages
of additional review of thin sections using multiplanar sliding-slab
averaging technique. Am J Roentgenol 205:W485–W491
20. Rosai J, Ackerman LV (2011) Appendix. In: Rosai J (ed) Rosai and
Ackerman's surgical pathology. Mosby, St. Louis, pp 714–718
21. Westfall PH, Krishen A (2001) Optimally weighted, fixed sequence
and gatekeeper multiple testing procedures. J Stat Plan Inference
99:25–40
22. Gallas BD (2006) One-shot estimate of MRMC variance: AUC.
Acad Radiol 13:353–362
23. Obuchowski NA, Gallas BD, Hillis SL (2012) Multi-reader ROC
studies with split-plot designs: a comparison of statistical methods.
Acad Radiol 19:1508–1517
24. Park JH, Kim B, Kim MS et al (2016) Comparison of filtered back
projection and iterative reconstruction in diagnosing appendicitis at
2-mSv CT. Abdom Radiol 41:1227–1236
25. Yang HK, Ko Y, Lee MH et al (2015) Initial performance of radi-
ologists and radiology residents in interpreting low-dose (2-mSv)
appendiceal CT. Am J Roentgenol 205:W594–W611
26. Chen W, Petrick NA, Sahiner B (2012) Hypothesis testing in non-
inferiority and equivalence MRMC ROC studies. Acad Radiol 19:
1158–1165
27. Alvarado A (1986) A practical score for the early diagnosis of acute
appendicitis. Ann Emerg Med 15:55–64
28. Andersson M, Andersson RE (2008) The appendicitis inflammato-
ry response score: a tool for the diagnosis of acute appendicitis that
outperforms the Alvarado score. World J Surg 32:1843–1849
29. Pearce MS, Salotti JA, Little MP et al (2012) Radiation exposure
from CT scans in childhood and subsequent risk of leukaemia and
brain tumours: a retrospective cohort study. Lancet 380:499–505
30. Mathews JD, Forsythe AV, Brady Z et al (2013) Cancer risk in 680,
000 people exposed to computed tomography scans in childhood or
adolescence: data linkage study of 11 million Australians. BMJ
346:f2360
31. McCollough CH, Chen GH, Kalender W et al (2012) Achieving
routine submillisievert CT scanning: report from the summit on
management of radiation dose in CT. Radiology 264:567–580
32. United Nations Scientific Committee on the Effects of Atomic
Radiation. Sources and effects of ionizing radiation. Medical radi-
ation exposures, annex A (2008) Report to the General Assembly
with annexes. United Nations, New York
33. Huber A, Landau J, Ebner L et al (2016) Performance of ultralow-
dose CT with iterative reconstruction in lung cancer screening: lim-
iting radiation exposure to the equivalent of conventional chest X-
ray imaging. Eur Radiol 26:3643–3652
34. Fontarensky M, Alfidja A, Perignon R et al (2015) Reduced radia-
tion dose with model-based iterative reconstruction versus standard
dose with adaptive statistical iterative reconstruction in abdominal
CT for diagnosis of acute renal colic. Radiology 276:156–166
35. Solomon J, Marin D, Roy Choudhury K, Patel B, Samei E (2017)
Effect of radiation dose reduction and reconstruction algorithm on
image noise, contrast, resolution, and detectability of subtle
hypoattenuating liver lesions at multidetector CT: filtered back pro-
jection versus a commercial model-based iterative reconstruction
algorithm. Radiology 284:777–787
36. Callahan MJ, Kleinman PL, Strauss KJ et al (2015) Pediatric CT
dose reduction for suspected appendicitis: a practice quality im-
provement project using artificial Gaussian noise—part 1, computer
simulations. Am J Roentgenol 204:W86–W94