Regional Brain Volumes and Their Later Neurodevelopmental Correlates in Term and Preterm Infants
Objective. To compare regional brain volumes measured in term and preterm infants, and to correlate regional volumes with measures of neurodevelopmental outcome.
Methods. High-contrast, high-resolution magnetic resonance imaging scans were acquired in 10 preterm and 14 term infants who were scanned near term. The cerebrum was segmented into cortical gray matter, white matter, cerebral ventricles, subcortical gray matter, cerebellum, and brainstem. The cortical gray matter, white matter, and ventricles were further divided into specific anatomic subregions, and the volumes were compared across groups. Measures of cognitive and motor development were acquired between 18 and 20 months of corrected age. Correlations of regional brain volumes with developmental outcome were assessed in the preterm group.
Results. Volumes in preterm infants were reduced in parieto-occipital gray matter and increased in the midbody, occipital horn, and temporal horns of the lateral ventricles. Gray matter volumes were also less prominently reduced in the sensorimotor and inferior occipital cortices. Normal lateralization of white matter volumes were altered in the parieto-occipital region in the preterm infants, who had significantly larger left-sided and smaller right-sided structures. White matter volumes in the sensorimotor and midtemporal regions correlated strongly with measures of neurodevelopmental outcome.
Conclusions. These findings of reduced volumes in sensorimotor and parieto-occipital regions in preterm infants, and the prospective correlations of regional volumes with cognitive outcome, confirm and extend findings previously reported in a cross-sectional study of 8-year-old prematurely born children. The data suggest that regional brain volumes near term are a promising marker for predicting disturbances of cognitive outcome in preterm infants. Further prospective, longitudinal studies of neonatal brain volumes and developmental indices into later childhood are required to confirm the utility of regional brain volumes as predictors of longer term outcome.
Previous neuroimaging studies have shown that children and infants born prematurely more often have anatomic brain abnormalities than do term controls. A broad range of abnormalities have been reported in qualitative studies of school-aged preterm children, including ventricular enlargement1 (especially of occipital horns2), white matter damage,1–3 and thinning of the corpus callosum.1,4 Qualitative studies in preterm neonates have reported high rates of periventricular leukomalacia (PVL),5,6 basal ganglia hemorrhage,5,7 cysts,5 white matter hemorrhage,5 delayed myelination,8,9 thalamic lesions,7 and brainstem abnormalities.7 More sophisticated, quantitative analyses have suggested that PVL in preterm newborns may be associated with the later development of reduced cortical gray matter volumes near term.10 In addition, preterm infants have been reported to have reduced cortical gray matter when treated with dexamethasone,11 reduced myelinated white matter volumes,11 and disturbances in the structural organization of white matter fibers.12 Several retrospective, qualitative studies have suggested an association of these brain lesions with neurodevelopmental outcome.1–3,8,13
We previously reported a detailed, quantitative comparison of regional brain volumes in 8-year-old children born prematurely and age-matched term controls.14 Preterm cortical volumes were significantly reduced in premotor, sensorimotor, midtemporal, and parieto-occipital regions. The degree of these morphologic abnormalities was strongly and inversely associated with measures of intelligence. Other morphologic abnormalities that were detected in the preterm group included dilated occipital and temporal horns of the lateral ventricles; reduced volumes of the basal ganglia, amygdala, and hippocampus; and thinning of the corpus callosum. Imaging studies of preterm infants using these same methodologies may be helpful not only to determine the nature and timing of the pathogenic insults, but also to predict developmental outcome.
We report here a study of regional brain morphology in 14 term and 10 preterm infants that employed many of the same image analytic methodologies as those in our previous study of school-aged, prematurely born children. Based on our findings in 8-year-old children, we hypothesized that in preterm infants we would detect significant, regionally specific abnormalities in volumes of cortical gray matter, white matter, and cerebral ventricles. We expected that posthoc analyses would detect these abnormalities primarily in the premotor, sensorimotor, temporal, and parieto-occipital regions of the cortex and white matter, and in the occipital and temporal horns of the lateral ventricles. We also predicted that posthoc analyses would detect correlations of regional brain volumes measured at term with measures of cognitive development obtained at 18 months of age.
The study protocol was approved by the Human Investigation Committee at the Yale University School of Medicine. Parents of all study infants provided written informed consent for their child’s participation.
Healthy term and medically stable preterm infants were recruited from the Well Infant Nurseries and the Newborn Intensive Care Unit of Yale New Haven Hospital. Birth weights for all infants were appropriate for gestational age, and none of the infants had known chromosomal or congenital anomalies. Perinatal data were obtained by maternal interview and chart review. Ultrasound-based designations for intraventricular hemorrhage and PVL are described elsewhere.15 Bronchopulmonary dysplasia was diagnosed as oxygen dependence at 36 weeks corrected age (CA; age from the obstetric due date).
All subjects were seen for neurodevelopmental assessment between 18 and 20 months CA. Standard neurologic examinations were performed, and each infant was evaluated using the Bayley Scales of Infant Development, which provided standardized subscale scores in both mental and motor domains. In the general population, these scales have a mean of 100 and standard deviation of 16.16
Magnetic Resonance Imaging (MRI) Scanning
Infants were swaddled, outfitted with earphones, and placed in the extremity (knee) coil of a 1.5 Telsa MR scanner (Signa LX; GE Medical Systems, Milwaukee, WI). Infant heads were lightly cushioned using towels and foam pads, and scanner sound was dampened with infant ear shields (Minimuffs; Natus Medical Inc, San Carlos, CA). All infants were monitored by electrocardiography and pulse oximetry (In vivo Research, Inc, Orlando, FL). A neonatal research nurse and pediatrician were present with the infant for the duration of the scan. All infants were scanned without sedation. Term infants were scanned within 3 days of their birth, and preterm infants were scanned as close as possible to the time of their hospital discharge. All MRI scans received clinical readings by a neuroradiologist, who was blind to the perinatal data.
An axial 3D Fast Spin Echo pulse sequence was acquired, with repetition time = 4700 msec, echo time = 150 msec, echo train length = 32, field of view = 20 cm, slice thickness = 1.3 mm, 8 locations per slab, 18 slabs, matrix 256 × 256, and 1 excitation. Scan time was 10 minutes 41 seconds. The long repetition time and echo time of this pulse sequence, necessitated by the high water content of infant brains,17,18 provided improved contrast (the ability to distinguish between tissue types19) compared with prior imaging studies of preterm infants.10,11,20 Resolution (.78 × .78 × 1.3 mm = .79 mm3/voxel) was also improved over those studies (.70 × .70 × 3 mm = 1.5 mm3/voxel).
Drift in image intensity because of inhomogeneity of the radio frequency pulse was detected along the anterior-posterior (A-P) and inferior-superior axes of the images. This drift was corrected by manually sampling the intensity of pixels representing pure cerebrospinal fluid (CSF; ie, CSF that contained no partial volume effects from adjacent gray matter) along each of these axes. Correction along the A-P axis was performed first by manually sampling CSF intensities independently at least 10 times in each coronal slice. These values were averaged for each slice, and a best fit, least squares regression function with a quadratic term for imaging slice was calculated for the entire A-P length of the brain. The intensity of each pixel in each coronal slice was then adjusted using this regression function. Identical CSF sampling and intensity correction procedures were performed in each axial slice to correct for drift in the inferior-superior direction. Sampling of CSF along each of the 3 axes was then performed to confirm the absence of any residual intensity drift. The brain was next rotated into a standard orientation using standard midline landmarks to correct for head rotation and tilt, and using the anterior and posterior commissures (AC and PC) to correct for head flexion and extension.
Tissue segmentation was performed by manually sampling pixels in the intensity-corrected data set that represented pure CSF, cortical gray matter, or white matter. Cortical CSF was segmented from cortical gray matter by invoking the average of the intensity values for CSF and gray matter in each axial slice as an intensity threshold. Cortical CSF was masked out of the image and then manually edited to yield an isolated cerebrum. Cortical gray matter was segmented from underlying white matter using an analogous intensity threshold, calculated as the average of gray and white matter intensities, and then manually edited. Cortical gray matter and CSF were masked out of the original intensity-corrected image to yield an image consisting of white matter, subcortical gray matter, ventricles, brainstem, and cerebellum. Subcortical gray matter, which included all of the basal ganglia and thalamus (these could not be differentiated from one another at the young age of these subjects), was defined and removed with manual tracing. The ventricles and cerebellum were defined and removed separately using an isointensity contour function with manual editing. The brainstem was transected at the pontomedullary junction. Pure white matter then remained (Fig 1).
Cortical gray and white matter tissues were parcellated into 8 subregions using a combination of 3 coronal planes—1 positioned tangent to the genu of the corpus callosum, and 1 each positioned through the AC and PC at their midline crossings—and an axial plane containing the AC-PC line. The lateral bodies of the cerebral ventricles were divided into frontal, midbody, occipital, and temporal subregions using these same planar divisions; the third and fourth ventricles were delimited manually from the rest of the ventricular system at the foramen of Monroe and the top of the cerebral aqueduct (Fig 2). The validity of related parcellation schemes have previously been documented.21–25 Moreover, this stereotactic method of subdivision would seem to be the most appropriate means of defining subregions of tissue types in infant brains, as gyral and sulcal anatomy at these young ages are too immature to permit their use in region definition.26–28 The interrater reliability of region and subregion measurements was assessed on 5 scans. Intraclass correlation coefficients calculated using a 2-way random effects model29 were >.90 for each of the region definitions.
All statistical procedures were performed in SAS version 8.0 (SAS Institute Inc, Cary, NC). A mixed models analysis (Proc Mixed) with repeated measures was performed across the spatial domain of the brain, with subregion volumes within a particular tissue type (ie, cortical gray, white matter, subcortical gray, etc) entered as the dependent measures. Each of the models included the 2 within-subjects factors, region and hemisphere, and their appropriate number of levels. The models for cortical gray and white matter, for example, included a region factor having 8 levels for each of the cortical subregions, and 2 levels of the factor hemisphere (left and right). A group factor with 2 levels (term and preterm) was a between-subjects factor. The rationale for selecting statistical covariates is described below, but for the final models included postmenstrual age (PMA) at the time of scanning, sex (male or female), and head circumference (to control for scaling effects within the brain). In addition to these covariates, we considered for inclusion in the model all 2- and 3-way interactions of group, sex, hemisphere, region, head circumference, and PMA, as well as the 2-way interactions of head circumference with hemisphere or region. Terms that were not statistically significant were eliminated via backward stepwise regression, with the constraint that the model at each step had to be hierarchically well-formulated (ie, all possible lower order terms had to be included in the model, regardless of their statistical significance).30 Least squares means and standard errors were calculated in the mixed models and plotted to assist in the interpretation of significant findings.
To identify the component terms that contributed most to the significance of group-by-region interactions, we examined the parameter estimates, 95% confidence intervals, and P values of the component terms in an analysis of fixed effects for the final mixed models. Because these analyses were performed relative to a reference region, tests for fixed effects were not possible for that reference region. Therefore, to assess the contribution of all relevant subregions of a given tissue type to significant interactions, the volumes of subcortical gray matter were used as the reference region in calculating fixed effects for other tissue types (only after determining, however, that subcortical gray matter volumes did not differ between term and preterm infants).
Additional analyses were conducted to assess the possible influence of infant characteristics such as minority status, weight at the time of scan, and maternal education on the stability of our findings. However, these variables had negligible effects on the P values and parameter estimates of the models, and thus were not included as covariates.
We tested our a priori hypotheses by assessing the significance of the region-by-group interaction for volumes of cortical gray matter, white matter, and cerebral ventricles. We also assessed the significance of this interaction for the analysis of other brain regions. To account for the multiple comparisons performed in testing a priori hypotheses, multivariate P values <.01 were considered statistically significant. P values <.05 in the posthoc fixed effects models were considered significant. All P values reported here are 2-sided and uncorrected for multiple comparisons.
Correlations of regional volumes with the Bayley measures of neurodevelopment at 18 to 20 months CA in the 9 preterm infants who had follow-up data available were assessed using Spearman or Pearson correlation coefficients, with and without head circumference and gestational age at birth entered as statistical covariates.
Usable scans were acquired in 14 term and 10 preterm infants. Scans were attempted in another 7 infants (3 term and 4 preterm), but motion artifact precluded their use. The preterm infants contributing to the usable data set, compared with the term infants, were of similar sex composition (term: 9 boys, 5 girls; preterm: 8 boys, 2 girls; Fisher exact P = .65), minority representation (infants were identified by their mothers as non-white; term = 8, preterm = 4; Fisher exact P = .68), and maternal education (t = 1.2, P = .27). The preterm compared with term infants at the time of scan were significantly younger (PMA preterm: 35.4 ± 1.1 week; term: 40.5 ± 1.5 weeks; t = 9.39, P < .001), they weighed less (preterm: 2259 ± 360 g; term: 3316 ± 441 g; t = 9.39, P < .001), and they had significantly smaller head circumferences (preterm: 32.3 ± 1.3 cm; term: 34.9 ± 32.3 cm; t = 4.6, P < .001). Of the 10 preterm infants, 9 received a single course of antenatal corticosteroids, 6 received standard doses of postnatal steroids, 9 were diagnosed with bronchopulmonary dysplasia, and 3 were products of multiple gestation pregnancies. Clinical readings of the MRI scans indicated that in the preterm infants, 2 had evidence of previous intraventricular hemorrhage and 2 had PVL. Term infants all had normal clinical assessments at the time of scanning. The term infants were all healthy and cared for in the newborn nursery.
Demographic Correlates and Determination of Statistical Covariates
Scatterplots indicated that head circumference, as well as both PMA and weight at the time of scanning, correlated significantly with cortical gray and white matter volumes (Fig 3). Head circumference, PMA, and weight were themselves significantly intercorrelated (data not shown). Inclusion of head circumference to control for scaling effects within the brain was deemed crucial from a theoretical perspective, as well as from the practical standpoint of accounting for the largest portion of variance in regional volumes. Preliminary models indicated that PMA accounted for additional variance in volume when head circumference was included in the model, whereas weight did not. Therefore, head circumference and PMA were included as covariates in all primary analyses, although we also repeated the analyses to ensure that group differences in PMA at the time of scanning were not influencing our findings unduly when this was included as a statistical covariate. Regional volumes did not correlate significantly with any other demographic features, including gestational age at birth or sex.
Testing of a priori hypotheses confirmed the presence of regionally specific differences in volume between the term and preterm groups within cortical gray matter, white matter, and lateral ventricles (Table 1). In white matter, these regional differences moreover varied by hemisphere. Volumes in subcortical gray matter, cerebellum, and brainstem did not differ between groups.
Analyses of fixed effects indicated that in gray matter, the significant group-by-region effect (P < .0001) derived primarily from group differences in volumes of the sensorimotor (t176 = 2.4, P = .02), parieto-occipital (t176 = 9.8, P < .0001), and inferior occipital (t176 = 3.5, P = .0006) cortices. In white matter, the significant group-by-hemisphere-by-region effect (P = .0004) derived primarily from hemisphere-specific differences between groups in volumes of the parieto-occipital (t176 = 2.9, P = .004) and inferior occipital (t176 = 1.8, P = .07) regions. In the ventricles, the significant group-by-region effect (P = .0003) derived primarily from group differences in volumes of the midbody (t134 = 8.8, P < .0001), occipital horns (t134 = 10.5, P < .0001), and temporal horns (t134 = 7.5, P < .0001), especially on the left (group-by-hemisphere F1,46 = 2.9, P = .09).
Examination of regional least squares means with head circumference and PMA as covariates revealed that parieto-occipital gray matter was significantly smaller in the preterm group (Fig 4). However, the overall group-by-region effect in cortical gray matter derived additionally from group differences in regional volume that were in opposing directions across differing brain regions—preterm volumes were similar to term volumes in sensorimotor and inferior occipital gray matter regions, whereas they were substantially larger in dorsal prefrontal, orbitoprefrontal, premotor, subgenual, and midtemporal regions (Fig 4). This pattern of larger anterior regions in the preterm group resulted from covarying for PMA in the statistical models, which in effect corrected for group differences in gestational age at the time of scanning. Because regional volumes continue to grow in the postnatal period as PMA increases (Fig 3), correcting statistically for PMA essentially enlarged most preterm cortical gray matter volumes, compared with volumes in the group when they were not adjusted for PMA. Thus, when cortical gray matter was reexamined without PMA as a covariate, volumes in these same anterior regions were comparable across term and preterm infants, whereas volumes in the preterm group were substantially smaller in the sensorimotor, inferior occipital, and parieto-occipital cortices (Fig 5). In other words, covarying for PMA in this data set tended to normalize sensorimotor and inferior occipital volumes, and it enlarged volumes in dorsal prefrontal and other anterior cortical regions in preterm compared with term infants. Therefore, in both analyses (ie, with and without PMA as a covariate), smaller sensorimotor, parieto-occipital, and inferior occipital regions, and larger prefrontal and premotor regions contributed together to the significant group-by-region interaction. We might best conclude from these analyses of cortical gray matter that volumes of parieto-occipital gray matter were significantly reduced in the preterm group, whether or not PMA was included as a covariate. Preterm volumes were, in addition, relatively larger in anterior cortices and relatively smaller in sensorimotor and posterior cortices; the emphasis on one or the other of these effects depended on whether the analysis corrected statistically for PMA at the time of scanning.
Examination of least squares means for white matter volumes with head circumference and PMA as covariates indicated that the hemisphere-specific effects in parieto-occipital regions resulted from larger left-sided and smaller right-sided volumes in the preterm group (Fig 4). A similar pattern was observed in the inferior occipital region. Mean volumes in the ventricles were larger in the midbody, occipital horns, and temporal horns of the lateral ventricles, with a left-sided predominance evident in the occipital and temporal horns of the preterm group (Fig 4). These findings persisted when PMA was not included in the analyses.
To determine if the lateralized findings in the preterm white matter were indeed specific to this tissue class, an additional repeated measures analysis was performed to test whether the group-by-hemisphere-by-region effect interacted with tissue class. The modeling procedure was similar to those already described, except for the inclusion of an additional within-subjects factor of tissue type (white or gray). A significant tissue-by-group-by-hemisphere-by-region effect (F7154 = 2.2, P = .04) confirmed that the lateralized abnormalities in the preterm group were restricted to the white matter.
Without the inclusion of statistical covariates, the mental subscale of the Bayley at 18 to 20 months CA correlated significantly with volumes of neonatal white matter in the left (Spearman’s ρ = .72, P = .03) and right (ρ = .73, P = 02) premotor region, left (ρ = .83, P = .005) and right (ρ = .95, P < .0001) sensorimotor region, the right subgenual region (ρ = .78, P = .01), and the left (ρ = .77, P = .02) and right (ρ = .92, P = .001) midtemporal regions (Fig 6). The mental subscale also correlated significantly with volumes of gray matter in the left sensorimotor (ρ = .73, P = .02) and left midtemporal cortices (ρ = .72, P = .03). The motor subscale score correlated with white matter volumes in the left (ρ = .82, P = .007) and right (ρ = .77, P = .02) subgenual regions.
Because gestational age at birth and head circumference are known to correlate with neurodevelopmental outcome,31,32 correlations of regional volumes with neurodevelopmental indices were assessed while controlling for these variables. Mental subscale scores still correlated significantly with white matter volumes in the right sensorimotor (β = .94, P = .003) and right midtemporal regions (β = .94, P = .003; Fig 6). Weaker associations at trend levels of significance (P < .10) were detected for the mental subscale scores in these same regions of the left cerebral hemisphere. Inclusion of PMA at the time of scanning as a covariate did not affect these results. No other regional volumes of white or gray matter, ventricles, or subcortical structures correlated with developmental indices in the preterm infants at the time of follow-up.
Significant differences were detected between term and preterm infants in subregions of cortical gray matter, white matter, and cerebral ventricles. Although other imaging studies of infant brains have segmented images into gray, white, and CSF tissue types,10,11,20 this is to our knowledge the first study to provide subregional measurements of those tissues in infant brains. This is also the first prospective study to demonstrate that brain volumes correlate with measures of neurodevelopmental outcome. Defining subregions proved to be important in confirming our hypotheses regarding regional specificity of abnormal brain development in preterm infants and in demonstrating the potential of these abnormalities to predict poor developmental outcome. Moreover, the regional subdivisions were the same as those employed in our previous imaging study of 8-year-old preterm children,14 thereby permitting a direct comparison of findings across age groups. Although these findings are intriguing, they are based on a relatively small number of clinically heterogeneous term and preterm infants; therefore, they warrant replication.
All cortical gray matter regions were significantly smaller in the preterm group when head circumference was not included as a covariate. However, covarying for head circumference is generally indicated in imaging studies to control for scaling effects within the brain, and it is especially important in the study of preterm infants, for whom head circumferences are known to be smaller than in same-age term controls.33 Controlling for head circumference in this study also helped to adjust for significant differences in age (PMA) across groups, because head circumference correlated with PMA at the time of scanning. Thus, when head circumference was included as a statistical covariate, parieto-occipital, sensorimotor, and inferior occipital cortices were significantly smaller in the preterm group, whereas other regions were relatively normal in size. However, PMA as a second covariate accounted for a significant additional portion of variance in volume, suggesting that PMA should be considered in the statistical model. With PMA included as a covariate, volumes of anterior cortical gray matter (dorsal prefrontal, orbitofrontal, subgenual, and premotor regions) tended to be larger, parieto-occipital volumes were significantly reduced, and sensorimotor and inferior occipital volumes were similar in preterm compared with term infants.
Thus, the findings in cortical gray matter depended to some extent on the statistical covariates employed. Parieto-occipital gray matter volumes were reduced bilaterally in all statistical models that we assessed, providing the clearest evidence for differences in cortical gray matter volumes across groups. Sensorimotor and inferior occipital gray matter volumes in this study are probably best regarded as being relatively smaller, and anterior cortical gray matter volumes as being relatively larger, in the preterm group. In our prior study of 8-year-old children,14 prefrontal regions of preterm children were nonsignificantly larger, and smaller parieto-occipital and sensorimotor regions were significantly smaller, than volumes in term controls, supporting the validity of similar findings in infants when volumes were adjusted for PMA and head circumference. To clarify group comparisons of sensorimotor, inferior occipital, and anterior cortical regions in term and preterm infants, future studies will need to match groups carefully on PMA at the time of scanning.
The midbody, occipital horns, and temporal horns of the lateral ventricles were significantly enlarged in preterm infants regardless of whether PMA was used as a statistical covariate, a finding that was also observed in our study of 8-year-olds. A left-sided predominance was especially evident in occipital and temporal horns of the preterm infants. In white matter regions, parieto-occipital volumes were larger on the left and smaller on the right in the preterm group, with a similar pattern observed in the inferior occipital region. This lateralized abnormality affected white matter uniquely and significantly compared with cortical gray matter volumes in the preterm group. The pathophysiological significance of the lateralized white matter abnormalities is unknown.
Prior published reports of quantitative brain volume measurements in term and preterm infants have come primarily from a single laboratory.10,11,20 In 1 of those previous studies, 11 preterm infants unexposed to postnatal steroid treatment did not differ significantly from 14 term infants in volumes of total cortical gray matter or ventricular CSF.11 In a second study of 10 preterm infants with PVL, 10 preterm infants without PVL, and 14 term controls, reduced cortical gray matter was found only in the preterm infants who had PVL.10 White matter volumes did not differ across any of the diagnostic groups, and CSF volumes were greater in preterm infants, whether or not they displayed evidence of PVL. Differences in MRI pulse sequences and the absence of subregional measurements in each of the tissue classes make comparison with our findings difficult. Nevertheless, our findings indicate that the reduced volumes of gray matter in parieto-occipital and sensorimotor regions in our sample were not caused by PVL, as only 2 of the preterm infants had PVL, and these subjects did not unduly influence the group comparisons.
Volumes of the sensorimotor and midtemporal regions, particularly the white matter portions of these regions in the right hemisphere, correlated strongly with measures of neurodevelopmental outcome at 18 to 20 months CA, whether or not gestational age at birth and head circumference were included as statistical covariates. These correlations were strongest for mental developmental indices. Cortical gray matter in these same regions also correlated with outcome when these variables were not included as covariates. The persistence of statistical significance after controlling for gestational age at birth indicates that white matter volumes in these regions predict outcome at 18 to 20 months CA beyond the degree to which gestational age and head circumference alone would predict. Volumes of these same sensorimotor and midtemporal regions, particularly in the right hemisphere, correlated significantly with cognitive measures in our cross-sectional study of 8-year-old prematurely born children.14 Taken together, findings across these studies suggest that volumes of these brain regions in preterm neonates may prove to be a useful marker to help identify children at risk for cognitive impairment if those preterm infants are followed into later childhood. If this possibility is confirmed, then the regional abnormalities detected in infancy may help to direct efforts to develop prevention and early intervention strategies for at-risk infants, based on the purported functional characteristics of the brain regions where the anatomic abnormalities are found. Involvement of parieto-occipital regions, and possibly sensorimotor, inferior occipital, and midtemporal regions, for example, may suggest that remediation efforts should be directed to assisting the development of visuospatial, attentional, and sensorimotor skills.
Despite the similarity in findings of the present study with those reported in 8-year-old preterm children, findings across the 2 studies differed in several important respects. No evidence for a significant association of volume abnormalities with gestational age at birth was detected in the infants, whereas this was a particularly strong finding in the sensorimotor regions of older children. Also, reductions in premotor, sensorimotor, and midtemporal volumes were much more prominent in 8-year-old children than in the infants studied here. If these age-specific differences in regional abnormalities are confirmed in infants and children born prematurely, they would suggest that the premotor, sensorimotor, and midtemporal regions of older preterm children may develop abnormal structural features sometime between infancy and later childhood. The later development of these structural abnormalities would in turn suggest that we may be able to identify the pathogenic influences that are operative in the intervening time, and then to devise strategies to prevent the development of these anatomic disturbances and preserve the functional capacities that those regions presumably subserve.
This work was supported in part by grants NS 27116 and NS 42027 (National Institute of Neurological Disorders and Stroke), RR 06022 (National Center of Research Resources), MH01232 and MH59139 (National Institute of Mental Health), and the Suzanne Crosby Murphy Endowment at Columbia University.
We thank Dr Robert Fulbright, Elizabeth Haldeman, Michael Kane, Ronald Whiteman, Lisa Perry, Karol Katz, Karen Schneider, Monica Konstantino, JoAnn Poulsen, Hedy Sarofin, and Terry Hicky for their technical assistance. Analyze Software was developed by the Biomedical Imaging Resource, Mayo Foundation (Rochester, MN), Richard A. Robb, PhD, Director.
- Received August 2, 2002.
- Accepted October 7, 2002.
- Reprint requests to (B.S.P.) Columbia College of Physicians and Surgeons and New York State Psychiatric Institute, Unit 74, 1051 Riverside Dr, New York, NY 10032. E-mail:
- ↵Olsén P, Vainionpää L, Pääkkö E, Korkman M, Pyhtinen J, Järvelin M-R. Psychological findings in preterm children related to neurologic status and magnetic resonance imaging. Pediatrics.1998;102 :329– 336
- ↵Cooke RWI, Abernethy LJ. Cranial magnetic resonance imaging and school performance in very low birth weight infants in adolescence. Arch Dis Child Fetal Neonatal Ed.1999;81 :F116– F121
- ↵Keeney SE, Adcock EW, McArdle CB. Prospective observations of 100 high-risk neonates by high-field (1.5 Tesla) magnetic resonance imaging of the central nervous system. II. Lesions associated with hypoxic-ischemic encephalopathy. Pediatrics.1991;87 :431– 438
- ↵Barkovich AJ, Sargent SK. Profound asphyxia in the premature infant: imaging findings. AJNR.1995;16 :1837– 1846
- ↵Guit GL, van de Bor M, den Ouden L, Wondergem JHM. Prediction of neurodevelopmental outcome in the preterm infant: MR-staged myelination compared with cranial US. Pediatr Radiol.1990;175 :107– 109
- ↵Murphy BP, Warfield S, Inder TE, et al. Impaired cerebral cortical gray matter growth following treatment with dexamethasone for neonatal chronic lung disease. Pediatrics.2001;107 :217– 221
- ↵Bayley N. The Bayley Scales of Infant Development. New York, NY: New York Psychological Corporation; 1993
- ↵Peterson BS, Ment LR. The necessity and difficulty of conducting MRI studies on infant brain development. Pediatrics.2001;107 :593– 594
- ↵Peterson BS. Brain imaging modalities and methodologies in the study of childhood neuropsychiatric disorders. Trends Evidence-Based Neuropsychiatry. In press
- Filipek PA, Richelme C, Kennedy DN, Caviness VS. The young adult human brain: an MRI-based morphometric analysis. Cereb Cortex.1994;4 :344– 360
- ↵Kennedy DN, Lange N, Makris N, Bates J, Meyer J, Caviness VSJ. Gyri of the human neocortex: an MRI-based analysis of volume and variance. Cereb Cortex.1998;8 :372– 384
- ↵Battin MR, Maalouf EF, Counsell SJ, et al. Magnetic resonance imaging of the brain in very preterm infants: visualization of the germinal matrix, early myelination, and cortical folding. Pediatrics.1998;101 :957– 962
- ↵Hack M, Friedman H, Fanaroff AA. Outcomes of extremely low birth weight infants. Pediatrics.1996;98 :931– 937
- ↵McCormick M, Workman-Daniels K, Brooks-Gunn J. The behavioral and emotional well-being of school-age children with different birth weights. Pediatrics.1996;97 :18– 25
- ↵Ehrenkranz RA, Younes N, Lemons JA, et al. Longitudinal growth of hospitalized very low birth weight infants. Pediatrics.1999;104 :280– 289
- Copyright © 2003 by the American Academy of Pediatrics