Forensic Science International 214 (2012) 167–172

Contents lists available at ScienceDirect

Forensic Science International

journal homepage: www.elsevier.com/locate/forsciint

Evaluation of pediatric skull fracture imaging techniques

Maura H. Mulroy a, Andre M. Loyd a, Donald P. Frush b, Terence G. Verla a, Barry S. Myers a,
Cameron R. ‘Dale’ Bass a,*
a
Duke University, Department of Biomedical Engineering, Injury Biomechanics Laboratory, Box 90281, Durham, NC 27708-0281, USA
b
Duke University, Department of Radiology, DUMC 3808, Durham, NC 27710, USA

A R T I C L E I N F O A B S T R A C T

Article history: Radiologic imaging is crucial in the diagnosis of skull fracture, but there is some doubt as to whether
Received 25 April 2011 different imaging modalities can accurately identify fractures present on a human skull. While studies
Received in revised form 5 July 2011 have been performed to evaluate the efficacy of radiologic imaging at other anatomical locations, there
Accepted 27 July 2011
have been no systematic studies comparing various CT techniques, including high resolution imaging
Available online 30 August 2011
with and without 3D reconstructions to conventional radiologic imaging in children, we investigated
which imaging modalities: high-resolution CT scan with 3D projections, clinical-resolution CT scans or
Keywords:
X-rays, best showed fracture occurrence in a pediatric human cadaver skull by having an expert pediatric
Skull
Fracture
radiologist examine radiologic images from fractured skulls. The skulls used were taken from pediatric
Injury cadavers ranging in age from 5 months to 16 years. We evaluated the sensitivity and specificity for the
Imaging imaging modalities using dissection findings as the gold standard. We found that high-resolution CT
Pediatric scans with 3D projections and conventional CT provided the most accurate fracture diagnosis (single-
X-ray fracture sensitivity of 71%) followed by X-rays (single-fracture sensitivity of 63%). Linear fractures
Radiologic outsider the region of the sutures were more identifiable than diastatic fractures, though the incidence of
false positives was greater for linear fractures. In the two cases where multiple fractures were present on
the same anatomical skull location, the radiologist was less likely to identify the presence of additional
fractures than a single fracture. Overall, the high-resolution and clinical-resolution CT scans had the
similar accuracy for detecting skull fractures while the use of the X-ray was both less accurate and had a
lower specificity.
ß 2011 Elsevier Ireland Ltd. All rights reserved.

1. Introduction radiologic imaging (CT scan or X-ray) [8]. Radiologic imaging is

especially important in the diagnostic procedure because it can
Head injury is the leading cause of disability and death in detect skull fractures or brain abnormalities that are not
children [1]. Pediatric head injury requires approximately 95,000 suspected during history-taking or physical examinations.
hospital admission and costs the United States $10 billion dollars Finding a skull fracture can be of particular importance, as
per year [2]. Further, traumatic brain injuries (TBI) can lead to brain the presence of a skull fracture may be an indicator of further
lesions, amnesia and lasting neurological abnormality [3]. These intracranial injuries. For example, Quayle et al. found that the
issues can affect memory, social functioning, learning and presence of a skull fracture predicted concurrent intracranial
emotional awareness [4]. For children, mortality is of particular injury with an odds ratio of 92.4 [1].
concern because the developing brain may be more vulnerable to Several studies have been published examining the accuracy of
shear and resulting diffuse swelling [5,6]. Severe pediatric TBI has radiologic imaging as a diagnostic tool. In a comparative study
been found to have similar mortality rates as adults but tends to be between X-ray and CT scan, Lloyd et al. evaluated 156 images of
higher for younger children (<4 years old) [5,7]. heads were both X-rayed and CT scanned and found that X-ray was
When head trauma is a consideration in a patient, the unable to detect 31% of the skull fractures identified by the CT
primary diagnostic tools used are neurological examination and scans [9]. Though other studies have investigated diagnoses using
CT scans against X-rays in other anatomical regions (cf. [10,11]).
However, these studies are not directly applicable to the diagnosis
of pediatric skull fracture since the specificity and sensitivity of the
* Corresponding author. Tel.: +1 919 681 9979.
imaging techniques depend on the anatomical location. Finally, in
E-mail addresses: [email protected] (M.H. Mulroy),
[email protected] (A.M. Loyd), [email protected] (D.P. Frush),
a study of paired radiographs of cadaver skulls with and without
[email protected] (T.G. Verla), [email protected] (B.S. Myers), fractures, Weber and Folio found that radiologists were only
[email protected] (C.R. ‘Dale’ Bass). correct 62% of the time and were unable to find a source for the

0379-0738/$ – see front matter ß 2011 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.forsciint.2011.07.050
168 M.H. Mulroy et al. / Forensic Science International 214 (2012) 167–172

Table 1
Age, sex, cause of death and injury information for all specimens that were mechanical tested and used for image technique analysis.

Age Sex Cause of death Fracture location Fracture type

5-month M Respiratory failure R. Parietal, R. Coronal suture; Linear, diastatic

Sagittal suture
9-month M Chronic obstructive Forehead, orbits, R. Coronal suture, Linear, diastatic
pulmonary disease Nasion R. Anterolateral suture,
metopic suture, both frontozyomatic sutures
11-month F SIDS, renal failure Coronal suture; forehead Linear, diastatic
22-month F Non-Hodgkins lymphoma L. Coronal suture Diastatic
9-year M End-stage renal disease, Forehead, R. Orbit Linear
secondary hyperkalemia
16-year F Seisure disorder Forehead, R. Orbit Linear

errors [12]. No study to date has compared CT scan and X-ray A blinded expert clinical pediatric radiologist (with 20 years experience)
examined the high-resolution CT scans with a 3D skull projection, clinical-
results for skull using dissection as the gold standard.
resolution CT scans, and equivalent plane film X-rays for fractures, marking the
Therefore, the objective of our study was to determine which fracture locations on sketches of a human skull. Using the dissection results as the
imaging modalities could be used to best identify skull fracture in gold standard for the true fracture types and locations, the accuracy of each image
children. In doing so, we used fractured cadaveric pediatric heads reading was calculated. These images were presented in randomized order. To
to judge the accuracy of X-ray and high-resolution and clinical CT calculate the sensitivity and specificity of the imaging methods, the head was
divided into 15 different anatomical locations (Table 2) and both single and
scans with dissection as the gold standard. It is hypothesized that multiple fracture presence were judged on a binary scale. The specificity and
the high-resolution CT scans will be the most effective imaging sensitivity were calculated for each image. Here, sensitivity is the chance that a
modality for indentifying fracture in the pediatric skull. fracture is correctly identified and specificity is the probability that a non-injury is
corrected noted as a non-injury.

2. Material and methods

3. Results
Six pediatric human cadaver skulls, ranging in age from 5 months to 16 years old,
were put through a battery of head drop tests onto a rigid aluminum surface (Table Example CT scans 3D projections and simulated plane-film X-
1). All post mortem human specimen (PMHS) testing was approved by the rays for the tested subjects are shown (Figs. 1–4).
institutional review board (IRB) of Duke University. Detailed descriptions of the test
methodology and head responses were reported by Prange et al. (2004) [19]. The
Single fractures were more likely to be identified in CT scans;
heads were subjected to ‘‘non-destructive’’ vertical drops of 15 cm and 30 cm onto the clinical and high resolution CT scans resulted in a sensitivity of
an aluminum plate. If no fractures occurred during these drops, the head was 71%, whereas X-rays had a sensitivity of only 63%. In the two
subjected to a 2 m drop. Before the initial test and after the final test, each head was locations that had multiple fractures, the radiologist was able to
examined by CT. After the final CT scans were performed, the heads were dissected,
identify one of multiple fractures in both the X-ray and clinical CT
and the skull was photographed, visually and tactually checked for fractures. In this
study, we define diastatic fracture as a fracture along the fused suture line. scans, but could not identify either of the multiple fracture in high-
High-resolution CT scans were conducted using a 64-slice multidetector array CT resolution CT. The single-fracture specificity for CT scans was 95%
scanner (LightSpeed GE Healthcare, Milwaukee, WI). A peak voltage of 120 kVp and for both clinical and high-resolution CT scans and 87% for X-ray.
a tube current of 310 milli-amperes with a one second gantry rotation time Linear fractures of the skull bones were more discernable than
(310 mAs) were used along with a configuration of 16 mm  0.625 mm and a slice
thickness/slice interval of 0.625 mm. The examination was performed using a small
diastatic fractures; the average sensitivities for all imaging modes
scan field-of-view, a 512  512 pixel transverse resolution and a standard were 80% and 64%, respectively. However, the incidence of false
reconstruction algorithm using a 0.1 mm reconstruction interval. The high- positives was greater for linear fractures than for diastatic
resolution scanning parameters produced images with 0.43 mm/pixel resolution. fractures. The former produced an average specificity of 88%,
Clinical-resolution CT scans with 5 mm slice thickness were simulated from the
while the later resulted in an average specificity of 98% (Table 3).
original high-resolution scans by reslicing the scans using ImageJTM.
To make the X-rays of the heads, the high-resolution CT scans were converted to In two skulls, multiple fractures were present on the same skull
equivalent plane film X-rays using ImageJTM 1.43 (NIH, Bethesda, MD). The bone. In these cases, any additional fractures present on one
threshold used was found by minimizing physical measurements of the mandible location of the skull were less identifiable (average sensitivi-
with measurements of the isosurface measurements of the mandible, thus ty = 33%) than any single fracture (average sensitivity = 68%).
eliminating the presence of soft tissue [13,14]. The intensities of each CT scan
Conversely, the specificity was greater for the identification of
slice were added onto orthogonal coronal and sagittal projections, resulting in a
simulated plane film X-ray. multiple fractures (97%) than for single fractures (92%) (Table 4).

Table 2
List of 15 anatomical skull locations judged for fracture.
4. Discussion

Location Type The goal of this study was to investigate the accuracy of
Frontal Bone simulated X-ray, clinical CT scans and high-resolution CT scan for
R. Parietal Bone use in detecting fractures of the pediatric skull. The study looked at
L. Parietal Bone
the specificity and sensitivity of each imaging modality and each
R. Temporal Bone
L. Temporal Bone fracture type using dissection as the gold standard.
R. Sphenoid Bone Previous studies of the X-ray have found a fracture identifica-
L. Sphenoid Bone tion sensitivity of 40% for the adult rib and 62% for the adult skull
Occipital Bone [11,12]. The X-ray accuracy found by Webber is close to the
Sagittal Suture
Coronal Suture
sensitivity of 63% found in the current study. However, the
R. Squamous Suture previous studies identified the presence of any fracture in the
L. Squamous Suture image whereas this study investigated the presence of fractures at
Metopic Suture specific anatomical locations. The authors were unable to find CT
R. Lambdoidal Suture
scan assessments of skull that used dissection as the gold standard.
L. Lambdoidal Suture
Further, as implied by previous studies, the sensitivity of the best
 M.H. Mulroy et al. / Forensic Science International 214 (2012) 167–172 169

Fig. 1. High-resolution CT scan 3D projection (top) and simulated plane-film X-rays (bottom) for the 9-month-old skull. The 9-month-old sustained fractures to the frontal
bones, the coronal, metopic, anterolateral and frontozygomatic sutures during the 2 m drop.

Fig. 2. High-resolution CT scan 3D projection (top) and simulated plane-film X-rays (bottom) for the 11-month-old skull. The CT scan was taken after the 2 m drop, in which
the 11-month-old sustained frontal bone and coronal suture factures.
170 M.H. Mulroy et al. / Forensic Science International 214 (2012) 167–172

Fig. 3. High-resolution CT scan 3D projection (top) and simulated plane-film X-ray (bottom) for the 22-month-old skull. The 22-month-old sustained a diastatic fracture at the
left coronal sutures during the 15 cm vertex drop.

modality, the diagnostic CT scans, left substantial room for fracture as the high-resolution CT scans. A further intriguing
improvement in identifying skull fractures. finding was that one of the multiple fracture injuries was not
The results of this study agree with literature suggesting that CT discernable from the high-resolution CT scans. This implies that
scans are more sensitive than X-rays at detecting fractures [10]. higher resolution CT scans may not lead to more effective
Either mode of CT scan was an improvement over X-ray in identification of pediatric skull fracture. This conclusion may have
specificity and sensitivity. An interesting finding was that far reaching consequences; a higher resolution imaging technique
conventional CT technique was just as effective at identifying may be unnecessary for optimal identification of pediatric skull

Fig. 4. High-resolution CT scan 3D projection (left) and simulated plane-film X-ray (right) of the 9-year-old skull. The 9-year-old sustained a linear fracture on the forehead
during the 2 m drop.
 M.H. Mulroy et al. / Forensic Science International 214 (2012) 167–172 171

Table 3 There are several limitations to this study: the use of cadavers
Sensitivity and specificity (the specificity is shown in parentheses) of each image
rather than live humans, the use of a single radiologist, the
modality for detecting fractures.
simulation of the clinical CT and X-ray from the high-resolution CT
Image modality Single fracture Multiple fracture scan, and the limited number of samples used. The simulated CTs
X-ray 63% (87%) 50% (96%) had the same mAs and kVp values. The identification of fractures in
Clinical CT 71% (95%) 50% (96%) a cadaver is more difficult than in living humans because there are
High-Res. CT 71% (95%) 0% (100%) no areas of acute swelling near the fracture. In addition, living
patients are typically able to indicate the location of their pain,
Table 4 which may simplify the injury diagnosis [11]. However, for acute
Sensitivity and specificity (the specificity is shown in parentheses) of each image
injury presentation, such as in the emergency setting subacute
modality of for detecting linear and diastatic fractures.
changes from healing will not be present so the accuracy of the
Image modality Linear fracture Diastatic fracture fracture diagnosis in the current study may be comparable (cf.
X-ray 80% (80%) 42% (96%) [17]). Another limitation involves the use of a single radiologist for
Clinical CT 80% (91%) 75% (100%) this study; using multiple radiologists could reduce error from
High-Res. CT 80% (93%) 75% (97%) human observation. The investigation was designed to only look
for the presence or absence of fracture, in contrast with the clinical
scenarios which involve additional information. This may poten-
fractures. Future studies should investigate the potential for tially alter search patterns and the potential for fracture detection.
trading intrinsic resolution with dose, even for conventional Moreover, the simulated imaging may not be an exact representa-
imaging resolutions. tion of the actual X-ray and clinical CT. Further, CT technique may
The current study shows that linear fractures are more miss fractures in the plane of the reconstructed slice, and the CT
identifiable than diastatic fractures. The reason for this may stem scans in this study use a single scanner. Finally, only 6 specimens
from the difficulty in identifying a diastatic fracture from the normal were available for this study, preventing further statistical
suture anatomy. The pediatric sutures continue to interdigitate and analyses.
do not attain the full strength of the skull until after 6 years of age [cf.
15, 16]. So, a diastatic fracture for a specimen 22 months or younger 5. Conclusions
could be interpreted as being a normal suture, particularly if the
diastatic fracture is not displaced. Linear fractures are not as likely to This study analyzes three different imaging techniques for
be masked by the natural anatomy of the skull; however, the detecting fractures in the pediatric skull. To the best of our
increase in specificity from linear fractures to diastatic fractures knowledge, there have been no carefully controlled pediatric cranial
implies that the incidence of false positives is lower for diastatic studies that examined imaging modalities using anatomic dissec-
fractures. This may suggest that non-injury related skull irregulari- tion as the gold standard, and there have only been a few studies that
ties might be mistaken for linear fractures. Though linear fractures have investigated the effectiveness of image modalities in detecting
are found to be more detectable, when they are very thin, they pediatric skull fractures [1,9,18].
cannot be discerned with conventional imaging techniques. For This investigation found that the likelihood of diagnosing a
instance, two thin linear fractures in the 16-year skull of lengths fracture correctly is dependent upon the imaging modality used
5 cm and 6 cm were not diagnosed as fractures regardless of the and the type of fracture. It was found that conventional and high
image modality (e.g. Fig. 5a) likely owing to the small width of the resolution CT techniques were more accurate at diagnosing skull
fractures The measured fracture width from necropsy photos fractures than X-ray in sensitivity and specificity. Also, unexpect-
(Fig. 5b) is approximately 0.04 mm, far smaller than the X-ray edly, the accuracy of detecting skull fracture did not increase with
resolution or CT voxel edge. resolution, and therefore radiation level, for CT scans. Linear

Fig. 5. (a) High-resolution CT scan reconstruction for the sixteen-year with two linear fractures in the forehead. The fractures were 5 cm and 6 cm long, but were very thin
(0.04 mm or less) and could not be identified in any image modality. (b) Necropsy photograph of the 5 cm fracture.
172 M.H. Mulroy et al. / Forensic Science International 214 (2012) 167–172

fractures were more identifiable than diastatic fractures, but also [7] T. Luerssen, M. Klauber, L. Marshall, Outcome from head injury related to patient’s
age. A longitudinal prospective study of adult and pediatric head injury, Journal of
had a higher incident of false negatives. This current data can be Neurosurgery 68 (1988) 409–416.
used in future analysis of image modality use when fractures are [8] L.H. Jorgen Borg, J.D. Cassidy, P.M. Peloso, L.J. Carroll, Hans von Holst, Kaj Ericson,
expected in the pediatric head. Diagnostic procedures in mild traumatic brain injury: results of the who collabo-
rating centre task force on mild traumatic brain injury, Journal of Rehabilitation
Medicine 36 (43) (2004) 61–75.
[9] D.A. Lloyd, et al., Pedictive value of skull radiography for intracranial injury in
Acknowledgements children with blunt head injury, The Lancet 349 (9055) (1997) 821–824.
[10] J.D. Harley, L.A. Mack, R.A. Winquist, CT of acetabular fractures: comparison with
The authors thank Dr. Mark Palmeri for assisting us with the use conventional radiography, American Journal of Roentgenology 138 (3) (1982)
413–417.
of ImageJ and production of the X-ray images. The authors [11] R.W. Kent, et al., Radiographic detection of rib fractures: a restraint-based study of
gratefully acknowledge funding from the US National Highway occupants in car crashes, Traffic Injury Prevention 3 (1) (2002) 49–57.
Traffic Safety Administration (NHTSA) for portions of this study [12] R.L. Webber, J. Folio, Radiographic detectability of occipital and temporal-parietal
fractures induced in Cadaver heads, The Journal of Trauma 16 (2) (1976) 115–124.
under cooperative agreement DTNH22-08-H-0187. [13] A.M. Loyd, et al., Thresholding techniques for developing geometrically accurate
pediatric skull and cervical spine models, in: Injury Biomechanics Research 32th
International Workshop, NHTSA National Transportation Biomechanics Research
References Center, Nashville, TN, 2004.
[14] A.M. Loyd, et al., Pediatric head contours and inertial properties for ATD design, in:
[1] K.S. Quayle, et al., Diagnostic testing for acute head injury in children: when are Stapp Car Crash Conference, Society of Automotive Engineers, Inc., Scottsdale, AZ,
head computed tomography and skull radiographs indicated? Pediatrics 99 (11) 2010.
(1997). [15] M.M. Cohen, Sutural biology, in: J.M. Michael Cohen, R.E. MacLean (Eds.), Cra-
[2] J. Langlosis, W. Rutland-Brown, K. Thomas, Traumatic Brain Injury in the United niosynostosis: Diagnosis, Evaluation, and Management, Oxford University Press,
States: Emergency Department Visits, Hopitalizations, and Deaths, National New York, 2000, pp. 11–22.
Center for Injury Prevention and Control, Atlanta, GA, 2004. [16] C.W. Kuffel, Orthotic modeling of the developing skull, Journal of Prosthetics &
[3] A. Marr, V. Coronado, Central Nervous System Injury Surveillance Data Submis- Orthotics 16 (4S) (2004) 15–17.
sion Standards-2002, Control Centers for Disease Control and Prevention National [17] A. Rosenberg, Bone, joints, and soft tissue tumors, in: R.S. Cotran, V. Kumar, T.
Center for Injury Prevention and Control, Atlanta, GA, 2004. Collins (Eds.), Robbins Pathologic Basis of Disease, W.B. Saunders Company,
[4] A.J. Schneier, et al., Incidence of pediatric traumatric brain injury and associated Philadelphia, 1999, pp. 1215–1268.
hospital resource utilization in the United States, Pediatrics 118 (2006) 483–492. [18] J.L. Maguire, et al., Should a head-injured child receive a head CT scan? A
[5] P.D. Adelson, P.M. Kochanek, Head injury in children, Journal of Child Neurology systematic review of clinical prediction rules, Pediatrics 124 (1) (2009) 145–154.
13 (2) (1998) 2–15. [19] M.T. Prange, J.F. Luck, A. Dibb, C. Van Ee, R.W. Nightingale, B.S. Myers, Mechanical
[6] D. Graham, A. Lawrence, J. Adams, Fatal head injury in children, Journal of Clinical Properties and Anthropometry of the Human Infant head, The Stapp Car Crash
Pathology 42 (1988) 18–22. Journal 48 (2004) 279–299.