The International Journal of Cardiovascular Imaging

https://doi.org/10.1007/s10554-018-1303-7

ORIGINAL PAPER

Normal pulmonary artery and branch pulmonary artery sizes

in children
S. Bruce Greenberg1   · Sean M. Lang1 · C. Heath Gauss1 · Shelly Y. Lensing1 · Sumera Ali2 · Karen A. Lyons1

Received: 1 November 2017 / Accepted: 11 January 2018

© Springer Science+Business Media B.V., part of Springer Nature 2018

Abstract
To establish standards for pulmonary artery and branch pulmonary artery (PA and BPA) effective diameter (ED) and cross-
sectional area (CSA) by using computed tomography (CT) data in children of a wide range of sizes and investigate the
roundness of arteries. The ED (average of short and long axes) and CSA for the PA and BPA were measured using 1-mm
collimation double-oblique reconstructions. Ordinary least squares regression was used to investigate models with various
functional forms that related ED and CSA to patient size. Aspect ratio (AR), the short axis divided by long axis, was meas-
ured to evaluate roundness. The ideal diameter derived from CSA measurements was compared to ED, short axis, and long
axis measurements. 108 CT examinations were analyzed in children without reason for abnormal PA size who ranged in age
from 0 to 18 years (mean, 10.9 years; SD, 5.9 years). Interrater reliability was excellent. Data were modeled using a natural
log-transformed response variable and a linear term for height as the independent variable. AR for the PA, right pulmonary
artery, and left pulmonary artery measured < 0.9 for 38, 55, and 37%, respectively, indicating that many arteries are not round.
Ideal diameter was not significantly different than ED but was for short- and long-axis diameter measurements. Normal ED
and CSA for PA and BPA were determined for children of different sizes. Measurements outside of the normal range are
consistent with dilatation or stenosis. Single diameter techniques are likely to introduce error.

Keywords  Pulmonary artery · Children · Normative · Computed tomography

Introduction
Normal standards for the pulmonary artery (PA) and branch
pulmonary arteries in children are necessary to determine
stenosis or dilatation. Published standards for infants
(1 month to 1 year old), children (birth to 12 years old), ado-
lescents (13–17 years old), and young adults (18–24 years
old) (hereafter, individuals in these age groups will be
referred to collectively as “children”) utilize echocardio-
graphic data. Echocardiographic standards for the PA and
branch pulmonary arteries rely on planar measurements
using a single parasternal, high left parasternal, or supra-
sternal short axis view rather than on transverse measure-
ments [1]. A single diameter measurement is performed, and
the assumption is made that the plane precisely bisects the
artery and that the measured vessel is perfectly round. Cur-
rently, no standards exist utilizing cross-sectional measure-
ments of the PA and branch pulmonary arteries in children.
Our primary purpose was to evaluate the effective diameter
(ED) and cross-sectional area (CSA) for the PA and branch
pulmonary arteries in children to establish normative data.
Our secondary purpose was to investigate if single diameter
measurements as currently performed by echocardiography
are justified. A single diameter measurement to characterize
vessel size assumes perfect roundness.
Materials and methods

* S. Bruce Greenberg The study was approved by the institutional review board.
greenbergsbruce@uams.edu Informed consent was waived by the institutional review
board. A retrospective study was performed to determine
1
Arkansas Children’s Hospital, University of Arkansas the normal EDs and CSAs for the PA and branch pulmonary
for Medical Sciences, Slot #105, 1 Children’s Way,
Little Rock, AR 72202, USA arteries in children of varying sizes.
2
Children’s Hospital of Boston, Boston, MA, USA

13
Vol.:(0123456789)

The inclusion criteria for children evaluated during a
10-month period between February and November 2016
included the following: (a) chest computed tomography
(CT) performed with intravenous contrast material during a
10-month period between February and November 2016 at
a tertiary care pediatric hospital, (b) record of patient height
determined by length measurement using a paper tape meas-
ure in infants and toddlers and a stadiometer for older chil-
dren, (c) patient age between birth and 18 years of age and
(d) CT images archived using 0.5-mm collimation.
Exclusion criteria included the following: (a) history
or CT evidence of co-morbidities that might predispose to
abnormal pulmonary arterial size such as congenital heart
disease, significant pulmonary disease, chronic hypoxemia,
pulmonary hypertension, scoliosis or neuromuscular abnor-
mality predisposing to restrictive lung disease, and chromo-
somal abnormality, (b) inadequate CT examination, and (c)
patient size less than the first percentile. Only the first CT
scan was included in the study if multiple examinations were
performed during the 10-month study period.
108 of 164 screened chest CT examinations met the
study inclusion criteria. Fifty-six studies were subsequently
excluded. Reasons for study exclusion included: severe pul-
monary or pleural disease (32), congenital heart disease (9),
patient size below the 1st percentile (9), and genetic disorder
(2). One each was excluded for sickle cell anemia, severe
Fig. 1  Sagittal (a) and axial
(b) reconstructions are used to
create a double oblique image
perpendicular to the pulmonary
artery, allowing for short and
long axis diameter measure-
ments (c) and cross-sectional
area measurements (d) using
electronic calipers
13
The International Journal of Cardiovascular Imaging
lymphadenopathy, pulmonary hypertension and extensive
injuries of non-accidental trauma. The BSA for each patient
was calculated by using the method of Haycock et al. [2].
The method was selected because a convenient BSA cal-
culator is available for free on the Internet (http://www.
ultradrive.com/bsac.htm) and has been used in our depart-
ment for many years to determine BSA.
All study examinations were performed with either of
two CT scanners (Aquilion 64 or Aquilion ONE; Toshiba,
Otawara-shi, Japan). The intravenous contrast material
used for children who weighed < 10 kg was iohexol (Omni-
paque 240; GE Healthcare, Little Chalfont, Buckingham-
shire, England), 240 mg of iodine/ml, at a dose of 3 ml/
kg of body weight. For children who weighed > 10 kg,
iohexol (Omnipaque 300; GE Healthcare), 300  mg of
iodine/ml, at a dose of 2 ml/kg, with a maximum dose
of 100 ml, was used. Images were reconstructed using
axial 0.5-mm collimation. Multiplanar reformations were
created by using a workstation (Vitrea FX; Vital Images,
Minneapolis, Minn). All multiplanar reformations were
double-oblique reconstructions using vascular window
(820–1000) and level (200–250) settings obtained perpen-
dicular to the PA (Fig. 1), right pulmonary artery (RPA),
and left pulmonary artery (LPA). The ED at each level
was determined by averaging the long and short diameter
measurements. CSA was measured by planimetry. The
The International Journal of Cardiovascular Imaging
PA was measured midway between the pulmonary valve
and PA bifurcation. The RPA was measured dorsal to the
ascending aorta and the LPA midway between the LPA
origin and bifurcation into left upper and lower pulmonary
arteries. Images from all studies were independently read
by one pediatric radiologist and one pediatric cardiologist.
Each reader independently created 1-mm thick double-
oblique multiplanar reconstructions from the original axial
data sets and used electronic calipers to measure long and
short diameters that were used to determine the ED. Elec-
tronic calipers were also used to perform planimetry to
determine the CSA.
Statistical methods
Intraclass correlation coefficients were estimated in order
to assess the interrater reliability of diameter measure-
ments and CSA measurements from two readers [3]. The
corresponding two measurements were then averaged for
further analysis. A modeling approach similar to the one
used in Hegde et al. [4] was utilized; three sets of ordinary
least square regression models were initially investigated,
and non-significant terms were dropped from the models.
Examples for ED of the PA follow:
(1) PA ED = β0 + β1 × height + β2 × height2 + β3 × height3 + 
β4 × height4
(2) loge (PA ED) = β0 + β1 × height + β2 × height2 + β3 × hei
ght3 + β4 × height4
(3) loge (PA ED) = β 0 + β 1 × loge (height) + β 2 × [loge
(height)]2 + β3 × [loge (height)]3 + β4 × [loge (height)]4,
where β0 is the intercept and β1–β4 are the slopes estimated
by the model for linear, quadratic, cubic, and quartic terms,
and height (cm) is used to predict l­ oge (PA ED), i.e., natural
log of ED in mm of the PA. ­R2 was used to determine the
best models. Parameter estimates from the resulting esti-
mated regression equations and the corresponding mean
squared error estimates were used to calculate z-scores in
order to ascertain normal PA and branch pulmonary artery
(BPA) sizes in children. Analyses were performed by using
software (SAS 9.4; SAS Institute, Cary, NC).
Additionally, to determine roundness, the aspect ratio
(AR) for PA and BPA measurements was defined as the
short axis diameter divided by the long axis diameter. A
perfectly round cross-section is defined by the upper bound
of AR = 1. The percentage of arteries with AR < 0.9 was cal-
culated. CSA measurements were used to calculate the ideal
diameters based on the area formula for a circle: A = πr2
where A is area and r is 1/2 of ideal diameter. Paired t tests
were then used to compare the calculated ideal diameters
with short, long, and ED measurements.
Results
Patient characteristics
The study population included 108 children, 64 boys and
44 girls. The median age was approximately 12 years and
8 months, and the age ranged from 3 months to 18 years
[mean 10.9 years; standard deviation (SD), 5.9  years].
The age distribution is summarized in Table 1. The mean
BSA was 1.34 ± 0.59  m 2 (SD). The mean height was
141.0 ± 34.9 cm (SD). The most common clinical indica-
tions for chest CT scan were oncology follow-up (71 chil-
dren), trauma (14 children), and mediastinal adenopathy (6
children).
Interrater reliability
Interrater reliability between the two readers regarding both
ED and CSA measurements was excellent, indicating that
the technique for measuring ED and CSA was reproduc-
ible. Intraclass correlation coefficients for the PA, RPA, and
LPA effective diameter measurements were 0.937, 0.956,
and 0.938, respectively. Intraclass correlation coefficients for
the PA, RPA, and LPA cross-sectional area measurements
were: 0.943, 0.924, and 0.920, respectively.
Regression models
The relationship of CSA and ED for the pulmonary and
branch pulmonary arteries to patient characteristics (age,
height, weight, and BSA) was considered separately since
age, height, weight, and BSA were highly correlated (all
pairwise correlations > 0.87). The model with a natural log-
transformed response variable and a linear term for height
as the only independent variable either had the largest ­R2 or
had an ­R2 within 3% of the largest ­R2 for applicable mod-
els. Therefore, for simplicity, that model was chosen as the
“best” model for each response variable.
Table 1  Age distribution of Age range (years) n (%)
patients (N = 108)
0–2 13 (12)
3–5 17 (16)
6–10 17 (16)
11–14 22 (20)
15–16 20 (19)
17–18 19 (18)
N total number of patients, n
number of patients in each age
category
13
 The International Journal of Cardiovascular Imaging

Next, an interaction effect between sex and height was
assessed, and there was a significant interaction effect
(p < 0.05) for the RPA, but not for the PA or the LPA. For the
RPA, as height increases, the increase in the estimated mean
CSA or ED is greater for females than for males. The for-
mulae to calculate the predicted CSA and ED on the natural
log scale for the pulmonary and branch pulmonary arteries
are summarized in Table 2.
To assess whether a normal child’s CSA or ED for the
PA or either of the branch pulmonary arteries is statistically
unusual or not, a z-score can be calculated for the artery of
interest using the following formula:
� � � � ��� √
Z = loge Xobs − loge Xpred MSE.
­ obs is the observed CSA or ED, and X
In this formula, X ­ pred
is the predicted CSA or ED. Since the formulae in Table 2
yield predicted values on the natural log scale, the applica-
ble predicted value can be substituted for the “loge ­(Xpred)”
expression in the z-score formula above. A statistical prop-
erty of the z-scores is that they will follow an approximate
normal distribution with a mean of 0 and SD of 1. A posi-
tive (or negative) z-score indicates the number of SDs that
the observed value is above (or below) the estimated mean.
As an example, a z-score of 1 would indicate that the cor-
responding observed value is 1 SD above the mean. Also,
it can be shown that for a normal distribution, one would
expect that 68.3% of the values from such a distribution
would fall in the interval of the mean + 1 SD, 95.4% of the
values would fall in the interval of the mean + 2 SDs, and
99.7% of the values would fall in the interval of the mean + 3
SDs. Rather than calculate z-scores, the data can be graphi-
cally displayed (Figs. 2, 3, 4).
Aspect ratios
The median AR along with the corresponding 25th and
75th percentiles for the PA, RPA, and the LPA were the
following: 0.921 (0.874, 0.954), 0.893 (0.828, 0.938), and
0.918 (0.875, 0.957), respectively. Additionally, the PA,
RPA, and LPA aspect ratios measured < 0.9 for 38, 55,
and 37% of the patient population, respectively. This indi-
cates that a single diameter measurement used to estimate
artery CSA, as commonly used in echocardiography, could
be associated with error. Mean ideal diameters, calculated
from CSA planimetry measures, did not significantly differ
from mean EDs for the PA (p = 0.269), RPA (p = 0.511), or
LPA (p = 0.515) indicating the ED is a practical proxy for
ideal diameter (Table 3). Mean short axis and mean long
axis measurements were significantly different (p < 0.001)
compared to mean ideal diameter for all locations indicating
that a single diameter measurement is not an effective proxy
for ideal diameter.
Discussion
A prerequisite for identifying abnormal is to first determine
normal. This study is the first to our knowledge to use multi-
planar reconstructions of CT scans to quantify normative PA
data. The normal range of PA and BPA size measurements
in children was determined in a wide range of child sizes in
our study. The regression equations derived from the meas-
urements allow for the calculation of z-scores. A z-score
calculator using a spreadsheet (Excel; Microsoft, Redmond,
Wash) is now used at our institution, with only the patient
height, sex, and either ED or CSA of the PA or BPA entered
to determine the z-score. Measurements outside of the nor-
mal ranges are consistent with dilatation or narrowing.
Significant gender differences were only identified for the
RPA. Perhaps gender differences for the PA and LPA could
have been identified by studying a larger cohort. It may also
be possible that when height rather than body surface area
is used to normalize size that gender is less important. Right
and left BPA size measurements were very similar. Slightly
greater flow to the larger right lung and expected slightly

Table 2  Formulae to calculate Measurement Formula √MSE R2

the predicted cross-sectional
area or effective diameter on the CSA Loge (PA CSA) = 4.1569 + 0.0120*height 0.2433 0.750
natural log scale (N = 108)
Loge (RPA CSA) = 2.7816 + 0.0162*height + 0.4989 0.2620 0.788
*male − 0.0032*height*male
Loge (LPA CSA) = 3.1379 + 0.0142*height 0.2399 0.811
ED Loge (PA ED) = 2.1844 + 0.0061*height 0.1216 0.758
Loge (RPA ED) = 1.4572 + 0.0084*height + 0.2921* 0.1283 0.803
male − 0.0019*height*male
Loge (LPA ED) = 1.6936 + 0.0071*height 0.1195 0.812

Loge = natural log, √MSE = square root of the mean square error; male = 1 when patient is male and
male = 0 when patient is female (applies only to RPA)
N total number of patients, CSA cross-sectional area, ED effective diameter, PA pulmonary artery, RPA
right pulmonary artery, LPA left pulmonary artery

13
The International Journal of Cardiovascular Imaging

Fig. 2  Pulmonary artery effec-

tive diameter versus height
(a) and cross-sectional area
versus height (b). Solid line:
estimated effective diameter
corresponding to the estimated
mean natural log-transformed
effective diameter for a given
height. Dotted lines: estimated
effective diameters correspond-
ing to natural log-transformed
effective diameters that are 1, 2,
or 3 SDs above (Z = 1, 2, or 3)
or below (Z = − 1, − 2, or − 3)
the estimated mean natural log-
transformed effective diameter
for a given height. Open circle:
observed effective diameters

larger size RPA size was not detected. The slight difference
in expected ED was too small in children to be detected.
A single diameter measurement derived from a short
axis plane echocardiogram is the current standard for
measuring PA, RPA, and LPA size. A recent review of
echocardiographic methods showed a general lack of stand-
ardization in technique and a paucity of infants included in
the studies [5]. These methods assume that the plane per-
fectly bifurcates the vessel of interest and that the vessel is
perfectly round. Our study found that, in fact, the PA and

13

Fig. 3  Right pulmonary artery effective diameter for females (a) and
males (b) and cross-sectional area of females (c) and males (d) versus
height. Solid line: estimated effective diameter corresponding to the
estimated mean natural log-transformed effective diameter for a given
height. Dotted lines: estimated effective diameters corresponding to
branch pulmonary arteries are frequently oval in shape. The
AR was < 0.9 in more than a third of PA and left pulmonary
arteries and < 0.9 in > 50% of right pulmonary arteries in the
study. Measurement of a single diameter to characterize PA
or BPA size has error inherent in the technique. Short and
long axis measurements were significantly different from
ideal diameter measurements indicating that the error inher-
ent in a single diameter measurement may be consequential.
Greater precision in vascular measurements may be in more
demand in the future since personalized medicine for patient
care is likely to include 3D printed phantoms and stents [6].
CT and magnetic resonance imaging three-dimensional
data sets are ideal for creating multiplanar reconstruc-
tions perpendicular to the PA and branch pulmonary arter-
ies for accurate measurements of diameter or CSA. Axial
13
The International Journal of Cardiovascular Imaging
natural log-transformed effective diameters that are 1, 2, or 3 SDs
above (Z = 1, 2, or 3) or below (Z = − 1, − 2, or − 3) the estimated
mean natural log-transformed effective diameter for a given height.
Open circle: observed effective diameters
reconstructions introduce error in measuring the true size
because of the normal oblique course of the PA and branch
pulmonary arteries. Few reports of the normal PA and BPA
size in children determined by CT exist. Akay et al. [7] and
Bayinder et al. [8] correlated measurements of the PA and
branch pulmonary arteries performed on axial reconstruc-
tions with child age. Measurements did not take into account
patient size, and only a single diameter was measured. The
PA diameter measurement was of the longest transverse
diameter. No measurements were performed using multipla-
nar reconstructions perpendicular to the vessels of interest.
We were unable to find reports of MRI to determine normal
PA and BPA size in children.
A limitation of our study was that electrocardiographic
gating is not routinely performed for chest CT imaging. As
The International Journal of Cardiovascular Imaging

Fig. 4  Left pulmonary artery

effective diameter (a) and
cross-sectional area (b) versus
height. Solid line: estimated
effective diameter corre-
sponding to the estimated
mean natural log-transformed
effective diameter for a given
height. Dotted lines: estimated
effective diameters correspond-
ing to natural log-transformed
effective diameters that are 1, 2,
or 3 SDs above (Z = 1, 2, or 3)
or below (Z = − 1, − 2, or − 3)
the estimated mean natural log-
transformed effective diameter
for a given height. Open circle:
observed effective diameters

such, the measurements reflect neither end-systole nor end-
diastole. Some image blurring is inherent in non-cardiac
gated studies. Intermediate ED or cross-section area rather
than maximum and minimum measurements are obtained.
The range of measurements will include this variation and
may increase the SD for the PA and branch pulmonary
arteries. However, this is also true of the technique most
likely to be used when performing chest CT imaging.
Another limitation of the study is that it only included 108
patients.
Interest in determining the range of normal sizes of the
PA and branch pulmonary arteries in children has centered

13
 The International Journal of Cardiovascular Imaging

Table 3  Comparisons to ideal diameter (N = 108) having objective data for a healthy population is important
Ideal diameter versus Mean difference (SD) p ­Valuea for clinical evaluation.

MPA Compliance with ethical standards 

 Minimum diameter 1.10 (0.97) < 0.001
 Maximum diameter − 1.20 (1.17) < 0.001 Conflict of interest  All authors declare that they have no conflict of
 Effective diameter − 0.05 (0.47) 0.269 interest.
RPA
 Minimum diameter 1.07 (0.99) < 0.001
 Maximum diameter − 0.99 (1.18) < 0.001 References
 Effective diameter 0.04 (0.67) 0.511
1. Lopez L, Colan SD, Frommelt PC, Ensing GJ, Kendall K, You-
LPA
noszai AK, Lai WW, Geva T (2010) Recommendations for
 Minimum diameter 0.68 (0.73) < 0.001 quantification methods during the performance of a pediatric
 Maximum diameter − 0.74 (0.65) < 0.001 echocardiogram: a report from the pediatric measurements writ-
 Effective diameter − 0.03 (0.46) 0.515 ing group of the American Society of Echocardiography Pediatric
and Congenital Heart Disease Council. J Am Soc Echocardiogr
Mean values represent mean differences between ideal diameter and 23:465–495. https://doi.org/10.1016/j.echo.2010.03.019
the parameter of interest such that negative values indicate the mean 2. Haycock GB, Schwartz GJ, Wisotsky DH (1978) Geometric
ideal diameter is smaller method for measuring body surface area: a height-weight formula
validated in infants, children, and adults. J Pediatr 93(1):62–66
N total number of patients, MPA pulmonary artery, RPA right pulmo-
3. Shrout PE, Fleiss JL (1979) Intraclass correlations: uses in assess-
nary artery, LPA left pulmonary artery, SD standard deviation
a
ing rate reliability. Psychol Bull 86(2):420–428
 Paired t test 4. Hegde SV, Lensing SY, Greenberg SB (2015) Determining the
normal aorta size in children. Radiology 274(3):859–865. https://
doi.org/10.1148/radiol.14140500
upon congenital heart disease. The McGoon ratio comparing
5. Cantinotti M, Sdcalese M, Molinaro S, Murzi B, Passino C
the combined diameters of the branch pulmonary arteries (2012) Limitations of current echocardiographic nomograms for
to the descending aorta at the diaphragm assumed a nor- left ventricular, valvular, and arterial dimensions in children: a
mal descending aorta diameter. The Nakata index combines critical review. J Am Soc Echocardiogr 25:142–152. https://doi.
org/10.1016/j.echo.2011.10.016
CSAs of the branch pulmonary arteries and normalizes the
6. Meess KM, Izzo RL, Dryjski ML, Curl RE, Harris LM, Springer
value by dividing by body surface area. The CSA measure- M, Siddiqui AH, Rudin S, Ionita CN (2017) 3D printed abdominal
ments are typically based on single diameter measurements. aortic aneurysm phantom for image guided surgical planning with
More recently, z-score measurements have replaced the a patient specific fenestrated endovascular graft system. Proc SPIE
Int Soc Opt Eng. https://doi.org/10.1117/12.2253902
McGoon ratio and Nakata index for clinical management.
7. Akay HO, Ozmen CA, Bayrak AH, Senturk S, Katar S, Nazaro-
The current study extends this trend by providing a more glu H, Taskesen M (2009) Diameters of normal thoracic vascular
precise method of measurement than previously reported. structures in pediatric patients. Surg Radiol Anat 31:801–807.
The formulas used to help provide a normal range of val- https://doi.org/10.1007/s00276-009-0525-8
8. Bayinder P, Bayraktaroglu S, Ceylan N, Savas R, Alper HH
ues for ED and CSA are only valid for the range of data (2016) Multidetector computed tomographic assessment of the
analyzed, but can be applied to most pediatric patients that normal diameters for the thoracic aorta and pulmonary arteries in
a clinician is likely to encounter. Although a small percent- infants and children. Acta Radiol 57(10):1261–1267. https://doi.
age of normal children will have values outside of 2 SDs, org/10.1177/0284185115622074

13