Note: This copy is for your personal non-commercial use only.

To order presentation-ready
Imaging copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

Potential for Change in US

Diagnosis of Hip Dysplasia
Original Research  n  Pediatric

Solely Caused by Changes in

Probe Orientation: Patterns of
Alpha-angle Variation Revealed by
Using Three-dimensional US1
Jacob L. Jaremko, MD, PhD, FRCPC
Purpose: To use three-dimensional (3D) ultrasonography (US) to
Myles Mabee, BSc
quantify the alpha-angle variability due to changing probe
Vimarsha G. Swami, BSc orientation during two-dimensional (2D) US of the infant
Lucy Jamieson, MD, FRCPC hip and its effect on the diagnostic classification of devel-
Kelvin Chow, BSc opmental dysplasia of the hip (DDH).
Richard B. Thompson, PhD
Materials and In this institutional research ethics board–approved pro-
Methods: spective study, with parental written informed consent,
13-MHz 3D US was added to initial 2D US for 56 hips in
35 infants (mean age, 41.7 days; range, 4–112 days), 26
of whom were female (mean age, 38.7 days; range, 6–112
days) and nine of whom were male (mean age, 50.2 days;
range, 4–111 days). Findings in 20 hips were normal at
the initial visit and were initially inconclusive but normal-
ized spontaneously at follow-up in 23 hips; 13 hips were
treated for dysplasia. With the computer algorithm, 3D
US data were resectioned in planes tilted in 5° increments
away from a central plane, as if slowly rotating a 2D US
probe, until resulting images no longer met Graf quality
criteria. On each acceptable 2D image, two observers
measured alpha angles, and descriptive statistics, includ-
ing mean, standard deviation, and limits of agreement,
were computed.

Results: Acceptable 2D images were produced over a range of

probe orientations averaging 24° (maximum, 45°) from
the central plane. Over this range, alpha-angle variation
was 19° (upper limit of agreement), leading to alteration
of the diagnostic category of hip dysplasia in 54% of hips
scanned.

Conclusion: Use of 3D US showed that alpha angles measured at rou-

tine 2D US of the hip can vary substantially between 2D
scans solely because of changes in probe positioning. Not
1
only could normal hips appear dysplastic, but dysplastic
 From the Departments of Radiology and Diagnostic
Imaging (J.L.J., V.G.S., L.J.) and Biomedical Engineering
hips also could have normal alpha angles. Three-dimen-
(M.M., K.C., R.B.T.), 2A2.41 WC Mackenzie Health Sciences sional US can display the full acetabular shape, which
Centre, University of Alberta, 8440-112 St, Edmonton, AB, might improve DDH assessment accuracy.
Canada T6G 2B7. Received February 22, 2014; revision
requested April 1; final revision received April 10; accepted q
 RSNA, 2014
April 21; final version accepted April 29. Address corre-
spondence to J.L.J. (e-mail: jjaremko@ualberta.ca).

q
 RSNA, 2014

870 radiology.rsna.org  n  Radiology: Volume 273: Number 3—December 2014

PEDIATRIC IMAGING: Alpha-angle Variation Can Change Diagnosis of Hip Dysplasia at US Jaremko et al

U
ltrasonography (US) is the pre- from 43° to 49° is classified as having rates (1). To avoid this problem, clin-
ferred imaging modality to eval- moderate dysplasia, and a hip with an ical practice often includes follow-up
uate developmental dysplasia of alpha angle of less than 43° is classi- US until initially borderline hips defi-
the hip (DDH) in infants younger than fied as having severe dysplasia. Other nitely appear normal, with associated
6 months (1). DDH is common, oc- groups prefer different thresholds (eg, costs and parental anxiety, or routine
curring in 1.6–28.5 per 1000 infants, classifying a hip with an alpha angle overtreatment until hips normalize. The
depending on definition criteria (2), greater than 55°as normal, a hip with poor diagnostic accuracy of current US
and one in 1000 infants is born with a an alpha angle of 45°–55° as border- of the hip may account for its lack of
dislocated hip (1). Imaging is needed line, and a hip with an alpha angle of beneficial effect on patient care, with
because clinical diagnostic tests such less than 45° as abnormal) (5). Treat- rates of late DDH diagnosis or surgery
as the Ortolani and Barlow maneuvers ment decisions are based on the de- not reduced by using universal or se-
lack sensitivity after the neonatal pe- gree of dysplasia and clinical factors, lective US, compared with rates with
riod and for mild disease (3). With US, particularly the age of the patient at clinical screening alone, despite higher
the most widely used index for evaluat- the time of diagnosis. treatment rates (2).
ing DDH is the acetabular alpha angle, Diagnosing dysplasia requires com- Three-dimensional US probes now
measured by using the Graf method bining clinical and imaging assessment, can allow acquisition of images rapidly
(4) on a standardized two-dimensional and, in current practice, the quantita- enough for scans to be feasible even in
(2D) coronal image that must contain tive component of imaging assessment a noncompliant infant. Because the en-
a flat horizontal iliac wing, labrum, relies heavily on the value of the alpha tire acetabular shape can be acquired
bony and cartilaginous acetabular angle. The alpha angle is subject to wide in a single 3D data set, the effect of
roof, and os ischium. These criteria variability of two main types: interscan differences in 2D probe orientation can
are intended to standardize the plane and interobserver. Interobserver vari- be explored by viewing sections cut
in which imaging is performed. With ability is easily quantified: two people along different planes through 3D data.
the Graf method, a hip with an alpha measuring the alpha angle differently Compared with obtaining repeated
angle greater than 60° is classified as on the same 2D US image. Interscan er- 2D measurements, this approach has
normal, a hip with an alpha angle from ror is more difficult to analyze. Because the advantage of holding all variables
50° to 59° is classified as having mild the US probe is handheld, each scan constant save for the 2D section orien-
dysplasia, a hip with an alpha angle inevitably is obtained with the probe tation, allowing the effects of this ori-
and the patient in a different position, entation to be evaluated independently.
which, because of the complex three- The purpose of this study was to quan-
Advances in Knowledge dimensional (3D) acetabular shape, re- tify alpha-angle variability due to chang-
nn Three-dimensional (3D) ultraso- sults in images showing different alpha ing probe orientation during 2D US of
nography (US) technology can be angles. Quantification of alpha-angle the infant hip and its effect on the diag-
used to show that interscan vari- variability due to probe orientation in nostic classification of DDH.
ability of alpha angles measured live patients or in dysplastic hips has
on two-dimensional (2D) US not been well established in the liter-
images, solely because of ature. This variability is important be- Published online before print
changes in probe positioning, is cause it could lead to misdiagnosis of 10.1148/radiol.14140451  Content code:
substantial (upper limit of DDH. Missed cases of DDH are partic-
Radiology 2014; 273:870–878
agreement, 19°) for infants with ularly of concern because DDH is usu-
normal and dysplastic hips. ally easily treated conservatively if di- Abbreviations:
agnosed early in infancy (6), whereas a CI = confidence interval
nn Changing probe orientation later diagnosis is associated with more DDH = developmental dysplasia of the hip
altered the alpha angle enough to invasive treatment and lower success
3D = three-dimensional
allow a change in the Graf diag- 2D = two-dimensional
nostic category in more than Author contributions:
50% of infant hips imaged and in Implications for Patient Care
Guarantors of integrity of entire study, J.L.J., M.M.;
72% of hips imaged in infants nn Assessing DDH in infants by study concepts/study design or data acquisition or data
younger than 1 month. using 2D US is limited by sub- analysis/interpretation, all authors; manuscript drafting or
stantial variation caused by manuscript revision for important intellectual content, all
nn Three-dimensional US is feasible authors; approval of final version of submitted manuscript,
in assessing patients for develop- probe positioning.
all authors; agrees to ensure any questions related to the
mental dysplasia of the hip nn Three-dimensional US demon- work are appropriately resolved, all authors; literature
(DDH) because a 3D sweep of strates the full acetabular shape research, J.L.J., V.G.S., L.J.; clinical studies, J.L.J., M.M.,
the hip can be performed in 3.2 and, with validation, might offer V.G.S., L.J., K.C.; statistical analysis, J.L.J., M.M., V.G.S.;
and manuscript editing, J.L.J., M.M., V.G.S., K.C., R.B.T.
seconds and with less difficulty in more reproducible DDH assess-
positioning than can 2D US. ment in the future. Conflicts of interest are listed at the end of this article.

Radiology: Volume 273: Number 3—December 2014  n  radiology.rsna.org 871

PEDIATRIC IMAGING: Alpha-angle Variation Can Change Diagnosis of Hip Dysplasia at US Jaremko et al

Materials and Methods protocol is to gather images showing Figure 1

the best Graf representation of each
Patients hip with and without measurements
This prospective study was approved for alpha angle, beta angle, and cover-
by the University of Alberta Health Re- age, as well as axial dynamic images.
search Ethics Board. Imaging was per- Two-dimensional US images were inter-
formed at a tertiary pediatric hospital preted by a pediatric radiologist, with
(Stollery Children’s Hospital Pediatric results and images made available to re-
Orthopedic Clinic, Edmonton, Alberta, ferring clinicians. In addition, two study
Canada) from October 2012 to Novem- team members, including a radiologist,
ber 2013. At the first routine clinical 2D technologist, or medical or graduate
US of the hip in each patient, written student trained by the study radiolo-
informed consent was obtained from a gists, used a high-resolution 13-MHz
parent to add 3D US of the hip at the 3D linear (13VL5; Philips Healthcare)
same visit. The imaging indication was transducer in the coronal orientation to
clinical suspicion of DDH because of obtain a 3D US data set at each hip.
laxity at examination, asymmetric hip With this transducer, we performed a
creases, or other risk factors such as a 3.2-second automated sweep through a
family history of a positive finding for range of 615° starting at 0° to gener-
DDH. Because dysplasia can be unilat- ate a 3D data set of 256 US sections
eral or bilateral, we included each hip that were 0.13 mm thick, each contain-
separately. Considering only the 3D ing 411 3 192 pixels measuring 0.11 3
scans showing the entire range of planes 0.20 mm. These 3D scans were not re-
necessary for this study, we had 56 hips leased for use in clinical treatment.
in 35 patients, and 26 (74%) of them Figure 1:  Generation of a 3D hip surface
were female. Patients underwent US at Image Processing
model from 3D US images. (a) The surfaces of
the mean age of 41.7 days (range, 4–112 Images were analyzed off-line by using the femoral head (green line), ilium (red line), and
days; for female infants, mean age was custom software (Matlab R2010a-2012; os ischium (blue line) are traced on sections of
38.7 days [range, 6–112 days]; for male MathWorks, Natick, Mass) that al- 3D data at regular intervals and interpolated into
infants, mean age was 50.2 days [range, lowed viewing of 3D US images fol- the intervening sections, with adjustments made
4–111 days]). They received routine care lowed by markup of points, lines, and manually by nudging contours that do not match
at a pediatric orthopedic clinic from one curves replicating the functionality of underlying anatomy into optimized position. (b) Re-
or more of five clinic surgeons who were standard cardiac magnetic resonance sulting surface model demonstrating the complex
blinded to 3D US images and findings. imaging workstation analysis of fea- 3D anatomy of the acetabulum (structure at left).
We observed clinical care for at least 6 tures such as ventricular contours.
months to classify each imaged hip as The major obstacle to performing this
normal at the first orthopedic assess- analysis was the proprietary 3D data a patch surface connecting the traced
ment (category 0; n = 20), questionably file format. A key function of the cus- and interpolated contours (Fig 1). The
abnormal initially but with findings that tom software was to extract manufac- training process involved initial com-
resolved spontaneously at follow-up im- turer-specific pixel spacing information bined review (by J.L.J. and M.M.) of
aging and clinical examination (category from the 3D US Digital Imaging and anatomy on multiplane and 3D-refor-
1; n = 23), or dysplastic and requiring Communications in Medicine data. For matted computed tomographic scans
treatment by using a Pavlik harness and/ each hip, an observer (M.M., a grad- of a normal infant pelvis and a physi-
or surgery (category 2; n = 13). uate student in biomedical engineer- cal model of this pelvis generated by a
ing and radiology) trained by the lead 3D printer. This process was followed
Imaging radiologist (J.L.J.) traced the relevant by a preliminary trial in which surface
All imaging was performed by us- anatomic contours (acetabular rim, models from 3D US were traced by
ing platforms (Philips iU22; Philips femoral head cartilage surface, os is- both observers in 15 hips in eight pa-
Healthcare, Andover, Mass). We per- chium) on the original 3D US data set tients ranging from normal to severely
formed conventional 2D US in both by using a customized semiautomated dysplastic, separate from the final
hips by using a 12-MHz linear trans- interactive interface that allowed con- study, and consensus review of models
ducer (12 L5; Philips Healthcare), tours traced on selected sections to be produced by two authors (J.L.J. and
including static coronal imaging in the interpolated to intervening sections, M.M.) to ensure anatomically appro-
Graf standard plane and axial dynamic then reviewed and corrected by using a priate appearance. Tracings in the final
imaging to assess for hip stability by nudging tool. A 3D model then was gen- study were reviewed and approved by
using our usual protocol. Our usual erated for each structure by creating the lead radiologist (J.L.J.).

872 radiology.rsna.org  n Radiology: Volume 273: Number 3—December 2014

PEDIATRIC IMAGING: Alpha-angle Variation Can Change Diagnosis of Hip Dysplasia at US Jaremko et al

As a reference orientation, we central plane in a simple way from sections in which this angle could be
required an initial central plane lo- two landmarks. The observer located defined, points A and P on Figure 2a.
cated in the middle of the acetabulum the corner point of the acetabular an- This central plane then was generated
in each patient. We generated this gle in the most anterior and posterior automatically perpendicular to the

Figure 2

Figure 2:  Model demonstrating the central plane and rotated planes. (a) Central plane is obtained through the middle of the
acetabulum as the plane perpendicular to the line joining the corner points of the most anterior (A) and posterior (P) positions in
which an acetabular angle can be drawn at the midpoint (M) of this line. (b) Planes were cut at angles rotated from the central
plane at 5° intervals in axis 1, representing probe tilt around a craniocaudal axis (ie, rotating the base of the transducer from
the table toward the ceiling). (c) Planes were similarly cut at 5° intervals around a transverse axis (axis 2), representing probe
tilt as if twisting the transducer around a line joining the two hips.

Radiology: Volume 273: Number 3—December 2014  n  radiology.rsna.org 873

PEDIATRIC IMAGING: Alpha-angle Variation Can Change Diagnosis of Hip Dysplasia at US Jaremko et al

midpoint of line A–P (Fig 2a). Three- same observer on each of 118 images Variability of Alpha Angle with Section
dimensional data were resampled along obtained in this fashion in a subset Orientation
this plane to produce the 2D US im- of 15 hips, 1 month after the initial The alpha angles recorded at each hip
age that would result from a 2D probe reading session, and this observer by the radiologist across all images with
placed in exactly that location and ori- was blinded to previously measured a quality score of 4 varied by a mean
entation. From this reference central alpha angles. of 6.9° 6 4.8 in different probe orien-
plane, 2D US images were extracted tations along axis 1 (range, 0°–21°),
along orientations rotated in 5° incre- Statistical Analysis 10.8° 6 7.1 in axis 2 (range, 0°–31°);
ments from 250° to 50° away, as if Descriptive statistics were recorded as combining both axes, the angles varied
slowly rotating a 2D US probe. Images the mean 6 standard deviation. We by a mean of 13.9° 6 7.1 (upper limit
were produced by rotation around axis computed differences in alpha angles of 95% CI, 28°). Ranges were similar
1 oriented craniocaudally (Fig 2b) and between observers and between obser- for observer 2. The proportion of this
axis 2 oriented transversely between vations at various plane orientations. variability occurring for alpha angles
hips (Fig 2c). We characterized variability in alpha greater than 60° may not be clinically
Each of the resulting 21 images angles as the upper limit of agreement relevant because those angles are con-
obtained at 5° increments of probe in these tests (7). For interobserver vari- sidered normal. When we considered
orientation for each axis was reviewed ability (assessment of the same image by any measured alpha angle greater than
for quality by two observers (J.L.J., different observers), this was the same 60° to be equal to 60°, the remain-
a radiologist with dual pediatric and as the repeatability coefficient (1.96 ing variability was 5.5° 6 6.9 (range,
musculoskeletal fellowship training multiplied by the standard deviation), 0°–24°; upper limit of 95% CI, 19°)
and 10 years of imaging experience as also measured by Gwynne Jones et across the two axes. For nearly all hips
[observer 1], and M.M., a biomedical al (5), whereas for interscan variability, and all axes, changes in virtual probe
engineering graduate student prepar- this was the mean difference plus 1.96 orientation produced 2D images show-
ing a thesis on US for hip dysplasia, multiplied by the standard deviation (ie, ing visually substantial variation in the
trained and supervised by J.L.J. [ob- the upper limit of the 95% confidence acetabular shape corresponding to the
server 2]). Image quality was scored interval [CI] for the alpha angle). In- numerical variation in alpha angle. For
on the basis of the number of ma- terobserver variability was assessed by example, the hip in Figure 3 had alpha
jor Graf standard plane criteria met using mean difference and intraclass angles measured at 52°–70° on images
(maximum, four): flat horizontal iliac correlation coefficients calculated by us- with a quality score of 4, depending on
wing, labrum visible, os ischium pre- ing a two-way mixed effects model. We the section orientation, an 18° range
sent, and midportion of femoral head used univariate analysis of variance to that was slightly higher than the mean
visible. As a nearly spherical struc- compare means in dysplastic and nor- range of 13.9°.
ture, the midpoint of the femoral head mal hips. Statistics were calculated with To assess the clinical significance of
is best determined by observing that software (SPSS 20; SPSS, Chicago, Ill). this variation, we considered whether
the image saved is that in which the observed differences in alpha angle
head has the largest possible radius. caused by simulated probe rotation led
We rated the femoral head acceptable Results to a change in the Graf diagnostic cat-
on an image if it was plausible from egory. Considering only 3D US images
review of that image that the middle Section Orientation with a quality score of 4, 26 (46%) of
one-third of the femoral head was in- In all 56 hips, at least one image had 56 hips had images with findings within
cluded, regardless of whether other an acceptable quality score of 4. The only one Graf category; 20 (36%) of 56
3D US images of the region (which range of simulated probe orientations hips had images with findings in two
would not be available to a radiolo- providing images with a quality score Graf categories (Fig 4); nine (16%)
gist reporting 2D US findings) might of 4, indicating that four main features of 56 hips had images with findings in
show a larger femoral head radius. required in a Graf standard-plane im- three Graf categories (Fig 5); and one
Only 18% (428 of 2352) of the images age are present (flat horizontal iliac (2%) had images with findings in all four
reviewed had a quality score of 4 from wing, labrum visible, os oschium pre- Graf categories. Overall, 54% of hips
both observers. On each of these, the sent, and midportion of femoral head had findings in more than one Graf cat-
observers separately measured the al- visible), was 24° 6 10 (standard devi- egory. Changes in diagnosis were signif-
pha angle between acetabular roof and ation) in axis 1 (maximum, 45°) and icantly more common at younger ages:
iliac lines, and these observers were 23° 6 8.4 in axis 2 (maximum, 40°). Findings on images in 18 (72%) of 25
blinded to the alpha angle measured The image with the central plane had hips scanned in infants younger than 31
on other images of that hip, as well a quality score of 4 in 38 (68%) of 56 days were classified in two or more Graf
as to each other’s work and to clinical hips and was within a 10° rotation from categories versus findings on images in
data. For intraobserver variability, the an image with a quality score of 4 in 50 just nine (29%) of 31 hips scanned in
alpha angle was remeasured by the (89%) of 56 hips. older infants (P = .002).

874 radiology.rsna.org  n Radiology: Volume 273: Number 3—December 2014

PEDIATRIC IMAGING: Alpha-angle Variation Can Change Diagnosis of Hip Dysplasia at US Jaremko et al

Figure 3

Figure 3:  Effect of probe tilt in two axes. Top: A and B, 3D US images obtained on cut planes 25° apart along axis 1. Top right,
Illustration of acetabular model shows cut planes A and B. Bottom: C and D, US images obtained on cut planes 20° apart along axis 2.
Bottom right, Illustration of 3D acetabular model shows cut planes C and D. Image C shows a trace of curvature at the junction of the
iliac wall and acetabular roof, rather than this being the junction of two perfectly straight lines; the os ischium is more clearly defined in
A and C than in the other images, and the femoral head is perhaps best seen in B or D, but, in our experience, each image potentially
could be accepted in real-world clinical practice. These images demonstrate an overall range of alpha angle variation of 18° (52°–70°),
which was typical in our data. In this hip, the images were in one of two possible diagnostic categories (normal or mildly dysplastic). The
hip was assessed as having findings inconclusive for dysplasia at initial clinical and 2D US examination and was assessed as normal at
follow-up without treatment (category 1). Note that the 3D US images were not released to clinicians and had no role in treatment.

Figure 4 Variability of Alpha Angle between and

within Observers
Four hundred twenty-eight images had
a quality score of 4 and were of accept-
able quality in 56 hips. In these im-
ages, interobserver variation in alpha
angle was a mean of 0.1° 6 5.1 (re-
peatability coefficient, 10°; coefficient
of variation, 0.048). The intraclass
correlation coefficient was 0.89. Alpha
angles recorded by the two observers
were within 5° and 10° of each other
in 73% (313 of 428) and 97% (415 of
428) of cases, respectively. Intraob-
server variability was lower, with the
mean difference between alpha angles
of 1.8° 6 3.7 (repeatability coefficient,
7.4°) and measurements within 5°
Figure 4:  (a, b) US scans in right hip in a 19-day-old girl show findings crossing two Graf diagnostic categories and 10° of each other in 82% (351 of
because of probe tilt. The images extracted along rotated cut planes from a single 3D US scan show alpha angles 428) and 96% (411 of 428) of cases,
ranging from 48° to 65° (ie, moderately dysplastic to normal), depending solely on probe tilt. Clinically, findings in respectively.
this hip were considered borderline (category 1) at initial visit and normalized eventually at follow-up.

Radiology: Volume 273: Number 3—December 2014  n  radiology.rsna.org 875

PEDIATRIC IMAGING: Alpha-angle Variation Can Change Diagnosis of Hip Dysplasia at US Jaremko et al

Figure 5

Figure 5:  (a–c) US scans in left hip in a 3-month-old girl show findings crossing three Graf diagnostic categories because of probe tilt. The images from cut planes
in a single 3D US scan across a 24° range of probe tilt show a grossly abnormal hip with severe acetabular rounding and alpha angles varying from 36° to 60°. Note
that the os ischium is seen poorly on the image best showing the midportion of the femoral head (a) and is more clearly visible on image c, where the femoral head is
obscured partly by artifact from the greater trochanter, suggesting that the femoral head probably is subluxed. Pavlik harness treatment was unsuccessful, and treatment
included spica casting. Reporting the finding as a normal 60° alpha angle in this case would have led to incorrect characterization of this obviously dysplastic hip. (b)
Intermediate image at an orientation between that of a and c.

Normal versus Dysplastic Hips Figure 6

We observed clinically normal hips
that appeared dysplastic in some
planes (Fig 3) and clinically dysplas-
tic hips that had normal 60° alpha
angles in certain planes, either with
the acetabular contours appearing
grossly abnormal (Fig 5) or essen-
tially normal (Fig 6). The alpha angles
we observed at 3D US, averaged from
acceptable images across all simu-
lated probe orientations in both axes,
clearly indicated (P , .001) clinical
diagnostic categories: 3D US images
in dysplastic hips requiring treatment
(category 2; n = 13) had a mean alpha
angle of 49.5° versus that in images
in initially borderline hips requiring
follow-up but no treatment (category
1; n = 23; mean angle, 63.2°) and that
in images in hips not requiring treat-
ment or follow-up (category 0; n = 20;
mean angle, 71.4°) (Fig 7).

Figure 6:  (a, b) US in left hip in a 6-day-old girl shows that a dysplastic hip can appear normal, depending
Discussion on probe tilt. The images from cut planes in a single 3D US scan show alpha angle variation from 52° to 61°.
We used 3D US technology in infants Other images lacking only the os ischium had alpha angles as low as 48°. Image a is within normal limits.
with normal and dysplastic hips to The hip was judged clinically dysplastic and normalized with treatment with a Pavlik harness. Image b at a
examine the potential alpha angle in- different orientation shows the alpha angle in dysplastic range.
terscan variation solely caused by US
probe orientation. Our most important

876 radiology.rsna.org  n Radiology: Volume 273: Number 3—December 2014

PEDIATRIC IMAGING: Alpha-angle Variation Can Change Diagnosis of Hip Dysplasia at US Jaremko et al

Figure 7 in some patients by as much as 45°, in move the probe through a range of ori-
each axis. This finding means that the entations to select the best image of the
sonographer often has a relatively wide hip, with the steepest possible alpha
range of probe positions in which ac- angle, deepest acetabular fossa, and
ceptable images can be generated and largest femoral head radius of all pos-
must rely on skill and experience to se- sible images meeting Graf standards. A
lect the best image to save within this sonographer typically would not select
range. The interscan variability of alpha an image at the margins of the accept-
angles measured on 2D US images in able range of probe orientations. How-
those different probe orientations was ever, in a squirming, crying infant, even
substantial (upper limit of 95% CI: 19°, the most experienced sonographer has
28°) and higher than interobserver var- a limited opportunity truly to assess all
iability (repeatability coefficient, 10°). possible imaging planes with the probe.
DDH diagnosis from the alpha an- The hazard for the interpreting radiol-
gle measured on a standardized 2D US ogist or surgeon is that the image may
image is based on the assumption that be of diagnostic quality but not actually
this is a one-of-a-kind clinically repre- represent the best image of the acetab-
sentative image (ie, that if any trained ulum that could have been obtained.
individual performs US in that hip that Our study results demonstrate the wide
day, any image obtained containing ap- possible range of interscan variability
propriate landmarks will have similar in alpha angle measurement. Careful
acetabular contours and alpha angle). sonographer training and experience,
The risk of misdiagnosis and overtreat- and thorough radiologist assessment of
ment because of incorrect probe angu- other factors such as acetabular round-
lation is a recognized hazard of 2D US, ing, coverage, and dynamic hip stabil-
with Graf et al (8) even advocating use ity, are critical to minimize the practical
of a probe guide-frame system to avoid effect of this variability, particularly in
tilt. Still, to our knowledge, investiga- the youngest infants. This is especially
Figure 7:  Point estimates show differences in tors in only one previous study directly true given the relatively high superim-
alpha angles according to clinical diagnosis (Dx). assessed the effect of changes in US posed interobserver variability in alpha
 = mean values, error bars = 95% CI of the transducer angle on alpha angle mea- angle measurement once an image is
mean for the alpha angles in all acceptable (quality surements (9). They used ex vivo 2D obtained. The reliability of US diagno-
score 4) images in each hip across all studied
US in five infant cadaver hips in a water sis of hip dysplasia is likely considerably
probe orientations, 0.0 = normal, 1.0 = hips with
bath and showed that the alpha angle greater in clinical practice, when the al-
initially inconclusive findings but that normalized
in a normal hip could vary from 48° to pha angle, acetabular coverage, acetab-
at follow-up, and 2.0 = dysplastic hips. Mean
values differed significantly (P , .001) between 65° (ie, from moderately dysplastic to ular morphology, and hip stability are
each category. Note that the normal hips had mean normal) by changing probe position. assessed together, than when the alpha
alpha angles substantially higher than the traditional Our 3D US study results help confirm angle alone is used for diagnosis. Test-
threshold of 60° for detection of DDH, whereas even these findings in vivo and demonstrate ing this reliability was outside the scope
the hips that were rated as having inconclusive find- that dysplastic hips can be shown to of our study, which focused on the limi-
ings on the basis of 2D US and clinical assessment have normal 60° alpha angles in probe tations of basing diagnosis primarily on
but that ultimately were determined to be normal orientations producing images that still the alpha angle.
had mean alpha angles slightly greater than 60° meet Graf standard plane criteria. The We did not assess acetabular cov-
when all acceptable 3D US images were considered. range of alpha angle variability in our erage formally in this study. Our pre-
study compares well with data in that liminary testing of this index showed
study, with an upper limit of agreement an even wider range of variation with
finding was that changing probe orien- of 19° versus a range of 17° in their plane orientation than the alpha angle,
tation altered the alpha angle enough to study (9). as expected, because this index involves
lead to a change in the Graf diagnostic The 19° upper limit of agreement the 3D geometric relationship between
category in more than one-half of the of interscan variability reported here is two bones (femoral head and acetabu-
imaged hips overall and in nearly three- greater than the approximately 8° re- lum) rather than only the shape of the
quarters of hips imaged in patients ported in a larger clinical study (10). acetabulum.
younger than 1 month. We also found This finding is expected because we in- Although the alpha angles differed
that a high-quality image satisfying Graf cluded all US images in each hip that significantly between dysplastic and
criteria could be obtained at orienta- met Graf standards, whereas in clinical normal hips in our study, it is evi-
tions varying by an average of 24°, or practice, sonographers are trained to dent from our analysis that there are

Radiology: Volume 273: Number 3—December 2014  n  radiology.rsna.org 877

PEDIATRIC IMAGING: Alpha-angle Variation Can Change Diagnosis of Hip Dysplasia at US Jaremko et al

intrinsic flaws in using the 2D Graf is the most clinically important period 3. Dezateux C, Rosendahl K. Developmental
standard plane as a diagnostic method because the decision to begin treating dysplasia of the hip. Lancet 2007;369(9572):
1541–1552.
for DDH. Simply by rotating the US DDH is generally within weeks of ini-
probe, the sonographer could change tial assessment. The probe movements 4. Graf R. Fundamentals of sonographic diag-
the diagnostic category for more than we assessed in 3D data were limited nosis of infant hip dysplasia. J Pediatr Or-
thop 1984;4(6):735–740.
one-half of the hips in this study by at to pure rotation in one of two planes,
least one, and often two, levels (eg, and we did not consider combinations 5. Gwynne Jones DP, Vane AGS, Coulter G,
from normal to moderately dysplas- of two rotations or rotation and trans- Herbison P, Dunbar JD. Ultrasound mea-
tic). Because the image on the screen lation. These assessments would have surements in the management of unstable
hips treated with the Pavlik harness: reli-
still appears diagnostic, and there is no been challenging to perform math-
ability and correlation with outcome. J Pedi-
record of the position of the 2D US ematically, and although they might atr Orthop 2006;26(6):818–822.
probe, this error invisibly degrades have revealed further increased vari-
6. Cashman JP, Round J, Taylor G, Clarke NM.
the accuracy and reliability of 2D US ation in alpha angles, they would not
The natural history of developmental dyspla-
assessment of DDH. Conceptually, 3D alter our conclusion that this variation sia of the hip after early supervised treat-
US of DDH could reduce or eliminate is an important hazard of 2D US as- ment in the Pavlik harness: a prospective,
this error by removing dependence on sessment of DDH. longitudinal follow-up. J Bone Joint Surg Br
the specific plane of image acquisition. In conclusion, we found that 3D US 2002;84(3):418–425.
Three-dimensional US ideally of the infant hip is feasible and images 7. Bland JM, Altman DG. Statistical methods for
would allow firm diagnosis of every can be analyzed quantitatively. After assessing agreement between two methods
hip as normal or dysplastic at the ini- an initial flurry of studies in the 1990s of clinical measurement. Lancet 1986;
tial scan. Although our study was not (11–14), 3D US was not pursued in hip 1(8476):307–310.
designed or powered to define these dysplasia assessment. Those early re- 8. Graf R, Mohajer M, Plattner F. Hip so-
diagnostic thresholds in 3D US, hips ports describe slow manual sweeps to nography update: quality-management, ca-
classified in different diagnostic cate- produce scans and cumbersome post- tastrophes—tips and tricks. Med Ultrason
gories showed substantially different processing. The probe we used can 2013;15(4):299–303.
mean alpha angles across all 3D US help obtain a 3D scan in 3.2 seconds 9. Falliner A, Hahne HJ, Hedderich J, Brossmann
imaging planes. We expect that diag- with less difficulty in positioning than J, Hassenpflug J. Comparable ultrasound
nostic accuracy of US for DDH could in 2D US. In our experience, almost measurements of ten anatomical specimens
of infant hip joints by the methods of Graf and
be improved in the future by consider- all infants are able to stay still for this
Terjesen. Acta Radiol 2004;45(2):227–235.
ing this additional acetabular shape in- brief time, especially if distracted by
formation available from 3D US, both toys, sucrose, a warmed blanket, and
10. Zieger M. Ultrasound of the infant hip.
visually and by measuring alpha angles warmed US gel. Computer analysis II. Validity of the method. Pediatr Radiol
1986;16(6):488–492.
and novel indexes being developed. time per hip is less than 10 minutes.
Our study had limitations. As a pi- As technology advances, 3D scanning 11. Gerscovich EO, Greenspan A, Cronan MS,
lot study, the sample size was relatively and analysis times probably will con- Karol LA, McGahan JP. Three-dimensional
sonographic evaluation of developmental
small, and some of our earlier scans tinue to shorten, potentially further
dysplasia of the hip: preliminary findings.
included a range of the acetabulum too improving the practical application of Radiology 1994;190(2):407–410.
narrow to be included; however, our this technology in the future.
12. Graf R, Lercher K. Experiences with a 3-D
data set demonstrates the full range
Disclosures of Conflicts of Interest: J.L.J. ultrasound system in infant hip joints [in
from normal to severely dysplastic disclosed no relevant relationships. M.M. dis- German]. Ultraschall Med 1996;17(5):218–
hips. The clinical diagnoses and treat- closed no relevant relationships. V.G.S. dis- 224.
ment decisions were made by one of closed no relevant relationships. L.J. disclosed
no relevant relationships. K.C. disclosed no rel- 13. Sohn C, Lenz GP, Thies M. 3-dimensional
five orthopedic surgeons on the basis evant relationships. R.B.T. disclosed no relevant ultrasound image of the infant hip [in Ger-
of their own experience and prefer- relationships. man]. Ultraschall Med 1990;11(6):302–305.
ences, and they were not validated
14. von Jan U, Overhoff HM, Lazovic D. 3-D

by any external reference standard. References visualization of the newborn’s hip joint using
However, this situation represents 1. Storer SK, Skaggs DL. Developmental dys- ultrasound and automatic image segmenta-
typical clinical practice, and our con- plasia of the hip. Am Fam Physician 2006; tion. Stud Health Technol Inform 2000;77:
clusions regarding the variability of 74(8):1310–1316. 1170–1174.
US measurements do not depend on
2. Shorter D, Hong T, Osborn DA. Cochrane Re-
the specific diagnosis in each patient. view: screening programmes for developmen-
The follow-up interval was relatively tal dysplasia of the hip in newborn infants.
short at more than 6 months, but this Evid Based Child Health 2013;8(1):11–54.

878 radiology.rsna.org  n Radiology: Volume 273: Number 3—December 2014