Sequential Cranial Ultrasound and Cerebellar Diffusion Weighted Imaging Contribute to the Early Prognosis of Neurodevelopmental Outcome in Preterm Infants

Margaretha J Brouwer; Britt J M van Kooij; Ingrid C van Haastert; Corine Koopman-Esseboom; Floris Groenendaal; Linda S de Vries; Manon J N L Benders

doi:10.1371/journal.pone.0109556

. 2014 Oct 20;9(10):e109556. doi: 10.1371/journal.pone.0109556

Sequential Cranial Ultrasound and Cerebellar Diffusion Weighted Imaging Contribute to the Early Prognosis of Neurodevelopmental Outcome in Preterm Infants

Margaretha J Brouwer ¹, Britt J M van Kooij ¹, Ingrid C van Haastert ¹, Corine Koopman-Esseboom ¹, Floris Groenendaal ¹, Linda S de Vries ¹, Manon J N L Benders ^1,^2,^*

Editor: Pierre Gressens³

PMCID: PMC4203729 PMID: 25329772

Abstract

Objective

To evaluate the contribution of sequential cranial ultrasound (cUS) and term-equivalent age magnetic resonance imaging (TEA-MRI) including diffusion weighted imaging (DWI) to the early prognosis of neurodevelopmental outcome in a cohort of very preterm infants (gestational age [GA] <31 weeks).

Study design

In total, 93 preterm infants (median [range] GA in weeks: 28.3 [25.0–30.9]) were enrolled in this prospective cohort study and underwent early and term cUS as well as TEA-MRI including DWI. Early cUS abnormalities were classified as normal, mild, moderate or severe. Term cUS was evaluated for ex-vacuo ventriculomegaly (VM) and enlargement of the extracerebral cerebrospinal fluid (eCSF) space. Abnormalities on T1- and T2-weighted TEA-MRI were scored according to Kidokoro et al. Using DWI at TEA, apparent diffusion coefficients (ADCs) were measured in four white matter regions bilaterally and both cerebellar hemispheres. Neurodevelopmental outcome was assessed at two years’ corrected age (CA) using the Bayley Scales of Infant and Toddler Development, third edition. Linear regression analysis was conducted to explore the correlation between the different neuroimaging modalities and outcome.

Results

Moderate/severe abnormalities on early cUS, ex-vacuo VM and enlargement of the eCSF space on term cUS and increased cerebellar ADC values on term DWI were independently associated with worse motor outcome (p<.05). Ex-vacuo VM on term cUS was also related to worse cognitive performance at two years’ CA (p<.01).

Conclusion

These data support the clinical value of sequential cUS and recommend repeating cUS at TEA. In particular, assessment of moderate/severe early cUS abnormalities and ex-vacuo VM on term cUS provides important prognostic information. Cerebellar ADC values may further aid in the prognostication of gross motor function.

Introduction

During the last decades, several neuroimaging modalities have been introduced to evaluate the preterm brain during the neonatal period in order to improve prediction of developmental impairments.

Cranial ultrasound (cUS) has proven to be an easily accessible neuroimaging technique, which can be performed at the bedside to detect a variety of brain lesions in the preterm infant. Significant associations have been reported between severe cUS abnormalities, such as germinal matrix-intraventricular hemorrhage (GMH-IVH) grade III or IV according to Papile et al. [1] or cystic periventricular leukomalacia (PVL) according to de Vries et al. [2], and an unfavourable cognitive and motor outcome. [3] Conventional magnetic resonance imaging performed around term-equivalent age (TEA-MRI) or even soon after birth has been shown to be superior to cUS in detecting more subtle brain lesions. [4] Scoring systems have been developed for white matter (WM) and gray matter (GM) abnormalities in order to better predict the risk of an adverse outcome. [5] An adjusted brain scoring system for TEA-MRI has recently been published, which also incorporates evaluation of the cerebellum [6], reflecting the increased recognition of the impact of cerebellar abnormalities on neurodevelopment [7], [8].

The application of diffusion weighted imaging (DWI) was introduced to assess the microstructural development of WM in the preterm brain. [9] Data on the correlation between WM apparent diffusion coefficients (ADCs) and brain injury [10]–[13] or neurodevelopmental outcome [14]–[16] remain, however, controversial. A few studies have addressed the effect of supratentorial brain injury on cerebellar ADC values, with inconclusive results. [12], [17] No data are available on the prognostic value of cerebellar ADC values.

With the increasing availability of neuroimaging modalities, the question arises to which extent they are complementary, at which point in time they are most predictive and whether TEA-MRI should be performed in every extremely low birth weight infant. The aim of this cohort study among very preterm infants was to examine the contribution of sequential cUS and TEA-MRI including DWI to the early prognosis of neurodevelopmental outcome at two years’ corrected age (CA).

Materials and Methods

Patients

From October 2006 to March 2008, very preterm infants (gestational age [GA] <31 weeks) admitted to the level three unit of the Wilhelmina Children’s Hospital, Utrecht, The Netherlands, were recruited for a prospective neuroimaging study approved by the Medical Ethics Committee of our institute. Neonates who were born outside our referral district or with congenital anomalies were excluded. Of the 168 consecutively admitted neonates, eighteen infants deceased before reaching TEA, no parental informed consent was obtained for fifteen infants and sixteen infants were examined on a 1.5 Tesla system, resulting in 119 infants with 3.0 Tesla MR imaging at TEA. From this cohort, we further excluded infants with congenital brain abnormalities on sequential cUS and TEA-MRI (n = 2; i.e., heterotopia, antenatal porencephalic cyst) as well as infants who were scanned at a postmenstrual age (PMA) ≥44 weeks (n = 2) or with motion artifacts on DWI (n = 18). Four infants were lost to follow up at two years’ CA, leaving 93 infants eligible for final inclusion. Written parental informed consent was obtained for all infants.

Neuroimaging

Early and term cUS

Ultrasound scans were performed within six hours of admission, at least three times in the first week after birth and then weekly till discharge to a level two hospital. These scans are referred to as early cUS. At TEA, within 24 hours following MRI, cUS was repeated. This scan is referred to as term cUS. Scanning was performed with a Toshiba Aplio (Toshiba Medical Systems, Zoetermeer, The Netherlands) or ATL-5000 ultrasound machine (Philips Medical Systems, Best, The Netherlands) with a transducer frequency of 5–8 MHz. Using the anterior fontanel as an acoustic window, standard views were taken in the coronal and sagittal planes. Comprehensive evaluation of the cerebellum through the mastoid and posterior fontanel was not performed at the time in all neonates. Hence, cerebellar cUS abnormalities could not be analyzed in this paper.

TEA-MRI

MR images were acquired on a 3.0 Tesla MR system (Philips Healthcare, Best, The Netherlands) using a sense head coil. Infants were sedated with 50–60 mg/kg chloralhydrate fifteen minutes prior to the examination. The protocol involved sagittal T1-weighted imaging (repetition time [TR] = 886 ms; echo time [TE] = 15 ms; slice thickness = 3.0 mm), axial 3DT1-weighted imaging (TR = 9.4 ms; TE = 4.6 ms; slice thickness = 2.0 mm, no gap), axial T2-weighted imaging (TR = 6293 ms; TE = 120 ms; slice thickness = 2.0 mm, no gap) and axial DWI (TR = 2407 ms; TE = 68 ms; voxel size = 0.9×0.9×4.0 mm; b-values = 0 and 800 mm²/s).

Assessment of brain injury

Early cUS

Assessment of brain injury was performed offline by two authors (MJNLB, LSV) using DicomWorks 1.3.5 (www.dicomworks.com). For each infant, a final classification was applied to the whole set of early cUS scans based on the most severe lesion observed. Early cUS findings were classified as 1) normal when no or just minor abnormalities were observed (e.g., germinal layer cysts, plexus cysts, subependymal pseudocysts and/or lenticulostriate vasculopathy); 2) mildly abnormal in the presence of a GMH-IVH grade I or II according to Papile et al. [1], mild subsequent ventricular dilatation and/or PVL grade I according to de Vries et al. [2]; 3) moderately abnormal if a GMH-IVH grade III was diagnosed and 4) severely abnormal in case of a GMH-IVH grade IV, progressive post-haemorrhagic ventricular dilatation (PHVD, i.e., ventricular index >97^th percentile according to Levene [18], anterior horn width >6 mm [19] and/or thalamo-occipital distance >24 mm [19]) requiring intervention, cystic PVL grade II or III, parenchymal lobar haemorrhage and/or focal infarction.

Term cUS

Term cUS scans were qualitatively scored by one of the authors (LSV) for the presence of 1) enlargement of the extracerebral cerebrospinal fluid (eCSF) space, defined as an increase in interhemispheric distance and/or subarachnoid space (Figure 1) [20] and 2) ex-vacuo ventriculomegaly (VM), defined as an increase in ventricular size – often most pronounced for the occipital horns – with irregularity of the ventricular margins (Figure 2A). [21] Ex-vacuo VM, considered as a sequel of WM injury, was distinguished from PHVD on the basis of ventricular shape. In contrast to ex-vacuo VM, PHVD tends to be associated with so-called “ballooning” of the ventricles with both anterior and posterior dilatation without an irregular ventricular shape (Figure 2B). [21] In infants with prior GMH-IVH, early cUS did further aid in the discrimination between these two entities.

Note the posterior dilatation and irregular ventricular margins in (A), in contrast to the more global ventricular enlargement in (B).

T1- and T2-weighted TEA-MRI

WM, cortical and deep GM and cerebellum were evaluated for the presence of brain abnormalities (assessed by LSV) and abnormal brain metrics (measured by MJB) according to the scoring system by Kidokoro et al. [6] for T1- and T2-weighted TEA-MRI. Measurements were obtained using OsiriX (32-bit version, www.osirix-viewer.com), which allowed for free conversion to all planes. A global brain abnormality score was calculated as the sum of these regional scores and further classified as normal (0–≤3), mildly abnormal (4–≤7), moderately abnormal (8–≤11) and severely abnormal (≥12) according to Kidokoro et al [6].

Term DWI – ADC values

ADC values were measured in ten regions of interest (ROIs) in the WM and cerebellum (Figure 3). Measurements were performed by two independent observers (MJB, BJMK) using a standalone MR work station (Achieva, Philips Medical Systems, Best, The Netherlands).

For the WM, ADC maps and ADC maps with enhanced contrast (e-ADC) were first derived from the diffusion weighted images. ROIs were then manually drawn on the e-ADC map using a seed growing tool and copied to the ADC map. The frontal and occipital WM were assessed at the level of the foramen of Monro and the trigone, respectively. Sagittal T1-weighted MRI was used to define the cross-sectional planes for ADC measurements of the parieto-occipital and parietal WM, in general two and three slices above the level of the trigone, respectively. Care was taken to avoid partial volume averaging from CSF or cortical GM on the slices above and below. Cerebellar ROIs were positioned at the level of the maximal cerebellar hemispheric width. Given the complex structure of the cerebellum and partial volume effects, large ROIs were placed in the center of both hemispheres instead of selecting a particular tissue type. In infants with profound supra- or infratentorial parenchymal lesions, ADC measurements were not performed in the areas involved.

For nineteen infants selected at random, ADC measurements were repeated by both observers after a time interval of at least 24 hours. After calculating the intra- and interobserver reliability, both observers’ ADC measurements were averaged. Means were calculated for bilateral ADC measurements of each WM region and the cerebellum and used in consecutive analyses.

Neurodevelopmental outcome

Neurodevelopmental outcome was assessed at 24 months’ CA by a single developmental specialist (ICH) using the cognitive, fine motor and gross motor subtest of the Bayley Scales of Infant and Toddler Development, third edition (BSITD-III). The composite and scaled scores corrected for premature birth were calculated (mean [SD] in a normative population: 100 [15] and 10 [3], respectively).

Statistical analysis

Data were analyzed using IBM SPSS Statistics version 20 (SPSS Inc, Chicago, Illinois).

The intra- and interobserver reliability for ADC measurements were assessed by calculating the intraclass correlation coefficients for single measures using the two-way random model for absolute agreement.

To evaluate the relationship between the different neuroimaging modalities, several statistical tests were conducted. Categorical variables were analyzed using either a Chi-square or Fisher’s exact test. The distribution of continuous variables between groups was compared using ANOVA and Mann-Whitney U tests for parametric and non-parametric data, respectively. Continuous data were compared by calculating a Spearman’s rho. The distribution of GMH-IVH and PHVD among infants with and without ex-vacuo VM on term cUS was compared using a Chi-square and Fisher’s exact test, respectively.

The correlation between neuroimaging and outcome was evaluated using linear regression analysis with outcome as dependent variable. Cognitive and fine motor outcome were corrected for maternal education; gross motor outcome was not related to maternal education. First, the association between the different neuroimaging modalities and outcome was explored. Next, neuroimaging parameters were combined into one model. This was done by hand in a forward manner; entrance criterion was a p-value <.10, exclusion criterion a p-value ≥.05. Finally, the additional influence of gender, GA and birth weight z-score on outcome was assessed.

Results

Descriptive results

A total number of 93 preterm neonates (50 males, 43 females) were enrolled (median [range] GA in weeks: 28.3 [25.0–30.9]; birth weight in grams: 1060 [630–1910]; birth weight z-score: 0.05 [−1.14–2.06]). The maternal educational level was low, intermediate and high for, respectively, 23 (24.7%), 38 (40.6%) and 30 (32.3%) infants; data for two infants were missing. TEA-MRI was performed at a median (range) PMA of 41.6 (39.6–43.6) weeks. Neurodevelopmental assessment (BSITD-III) was conducted at a median (range) CA of 24.1 (23.6–27.6) months. The infants’ neuroimaging findings and neurodevelopmental outcome are summarized in Table 1.

Table 1. Descriptive results (n = 93).

Early cUS; n (%)
Normal		26 (28.0)
Mild abnormalities		55 (59.1)
Moderate abnormalities		3 (3.2)
Severe abnormalities		9 (9.7)
Term cUS; n (%) ^$
Ex-vacuo VM		18 (19.4)
Enlargement eCSF space		41 (44.1)
T1- and T2-weighted TEA-MRI; median (IQR; range)
Global brain abnormality score		3 (2–6; 0–15)
WM score		3 (2–4; 0–11)
Cortical GM score		0 (0; 0–3)
Deep GM score		0 (0; 0–3)
Cerebellum score		0 (0–1; 0–7)
*Term DWI – ADC values 10 ⁻³ mm²/s; median (IQR; range)** ^Δ
Frontal WM	R	1.611 (1.516–1.666; 1.292–1.840)
	L	1.626 (1.526–1.679; 1.324–1.837)
Occipital WM	R	1.579 (1.508–1.661; 1.367–1.794)
	L	1.607 (1.523–1.687; 1.356–1.843)
Parieto-occipital WM	R	1.654 (1.581–1.716; 1.409–1.843)
	L	1.662 (1.579–1.727; 1.434–1.870)
Parietal WM	R	1.610 (1.538–1.671; 1.382–1.923)
	L	1.600 (1.524–1.664; 1.408–1.906)
Cerebellum	R	1.075 (1.021–1.121;.920–1.242)
	L	1.078 (1.025–1.125;.911–1.271)
Neurodevelopmental outcome (BSITD-III) at 24 months’ CA; median (IQR; range)
Cognitive composite score		105 (95–110; 70–145)
	Cognitive scaled score	11 (9–12; 4–19)
Motor composite score		107 (100–112; 73–142)
	Fine motor scaled score	13 (11–14; 5–19)
	Gross motor scaled score	9 (9–10; 6–15)

Open in a new tab

IQR: interquartile range; L: left hemisphere; R: right hemisphere;

term cUS data were missing for ten infants;

^Δ

presented are ADC values uncorrected for PMA.