Magnetic resonance imaging and developmental outcome following preterm birth: review of current evidence


  • Anthony R Hart MRCPCH,

    Corresponding author
    1. Clinical Research Fellow, Neonatal Intensive Care Unit, Jessop Wing, Sheffield Teaching Hospitals NHS Trust
      * Correspondence to first author at Neonatal Intensive Care Unit, Jessop Wing, Sheffield Teaching Hospitals NHS Trust, Tree Root Walk, Sheffield S10 2SF, UK.
    Search for more papers by this author
  • Elspeth W Whitby PhD FFDRCSI,

    1. Senior Lecturer in Radiology
    Search for more papers by this author
  • Paul D Griffiths PhD FRCR,

    1. Professor of Radiology, Academic Unit of Radiology, University of Sheffield, Royal Hallamshire Hospital
    Search for more papers by this author
  • Michael F Smith FRCP FRCPCH

    1. Consultant Neonatologist, Neonatal Intensive Care Unit, Jessop Wing, Sheffield Teaching Hospitals NHS Trust, Sheffield, UK.
    Search for more papers by this author

  • See end of paper for list of abbreviations.

* Correspondence to first author at Neonatal Intensive Care Unit, Jessop Wing, Sheffield Teaching Hospitals NHS Trust, Tree Root Walk, Sheffield S10 2SF, UK.


Preterm birth is associated with an increased risk of developmental difficulties. Magnetic resonance imaging (MRI) is increasingly being used to identify damage to the brain following preterm birth. It is hoped this information will aid prognostication and identify neonates who would benefit from early therapeutic intervention. Cystic periventricular white matter damage has traditionally been associated with abnormal motor developmental and cerebral palsy, but its presence on MRI does not preclude normal cognitive development. This has led to increasing interest in the identification of diffuse periventricular white matter damage with conventional and sophisticated MRI. However, the correlation between these appearances and developmental outcome remains unclear. Measurements of the size, volumes, and growth rates of many regions of the brain, such as the corpus callosum, ventricular system, cortex, deep grey matter, and cerebellum, are all also altered following preterm birth, but there is insufficient evidence to use this data in the clinical setting. This article is a review of the current evidence on MRI and developmental outcome, suggesting possible indications for the use of MRI following preterm birth.

As neonatal intensive care techniques develop, infants born at the extreme of prematurity increasingly survive. However, preterm infants can experience developmental and neurological difficulties later in life. For example, neonates born under 33 weeks’ gestation have lower developmental scores and an increased risk of cerebral palsy (CP) between 2 and 6 years of age than term controls. Infants born under 25 weeks’ gestation are most likely to be affected.1–3 Ex-preterm individuals also display an increased incidence of behavioural and emotional problems at 5 years of age, such as attention-deficit–hyperactivity disorder (ADHD).4,5 Even neonates thought to be at low risk of developmental difficulties (such as those born between 30–34 weeks’ gestation with uncomplicated perinatal histories, normal cranial ultrasound scans, and no obvious neurodevelopmental deficits), can have subtle neuropsychological abnormalities later in life.6

Preterm birth is also associated with injury to the developing brain, which may cause permanent damage. The most common form of damage is periventricular white matter (PVWM) damage, which can be focal and cystic (also called cystic periventricular leukomalacia) or diffuse and non- cystic.7–13 The mechanism of damage to the PVWM is not fully understood. One possible mechanism is hypoxia–ischemia.7–9 Volpe suggested that PVWM tissue is susceptible to hypoxia–ischemia because it lies within the vascular end zones of immature penetrating arteries;7,8 although others argue there is insufficient evidence to support this theory.9 Either way, hypoxia–ischemia occurs when cerebral blood flow is reduced during periods of hypotension or hypocarbia.7–9 Neonates with a ‘pressure-passive cerebral blood flow’ are thought to be particularly susceptible to hypoxia–ischemia, as they are unable to auto-regulate their cerebral blood flow in response to changes in blood pressure.7,9

Hypoxia–ischemia injury is thought to cause necrosis and/or apoptosis of late precursor oligodendrocytes in the PVWM, which are extremely sensitive to this insult.7–9 Back et al. suggested that the predilection of PVWM to damage is not related to vascular development, but to the large number of late precursor oligodendrocytes in this area around the time of preterm birth.9 The mode of cell death is currently under investigation, but probably involves the influx of calcium into the cell and nucleus, release of glutamate and adenosine into the extracellular compartment, and production of oxygen and nitrogen free radicals by activated microglia.7,8,14–16

A second proposed mechanism of injury involves cytokines, produced in response to neonatal or fetal infection. These may directly cause oligodendrocyte precursor death or induce hypoxic–ischemic damage by making the neonate unstable cardiovascularly.17–19 Research into lipopolysaccharides has shown that bacterial endotoxins also make the preterm brain highly sensitive to hypoxic–ischemic injury.7,8 A third cause of PVWM damage is atrophy or poor growth, development, and myelination of the brain secondary to the initial insult.8,9

Aside from PVWM damage, other forms of injury to the preterm brain include germinal matrix and intraventricular haemorrhage (IVH), periventricular haemorrhagic infarction (PVHI), and cerebellar haemorrhage.

The ability to identify neonates at high risk of sub-optimal developmental would allow timely referral for therapy and treatment. A recent Cochrane review showed that early intervention following preterm birth can improve cognitive outcome up to preschool age, but not motor outcome up to 2 years of age.20 Currently, there is no accurate method of identifying all preterm neonates at risk of poor developmental outcome. Clinical risk factors predict long-term outcome poorly, even within a classification tree analysis.21

Cranial ultrasound is routinely used in neonatal units to identify major brain pathologies because it is accessible, inexpensive, non-invasive, and can be performed by a variety of trained professionals. However, the sensitivity of abnormalities detected by ultrasound to predict developmental outcome is poor (26–86%), even though the specificity is good (82–100%; Table I). These observations may be related to the inability of cranial ultrasound to identify subtle, diffuse PVWM damage.11,22–26 The importance of diffuse PVWM damage is emphasized by the observation that cognitive deficits remain common while the incidence of cystic PVWM damage is decreasing, suggesting that pathologies undetected by ultrasound play role in developmental difficulties.11,22,27–29

Table I.   Specificity and sensitivity of cranial ultrasound at predicting developmental outcome
StudySample size, nAge at time of cranial ultrasoundDefinition of abnormal cranial ultrasoundAge at developmental assessmentDevelopmentassessedSpecificitySensitivity
  1. GA, gestational age; IVH, intraventricular haemorrhage; PVWM, periventricular white matter damage; PVHI, periventricular haemorrhagic infarction; PPV, positive predictive value;NPV, negative predictive value.

Valkama et al.715138.6wks post-GA (SD 2.4wks)‘Parenchymal lesions’: IVH, subependymal and parenchymal haemorrhages, cystic PVWM damge, PVHI, reduction of white matterMean age 18.6mo (SD 1)Neurological examination, including muscle tone, presence of CP and ambulation85% (100% if grade I–III IVH excluded)67% (58% if grade I–III IVH excluded)
De Vries et al.992139 (1929 survivors)After birth, once a week during admission, and at 40wks GA‘Major abnormalities’: IVH grade III–IV, cystic PVWM damage, and subcortical leukomalacia, basal ganglia lesions, PVHIBy 24mo chronological agePresence of CP, locomotor function, development quotient using Griffiths Mental Development Scale<32wks GA: 95% (PPV 48%, NPV 99%).
33–36wks GA: 99% (PPV 83%, NPV 99%)
<32wks GA: 76%.
33–36wks GA: 86%
Mirmiran et al.766136–40wks GAIVH grade III–IV, PVHI, mineralization, atrophy, ventricular dilation >1cm at the midbody of the lateral ventricle, cystic PVWM damage, cavitations 20 and 31moAmiel-Tison standardized neurological examination20mo presence of CP: 86% (PPV 22%, NPV 90%).
31mo presence of CP: 82% (PPV 33%, NPV 87%)
20mo presence of CP: 29%.
31mo presence of CP: 43%
Stewart and Kirkbride1001142Discharge from neonatal unitCystic PVWM damage, PVHI8yDisabling impairment; extra educational provision; IQ <7087%

Fortunately, MRI is superior to ultrasound at detecting subtle abnormalities.11,22–26 However, MRI is not as accessible as ultrasound. Neonates have to be moved to a scanner that may be in another part of the hospital, or a different hospital altogether. This is dangerous for unstable neonates. MRI also requires specialist interpretation and is susceptible to movement artefact. Therefore, although MRI may detect subtle abnormalities more accurately than ultrasound, considerable logistical difficulties exist in obtaining images. Whether MRI is good at predicting developmental outcome is a different question altogether. This article reviews current evidence for the association between MRI results and developmental outcome in preterm neonates. This review also discusses whether MRI is justified in every preterm neonate.

Appearances of cerebral white matter on MRI

Cystic periventricular white matter damage

Histologically, the initial stages of cystic PVWM damage are characterized by bilateral areas of coagulative necrosis within the subventricular lamina of the PVWM. Macrophages are found in the centre and hypertrophic lipid-laden astrocytes around the edges of these lesions. Degeneration of all cellular elements follows, and cysts or glial scars form.9,30 Large cysts may disappear if their walls collapse.30 Secondary ventriculomegaly can develop because of white matter atrophy, impaired growth, or impaired myelination.9,30

MRI visualizes cystic PVWM damage well, and has been used to estimate the severity of cystic PVWM damage depending on its site, laterality, and the size of the ventricles.31–37 Increasing severity of cystic PVWM damage and the presence of cysts at term-equivalent age is associated with motor delay and CP at 2 years’ corrected age, but specificities and sensitivities are not reported.36 MRI of the brain later in childhood confirm this association,31–34 although around 25% of children with cystic PVWM damage may not develop CP.34

Increasing severity of cystic PVWM damage at term-equivalent age also correlates with worse cognitive outcome and object memory deficits between 18 and 24 months of age.35–37 The specificities and sensitivities make clinical application limited (Table II), as does the observation that severe cystic PVWM damage does not preclude normal developmental scores at 2 years.36 These findings have led researchers to turn their attention towards diffuse PVWM damage as a possible cause of cognitive impairment.7–9,26,29

Table II.   Sensitivity and specificity of white matter damage (defined as mild, moderate, or severe using combinations of scores for signal abnormality, white matter volume reduction, presence and severity of cystic PVWM damage, ventricular dilatation, and presence and severity of corpus callosum thinning) at detecting developmental outcomea
DevelopmentaloutcomeModerate to severe white matter damage (n=35)Any white matter damage(n=120)
  1. aAdapted from Woodward et al.36

Severe cognitive delay41848931
Severe motor delay65858830
Presence of cerebral palsy65849431
Neurosensory impairment82828930
Any neuro-developmental impairment38898434

Diffuse periventricular white matter damage

Histologically, diffuse PVWM damage is characterized initially by reactive microglia, with hypertrophic astrocytes and amphophilic globules present later on.9,30 The number of late precursor oligodendrocytes in diffuse PVWM damage is markedly decreased, suggesting this is the most likely affected cell.9 MRI detects subtle white matter abnormalities better than cranial ultrasound,11,22–26 although whether these reflect diffuse PVWM damage is controversial. Hyperintense foci in the PVWM observed on T1-weighted MRI is one form of possible diffuse PVWM damage.25,37–39 The relationship between these lesions and developmental outcome is unclear. Some studies have found no association,25,37,38 another has found that their presence in the corticospinal tracts of the corona radiata on coronal images, predicts CP aged 4 years with a sensitivity of 100% and specificity of 97%.39 Increasing size and irregular shape of the ventricles also helps predict CP, as lesions associated with ventriculomegaly (defined as a ventricular brain ratio of greater than or equal to 0.35) predict Gross Motor Function Classification Score Level V with a sensitivity and specificity of 100%.39

Diffuse excessive high signal intensity (DEHSI) is another appearance on MRI that may represent diffuse PVWM damage. DEHSI is defined as high signal intensity in the white matter on T2-weighted fast spin echo images. The amount of high signal intensity is more than expected, and extends from the immediate periventricular region, where high signal can normally be seen, into the adjacent periventricular and subcortical regions.10 DEHSI appears after 27 weeks’ gestation and is reported in 76% of preterm infants at term-equivalent age, but is not described in term controls. It can be associated with ventricular dilatation and squaring of the margins of the lateral ventricles, which has been used to support suggestions that it represents pathology.10

Developmentally, a single, small study found no correlation between the presence and severity of DEHSI and outcome at 18 months’ corrected age. However, a correlation was found if children with overt, focal white matter lesions were excluded; the authors do not suggest why this should be the case.28 Overall, DEHSI remains a controversial finding on MRI. This is partly because it is a visual, rather than an objective phenomenon whose appearances are affected by image windowing.12 Additionally, other researchers have been unable to identify DEHSI in preterm neonates,25 and to our knowledge there are no histological studies of its characteristics. However, diffusion MRI techniques support suggestions that DEHSI is a pathological entity,40 as described in the next section.

Quantitative diffusion weighted and tensor imaging of white matter

Diffusion weighted MRI uses water diffusion within tissues for its image contrast. Tissues in which water molecules diffuse freely have different contrast to those with restricted diffusion. Numerical values, called apparent diffusion coefficients (ADCs), quantify the degree of water molecule diffusion. ADC values in the central white matter at 28 weeks’ gestation are high, reflecting the large amount of water in the extracellular matrix able to diffuse, and then decrease towards term.41 These changes in ADCs correspond to reductions in extracellular matrix water content and the development of physical barriers to diffusion, such as axonal thickening and the wrapping of oligodendrocytes around axons.8

These histological changes present a barrier, in particular to water molecule diffusion across axons, causing them to diffuse preferentially along them instead.41–44 This can be observed using diffusion tensor imaging (DTI). DTI measures diffusion from many more directions than diffusion weighted MRI, generating vectors of the major direction of water diffusion. These vectors are generally assumed to represent white matter tracts in the brain,45 as seen when they are colour coded to produce maps. DTI can also generate numerical values to quantify the directionality of water diffusion, with high values representing a high degree of directionality. These values are called anisotropy, which is either fractional or relative depending on the mathematical equation used to generate them. Thus, relative anisotropy increases as the white matter develops and water diffusion increasingly occurs along axons.41,43,44

Damage to the axons or late precursor oligodendrocytes may prevent the normal development of these barriers, altering the degree and direction of water molecule diffusion: potentially, MRI may identify damage before it is visible on conventional imaging. ADC values, for example, are higher in preterm neonates with cystic PVWM damage and PVHI than normal MRI appearances.46 Following damage to the white matter, serial ADC values in the central white matter are static or increase, and anisotropy also fails to increase towards term.47 This demonstrates that normal development of white matter is impaired following initial damage.

Diffusion weighted imaging also detects microstructural changes in the white matter of preterm neonates with normal MRI appearances. ADC values are higher and relative anisotropy values are lower at term in the central white matter of preterm infants than controls.41,43 DTI has also shown that term infants have multiple aligned vector bundles in their central white matter.43 Vectors are smaller and have a more diffuse orientation in preterm neonates, and those with cystic PVWM damage have further abnormalities around the lateral ventricle, internal capsule, and posterior limb of internal capsule (PLIC).43

Diffusion weighted imaging also supports the suggestion that DEHSI represents damage to oligodendrocytes or axons. ADC values are higher in the central white matter of infants with DEHSI compared with those with normal appearances, but are not significantly different from neonates with overt damage.46 Radial water diffusivity (a measurement of water molecule diffusion perpendicular to white matter tracts) on DTI is also statistically higher in the posterior part of the PLIC and splenium of the corpus callosum in preterm infants with DEHSI compared with term controls. Axial diffusivity (a measurement of water molecule diffusion parallel to white matter tracts) and radial diffusivity are both elevated in the centrum semiovale, and the occipital, frontal, and posterior PVWM in the same infants.40

If diffusion imaging is detecting and quantifying diffuse PVWM damage, does it help to predict developmental outcome? Mean ADC values of the centrum semiovale at term are negatively associated with developmental quotients at 2 years of age in preterm neonates without cystic PVWM damage, PVHI, or post-haemorrhagic hydrocephalus. However, 74% had DESHI.48 Therefore, diffusion weighted MRI and DTI of the white matter are techniques that potentially offer a method of accurately quantifying diffuse PVWM damage, and may help to identify neonates at high risk of developmental difficulties. Further work is still needed to clarify the relationship with developmental outcome and establish sensitivities and specificities.

Appearances of posterior limb of internal capsule on MRI

Appearances of PLIC on conventional imaging have been studied following the observation that an abnormal or absent signal predicts poor motor outcome in term neonates with hypoxic–ischemic encephalopathy.49 In preterm infants, an abnormal or absent signal in PLIC is seen following unilateral periventricular haemorrhagic infarction and cystic PVWM damage. These are associated with poor motor outcome in small studies.50,51

ADC and relative anisotropy values in the PLIC are not lower at term-equivalent age in preterm neonates compared with controls. However, fractional anisotropy values are lower in neonates who subsequently developed CP.52,53 In a single small study, fractional anisotropy was more accurate than conventional MRI at predicting development of CP.53

Corpus callosum measurements

Children and adults born preterm have a thinner corpus callosum with a smaller cross-sectional area than controls. The posterior part is particularly affected.32,54–59 Cystic PVWM damage is also associated with corpus callosum thinning.60–62

Decreased thickness and cross-sectional surface area of the corpus callosum in childhood are associated with CP and motor delay.32,54 The ratio between splenium thickness and corpus callosum length is also smaller in children with CP.59 The total cross-sectional surface area has a positive linear association with visual motor integration tests54 and verbal fluency in boys,58 but the surface area, ratio of the thickness of the midbody:length, and thickness of the splenium:length do not correlate to IQ.57–59 Similarly, the cross-sectional surface area does not correlate to the presence of ADHD.57

Growth rates of corpus callosum length and height (measured one third of the distance from the extreme margins of the genu and the splenium on mid sagittal views) have been calculated with serial MRI following preterm birth. These growth rates have then been compared to in utero rates measured with transvaginal ultrasound.

Corpus callosum growth rates are normal for the first 2 weeks of life after preterm birth, as long as there is no intrauterine growth restriction. However, these rate decrease to half the interuterine rate between 2 and 6 weeks’ postnatal age.63,64 Motor deficits are more prevalent in individuals with the slowest growth rates between 2 and 6 weeks of age,63 as well as those with a shorter corpus callosum at term.64

There are few prospective studies of corpus callosum appearances at term-equivalent age and later developmental outcome. From the work published so far, these measurements may prove useful in the future for predicting motor but not cognitive impairment.

Volumetric measurements of cerebral white matter

If initial damage causes secondary impaired growth and development of the white matter, this may affect tissue volume. Volumetric studies have shown that preterm neonates have smaller white matter volumes at term-equivalent age than term controls.65–69 Myelinated white matter volumes on MRI are particularly affected.65 Regionally, unmyelinated white matter volumes on MRI are smaller in the sensorimotor and orbitofrontal regions. Myelinated and unmyelinated white matter volumes are smaller in the right parieto-occipital and inferior occipital regions.69 These observations are of uncertain developmental significance. In a single small study, volumes of myelinated and unmyelinated white matter on MRI in the right sensorimotor and midtemporal regions correlated to cognitive development; left and right subgenual regions correlated to motor development at 18 to 20 months’ corrected age.66

Ventricular size and volumes

The ventricular system enlarges following cerebral white matter atrophy or abnormal development. Therefore, ventricular volume and size have been used as a surrogate assessment of the severity of white matter damage. Of preterm infants at term-equivalent age, 39 to 75% have either larger ventricles or increased ventricular volumes than controls and previously published normal references.28,69,70 However, the range of ventricular volumes overlap considerably between preterm and term groups.69 Ventricular dilatation at term (defined as an axial diameter >10mm) also does not correlate clearly with developmental outcome between 18 and 34 months’ corrected age, unless infants with IVH are studied separately.28

In older individuals, the ventricles are dilated with increased volumes in infancy, at 6 years of age, in adolescence, and adulthood following preterm birth.55–57,66,71–73 Unfortunately, there is no consistent definition of ventricular dilatation across these studies.56–58,71,73 The presence of ventricular dilatation does not correlate clearly with motor or cognitive outcome, IQ, or the presence of ADHD.57,58,71,73 Increased ventricular volumes similarly do not correlate with cognitive or motor developmental scores.58,66

Dilated occipital horns are seen in 44.4% of 1-year-old infants born with very low birthweight (although the exact definition of ‘dilatation’ is not reported), and are associated with lower motor, cognitive, and psychomotor developmental scores.74 However, these results should be interpreted cautiously because of the early age of developmental assessment.

In summary, larger ventricles at term may help to identify neonates at risk of experiencing developmental difficulties, but normal outcome is not precluded. Although few groups have reported the specificity and sensitivity of ventricular volumes at predicting developmental outcome, it appears that they do not provide an accurate indication of developmental prognosis on their own.

Germinal matrix and intraventricular haemorrhage

Germinal matrix and IVH can have secondary effects on white matter. These lesions are well recognized on serial cranial ultrasound imaging, and severe haemorrhage is related to worse developmental outcome.36,75,76 Even IVH grade I and II, previously thought to be of little developmental significance, are associated with neurological abnormality, CP, and poor motor and psychomotor development at 20 months’ corrected gestational age.77 There is no study, to the authors’ knowledge, examining the ability of MRI to detect IVH, and we do not think MRI has a role in the acute setting when infants are unstable and ultrasound is available.

Periventricular haemorrhagic infarction

PVHI is a venous infarction to the PVWM caused by obstruction to terminal vein blood flow by germinal matrix and IVH.26 The incidence of PVHI following preterm birth on ultrasound imaging is 1%, increasing as gestational age at delivery and birth weight decrease.78–81 Mortality following PVHI is between 40% and 59%. Survivors have a high risk of developmental problems and require therapeutic intervention (Table III).26,80 The presence and severity of developmental difficulties differ according to the size of the PVHI and the affected area. For example, extensive PVHI (i.e. lesions affecting more than one region of the brain) is nearly always followed by motor and/or cognitive impairments, whereas neonates with focal lesions, particularly if unilateral, are affected much less severely.26,78,80

Table III.   Neurodevelopmental outcome in survivors following detection of periventricular haemorrhagic infarction on cranial ultrasound80
Developmental difficulty ortherapeutic interventionFrequency, %
Gross motor delay73
Cognitive impairment50
Impairments in daily living33
Abnormal neurological examination or cognitive impairment72
Abnormal visual acuity in at least one eye44
Visual field defects33
Abnormal visual acuity or visual field55
Need for ventriculo-peritoneal shunt37
Physiotherapy input90
Occupational therapy input87
Speech and language input43
Received botulinum toxin treatment33

Cranial ultrasound detects PVHI well,11 and to the authors’ knowledge, there is no large study comparing the ability of ultrasound and MRI to predict developmental outcome following PVHI. However, it seems likely that MRI gives a better understanding of the extent and anatomical region of the brain affected, which may help to predict outcome.

Appearances of grey matter on MRI

Cortical grey matter

Preterm birth is associated with decreased volumes of cortical grey matter at term-equivalent age.65,66 Regional differences have been noted, with parieto-occipital, sensorimotor, and inferior occipital regions being significantly reduced. These findings are associated with a relative increase in grey matter volumes within the anterior cortices.66 Volumes in the left sensorimotor and left mid-temporal regions at term correlate with motor outcome 18 to 20 months later.66 Volumes of cortical and deep grey matter combined are also smaller, and associated with ‘moderate to severe disability’ at 1 year of age.65 Reduced cortical volumes are also seen in preterm adolescents, although disagreement exists about whether they correlate to lower IQ.72,82

Preterm birth is associated with smaller cortical surface area at term than controls,83 and a reduced growth rate of the surface area.84 The disrupted growth in cortical surface area in preterm infants parallels the presence of developmental impairments.84 Adolescents who had very low birthweight also have decreased cortical thickness, which strongly correlates to IQ.82 Unfortunately, the authors do not report if there was a point at which smaller surface area was always associated with lower IQ, and sensitivity and specificity are not reported.

Subcortical grey matter

The volume of subcortical grey matter is also smaller following preterm birth,65,85,86 with the lentiform and thalamus particularly affected.85,86 At term, these volumes are particularly reduced in preterm neonates who have overt supratentorial brain damage such as IVH, cystic PVWM damage, and PVHI.85 Preterm neonates with DEHSI (identified visually and with raised ADC values) also have smaller lentiform and thalamic volumes than preterm neonates with normal MRI and term controls.86 One study has examined the relationship between subcortical grey matter volumes and development, showing reduced volumes at term are associated with ‘moderate to severe disability’ at 1 year of age.65

In the current authors’ experience, the boundaries between the grey and white matter are poorly delineated in the neonatal brain. Therefore, it is difficult to measure the volume of grey matter accurately, leading to questions about the practicality, reliability, and reproducibility of this approach in the clinic.

The cause of reduced grey matter volume is debatable. Volpe proposes that grey matter reductions are a secondary effect of white matter damage, citing histopathological findings that abnormality is only found within the white matter of neonates with reduced white and grey matter volumes.8 This is supported by MRI, which has shown reduced cortical volumes are seen in children with overt white matter damage87 and isolated IVH.88 Grey matter volumes are not reduced at term in low risk preterm infants with normal MRI, despite moderately reduced white matter volumes.89

Quantitative measurements of grey matter damage

ADC values in the thalamus and lentiform nucleus do not differ between preterm neonates at term and controls, although none of the infants studied had visible subcortical grey matter abnormalities on conventional MRI.86

Appearances of cerebellum on MRI

Cerebellar haemorrhage

Cerebellar haemorrhages can be visualized on cranial ultrasound through the mastoid or posterior fontanelle.90,91 They occur in 1.2 to 4.5% of preterm infants; those with birth weights under 750g are most affected.90 Cerebellar haemorrhages are primarily unilateral, more common on the right, involve the vermis in 20% of cases, and are frequently associated with supratentorial pathology.90 They are associated with motor, cognitive, expressive, and receptive language, internalization behaviour, and social/communication difficulties later in life. These problems are particularly prevalent and profound following bilateral injury, whereas autistic traits are common if the vermis is involved.92 There is no study, to the authors’ knowledge, comparing the ability of cranial ultrasound and MRI to detect cerebellar haemorrhages, or their abilities to predict outcome. It seems logical that MRI will be better than ultrasound at delineating the size and extent of cerebellar haemorrhages.

Volumetric measurements of the cerebellum

Damage to the cerebrum is associated with poor development secondarily of the cerebellum. Primary damage to the cerebellum affects the cerebrum in the same manner. For example, unilateral and bilateral cerebral brain injury leads to smaller contralateral and bilateral cerebellar hemisphere volumes respectively. Similarly, primary cerebellar damage leads to decreased cerebral volumes.93 At term-equivalent age, cerebellar volumes are not decreased in preterm neonates unless white matter damage is present.93–96 No significant correlation has been found between cerebellar volumes at term and outcome at 2 years of age.95

Appearances of the hippocampus on MRI

A single, small study noted that hippocampal volumes are smaller in adolescents born preterm than controls. Reduced hippocampal volumes are associated with poor everyday memory, lower freedom from distractibility, and lower arithmetic scores on psychological testing.97 However, other studies have not found similar reductions in hippocampal volume of 20-year-olds born preterm.55 T2 relaxometry has also been used to study the hippocampus in 7-year-old children who were born preterm, finding no correlation between T2 relaxation times and IQ or motor scores.98

In theory, damage to the hippocampus may explain some of the subtle cognitive defects associated with preterm birth. However, the practical applications for neonatal care are doubtful because accurately measuring the hippocampus at term-equivalent age may prove difficult.


There is no doubt that MRI identifies structural changes in the preterm brain better than cranial ultrasound, particularly if the changes are subtle. However, histological proof that subtle magnetic resonance appearances, such as DEHSI, T1 hyperintensities, and changes in ADC and anisotropy values reflect pathology is still awaited. Furthermore, many of these MRI abnormalities are of doubtful developmental significance. Therefore, identifying subtle lesions with MRI may unnecessarily increase parents’ anxiety, while normal MRI could be falsely reassuring. The logistics of obtaining MRI also limits its clinical use, especially in unstable neonates. For all these reasons, serial cranial ultrasound is still the most practical imaging modality for sick preterm infants. Instead, MRI is more practical around term-equivalent age when neonates are stable, cystic damage has developed, and myelination in the PLIC can be seen on MRI. In such circumstances, MRI could delineate the site and extent of focal cystic PVWM damage, PVHI, and cerebellar haemorrhages, as well as to investigate other neurological abnormalities (Table IV).

Table IV.   Authors’ suggested indications for magnetic resonance following preterm birth
Suggested indications
 Grade III to IV intraventricular haemorrhages
 Periventricular haemorrhagic infarction
 Cystic periventricular white matter damage
 Cerebellar haemorrhage or other abnormalities on ultrasound imaging
 Suspected white matter abnormalities on cranial ultrasound (echodensities/echolucencies)
 Post-haemorrhagic hydrocephalus
 Abnormal neurological examination
 Investigation of other conditions warranting detailed neuroimaging, such as metabolic disorder or suspected congenital structural abnormality

Whether MRI will ever be able to predict neurodevelopemental outcome accurately is uncertain, particularly when environmental and genetic effects are considered. However, it is likely that MRI will become a useful tool as the nature of damage to the preterm brain and developmental significance of abnormal MRI become clearer. Until that time, MRI of all preterm neonates for prognostic reasons cannot be justified, and should be reserved for the individual investigation of known or suspected cerebral pathology.


Dr Hart’s post is funded by a grant from the Jessop Baby Fund.

List of abbreviations

Apparent diffusion coefficient


Diffuse excessive high signal intensity


Diffusion tensor imaging


Intraventricular haemorrhage


Posterior limb of internal capsule


Periventricular haemorrhagic infarction


Periventricular white matter