www.elsevier.com/locate/ynimg NeuroImage 27 (2005) 579 – 586
Comparing microstructural and macrostructural development of the cerebral cortex in premature newborns: Diffusion tensor imaging versus cortical gyration Amy R. deIpolyi,a,b Pratik Mukherjee,c,* Kanwar Gill,c Roland G. Henry,c Savannah C. Partridge,c Srivathsa Veeraraghavan,c Hua Jin,c Ying Lu,c Steven P. Miller,d,e Donna M. Ferriero,d,e Daniel B. Vigneron,c and A. James Barkovichc,d,e a
Gladstone Institute of Neurological Disease, University of California, San Francisco, CA 94143, USA Neuroscience Program, University of California, San Francisco, CA 94143, USA c Department of Radiology, Neuroradiology Section, University of California, San Francisco, Box 0628, 505 Parnassus Avenue, San Francisco, CA 94143, USA d Department of Neurology, University of California, CA 94143, USA e Department of Pediatrics, University of California, CA 94143, USA b
Received 21 October 2004; revised 6 February 2005; accepted 8 April 2005 Available online 25 May 2005 This study assessed microstructural development in four regions of the human cerebral cortex during preterm maturation using diffusion tensor imaging (DTI), compared to the macrostructural development of cortical gyration evaluated using three-dimensional volumetric T1-weighted MR imaging. Thirty-seven premature infants of estimated gestational age (EGA) ranging from 25 to 38 weeks were prospectively enrolled and imaged in an MR-compatible neonatal incubator with a high-sensitivity neonatal head coil. Cortical gyration was measured quantitatively as the ratio of gyral height to width on the volumetric MR images in four regions bilaterally (superior frontal, superior occipital, precentral, and postcentral gyri). Mean diffusivity (D av), fractional anisotropy (FA—the fraction of D av that is anisotropic), and the three DTI eigenvalues (components of diffusivity radial and tangential to the pial surface of cortex) were measured in the same cortical regions. Cortical gyration scores, FA, and radial diffusivity were all significantly correlated with EGA ( P < 0.0001). However, in multivariate analysis, no significant relationship ( P > 0.05) was found between DTI parameters and cortical gyration beyond their common association with estimated gestational age. Pre- and postcentral gyri had significantly lower anisotropy than the superior occipital and superior frontal gyri ( P < 0.05), indicating that DTI is sensitive to regional heterogeneity in cortical development. Maturational changes in the DTI eigenvalues of cortical gray matter were found to differ from those that have previously been described in developing white matter, with a significant age-related decline in the radial diffusivity ( P < 0.0001) but not in the tangential diffusivities ( P > 0.05). D 2005 Elsevier Inc. All rights reserved. Keywords: Cerebral cortex; Diffusion tensor imaging; Cortical gyration; Magnetic resonance imaging
* Corresponding author. Fax: +1 415 353 8593. E-mail address: [email protected]
(P. Mukherjee). Available online on ScienceDirect (www.sciencedirect.com). 1053-8119/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2005.04.027
Introduction Premature birth is an increasingly common public health problem. The rate of prematurity exceeded 11% of all live births in the United States in 2000 (Martin et al., 2002), and its incidence continues to rise. Premature infants are vulnerable to a host of neurodevelopmental deficits, including spastic diplegia or quadriplegia, visual impairment, cognitive deficiencies, and behavioral disorders such as attention deficit hyperactivity disorder (Hack and Taylor, 2000; Hack et al., 2002; Wood et al., 2000). Developing reliable methods of assessing human cortical development may enable the prediction of clinical outcomes and selection of patients who are good candidates for emerging neuroprotective treatments. MR imaging is especially important in the evaluation of preterm neonates because few clinical or laboratory tests presently available in the early neonatal period have proven useful in predicting the neurological outcomes of preterm newborns. Asymmetrical signal intensity in the posterior limb of the internal capsule on MRI performed at 40 weeks estimated gestational age (EGA) is associated with future hemiplegia, though a radiological marker would be more useful in preterm infants during the first few days or weeks after birth rather than at 40 weeks EGA, especially once more effective preventative treatments emerge (De Vries et al., 1999). Sequential cranial ultrasound has been used successfully in one study to detect major abnormalities, including severe hemorrhage, focal infarction, and cystic white matter injury, that are strong risk factors for cerebral palsy (De Vries et al., 2004). However, in this study, the positive predictive value of major cranial ultrasound abnormalities for cerebral palsy for infants born at less than 32 weeks EGA was only 48%, highlighting the need for other sensitive assays of impaired cortical development.
A.R. deIpolyi et al. / NeuroImage 27 (2005) 579 – 586
In spite of the advances in the capability of neuroimaging to predict future motor deficits, there is still a paucity of tests highly predictive of learning disabilities and sensorineural impairments (De Vries et al., 2004). Premature infants with white matter injury studied with MRI at or near-term-equivalent age have reduced gray matter volumes compared with premature infants without white matter injury or term-born infants (Inder et al., 1999). Furthermore, the mental subscale of the Bayley Scales of Infant Development correlates inversely with white matter volumes in sensorimotor and midtemporal regions of premature infants assessed by MRI near term (Peterson et al., 2003). These studies suggest that MR imaging may be a powerful tool to assess developmental changes that predict neurodevelopmental and cognitive outcome, though we still lack tests administered within days of birth that have high predictive value for specific cognitive deficits. The routine clinical evaluation of cortical maturation in neonates with T1- and T2-weighted MR imaging has been through the assessment of macroscopic gyration and sulcation patterns, which is not quantitative and which also may be insensitive to the complex microstructural changes that take place during cortical development. Qualitative gyral scoring systems take advantage of the fact that gyri become taller and narrower during development, and scores reflect the relationship between the height and width (van der Knaap et al., 1996). Though qualitative scoring methods have the benefit of rapid assessment in the clinical setting, there are several problems with the qualitative approach. First, subjective assessments may have inter-subject and even intra-subject variability. Second, since clinical evaluation is based on two-dimensional images of the complex three-dimensional contours of the developing cortical surface, variation in head position on different MR examinations may alter the appearance of gyri and sulci, decreasing the reliability of qualitative scoring. Finally, qualitative scores produce discontinuous variables that may not adequately capture the continuous process of development. Macroscopic measures of gyral development detect differences between term and extremely preterm infants (Ajayi-Obe et al., 2000; Battin et al., 1998) and between healthy newborns and newborns with severe brain insults (Slagle et al., 1989). However, gyral scores do not detect developmental alterations in moderately preterm infants (Slagle et al., 1989) or in newborns with mild brain injury (Battin et al., 1998), reinforcing the need for a more sensitive assay of cortical maturation in premature neonates. Unlike conventional T1- and T2-weighted sequences, DTI is sensitive to microstructural changes in the cerebral cortex and has emerged as a sensitive assay of cortical development in fetal mice (Mori et al., 2001) and premature human newborns (McKinstry et al., 2002). DTI measures the diffusion of water in each voxel and the extent to which water diffuses in particular directions as a result of the microstructural characteristics of the tissue imaged. High anisotropy indicates that the magnitude of diffusion is very unequal in different directions. The principal eigenvector specifies the direction in which water diffusion is greatest. The principal eigenvalue of the diffusion tensor is denoted k 1 and represents the magnitude of diffusion in the preferred direction given by the principal eigenvector. Early in preterm development, cortical gray matter is characterized by a microarchitecture oriented radial to the pial surface of the cortex, thought to be due to the presence of radial glial cells that guide neuronal migration as well as the early developing apical dendrites of immature neurons in the cortical plate. The radial microstructure preferentially hinders the diffusion of water in directions tangential to the pial surface, resulting in
predominantly radial diffusivity manifested by high values of diffusion anisotropy and by radial orientation of the principal eigenvector of the diffusion tensor in the developing brains of mice (Mori et al., 2001) and humans (Maas et al., 2004; McKinstry et al., 2002; Mukherjee et al., 2003a,b). Thus, in the developing preterm cortex, k 1 has relatively high values and corresponds to the radial component of diffusivity (McKinstry et al., 2002; Mukherjee et al., 2003a,b). As yet, there have been no DTI studies of the developing human cerebral cortex in regions other than the parieto-occipital cortex (McKinstry et al., 2002). In this report, we compare four different regions of the cerebral cortex to test the hypothesis that DTI can reveal regional heterogeneity in their development. We also perform the first quantitative investigation of maturational changes in the DTI eigenvalues of the cerebral cortex to test the hypothesis that the previously observed decrease in cortical anisotropy during preterm maturation (McKinstry et al., 2002) is largely due to a decline in the radial component of diffusivity. In order to determine how DTI measurements of microstructural cortical development compare to gyral maturation on a macrostructural scale, we use a variant of a previously used qualitative scoring technique to assess gyration and sulcation quantitatively. We expect that DTI parameters of cortical development and gyration scores would both correlate with the estimated gestational age of premature newborns at the time of their MR scans. Moreover, assuming that intrinsic microstructural factors measured by DTI drive the process of cortical folding, we hypothesized that DTI parameters may correlate with gyration scores even beyond their common association with EGA.
Materials and methods Subjects All premature infants in the neonatal intensive care unit at our institution who were born at or before 35 weeks EGA were screened for this study as part of an ongoing prospective investigation of preterm brain development approved by the institutional review board at our medical center. We obtained informed consent from all participants’ parents and calculated each subject’s EGA according to an early ultrasound or the last menstrual period, using the ultrasound date if the difference between the two estimates exceeded 1 week. Infants underwent the MR examination once they were stable for transport to the MR imager, and, if possible, scanned again near term-equivalent age but before hospital discharge. We excluded subjects with any cerebral cortical abnormality on T1- or T2-weighted images or motion artifact on MR imaging. Experimental data acquisition and MR image analysis We used a custom-built MR-compatible neonatal incubator and a high-sensitivity specialized neonatal head coil to reduce patient motion, increase patient safety and comfort, and improve signal-tonoise ratio of the MR images (Dumoulin et al., 2002). During scanning, neonatologists monitored the infants and hand-ventilated intubated newborns. Infants were fed prior to scanning to avoid sedation, but intravenous pentobarbital was given to some infants when sedation was necessary. Using a 1.5 T Signa EchoSpeed scanner (GE Medical Systems, Milwaukee, WI), we performed an MR examination consisting of
A.R. deIpolyi et al. / NeuroImage 27 (2005) 579 – 586
axial T2-weighted dual spin echo (TR 3000 ms, TE 60/120 ms, 4 mm slice thickness), and coronal T1-weighted 3D spoiled gradient echo (SPGR) imaging (TR 35 ms, TE 9 ms, flip angle 35-, 1.5 mm thickness), the latter with an 18 cm FOV and a 192 256 acquisition matrix and an in-plane resolution of 0.7 0.7 mm2. We used a multislice spin echo single-shot echoplanar sequence to perform DTI (TR 7000 ms, TE 99.5 ms, 3 mm slice thickness, no gap, 3 repetitions per image, with 18 36 cm FOV and 128 256 acquisition matrix), acquiring axial images through the whole brain with an in-plane resolution of 1.4 1.4 mm2 (Mukherjee et al., 2003a,b). We acquired 7 images per axial section, including a T2weighted reference image (b = 0 s/mm2) and 6 diffusion-weighted images (b = 600 s/mm2) in noncollinear gradient directions. MR data were transferred to off-line workstations for postprocessing. From the DTI data, we computed the D av, FA, and the three eigenvalues of the diffusion tensor (k 1, k 2, and k 3, in decreasing order of magnitude) using previously described methods (Partridge et al., 2004; Fig. 1). FA measures the fraction of overall diffusivity that is anisotropic and was calculated using the following equation: rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ X3 pﬃﬃﬃ ðk Dav Þ2 i¼1 i 3 qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ FA ¼ pﬃﬃﬃ 2 k21 þ k22 þ k23 qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ ðk1 k2 Þ2 þ ðk2 k3 Þ2 þ ðk3 k1 Þ2 ¼ pﬃﬃﬃqﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 2 k21 þ k22 þ k23 The T1-weighted coronal SPGR images were reformatted and rotated using three-dimensional imaging software to generate axial images in a consistent orientation across subjects. Coronal images were first rotated until the interhemispheric fissure was vertical then were reformatted to sagittal images. The sagittal images were rotated so that the frontal pole and occipital pole were oriented along a horizontal line then were reformatted to axial images that were used to score gyration. White matter injury of prematurity, often referred to as periventricular leukomalacia, was scored on T1- and T2-weighted images by two board-certified attending pediatric neuroradiologists
experienced in neonatal imaging who were blind to the DTI data and who resolved discrepancies by consensus. Newborns were given a white matter injury score of 0 if there were no abnormalities in the white matter; a score of 1 if there were three or fewer areas of T1 signal abnormality of less than 2 mm; a score of 2 if there were more than three areas of T1 signal abnormalities of less than 2 mm or if the areas were more than 2 mm but involved less than 5% of the hemisphere; and a score of 3 if the injury involved more than 5% of the hemisphere (Miller et al., 2002). Region of interest analysis and gyration scores We scored gyration on the reformatted SPGR axial images. Four gyri were assessed bilaterally, including the precentral and postcentral gyri, the superior frontal gyrus, and the superior occipital gyrus. Each gyrus was assessed on a consistent plane across subjects. Precentral and postcentral gyri were scored on the axial slice just above the roof of the lateral ventricles; the superior frontal gyrus was scored on a slice one third of the distance from the roof of the ventricles to the vertex of the brain; the superior occipital gyrus was scored an equivalent distance down from the roof of the ventricles. Our cortical gyration score is a quantitative version of the scoring system of van der Knaap et al. (1996). Instead of a qualitative index for the extent of cortical folding, we measured the height and width of each gyrus of interest then calculated the ratio of gyral height to gyral width, generating a continuous quantitative variable that increases monotonically with maturation during the preterm period (Fig. 2). On DTI images, we identified corresponding regions of interest in three consecutive axial images. Regions of interest were defined on the b = 0 s/mm2 T2-weighted images as five contiguous voxels (10 mm2 total area) within a one-voxel thick band of cortex on the surface of the gyrus that had been scored on the SPGR images (Fig. 3). The D av, FA, k 1, k 2, and k 3 were calculated for each region of interest. Statistical methods Mixed random-effects models (Laird and Ware, 1982) were employed to assess the differences in DTI parameters between
Fig. 1. Diffusion tensor imaging of a 30-week EGA premature newborn. (A) Mean diffusivity (D av) image shows lower D av in the cortical plate compared to the overlying CSF and the underlying subplate, however, the cortical plate itself appears regionally homogeneous. (B) Fractional anisotropy (FA) image shows higher FA in the cortical plate (white arrow and black arrow) compared to the overlying CSF (black arrowhead) and underlying subplate; note that the FA in the perirolandic cortex (white arrowhead) is less than in frontal (white arrow) or occipital cortex (black arrow), indicating more advanced maturation. FA was pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ ðk1 k2 Þ2 þ ðk2 k3 Þ2 þ ðk3 k1 Þ2 calculated as FA ¼ . (C) Maximum eigenvalue (k 1) image. For the cortical plate, this eigenvalue corresponds to the radial pﬃﬃpﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 2 2 2 2
k1 þ k2 þ k3
component of diffusivity. Like FA, the radial diffusivity is less in the perirolandic cortex than in frontal or occipital cortex, suggesting more advanced maturation. (D) Whisker plot shows the orientation of k 1 in each voxel of the right frontal lobe as red dashes, overlaid on the FA map of panel (B). The whisker plot shows that, in the cortical plate (CP), k 1 is consistently oriented radial to the pial surface. No such radial organization is observed in deeper cerebral layers, such as the subplate (SP), which has very low FA, or the intermediate zone (IZ), which has FA greater than the subplate but less than the cortical plate.
A.R. deIpolyi et al. / NeuroImage 27 (2005) 579 – 586
Fig. 2. Quantitative gyration scores. We measured the height and width of each gyrus of interest on the reformatted SPGR images. The gyral ratio was calculated as the height divided by the width. Example from the postcentral gyrus of a 32-week EGA premature infant.
different regions of interest. We controlled for possible confounders including EGA at the time of birth, EGA at the time of the scan, and laterality (i.e. left versus right hemisphere). We included all 37 subjects in the analyses and accounted for repeated measures from the same infant at different ages across serial examinations in the mixed models. We calculated the least squares means and standard errors of DTI values for each region as summary descriptions of the data. Statistical significance of the difference between regions was computed based on analysis of ranks of the DTI parameters using Tukey – Kramer adjustment for multiple comparisons. Estimating the correlation coefficients based on the longitudinal data should take into consideration both individual and design effects. We used the mixed effects model to estimate the effects of design parameters including EGA at the time of birth, EGA at the time of the scan, laterality, region of interest, white matter injury score, and repeated measures from the same subjects on the DTI parameters and gyral score. Similar to the theory of partial correlation coefficients for normal distributed data, we used residuals of mixed effects models to calculate the partial Pearson’s correlation coefficients between the parameters of interest and age after removing the design effects. The corresponding 95% confidence interval was calculated based on Fisher’s z transformation (Daniel, 1999). For an estimated correlation coefficient r using n pairs of data, a Fisher’s z transformation is z r = 1/2 ln[(1 + r) / (1 r)], which approximately follows a normal distribution with a mean of 1/2 ln[(1 + q) / (1 q)] and a variance of 1 / (n 3), where q is the true correlation coefficient. Instead of using the number of observed pairs to calculate variance, we used degrees of freedom for the residuals to replace n to reflect the effect of removing design parameters in our calculation of the partial correlation coefficients and then calculated a 95% confidence interval of q. The statistical calculations were performed using SAS 8.2 (SAS Institute, Cary, NC) and S-Plus 6 (Insightful Corp, Seattle, WA).
(range: 25.3 to 38.1 weeks). Nine newborns had a white matter injury score of 0, 18 had a white matter injury score of 1, 7 had a white matter injury score of 2, and 3 were scored as 3. On the conventional T1- and T2-weighted examinations, there were no cortical abnormalities in any of the examinations. The least squares means and standard errors of DTI parameters for each region are given in Table 1. The FA values in the developing cortex are low compared with the FA of adult white matter but are greater than neonatal white matter FA (see Fig. 1B). Our data are consistent with previous studies demonstrating that the primary eigenvector in the developing cortex is oriented radial to the pial surface, as shown by the whisker plot in Fig. 1D (Maas et al., 2004; McKinstry et al., 2002; Mori et al., 2001). Table 2 shows the results of pair-wise comparisons between the different regions, showing the statistically significant differences in DTI parameters between regions. FA was highest in the superior frontal gyrus followed by the superior occipital gyrus. FA was significantly lower in both the pre- and postcentral gyri than the other regions of interest. The major eigenvalue k 1 showed the same pattern of regional heterogeneity as FA. In contrast, D av was higher in the superior frontal and occipital gyri than in the pre- and postcentral gyri. There were no statistically significant differences in the minor eigenvalues, k 2 and k 3, among the four regions. In our mixed effects models, cerebral cortical FA and k 1 each showed a statistically significant inverse correlation with EGA (FA: r = 0.57; k 1: r = 0.53; P < 0.0001), as did D av to a lesser extent (r = 0.18, P = 0.028). Figs. 4 and 5 plot the relationship between FA and EGA and between k 1 and EGA. As illustrated in Fig. 6, gyration score correlated positively with EGA (r = 0.83, P < 0.0001). The minor eigenvalues k 2 and k 3 were not significantly correlated with EGA (k 2: r = 0.09; k 3: r = 0.05; P > 0.2). Cortical gyration also correlated inversely with cortical FA and k 1 (FA: r = 0.44; k 1: r = 0.41; P < 0.0001), but not D av, k 2, or k 3 (D av: r = 0.14, P = 0.09; k 2: r = 0.08, P = 0.31; or k 3: r = 0.03, P = 0.74). However, none of the 5 DTI parameters (D av, FA, k 1, k 2, k 3) showed any statistically significant residual correlation with the gyration ratio ( P > 0.05) after adjustment for EGA, white matter injury score, and repeated measures across serial scans. Furthermore, there was no significant relationship between white matter injury score and DTI parameters or gyral score, though there may
Results A total of 37 subjects, with a mean EGA at the time of birth of 28.1 weeks (range: 24.0 to 33.9 weeks), underwent MR imaging with DTI. Seven infants were scanned twice at different ages, for a total of 44 scans; mean EGA at the time of scan was 32.8 weeks
Fig. 3. Diffusion tensor imaging region of interest analysis. Regions of interest were drawn on T2-weighted (b = 0) images from the DTI sequence then overlaid on the corresponding DTI parametric images. Example from the superior occipital gyrus of a 32-week EGA premature infant, displayed on the FA image.
A.R. deIpolyi et al. / NeuroImage 27 (2005) 579 – 586
Table 1 DTI parameters and gyral ratios, presented as least squares means (standard errors) Region
FA D av (103 mm2/s) k 1 (103 mm2/s) k 2 (103 mm2/s) k 3 (103 mm2/s) Gyral ratio
Superior frontal gyrus
Superior occipital gyrus
0.155 1.148 1.330 1.129 0.982 0.860
0.148 1.149 1.319 1.142 0.986 1.070
0.224 1.173 1.465 1.127 0.993 0.920
0.205 1.175 1.433 1.121 1.001 1.030
(0.048) (0.096) (0.115) (0.103) (0.114) (0.027)
(0.048) (0.096) (0.115) (0.103) (0.114) (0.037)
have been too few infants with more than minimal injury (10 neonates) to test such a relationship.
Discussion Our study investigated whether DTI and quantitative scores of gyration can reliably assess human cortical maturation in premature infants. We used a technique to measure cortical folding quantitatively. By carefully controlling the orientation of images with 3-D reformatting and precisely measuring the height and width of gyri, we eliminated the possible sources of bias and confound that are inherent in two-dimensional qualitative scoring approaches. The spatial resolution of the reformatted 3-D images in the throughplane dimension is less than in the two in-plane dimensions. However, the through-plane resolution of 1.5 mm is still sufficient for evaluation of the four cortical regions chosen for scoring. As expected, calculated gyration scores increased with EGA in all regions of interest, a highly significant relationship. Similarly, we found that diffusion anisotropy decreases in the cortical gray matter during the period from 25 to 40 weeks EGA, confirming a previous study (McKinstry et al., 2002). The observed age-related reduction in cortical FA cannot be attributed to partial volume effects with the overlying cerebrospinal fluid within sulci or with the subjacent white matter since such partial volume effects would also increase D av and the three eigenvalues. Moreover, since cortical thickness increases dramatically during late preterm development (Huppi et al., 1998), partial volume effects would be expected to decrease with increasing EGA. Furthermore, one would not expect partial volume effects to be regional but would expect them to be uniform throughout the brain. One major finding of our study was that comparisons of DTI parameters in 4 functionally distinct regions of the cerebral cortex revealed evidence of regional heterogeneity in cortical development. The perirolandic cortex surrounding the central sulcus had consistently lower FA than the occipital cortex, and the superior frontal gyrus had consistently higher anisotropy values than the occipital cortex. One interpretation of these data is that the
(0.049) (0.097) (0.116) (0.104) (0.116) (0.060)
(0.049) (0.097) (0.117) (0.105) (0.116) (0.082)
perirolandic region may develop sooner, and the prefrontal cortex later, than the occipital cortex. Converging evidence for a regional pattern of cortical development emerged in a prior study using qualitative gyral scoring that showed an identical sequence of cortical development (van der Knaap et al., 1996). Furthermore, subcortical white matter myelination occurs earlier in the perirolandic region and later in the prefrontal region than in the occipital lobe during the first 2 years of postnatal life (Barkovich et al., 1988). This regional pattern of maturation also correlates with changing patterns of glucose uptake and regional blood flow (Chugani et al., 1987; Tokumaru et al., 1999), and with behavioral changes during neonatal and infant development. Sensorimotor processes mediated by the perirolandic cortex begin to develop earlier than visuospatial abilities mediated primarily by posterior cortical structures; logic and problem-solving abilities dependent on prefrontal lobe function are last to develop (Houde and TzourioMazoyer, 2003; Houde et al., 2000). The results of our DTI study are consistent with the idea that distinct sensorimotor, posterior, and anterior stages of pediatric behavioral maturation may be predetermined by the temporal pattern of regional cortical development prior to term-equivalent age. An alternative interpretation of the regional heterogeneity in cortical anisotropy during preterm development is that the different regions initially started out with different FA values prior to 25 weeks EGA, the youngest age examined in this study, and that the regional differences in cortical
Table 2 DTI parameters that differ significantly between regions of interest (P < 0.05) Postcentral gyrus Precentral gyrus Postcentral gyrus Superior frontal gyrus
Superior frontal gyrus
Superior occipital gyrus
FA, D av, k 1 FA, D av, k 1
FA, D av, k 1 FA, D av, k 1 FA, k 1
Fig. 4. Relationship between FA and EGA. Each data point represents the average of the three measurements taken bilaterally in each region of interest for each infant. For all four regions of interest, there is a significant negative correlation between EGA and FA. The perirolandic region (preand postcentral gyri) has consistently lower anisotropy levels than the frontal and occipital regions.
A.R. deIpolyi et al. / NeuroImage 27 (2005) 579 – 586
Fig. 5. Relationship between k 1 and EGA. Each data point represents the average of the three measurements taken bilaterally in each region of interest for each infant. For all four regions of interest, there is a significant negative correlation between k 1 and EGA. The perirolandic region (preand postcentral gyri) has consistently lower k 1 levels than the frontal and occipital regions. Here, the difference between the occipital and frontal regions is clearer, showing that the superior frontal gyrus has consistently higher k 1 values.
FA measured herein simply reflect the persistence of these initial differences. These two alternative explanations for the regional heterogeneity of developing cortex revealed by DTI are not mutually exclusive, however. Regional developmental heterogeneity could not be easily determined from the quantitative gyral ratios in this study because gyri in different regions of the brain mature to different final heights and widths and thus could not be directly compared to each other using an absolute scale. In contrast to gyration scores, we were able to compare quantitative anisotropy values in different cortical regions because FA decreases to the same noise levels throughout the cerebral cortex during preterm development. Another major finding of our study was that analysis of DTI eigenvalues can elucidate the mechanisms of microstructural development in the cerebral cortex. Specifically, we found that, while k 1 was inversely related to EGA, k 2 and k 3 were not correlated with EGA, indicating that the reduction in the major eigenvalue k 1 is the predominant component of the maturational loss of cortical anisotropy. Interestingly, this is exactly the reverse of the pattern of eigenvalue changes that have been previously observed during white matter maturation, wherein the greatest decreases are in k 2 and k 3 with comparatively little change in k 1 (Mukherjee et al., 2002). In the developing white matter, premyelination and myelination preferentially hinder water mobility orthogonal to the long axis of axons (decreasing k 2 and k 3), with little reduction in water diffusion parallel to axons (no change in k 1). In the cortical gray matter, the major eigenvalue k 1 corresponds to the magnitude of diffusion radial to the pial surface, as indicated in Fig. 1D and described previously (Maas et al., 2004; McKinstry et al., 2002; Mori et al., 2001). Thus, the relatively large decline in k 1 with EGA in the preterm cortex means that the maturational loss of cortical anisotropy is chiefly due to a preferential reduction in the radial component of water diffusivity. Biologically, this may correspond to the arborization of basal dendrites from immature neurons, the innervation of the cortical
plate by thalamocortical and corticocortical axons, and early synaptogenesis (Marin-Padilla, 1988; Mrzljak et al., 1988). The causes of cortical gyration are not well understood, though intrinsic and extrinsic mechanisms have been postulated. Intrinsic mechanisms include cellular and molecular processes such as differential growth of cortical layers, whereas extrinsic mechanisms induce cortical folding through structural and mechanical forces such as axonal tension (Van Essen, 1997). If intrinsic forces were driving cortical gyration, one might expect that anisotropy, a microstructural characteristic intrinsic to the cortex, would correlate with gyration scores independent of age. Our study failed to find such an association. Though anisotropy and gyration are both dependent on EGA, we found that anisotropy and gyration were not correlated with each other independent of their common association with EGA. One possible reason for this result is that our methods were not sensitive enough to detect a relationship between anisotropy and gyration. Alternatively, it is possible that such a relationship does not exist because the microstructural changes of cortical development assessed by DTI do not play a causative role in cortical folding. Previous qualitative gyral scoring methods have distinguished infants with substantial cortical injury from healthy infants (Slagle et al., 1989), but no previous study has been able to detect mild injury. DTI may be more sensitive than gyral assessment because microstructural processes may be disturbed by pathology to a greater extent than macrostructural development. Our study had too few subjects with white matter injury to test this possibility, and more work with larger numbers of subjects will be necessary to determine whether DTI is more sensitive to cortical injury and whether it can be applied to assessing repair and recovery after injury. Furthermore, long-term follow-up studies should determine which measures correlate best with long-term functional outcome to determine the prognostic significance of these assays of cortical maturation and whether combining microstructural and macrostructural assays may achieve better results than individual measures alone.
Fig. 6. Relationship between gyration score and EGA. Each data point represents the average of each measurement taken bilaterally for each gyrus of interest for each infant. For all four regions of interest, there is a significant positive correlation between gyral score and EGA. Because each gyrus begins at and matures to different heights and widths, it is not possible to compare the regions directly.
A.R. deIpolyi et al. / NeuroImage 27 (2005) 579 – 586
In conclusion, this quantitative investigation of macrostructural and microstructural features of preterm human cortical development confirms that diffusion anisotropy is strongly negatively correlated with EGA and that the extent of cortical gyration is strongly positively correlated with EGA, though anisotropy was not found to be significantly correlated with gyration beyond their common dependence on EGA. Analysis of the DTI eigenvalues showed that a decline in the radial component of diffusivity was largely responsible for the observed loss of cortical anisotropy during preterm development. Furthermore, DTI parameters were found to differ in functionally distinct cortical regions, providing evidence for regional heterogeneity in the microstructural characteristics of human cortical development. These results suggest that DTI is a powerful tool for studying cortical development by revealing microstructural processes to which conventional macrostructural imaging techniques are insensitive. Further research is ongoing to discover whether DTI may be superior to conventional MR imaging for detecting and characterizing abnormal cortical maturation and cortical injury in premature newborns.
Acknowledgments The authors thank the neonatal nurses of the Pediatric Clinical Research Center (PCRC, RR01271) for their help screening newborns and for their help in transporting newborns to the MR scanner. The authors also acknowledge the following grant support: NIH R01 NS40117, R21 NS40382, RR01271, and P50 NS35902.
References Ajayi-Obe, M., Saeed, N., Cowan, F.M., Rutherford, M.A., Edwards, A.D., 2000. Reduced development of cerebral cortex in extremely preterm infants. Lancet 356, 1162 – 1163. Barkovich, A.J., Kjos, B.O., Jackson Jr., D.E., Norman, D., 1988. Normal maturation of the neonatal and infant brain: MR imaging at 1.5 T. Radiology 166, 173 – 180. Battin, M.R., Maalouf, E.F., Counsell, S.J., Herlihy, A.H., Rutherford, M.A., Azzopardi, D., Edwards, A.D., 1998. Magnetic resonance imaging of the brain in very preterm infants: visualization of the germinal matrix, early myelination, and cortical folding. Pediatrics 101, 957 – 962. Chugani, H.T., Phelps, M.E., Mazziotta, J.C., 1987. Positron emission tomography study of human brain functional development. Ann. Neurol. 22, 487 – 497. Daniel, W.W., 1999. Biostatistics: A Foundation for Analysis in the Health Sciences, 7th edR Wiley, New York. De Vries, L.S., Groenendaal, F., van Haastert, I.C., Eken, P., Rademaker, K.J., Meiners, L.C., 1999. Asymmetrical myelination of the posterior limb of the internal capsule in infants with periventricular haemorrhagic infarction: an early predictor of hemiplegia. Neuropediatrics 30, 314 – 319. De Vries, L.S., Van Haastert, I.L., Rademaker, K.J., Koopman, C., Groenendaal, F., 2004. Ultrasound abnormalities preceding cerebral palsy in high-risk preterm infants. J. Pediatr. 144, 815 – 820. Dumoulin, C.L., Rohling, K.W., Piel, J.E., Rossi, C.J., Giaquinto, R.O., Watkins, R.D., Vigneron, D.B., Barkovich, A.J., Newton, N., 2002. An MRI compatible neonate incubator. Concepts Magn. Reson. (Magn. Reson. Eng.) 15, 117 – 128. Hack, M., Taylor, H.G., 2000. Perinatal brain injury in preterm infants and later neurobehavioral function. JAMA 284, 1973 – 1974.
Hack, M., Flannery, D.J., Schluchter, M., Cartar, L., Borawski, E., Klein, N., 2002. Outcomes in young adulthood for very-low-birth-weight infants. N. Engl. J. Med. 346, 149 – 157. Houde, O., Tzourio-Mazoyer, N., 2003. Neural foundations of logical and mathematical cognition. Nat. Rev., Neurosci. 4, 507 – 514. Houde, O., Zago, L., Mellet, E., Moutier, S., Pineau, A., Mazoyer, B., Tzourio-Mazoyer, N., 2000. Shifting from the perceptual brain to the logical brain: the neural impact of cognitive inhibition training. J. Cogn. Neurosci. 12, 721 – 728. Huppi, P.S., Warfield, S., Kikinis, R., Barnes, P.D., Zientara, G.P., Jolesz, F.A., Tsuji, M.K., Volpe, J.J., 1998. Quantitative magnetic resonance imaging of brain development in premature and mature newborns. Ann. Neurol. 43, 224 – 235. Inder, T.E., Huppi, P.S., Warfield, S., Kikinis, R., Zientara, G.P., Barnes, P.D., Jolesz, F., Volpe, J.J., 1999. Periventricular white matter injury in the premature infant is followed by reduced cerebral cortical gray matter volume at term. Ann. Neurol. 46, 755 – 760. Laird, N.M., Ware, J.H., 1982. Random-effects models for longitudinal data. Biometrics 38, 963 – 974. Maas, L.C., Mukherjee, P., Carballido-Gamio, J., Veeraraghavan, S., Miller, S.P., Partridge, S.C., Henry, R.G., Barkovich, A.J., Vigneron, D.B., 2004. Early laminar organization of the human cerebrum demonstrated with diffusion tensor imaging in extremely premature infants. NeuroImage 22, 1134 – 1140. Marin-Padilla, M., 1988. Cerebral cortex. In: Peters, A., Jones, E.G. (Eds.), Cerebral Cortex. Plenum Press, New York, pp. 1 – 30. Martin, J.A., Hamilton, B.E., Ventura, S.J., Menacker, F., Park, M.M., 2002. Births: final data for 2000. Natl. Vital. Stat. Rep. 50, 1 – 101. McKinstry, R.C., Mathur, A., Miller, J.H., Ozcan, A., Snyder, A.Z., Schefft, G.L., Almli, C.R., Shiran, S.I., Conturo, T.E., Neil, J.J., 2002. Radial organization of developing preterm human cerebral cortex revealed by non-invasive water diffusion anisotropy MRI. Cereb. Cortex 12, 1237 – 1243. Miller, S.P., Vigneron, D.B., Henry, R.G., Bohland, M.A., Ceppi-Cozzio, C., Hoffman, C., Newton, N., Partridge, J.C., Ferriero, D.M., Barkovich, A.J., 2002. Serial quantitative diffusion tensor MRI of the premature brain: development in newborns with and without injury. J. Magn. Reson. Imaging 16, 621 – 632. Mori, S., Itoh, R., Zhang, J., Kaufmann, W.E., van Zijl, P.C., Solaiyappan, M., Yarowsky, P., 2001. Diffusion tensor imaging of the developing mouse brain. Magn. Reson. Med. 46, 18 – 23. Mrzljak, L., Uylings, H.B., Kostovic, I., Van Eden, C.G., 1988. Prenatal development of neurons in the human prefrontal cortex: I. A qualitative Golgi study. J. Comp. Neurol. 271, 355 – 386. Mukherjee, P., Miller, J.H., Shimony, J.S., Philip, J.V., Nehra, D., Snyder, A.Z., Conturo, T.E., Neil, J.J., McKinstry, R.C., 2002. Diffusion-tensor MR imaging of gray and white matter development during normal human brain maturation. Am. J. Neuroradiol. 23, 1445 – 1456. Mukherjee, P., Gill, K., Veeraraghavan, S., Henry, R.G., Miller, S.P., Vigneron, D.B., Barkovich, A.J., 2003a. Region-specific maturation of the cerebral cortex in premature newborns revealed by high-resolution diffusion tensor imaging. Proceedings of the International Society for Magnetic Resonance in Medicine, Toronto. Mukherjee, P., Gill, K., Veeraraghavan, S., Henry, R.G., Miller, S.P., Vigneron, D.B., Barkovich, A.J., 2003b. Anisotropy loss during maturation of the cerebral cortex in premature newborns is due to decreasing water diffusion perpendicular (but not parallel) to the cortical surface. Proceedings of the International Society for Magnetic Resonance in Medicine, Toronto. Partridge, S.C., Mukherjee, P., Henry, R.G., Miller, S.P., Berman, J.I., Jin, H., Lu, Y., Glenn, O.A., Ferriero, D.M., Barkovich, A.J., Vigneron, D.B., 2004. Diffusion tensor imaging: serial quantitation of white matter tract maturity in premature newborns. NeuroImage 22, 1302 – 1314. Peterson, B.S., Anderson, A.W., Ehrenkranz, R., Staib, L.H., Tageldin, M., Colson, E., Gore, J.C., Duncan, C.C., Makuch, R., Ment, L.R., 2003.
A.R. deIpolyi et al. / NeuroImage 27 (2005) 579 – 586
Regional brain volumes and their later neurodevelopmental correlates in term and preterm infants. Pediatrics 111, 939 – 948. Slagle, T.A., Oliphant, M., Gross, S.J., 1989. Cingulate sulcus development in preterm infants. Pediatr. Res. 26, 598 – 602. Tokumaru, A.M., Barkovich, A.J., O’Uchi, T., Matsuo, T., Kusano, S., 1999. The evolution of cerebral blood flow in the developing brain: evaluation with iodine-123 iodoamphetamine SPECT and correlation with MR imaging. Am. J. Neuroradiol. 20, 845 – 852.
van der Knaap, M.S., van Wezel-Meijler, G., Barth, P.G., Barkhof, F., Ader, H.J., Valk, J., 1996. Normal gyration and sulcation in preterm and term neonates: appearance on MR images. Radiology 200, 389 – 396. Van Essen, D.C., 1997. A tension-based theory of morphogenesis and compact wiring in the central nervous system. Nature 385, 313 – 318. Wood, N.S., Marlow, N., Costeloe, K., Gibson, A.T., Wilkinson, A.R., 2000. Neurologic and developmental disability after extremely preterm birth. EPICure Study Group. N. Engl. J. Med. 343, 378 – 384.