DEVELOPMENTAL MEDICINE & CHILD NEUROLOGY REVIEW

A systematic review of tests to predict cerebral palsy in young

children
MARGOT BOSANQUET 1,2,3 | LISA COPELAND1 | ROBERT WARE 3,4 | ROSLYN BOYD 2,3,5
1 Department of Rehabilitation Medicine, Royal Children’s Hospital; 2 Department of Child Health, The University of Queensland; 3 Queensland Children’s Medical
Research Institute, University of Queensland; 4 School of Population Health, University Of Queensland; 5 Queensland Cerebral Palsy and Rehabilitation Research
Centre, The University of Queensland, Brisbane, QLD, Australia.
Correspondence to Dr Margot Bosanquet, Department of Paediatric Rehabilitation Medicine, Level 2, Coles Building, Royal Children’s Hospital, Herston Road, Herston, QLD 4029, Australia.
Email: margot_bosanquet@health.qld.gov.au

This article is commented on by Hanna on pages 397–398 of this issue.

PUBLICATION DATA AIM This systematic review evaluates the accuracy of predictive assessments and
Accepted for publication 27th October 2012. investigations used to assist in the diagnosis of cerebral palsy (CP) in preschool-age children
(<5y).
ABBREVIATIONS METHOD Six databases were searched for studies that included a diagnosis of CP validated
CUS Cranial ultrasound after 2 years of age. The validity of the studies meeting the criteria was evaluated using the
GMA General movements Standards for Reporting Diagnostic Accuracy criteria. Where possible, results were pooled
assessment and a meta-analysis was undertaken.
IVH Intraventricular haemorrhage RESULTS Nineteen out of 351 studies met the full inclusion criteria, including studies of
STARD Standards for Reporting general movements assessment (GMA), cranial ultrasound, brain magnetic resonance
Diagnostic Accuracy imaging (MRI), and neurological examination. All studies assessed high-risk populations
including preterm (gestational range 23–41wks) and low-birthweight infants (range 500–
4350g). Summary estimates of sensitivity and specificity of GMA were 98% (95% confidence
interval [CI] 74–100%) and 91% (95% CI 83–93%) respectively; of cranial ultrasound 74% (95%
CI 63–83%) and 92% (95% CI 81–96%) respectively; and of neurological examination 88% (95%
CI 55–97%) and 87% (95% CI 57–97%) respectively. MRI performed at term corrected age (in
preterm infants) appeared to be a strong predictor of CP, with sensitivity ranging in
individual studies from 86 to 100% and specificity ranging from 89 to 97% There was
inadequate evidence for the use of other predictive tools.
SUMMARY This review found that the assessment with the best evidence and strength for
predictive accuracy is the GMA. MRI has a good predictive value when performed at term-
corrected age. Cranial ultrasound is as specific as MRI and has the advantage of being
readily available at the bedside. Studies to date have focused on high-risk infants. The
accuracy of these tests in low-risk infants remains unclear and requires further research.

Cerebral palsy (CP) occurs in 1 to 3% of live-born
infants.1–3 This increases to 8 to 40% in high-risk popula-
tions such as extremely preterm infants.4 CP is an umbrella
term for conditions that are characterized by a non-
progressive, but not unchanging, motor impairment related
to brain injury early in development.5–7 The brain injury is
characteristically static and results in activity limitation.7
Given the diversity of the clinical picture and implications
on management, CP is a diagnosis that is further described
by motor type, distribution, functional severity, and co-
morbidities.8
Metabolic, syndromic, and genetic conditions that
appear similar to CP may present in early life, with a pic-
ture of motor impairment. Differentiating these from CP
early on has genetic and treatment implications.8 When
CP is diagnosed early, surveillance programmes for muscu-
loskeletal and other comorbidities can be instigated and
early intervention programmes initiated.9–11 From the child
and family’s perspective, receiving a diagnosis of CP not
only provides ‘an answer’ but also streamlines appropriate
funding and social support. Diagnosis before school age
maximizes opportunities for appropriate physical and
learning support to be in place for the child’s transition to
schooling.
The age at which an accurate diagnosis of CP can be
made is, however, contentious. There is a paucity of data
on the rates of misdiagnosis of CP within CP registers.3
The data that are available suggest that misdiagnoses or
false-positive diagnoses of CP are more commonly made
before 18 months of age.3,12,13 Commonly reported false-
positive diagnoses include transient neurological signs of
infancy (usually dystonia and hypotonia) and other neuro-

418 DOI: 10.1111/dmcn.12140 © The Authors. Developmental Medicine & Child Neurology © 2013 Mac Keith Press
developmental disorders such as global developmental
delay and cognitive impairment.3,4,14 A proposed age at
which CP can be diagnosed is 36 months, when motor
capacity is more reliably assessed.15 CP can be difficult to
differentiate from progressive conditions such as the
leukoencephalopathies and hereditary causes of spastic
paraparesis, which have variable phenotypes and rates of
motor deterioration. The progression of these conditions
may not be evident in early life, and, therefore, differentia-
tion from CP at an early age may be clinically impossible.
There are many investigatory modalities that may assist
in early accurate diagnosis of CP, including early motor
assessment tools, brain imaging (cranial ultrasound [CUS]
and magnetic resonance imaging [MRI]) and neurological
examination (more commonly the standardized Lacey
Assessment of Preterm Infants and the Hammersmith
Infant Neurologic Examination).16,17
To date there has been no systematic review that has
examined the predictive accuracy of all available tools used
to assist in diagnosing CP in young children. There are
three systematic reviews on single modalities including
neuromotor assessments (outcome at 12mo),18 General
movements assessment alone (outcomes from 1–18y),19 and
brain MRI (diagnosis validated after age 2y).20 There are
also existing broader reviews of GMA.21
The purpose of this review was to search the literature
systematically for investigations and assessments that are
used to assist in the accurate diagnosis (diagnosis validation
after age 2y) of CP in children aged less than 5 years.
METHOD
Search strategy
A comprehensive search was undertaken of computerized
databases including MEDLINE via Ovid (1966–April
2012), CINAHL (1982–April 2012), PubMed (1980–April
2012), EMBASE (1988–June 2011), Cochrane Library
(1972–April 2012), PsycINFO (1806–April 2012), and
PEDRO (1999–April 2012). The search strategy comprised
the MeSH headings or keywords detailed in Appendix SI
(online supporting information).
Inclusion criteria
Studies were eligible for inclusion if they were written in
the English language and human based. We included ran-
domized-controlled trials and cohort trials (retrospective or
prospective) with an acceptable prevalence of CP (2–40%
[SD 10%] in the preterm population and 1–5% in the term
population).1,22–24 To be included, studies needed to con-
firm a diagnosis of CP after the age of 2 years, when
motor function is considered more stable and myelination
is largely complete.25 As males are known to be at higher
risk of CP (accounting for 55% of the total population),26
a sex bias of 30% towards males was allowed for.
Exclusion criteria
Studies were excluded if they were case reports or if they
assessed progressive and/or neurodegenerative disease or
What this paper adds
• This is the first systematic review to evaluate the accuracy of multiple
modalities used to predict a diagnosis of CP
• The strongest evidence and highest predictive accuracy is for general move-
ments assessment.
• The studies included in this review provide evidence for predictive accuracy
of tests in high-risk infants only.
had unusual population demographics, including a signifi-
cant sex imbalance or unusually high or low rates of CP.
Data extraction and analysis
The title and abstracts of the studies were independently
examined for suitability by two reviewers (MB, LC) and
were checked by a third independent reviewer (RB). The
full text of suitable studies was then retrieved and further
assessed for conformity to inclusion criteria. Study validity
was analysed using the Standards for Reporting Diagnostic
Accuracy (STARD) criteria with added specific questions
for this review (Appendix SIIa, online supporting informa-
tion).27 Demographic variables collected included median
and range of gestational age, birthweight, sex, age at diag-
nosis, age at confirmed diagnosis, classification and distri-
bution of CP, identified differential diagnosis, and how
differential diagnoses were handled (Table I and Table SI,
online supporting information). For analyses of diagnostic
accuracy, the investigatory modalities were separated. For
each study the sensitivity (the proportion of true positives,
i.e. the proportion of children with CP correctly identified
by the test), specificity (the proportion of true negatives, i.e.
the proportion of children without CP correctly identified
by the test), positive likelihood ratio (how the odds of hav-
ing CP change when a test is positive), and negative likeli-
hood ratio (how the odds of having CP change when a test
is negative) were either directly extracted or calculated from
extracted data (Table SII, online supporting information).
Pooling of results for meta-analysis was undertaken
where possible. For each meta-analysis, the pooled sensitiv-
ity, specificity, positive and negative likelihood ratios, and
the diagnostic odds ratio (DOR) were calculated. The
DOR is an overall single measure of test accuracy, com-
puted as the positive likelihood ratio divided by the nega-
tive likelihood ratio. A large DOR suggests that the
accuracy of the test is high; when a test is no better at
detecting CP than chance, the DOR is 1. Basic statistical
summary data were calculated using Microsoft Excel soft-
ware (Microsoft, San Diego, CA USA) and Descriptive
Sampling Statistics (Baltimore University; http://www.stat-
pages.org). Meta-analysis was undertaken using the ‘met-
andi’ command in Stata Statistical Software (StataCorp,
College Station, TX, USA).
RESULTS
A total of 351 articles were initially identified (Fig. 1). Nine-
teen articles, comprising a total of 4778 children (163 born
at term, 4462 born preterm, and 153 of unknown gestational
age at birth), met the inclusion criteria. The characteristics
of the included studies and their population demographics
Review 419
are shown in Table I. All studies evaluated patients at high
risk for developing CP. All studies except one28 were based
at tertiary neonatal units. Three studies reporting a high
prevalence of CP were included and were explainable by
biased population selection from high-risk groups.31,35,40
Study participants who were born preterm at gestational
ages varying from 23 to 38 weeks postmenstrual age with
reported median or mean birthweights ranging from 502
to 2184g had a pooled CP prevalence of 9%. The term-
born population had gestational ages of 37 to 41 weeks
and birthweights of 1501 to 4350g. None of the studies
which assessed both term and preterm infants described
CP prevalence. In the only study of solely term-born
infants, the reported prevalence of CP was 23%. This
study was of infants with hypoxic–ischaemic encephalo-
pathy. One study did not report on birthweight or sex.36
Sex distribution was mildly skewed to males, with the med-
ian being 53% (range 43–90). Sex was reported in 11 out
of 19 studies.32–34,36,37,39,43
Motor distribution was reported in nine studies (total
644 patients; Table I). It was described as monoplegia in
one patient (0.2%), diplegia in 267 patients (41.5%), hemi-
plegia in 180 (28%), quadriplegia/tetraplegia in 67
(10.4%), and of non-specified distribution in 129 patients
(20%). Eleven studies reported on the severity of CP (total
666 patients). Three studies reported using the Gross
Motor Function Classification System (GMFCS),45 result-
ing in 26 participants classified as GMFCS level I, 12 as
GMFCS level II, 11 as GMFCS level III, 18 as GMFCS
level IV, and 13 as GMFCS level V.34,35,39 Eight studies
used a mild, moderate and severe classification; in all cases
the study was undertaken before the validation of the
GMFCS.16,18,28,32,33,38,41,43 In total, CP was classified in
180 participants as mild, in 132 as moderate, in three as
moderately severe, and in 146 as severe; in 113 partici-
pants, severity of disease was not specified, more com-
monly in those with dyskinesia. All eight studies included a
clear definition of mild, moderate, and severe CP. We con-
sidered this classification alongside the validated and more
currently used GMFCS, with participants with mild CP
being defined as independent ambulators (GMFCS level I),
those with moderate CP as able to ambulate with assis-
tance (GMFCS levels II and III), and those with severe CP
being non-ambulant (GMFCS levels IV and V).46,47 Col-
lectively this resulted in CP being classified as mild in 209
participants (31.4%), as moderate in 155 (23.3%), and as
severe in 189 (28.4%) participants, with severity not speci-
fied in 113 (17%) participants. The timing of assessment
was highly variable between investigative modalities (from
day 1 of life to 5y 6mo).
Participant demographics including age at diagnosis,
duration of follow-up, and identification of differentials
from the individual studies are presented in Table SI.
Quality
The quantitative validity of the individual studies was eval-
uated according to the STARD criteria (Appendix SIIa).
420 Developmental Medicine & Child Neurology 2013, 55: 418–426
The scores for each study are shown in Appendix SIIb
(online supporting information) and Table SII. The med-
ian score on the STARD criteria was 16.6 from a maxi-
mum of 25 with a range of 9.0 to 22.5. The highest
quality was found for MRI studies (median 19.5, range
16.6–20.3) with the weakest quality shown for ‘individual
or other’ modality studies (median 12.9; range 10.9–15.5).
Accuracy of prediction of CP
The quantitative outcomes of accuracy of prediction of CP
(sensitivity, specificity, and positive and negative likelihood
ratios) from the studies are illustrated in Table SII,
grouped by investigatory modality.
General movements assessment
Six studies, with a total of 1358 participants, evaluated
GMA for diagnostic accuracy. Two studies described over-
all accuracy of GMA with estimates of sensitivity of 95 to
100% and estimates of specificity of 96 to 98%.41,42
One study described the predictive accuracy of cramped
synchronized general movements in preterm infants.
Cramped synchronized general movements may be defined
as abnormal movements which are characteristically rigid
without fluency or smoothness and involve almost simulta-
neous relaxation and contraction of all limb and trunk
muscles.6,50 The specificity of assessing these movements
was higher than that of assessing for abnormal movements
alone (92% vs 38% at 28wks postmenstrual age increasing
to 100% vs 83% at 47–60wks postmenstrual age).
At any age, an abnormal GMA has a very high negative
predictive value of 95 to 100%, with a negative likelihood
ratio of 0 (a negative test is excellent at predicting absence
of spastic CP.)18,35 One study provided examples of indi-
viduals in whom a normal GMA and mildly abnormal neu-
rological examination resulted in a diagnosis of mild CP.31
Furthermore, cramped synchronized general movements
alone were shown to have a very high negative predictive
value (100%) and positive predictive value (87–100%) for
later spastic CP31 (Table SII).
Two studies evaluated GMA in the fidgety period (46–
60wks postmenstrual age).18,29 Fidgety movements are
characteristically of small amplitude and moderate speed
with variable acceleration of neck, trunk, and limbs in all
directions, and are seen continually in the awake infant,
except during fussing and crying.6,50 Sensitivity and speci-
ficity in the preterm infant were shown to range from 87
to 100% and 82 to 95% respectively,18,42 and 94.4% and
82.5% respectively in the term infant.31 The correlation
between GMA and neurological examination was given in
one study as 81.5%.6 The interrater reliability, reported by
three studies, was 0.88 to 0.91.29,35,42 Test–retest agree-
ment was shown to be excellent in two studies, both
reporting it as 100%.18,42
Sufficient data to conduct a meta-analysis could be
extracted from four studies comprising 326 children, of
whom 93 (29%) had been diagnosed with CP.18,29,31,35
The pooled sensitivity and specificity were 98% (95% CI
 Table I: Study characteristics – population demographics
Population characteristics

Motor distribution
Age at
assessment Sex, Gestation Tetra/ Motor Severity of motor
Study ID Study design Modality n (PMA) male (%) PMA (wks) Birthweight (g) Monoplegia Diplegia Hemiplegia quadriplegia Other type functiona

29
Adde et al. P GM 74 50–58wks 33 (44.59) 30.5 (24–36)b 1367 (540–3800)b – – 5 4 1 9 Sp, 1 U x
Anderson P Corpus 41 2 USS >7d apart, 24/55 (43.63) 28 (23–33)c 1086 (630–1475)c x x x x x x x
et al.30 callosum <2wks, 6wks, term
growth per (CA). MRI at
ultrasound term (CA)
Bax et al.28 C, M population MRI 351 46mo (12–91)d 267 (76.1) 235 term, 47 <28wks, 19.1% <10th – 148 113 – 90 62 Dk, 17 A, 124 mild, 115
-based 69 28–31wks, centile’ 261 Sp, 11 U moderate, 95 severe,
79 32–36wks 17 not specified
Cioni et al.31 P, M (2 sites) GM, CUS, N 58 38–42, 43–47, 34 (58.62) 391 (37–41)d 3166 (SD 540) x x x x x x x
48–56, 57–65 (1870–4350)d
De Vries P, S CUS 1929 Day0–3, then weekly, x Group A: <32 x – 27 23 6 20 3 A, 73 Sp 24 mild, 1 moderate,
et al.32 (1460, 469)j and at term 22 severe, 29 not
specified
Group B: 33–36 – 12 9 3 2 2 Dk, 24 Sp 7 mild, 9 severe, 10
not specified
Denays et al.33 P Spectroscopy 88 33–44 (38.4) x 43 preterm, 45 term 17 <1501g x x x x x 8 Sp, 1 D 2 mild, 5 moderate, 2
severe
Eriksson P MAI 175 5mo (CA), and 5.5y x Mean 28.2, (23–36) 1046 (519–1900) – 5 1 2 – x 5 GMFCS I, 1 III,
et al.34 1 IV, 1 V
Ferrari et al.35 P, M (2 centres) GM, N 84 Birth to 60, at 3–5wks 42 (50.00) 30.2 (SD 2.7)c 1410.14 (SD 456.71)c – 22 8 14 – 44 Sp 15 GMFCS I, 5 II,
intervals 5 III, 9 IV, 10 V
Harris36 R MAI,e Bayleys 153 3mo 15d–4mo 15d x x x x x x x x x x
Hope et al.37 R, Sf CUS 85 (17, 68) Day 1 (0–2)g x 27 (26–28)g 925 (800–1100)g x x x x x x x
1050 (825–1150)g
Lacey et al.16 P, S CUS, N 203 (65,138) <33 36 (55.38) 27.6 (SD 1.3)c 1132 (SD 194)c x x x x x x 3 mod severe, 1
moderate, 3 severe
>33 77 (55.80) 27.3 (SD 1.7)c 1012 (SD 245)c 13 mild, 4 moderate,
12 severe
Mirmiran P MRI 61 USS92 <30, 31 (50.82) 23–29 502–1240h x x x x x Sp 6, H1 4 moderate., 3 mild
et al.38 MRI 36–39
Nanba et al.39 P MRI 289 36–43 x 23–34 – x x x x x 37 Sp 6 GMFCS I, 7 II,
5 III, 8 IV, 11 V
Pierrat et al.40 P CUS, SEP 39 (SEP) 36+6 (30–40)d x 29+5 (SD 05) 1110 x x x x x x x
(USS) not given (25–36)d (SD 230)c
Prechtl et al.41 P, M GM, CUS 130 32 (29–38)g 78 (60.00) 32 (29–38)g 1660 (1245–2730)g 1 18 5 20 7 1 Dk, 50 Sp 2 mild, 49 not specified
Romeo et al.42 P GM, CUS, N 903 HINE 3, 6, 9, 12 PMA. 496 (54.93) 34.5 (SD 2.3)c 2184 (SD 543)c – 25 13 18 1 1 Dk, 56 Sp x
GM <4wks, every
34wks thereafter
Seme- P GM, CUS 112 CUS d0-d90, GM 57 (50.89) 33 (26–37)b 1975 (660–3820)b x 7 1 x 8 1 H, 8 Sp, 7 U 5 mild, 2 moderate, 1
Ciglenecki18 3mo (CA) severe, 8 not
specified
Szymonowicz P CUS 32 Day 0, 1, 2, 3, 4, x 27 (SD 4) (24–32)c 880 (SD 204) – 3 2 – – 5 Sp 3 mild, 2 severe
et al.43 then weekly (430–1250)c
Tudehope P CUS 146 x 63 (43.15) 29i 1170i x x x x x x x
et al.44
Total/range 16 P, 6 GM, 9 CUS, 4778 Day 0–91mo 52.91 23–41 800–4350 1 (0.2%) 267 180 (28.0%) 67 (10.4%) 129 (20.0%) 689k 666
1 C, 2 R 4 N, 3 MRI, (43.15–76.10)%b (41.5%)
5 other (coefficient of
variation 0.17)

a
Mild, no interference with function/GMFCS I; moderate, has ability to mobilize with assistance/GMFCS II or III; severe, non-ambulant,/GMFCS IV or V.52. bMedian (range). cMean SD. dMean (range). eMost had mild to severe hypoxic–ischaemic encephalopathy. fAll known to have
cranial ultrasonography abnormality in neonatal period. gMedian (interquartile range). hRange. iMedian. jOwing to change in available machinery, group 1 1990–1991, group 2 1992–2000. k580 (84.2%) spastic, 20 (2.9%) ataxic, 69 (10%) dyskintetic, 20 (2.9%) unknown. PMA, postmen-
strual age; P, prospective; GM, general movement assessment; Sp, spastic; U, unknown; x, not reported; USS, ultrasonography; CA, corrected Age; MRI, magnetic resonance imaging; C, cross-sectional; M, multicentre; Dk, dyskinetic; A, ataxia; CUS, cranial ultrasound; N, neurologi-
cal examination; S, stratified; D, dystonic; MAI, Motor Assessment of Infancy; GMFCS, Gross Motor Function Classification System; R, retrospective; SEP, sensory evoked potential; HINE, Hammersmith Infant Neurologic Examination; H, hypotonia.

Review
421
73–100%) and 91% (95% CI 83–95%) respectively. The
DOR was 453 (95% CI 18–11 495).
Cranial ultrasound (CUS)
Ten studies, with a total of 2644 patients, met the criteria
for evaluation of CUS (Table SII and Appendix
SII).16,18,31,37–44,51 There was variability in the specific
focus of this investigation and in the definition of normal
or low-risk versus high-risk infants. Two studies included
any abnormality, except congenital malformations, as a
risk.32,44 Seme-Ciglenecki18 defined low risk as normal,
grade 1 intraventricular haemorrhage (IVH), or transient
periventricular echodensities lasting less than 7 days, and
based measures of predictive accuracy on low versus high
risk. This study found that the risk attributed to grade 1
IVH did not significantly differ from normal CUS.
Studies that broadly examined the outcomes of normal
versus abnormal CUS had predictive accuracy of sensitivity
of 70 to 96% and specificity of 69 to 99%.18,31,38,40,41,44 Pre-
dictive accuracy was best when duration of CUS data collec-
tion was longer (sensitivity 70–76%, specificity 90–95%).18,32
Two studies evaluated the accuracy of CUS within the
first days of life (before day 2 and before day 7), during
which time 84% and 99% respectively, of cases of IVH
will occur.16,38,52,53 Sensitivity (43–44%) and specificity (82
–87%) were similar in these two studies (Table SII).
Central nervous system (CNS) malformation was
excluded from analysis in three studies with reported pre-
dictive specificity of 82 to 90%.16,18,31 Although the nega-
tive predictive value from these studies (88–89%) was less
than when CNS malformation was included (88–89% vs
97.6%), the likelihood ratios did not differ between studies
regardless of inclusion or exclusion of CNS malforma-
tion.16,18,44
A study by Hope et al.37 was unique in assessing com-
puter measures of texture in the white matter and choroid
plexus (one image only) as a predictor of CP. Notably, the
sensitivity and specificity of these measures were lower
than the sensitivity and specificity of measures used in
other studies.
Only one study evaluated the benefit of CUS in high-
risk infants born at term, of whom the majority had
hypoxic–ischaemic encephalopathy. In this study, CUS
predicted CP with a sensitivity of 88% and specificity of
83%.31
We were able to include in a meta-analysis data from six
studies, including a total of 2404 children, of whom 226
(9.4%) were diagnosed with CP.18,31,32,37,38,41 The pooled
sensitivity and specificity were 74% (95% CI 63–83%) and
92% (95% CI 81–96%) respectively. The DOR was 31
(95% CI 8–117).
Neurological examination
Four studies met the inclusion criteria for neurological
examination to identify CP (total n=1190).16,31,35,42 Partici-
pants in all studies were born preterm. In all studies,
except that by Lacey et al.,16 sequential assessments
422 Developmental Medicine & Child Neurology 2013, 55: 418–426
were performed until 5 to 12 months corrected age (four
assessments at age <12mo and reassessed at 2y). When per-
formed before term age (<37wks), the sensitivity of neuro-
logical examination ranged from 57 to 86%, with a
specificity of 45 to 83%.16,35 Neurological examination in
the post-term age group ranged from 68 to 96%, with spec-
ificity of 52 to 97%. Romeo et al.42 investigated the Ham-
mersmith Infant Neurologic Examination and found that
there was inconsistency in the cut-off between ‘normal’ and
‘abnormal’, with scores of less than 50 and less than 57
both being used. Outcomes, sensitivity, and specificity for
children with scores of less than 50 were not disclosed. One
study reported on the value of transient neurological abnor-
malities, with 34% of patients developing motor impair-
ment with an overall negative predictive value of 95%.16
Likelihood ratios could not be calculated, and sensitivity
and specificity were not reported in this study.
We were able to include data from two studies in a
meta-analysis, covering a total of 142 children, of whom
58 (41%) were diagnosed with CP.31,35 The pooled sensi-
tivity and specificity were 88% (95% CI 59–97%) and
87% (95% CI 57–97%) respectively. The DOR was 49
(95% CI 15–158).
Magnetic resonance imaging
Three studies met inclusion criteria (n=702). Two studies
were performed at term corrected age in the preterm
infant38,39 and the other at a median age of 46 months in a
population-based cohort.28 The sensitivity and specificity
of term MRI in predicting CP at 31 months, excluding
those with CNS malformation, was 86% and 89% respec-
tively.38 One study focused assessment of abnormality of
the corona radiata and corticospinal tract (i.e. deep white
matter) and found 100% sensitivity and 97% specificity in
prediction of CP,39 exceeding the predictive value of peri-
ventricular cystic lesions (Table SII). However, the exclu-
sion criteria in this study were extensive, with it essentially
restricted to analysis of white matter damage of immatu-
rity, periventricular leukomalacia, and cystic change.39
A large European population-based study (n=351) by
Bax et al.28 aimed to describe patterns of injury and CP
distribution and severity and also described those patients
with CP in whom MRI revealed no findings. This study
reported exclusively on children with a diagnosis of CP
validated before the age of 3 years and excluded those
without an established diagnosis. Magnetic resonance
imaging changes were seen in 88% of children, with MRI
reported as normal in 12%. In this study the most com-
mon finding in children with CP was white matter damage
of immaturity (including periventricular leukomalacia,
affecting 43% of the cohort), with 46% of the study popu-
lation (total n=351) having been born preterm. Other MRI
findings were basal ganglia lesions (13%), cortical/subcorti-
cal lesions (9%), malformations (9%), focal infarct (7%),
and miscellaneous lesions (7%). The highest rate of normal
MRI was found in children with ataxic CP (8 out of 17);
the prevalence of normal MRI in the other subtypes was

351 Studies met search
criteria based on title and
abstract
60 Articles warranted close
assessment of the full
publication
19 Articles met inclusion
criteria
Figure 1: Flow chart for process of article inclusion.
approximately equal. MRI was more often normal in
children with mild CP (61%).28 This was the only MRI
study that included CNS malformations.28 Further results
available from this study included correlates of MRI brain
injury with clinical disease distribution.28
All the MRI studies in this review excluded IVH in the
knowledge that the investigation as being performed after
the time when IVH commonly occurs. No studies were
suitable for pooling for a meta-analysis.
Other assessments
Two studies examined the value of the Motor Assessment
of Infancy, which was performed between the ages of 3
months and 5 months.34,36 Both studies were of relatively
low quality, with STARD values of 15.5 and 11.5.34,36
Only one of the studies analysed data for predictive accu-
racy. The studies were not suitable for pooling for a meta-
analysis owing to the diversity of the predictive modalities
and variable primary data content.
DISCUSSION
This systematic review has shown that there is good evi-
dence that GMA can accurately predict the development of
CP.18,31,35,41,42 There is reasonable evidence to support the
use of MRI at term corrected age, neurological examina-
tion in the older infant, and, to a lesser extent, ultrasound
in infants of preterm age.
All included studies were performed in high-risk popula-
tion groups with a predominance of preterm infants. Only
one study evaluated term-born patients, who are under-
represented in predictive analysis, despite accounting for
68% of the CP population.54 Thus, overall predictive accu-
290 Articles excluded that did not meet the
inclusion criteria
41 Excluded not meeting strict inclusion criteria
(Appendix SIII, online supporting information):
• Inadequate duration of follow-up (18)
• Not examining CP as an individual
outcome, or using a predictive modality
(14)
• Inadequate study quality (5)
References (first author): Adde et al.,29
Anderson,30 Bax et al.,28 Cioni et al.,31 De Vries
et al.,32 Denays et al.,33 Eriksson et al.,34 Ferrari et
al.,35 Harris,36 Hope et al.,37 Lacey et al.,16 Mirmiran
et al.,38 Nanba et al.,39 Pierrat et al.,40 Prechtl et
al.,41 Romeo et al.,42 Seme-Ciglenecki,18
Szymonowicz et al.,43 Tudehope et al.44
racy for these tests cannot be extrapolated to the general
population. Other than this review, there is a paucity of
published literature on tests of accurate prediction of CP
in ‘low-risk’ populations.
Collective evidence confirms that GMA is the most sen-
sitive and specific test currently available to allow early
prediction of spastic CP in high-risk infants. There is
evolving evidence that GMA may predict distribution of
CP but not predict severity, in which case it cannot replace
accurate history, examination, and long-term follow-up.55
GMA can be performed at the bedside or remotely, includ-
ing retrospectively by video recording. However, assessors
need to complete standardized training and certification to
achieve maximum accuracy in assessments.21 Access to
training courses is readily available in Europe, where GMA
is more often routinely incorporated in infant follow-up;
however, it may be difficult to access elsewhere. This tool
requires minimal equipment and staff time to be under-
taken, and for that reason provides exceptional value for
effort in the early accurate prediction of CP.
General movements assessment has benefit over the
Motor Assessment of Infancy, with significantly better pre-
dictive accuracy (sensitivity 87–100% vs 74% respectively;
specificity 82–100% vs 63% respectively). GMA is also a
more clinically feasible tool. GMA takes 15 minutes of
gestalt observation compared with 1 hour of directed task
requiring cognitive, behavioural ability, and compliance,
for the Motor Assessment of Infancy. GMA at a minimum
requires a single episode of 15 minutes of observation in
the fidgety movement stage (2–4mo corrected age) and
has been validated in both term-born infants with hypoxic–
ischaemic encephalopathy and preterm-born infants.
Review 423
 Evidence for neurological examination is strong in qual-
ity for the early accurate diagnosis of CP. Neurological
examination and tools that incorporate it are clinically
influenced by the experience of an examiner and a child’s
state of rest. There are a number of proposed neurological
examination protocols, which share features of assessment
of tone and paucity of movement. Such protocols include
the Hammersmith Infant Neurologic Examination,17 the
Touwen Infant Neurological Examination56 and the Lacey
Assessment of Preterm Infants;16 there remains no consen-
sus on which is the best one to use. This review supports
the view that neurological examination is less valuable at
preterm age, when GMA has better predictive accuracy.
Neurological examination detects signs that may be tran-
sient and which have been more commonly observed in
preterm-born infants, including jitteriness and dystonia.
Such signs are poor predictors of CP.14,32 There is a lack
of evidence on the reproducibility and interuser reliability
of neurological examination. Neurological examination
remains invaluable in assessment of clinical disease distri-
bution and comorbid neurological impairment, which may
impact on motor development and function (e.g. vision
and balance). There is possible value in the prediction of
milder cases of CP, in which GMA is less sensitive, with
two such cases being described in a single study in this
review.31
Magnetic resonance imaging shows usefulness at term
age for accurate prediction of CP (sensitivity 86–100%,
specificity 87–97%).38,39 It is better at detecting changes in
white matter at term age than CUS in high-risk infants
(sensitivity for MRI at term 97–100%).39 Before term age
in high-risk infants, there remains insufficient evidence to
support use of MRI over CUS.
There is an additional benefit of MRI with prediction of
distribution, with established evidence on clinical corre-
lates.20,28 There is also evolving evidence, although more
scarce, for MRI in predicting the severity of CP.39 There
is no evidence as yet for MRI used before term corrected
age. MRI requires the infant to remain still for up to 30
minutes (longer if the spine is also imaged), which affects
the clinical usefulness of this test, especially in the preterm
infant who may be clinically unstable. The provision of
MRI-compatible ‘humidicribs’ may help to increase the
number of infants at risk of CP who can be imaged with
minimal disruption to the patient. MRI has possible future
benefit in predicting the broader clinical picture associated
with CP in which patterns of injury may provide descrip-
tive risk of visual dysfunction, learning, and behavioural
disturbance, and likelihood of responding to antispasticity
medications.9,28
Magnetic resonance imaging should always be consid-
ered when clinical findings do not match the history, as it
can show changes pathognomonic of metabolic and genetic
disease or incidental findings, such as ventricular enlarge-
ment. Suspicion regarding aetiology should remain for
those children in whom MRI is normal, who can account
for 11.7 to 15% of children with CP. It is important that
424 Developmental Medicine & Child Neurology 2013, 55: 418–426
both groups of patients receive close follow-up with neuro-
logical assessment and ongoing review of developmental
progression. No article in this review reported on or
included spinal MRI findings in its analysis.
No study investigating advanced neuroimaging tech-
niques, including diffusion-weighted imaging, spectros-
copy, or tractography, met the criteria for inclusion in this
systematic review. Diffusion-weighted MRI may have a
role in early prediction of CP, including disease distribu-
tion and motor type; however, this needs further investiga-
tion. So far, evidence for this comes only from case
studies, a number of which show that diffusion-weighted
MRI provides earlier and more sensitive detection of white
matter change,50,57–59 including neonatal stroke.60 Diffu-
sion-weighted imaging has also been shown to assist in the
earlier differentiation of metabolic leukodystrophies.61
Cranial ultrasound studies included in this review were
overall of weaker quality than MRI studies; however, they
reported a similarly high specificity (87–97% and 90–99%
respectively).18 CUS is a widely used and invaluable tool in
the neotnatal intensive care setting as it is readily available
and well tolerated in high-risk infants. Best predictive
accuracy of CUS is available from sequential scanning of
preterm infants up to term age.16,18,38
There was no evidence for the predictive value of CUS
in the term-born infant in relation to outcome related to
CNS malformations. The CUS and MRI studies in pre-
term-born infants also did not report on the outcomes of
CNS malformation but included it as a possible aetiology.
This may be of little value, as CNS malformation is a less
common cause of CP in the preterm population (3% vs
16% in term-born infants).20,62 The predictive value of
early CUS may be better than reported in this review as
half of the included studies were performed more than a
decade ago and technological advancement since that time
is likely to give more accurate results.
Missed diagnoses occurred more commonly in children
with mild CP.31 No included study reported the rate of
misdiagnosis. One study evaluated the rate of misdiagnosis
in a population-based cohort.3 The rate was found to be
higher in those diagnosed before 24 months of age (71%),
with a total rate of misdiagnosis of 5% (total n=402). As
most studies excluded patients without a diagnosis of CP,
there were also insufficient data to report on the total rate
of false-positive diagnoses. Differential diagnoses which
were evident after validation of diagnosis after age 2 years,
were developmental retardation, developmental delay,
hypotonia, myotonic dystrophy, intellectual disability, mild
disability, developmental coordination disorder, and minor
neurological dysfunction (Table SI). There was also poor
reporting of the initial age at diagnosis of CP.
Clinical implications
The evidence from this review comes from high-risk
infants, predominantly those born preterm. This review
supports the notion that the tools discussed can assist in
the prediction of CP, which is valuable in prognostic
discussions with families and also allows for the benefits of
early intervention to be provided. GMA stands out as one
of the most cost-effective and sensitive tools available for
early prediction of CP. However, no single tool can result
in a clinical diagnosis and there is no substitute for the
combination of frequent follow-up, clinical assessment,
consideration of differential diagnoses, and considered
investigation. The use of any investigative tool should
remain sensitive to the child’s age and ability. This is per-
haps overlooked by the structure of some of the neonatal
follow-up studies, in which young children are subject to
prolonged assessments, which may confound results.
Research implications
This review assessed only those studies assessing diagnostic
accuracy after the age of 2 years. Given that reported mis-
diagnosis is more commonly not picked up until after age
5 years, there would be benefit in assessing the longer-
term predictive accuracy of all of these measures.3 There is
also a distinct lack of description of diagnostic accuracy,
including classification, within the general population,
which would also benefit from further assessment.
Limitations
The studies in this review included high-risk populations
only.
There is a greater volume of evidence for predictive
tests, including many other motor assessment tools, which
have been used in the under 2 years age group; however,
in these studies, diagnostic confirmation after the age of 2
was not performed, and therefore these studies could not
be included. Many papers look at neurodevelopment out-
comes and may assess tools of value in predicting CP;
however, as CP was not described as an outcome, these
studies also needed to be excluded from this review.
CONCLUSION
This study provides evidence predominantly to support the
use of predictive tools in the high-risk preterm population,
including GMA, CUS, neurological examination, and
MRI. It does not tell us the value of these tools have in
the general population or those with underlying CNS mal-
formation, as this was excluded from most studies. There
is evidence that milder cases of CP are more commonly
missed by the more common predictive or screening tools
used in follow-up programmes.54
A CK N O W L E D G E M E N T S
The authors acknowledge the Royal Children’s Hospital Founda-
tion/Golden Casket Clinical Fellowship Grant (No. 10286), the
Queensland Children’s Medical Research Institute, the NHMRC
Career Development Award (Number 473840), and the QLD
Government Smart State Fellowship.
SUPPORTING INFORMATION
Additional supporting information may be found in the online
version of this article:
Table SI: Confirmation of diagnosis and differential diagnosis
Table SII: Qualitative study validity (Standards for Reporting
Diagnostic Accuracy).
Appendix SI: Search terms used (MeSH or headings)
Appendix SII: (a) Standards for Reporting Diagnostic Accu-
racy questionnaire. (b) Quantitative validity of studies per Stan-
dards for Reporting Diagnostic Accuracy criteria.
Appendix SIII: Excluded articles.

REFERENCES
1. Pharoah PO, Cooke T, Johnson MA, King R, Mutch L.
Epidemiology of cerebral palsy in England and Scot-
land, 1984–9. Arch Dis Child Fetal Neonatal Ed 1998; 79:
F21–5.
2. Cans C. Surveillance of cerebral palsy in Europe: a col-
laboration of cerebral palsy surveys and registers. Dev
Med Child Neurol 2000; 42: 816–24.
3. Zarrinkalam R, Russo RN, Gibson CS. CP or not CP?
A review of diagnoses in a cerebral palsy register. Paedi-
atr Neurol 2010; 42: 177–80.
4. Stephens B, Vohr BR. Neurodevelopmental outcome of
the premature infant. Pediatr Clin N Am 2009; 56: 631–
46.
5. Bax M. Terminology and classification of cerebral palsy.
Dev Med Child Neurol 1964; 6: 295–307.
6. Mutch L. Cerebral palsy epidemiology: where are we
now and where are we going? Dev Med Child Neurol
1992; 34: 547–51.
7. Rosenbaum P, Paneth N, Leviton A. The definition and
classification of cerebral palsy. Dev Med Child Neurol
2007; 49(Suppl. 109): 8–14.
8. Sanger T. The definition and classification of cerebral
palsy: a neurological perspective on definition and classi-
fication. Dev Med Child Neurol 2007; 49(Suppl. 109): 30
–1.
9. Spittle AJ, Boyd RN, Inder TE, Doyle LW. Predicting
motor development in very preterm infants at 12
months’ corrected age: the role of qualitative magnetic
resonance imaging and general movements assessments.
Paediatrics 2009; 123: 512–7.
10. Spittle AJ, Doyle LW, Boyd RN. A systematic review of
the clinimetric properties of neuromotor assessments for
preterm infants during the first year of life. Dev Med
Child Neurol 2008; 50: 254–66.
11. Blauw-Hospers C, Hadders-Algra M. A systematic
review of the effects of early intervention on motor
development. Dev Med Child Neurol 2005; 47: 421–32.
12. Nelson KE, Ellenberg JH. Children who ‘outgrew’ cere-
bral palsy. Pediatrics 1982; 69: 529–36.
13. Nelson KE, Ellenberg JH. Antecedents of cerebral palsy:
multivariate analysis of risk. N Engl J Med 1986; 315: 81
–6.
14. Palmer F. strategies for the early diagnosis of cerebral
palsy. J Paediatr 2004; 145(Suppl. 2): S8–11.
15. McLaughlin J. The definition and classification of cere-
bral palsy: definition of cerebral palsy – clinical perspec-
tive. Dev Med Child Neurol 2007; 49(Suppl. 109): 27–8.
16. Lacey J, Rudge S, Rieger I, Osborn DA. Assessment of
neurological status in preterm infants in neonatal inten-
sive care and prediction of cerebral palsy. Aust J Physio-
ther 2004; 50: 137–44.
17. Haataja L, Mercuri E, Regev R, et al. Optimality score
for the neurological examination of the infant at 12 and
18 months of age. J Pediatr 1999; 135: 153–61.
18. Seme-Ciglenecki P. Predictive values of cranial ultra-
sound and assessment of general movements for neuro-
logical development of preterm infants in the Maribor
region of Slovenia. Wien Klin Wochenschr 2007; 119: 490
–6.
19. Darsaklis V, Snider L, Majnemer A, Mazer B. Predictive
validity of Prechtl’s Method on the qualitative assess-
ment of general movements: a systematic review of the
evidence. Dev Med Child Neurol 2011; 53: 896–906.

Review 425
20. Krageloh-Mann I, Horber V. The role of magnetic reso-
nance imaging in elucidating the pathogenesis of cere-
bral palsy: a systematic review. Dev Med Child Neurol
2007; 49: 144–51.
21. Einspieler C, Prechtl HFR. Prechtl’s assessment of gen-
eral movements: a diagnostic tool for the functional
assessment of the young nervous system. Ment Retard
Dev Disabil Res Rev 2005; 11: 61–7.
22. Meberg A, Broch H, Irgens LM. Cerebral palsy as indi-
cator of quality of neonatal care. Tidsskr Nor Laegeforen
2001; 121: 2917–22.
23. Wu YW, Croen LA, Shah SJ, Newman TB, Najjar DV.
Cerebral palsy in a term population: risk factors and
neuroimaging findings. Pediatrics 2006; 118: 690–7.
24. Liu JM, Li S, Lin Q, Li Z. Prevalence of cerebral palsy
in China. Int J Epidemiol 1999; 28: 949–54.
25. Paus M, Zijdenbos A, Worsley K, et al. Structural matu-
ration of neural pathways in children and adolescents: in
vivo study. Science 1999; 283: 1908–11.
26. Jarvis S, Glinianaia S, Arnaud C, et al. Case gender and
severity in cerebral palsy varies with intrauterine growth.
Arch Dis Child 2005; 90: 474–9.
27. Bossuyt P, Reitsma JB, Bruns DE, et al. Towards com-
plete and accurate reporting of studies of diagnostic accu-
racy: the STARD initiative. Clin Chem 2003; 49: 1–6.
28. Bax M, Tydeman C, Flodmark O. Clinical and MRI
correlates of cerebral palsy. JAMA 2006; 296: 1602–8.
29. Adde L, Rygg M, Lossius K, Oberg GK, Stoen R. Gen-
eral movement assessment: predicting cerebral palsy in
clinical practice. Early Hum Dev 2007; 83: 13–8.
30. Anderson NG, Laurent I, Woodward LJ, Inder TE.
Detection of impaired growth of the corpus callosum in
premature infants. Pediatrics 2006; 118: 951–60.
31. Cioni G, Prechtl HFR, Ferrari F, Paolicelli P, Einspieler
C, Roversi MF. Which better predicts later outcome in
full term infants: quality of general movements or neu-
rological examination? Early Hum Dev 1997; 50: 71–85.
32. De Vries LS, Van Haastert I-LC, Rademaker KJ, Koop-
man C, Groenendaal F. Ultrasound abnormalities pre-
ceding cerebral palsy in high-risk preterm infants.
J Pediatr 2004; 144: 815–20.
33. Denays R, Van Pachterbeke T, Toppet V, et al. Predic-
tion of cerebral palsy in high-risk neonates: a techne-
tium-99m-HMPAO SPECT study. J Nucl Med 1993;
34: 1223–7.
34. Eriksson C, Katz-Salamon M, Brogren Carlberg E.
Early motor assessment in very preterm born infants as
predictor of performance at 5.5 years. Adv Physiother
2006; 8: 175–81.
35. Ferrari F, Cioni G, Einspieler C, et al. Cramped syn-
chronized general movements in preterm infants as an
early marker for cerebral palsy. Arch Pediatr Adolesc Med
2002; 156: 460–7.
426 Developmental Medicine & Child Neurology 2013, 55: 418–426
36. Harris SR. Early detection of cerebral palsy: sensitivity
and specificity of two motor assessment tools. J Perinatol
1987; 7: 11–5.
37. Hope TA, Gregson PH, Linney NC, Schmidt MH, Ab-
dolell M. Selecting and assessing quantitative early ultra-
sound texture measures for their association with
cerebral palsy. IEEE Trans Med Imaging 2008; 27: 228–
36.
38. Mirmiran M, Barnes PD, Keller K, et al. Neonatal brain
magnetic resonance imaging before discharge is better
than serial cranial ultrasound in predicting cerebral palsy
in very low birth weight preterm infants. Pediatrics 2004;
114: 992–8.
39. Nanba Y, Matsui K, Aida N, et al. Magnetic resonance
imaging regional T1 abnormalities at term accurately
predict motor outcome in preterm infants. Pediatrics
2007; 120: e10–9.
40. Pierrat V, Eken P, de Vries LS. The predictive value of
cranial ultrasound and of somatosensory evoked poten-
tials after nerve stimulation for adverse neurological out-
come in preterm infants. Dev Med Child Neurol 1997;
39: 398–403.
41. Prechtl HF, Einspieler C, Cioni G, Bos AF, Ferrari F,
Sontheimer D. An early marker for neurological deficits
after perinatal brain lesions. Lancet 1997; 349: 1361–3.
42. Romeo DMM, Guzzetta A, Scoto M, et al. Early neuro-
logical assessment in preterm-infants: integration of tra-
ditional neurological examination and observation of
general movements. Eur J Paediatr Neurol 2008; 12: 183
–9.
43. Szymonowicz W, Yu VY, Bajuk B, Astbury J. Neuro-
developmental outcome of periventricular haemorrhage
and leukomalacia in infants 1250 g or less at birth. Early
Hum Dev 1986; 14: 1–7.
44. Tudehope DI, Masel J, Mohay H, et al. Neonatal cra-
nial ultrasonography as predictor of 2 year outcome of
very low birthweight infants. Aust Paediatr J 1989; 25:
66–71.
45. Palisano R, Rosenbaum P, Walter S, Russell D, Wood
E, Galuppi B. Development and reliability of a system
to classify gross motor function in children with cerebral
palsy. Dev Med Child Neurol 1997; 39: 214–23.
46. Rosenbaum P, Walter S, Hanna S, et al. Prognosis for
gross motor function in cerebral palsy. JAMA 2002;
288: 1357–63.
47. Nagae LM, Hoon AH Jr, Stashinko E, et al. Diffusion
tensor imaging in children with periventricular leuko-
malacia: variability of injuries to white matter tracts.
AJNR Am J Neuroradiol 2007; 28: 1213–22.
48. Hagberg B, Hagberg G, Olow I. The changing pano-
rama of cerebral palsy in Sweden 1954–1970. II. Analysis
of the various syndromes. Acta Paed Scand 1975; 64: 193
–200.
49. Ellenberg JH, Nelson KB. Early recognition of infants
at high risk for cerebral palsy: examination at age 4
months. Dev Med Child Neurol 1981; 23: 705–16.
50. Son SM, Ahn YH, Sakong J, et al. Diffusion tensor
imaging demonstrates focal lesions of the corticospinal
tract in hemiparetic patients with cerebral palsy. Neurosci
Lett 2007; 420: 34–8.
51. de Vries LS, Eken P, Pierrat V, Daniels H, Casaer P.
Prediction of neurodevelopmental outcome in the pre-
term infant: short latency cortical somatosensory evoked
potentials compared with cranial ultrasound. Arch Dis
Child 1992; 67: 1177–81.
52. Ment L, Ehrenkranz R, Duncan C, et al. Low-dose
indomethacin and prevention of intraventricular hemorr-
hare: a multicenter randomized trial. Pediatrics 1994; 93:
543–50.
53. Ment L, Duncan C, Ehrenkranz R, et al. Intraventricu-
lar haemorrhage in the preterm neonate: timing and
cerebral blood flow changes. J Pediatr 1984; 104: 419–
25.
54. Australian Cerebral Palsy Register Group. Report of the
Australian Cerebral Palsy Register, Birth Years 1993–
2003. 2009.http://secure.cpregister-aus.com.au/pubs/pdf/
ACPR-Report.pdf, (accessed February 2012)
55. Guzetta A. Hand movements at 3 months predict later
hemiplegia in term infants with neonatal cerebral infarc-
tion. Dev Med Child Neurol 2010; 52: 767–72.
56. Touwen BCL. Neurological Development in Infancy.
Clinics in Developmental Medicine, No. 58. London,
England: William Heinemann Medical Books, 1976.
57. Son SM, Park SH, Moon HK, et al. Diffusion tensor
tractography can predict hemiparesis in infants with high
risk factors. Neurosci Lett 2009; 451: 94–7.
58. Berman J, Mukherjee P, Partridge S, et al. Quantitative
diffusion tensor MRI fibre tractography of sensorimotor
white matter development in premature infants. Neuro-
Image 2005; 27: 862–71.
59. Hart AAM. Diffusion-weighted imaging of cerebral
white matter and the cerebellum following preterm
birth. Dev Med Child Neurol 2010; 52: 652–8.
60. Kirton A, Shroff M, Visvanathan T, deVeber G. Quanti-
fied corticospinal tract diffusion restriction predicts neo-
natal stroke outcome. Stroke 2007; 38: 974–80.
61. Engelbrecht V, Scherer A, Rassek M, Witsack H, Mod-
der U. Diffusion-weighted MR imaging in the brain in
children: findings in the normal brain and in the brain
with white matter diseases. Radiology 2002; 222: 410–8.
62. Bax M, Flodmark O, Tydeman C. From syndrome
towards disease. Dev Med Child Neurol 2007; 49 (Suppl.
109): 39–41.