NeuroImage 112 (2015) 30–42

Contents lists available at ScienceDirect

NeuroImage journal homepage: www.elsevier.com/locate/ynimg

Cerebral maturation in the early preterm period—A magnetization transfer and diffusion tensor imaging study using voxel-based analysis Revital Nossin-Manor a,b,⁎, Dallas Card a, Charles Raybaud a,d, Margot J. Taylor a,b,d, John G. Sled c,e a

Diagnostic Imaging, Hospital for Sick Children, Toronto, ON M5G 1X8, Canada Neurosciences and Mental Health, Research Institute, Hospital for Sick Children, Toronto, ON M5G 1X8, Canada Physiology and Experimental Medicine, Research Institute, Hospital for Sick Children, Toronto, ON M5G 1X8, Canada d Medical Imaging, University of Toronto, Toronto, ON M5S 3E2, Canada e Medical Biophysics, University of Toronto, Toronto, ON M5G 2M9, Canada b c

a r t i c l e

i n f o

Article history: Accepted 22 February 2015 Available online 27 February 2015 Keywords: Quantitative MRI MTR DTI Preterm neonates Brain maturation Voxel based analysis

a b s t r a c t The magnetization transfer ratio (MTR) and diffusion tensor imaging (DTI) correlates of early brain development were examined in cohort of 18 very preterm neonates (27–31 gestational weeks) presenting with normal radiological findings scanned within 2 weeks after birth (28–32 gestational weeks). A combination of non-linear image registration, tissue segmentation, and voxel-wise regression was used to map the age dependent changes in MTR and DTI-derived parameters in 3D across the brain based on the cross-sectional in vivo preterm data. The regression coefficient maps obtained differed between brain regions and between the different quantitative MRI indices. Significant linear increases as well as decreases in MTR and DTI-derived parameters were observed throughout the preterm brain. In particular, the lamination pattern in the cerebral wall was evident on parametric and regression coefficient maps. The frontal white matter area (subplate and intermediate zone) demonstrated a linear decrease in MTR. While the intermediate zone showed an unexpected decrease in fractional anisotropy (FA) with age, with this decrease (and the increase in mean diffusivity (MD)) driven primarily by an increase in radial diffusivity (RD) values, the subplate showed no change in FA (and an increase in MD). The latter was the result of a concomitant similar increase in axial diffusivity (AD) and RD values. Interpreting the in vivo results in terms of available histological data, we present a biophysical model that describes the relation between various microstructural changes measured by complementary quantitative methods available on clinical scanners and a range of maturational processes in brain tissue. © 2015 Elsevier Inc. All rights reserved.

Introduction During the second and third trimesters of pregnancy, a sequence of maturation events establish the foundations for normal brain structure and function including neuronal proliferation and migration, the formation of axonal pathways, programmed cell death and, toward the end of gestation, myelination (Volpe, 2008). These events proceed within Abbreviations: ACR, Anterior corona radiata; AD, Axial diffusivity; ALIC, Anterior limb of the internal capsule; CP, Cortical plate; DTI, Diffusion tensor imaging; EC, External capsule; ECM, Extracellular matrix; FA, Fractional anisotropy; FDR, False discovery rate; gCC, Genu of the corpus callosum; GM, Gray matter; GP, Globus pallidus; ILF, Inferior longitudinal fasciculus; IZ, Intermediate zone; MD, Mean diffusivity; MTI, Magnetization transfer imaging; MTR, Magnetization transfer ratio; MZ, Marginal zone; OR, Optic radiation; PLIC, Posterior limb of the internal capsule; RD, Radial diffusivity; sCC, Splenium of the corpus callosum; SFO, Superior fronto-occipital fasciculus; SLF, Superior longitudinal fasciculus; SP, Subplate; SS, Sagittal stratum; SVZ, Subventricular zone; VLN, Ventolateral thalamic nucleus; WM, White matter. ⁎ Corresponding author at: Diagnostic Imaging, Neuroscience and Mental Health program, The Hospital for Sick Children, 555 University Ave, Toronto, ON M5G 1X8, Canada. Fax: +1 416 813 7362. E-mail address: [email protected] (R. Nossin-Manor).

http://dx.doi.org/10.1016/j.neuroimage.2015.02.051 1053-8119/© 2015 Elsevier Inc. All rights reserved.

laminarly arranged cellular zones, not found in the adult brain, which set the stage for development through to adulthood (Haynes et al., 2005; Kostovic et al., 2002; Rados et al., 2006). This transient laminar organization develops during the mid-fetal period (17–24 gestational weeks) and attains its developmental peak during the early preterm period (26–34 gestational weeks) between 29 and 32 weeks gestational weeks. It consists of (from pia to ventricle): (a) a marginal zone (MZ), (b) the cortical plate (CP) with high cell-packing density, (c) the subplate (SP) zone, the most prominent zone rich in hydrophilic extracellular matrix (ECM) and subplate neurons, and the location of accumulation of ‘waiting’ thalamic afferent axons, (d) the intermediate zone (IZ; future white matter (WM)), containing migratory neurons, immature glial cells, large bundles of growing axons and their periventricular crossroads, (e) the subventricular zone (SVZ), a (callosal) fiber-rich zone and (f) the ventricular zone. The developing connections of thalamocortical axons, with their synaptic engagement in the CP after the ‘waiting’ period in the transient SP zone, followed by the similarly developing connections of the long association axons are the main connectivity events in the brain in the late fetal (22–25 gestational weeks) and early preterm (26–34 gestational weeks)

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

periods. These events are accompanied by a gradual decrease in the SP thickness. Established earlier, the radial glial system is organized in fascicles of fibers traversing the cerebral wall radially and facilitating the proliferation and migration of cortical neurons and glial cells, such as astrocytes and oligodendrocytes (Gadisseux et al., 1989; Rakic, 2003). All layers but the MZ are visible on in vitro and in vivo MRI (Brisse et al., 1997; Girard et al., 1995; Huang et al., 2006; Maas et al., 2004; Rados et al., 2006). The decrease in thickness and increase in signal intensity (on structural T1-weighted images) of the SP towards the end of the early preterm period is a prominent feature of cerebral MRI in preterm infants and is closely linked to the normal development of axonal connectivity in the brain (Kostovic and Judas, 2002). In the SVZ and IZ, the decrease in T1-weighted signal during the preterm period has been shown to correspond with areas of migrating radial glial cells (Battin et al., 1998; Childs et al., 1998) and developing crossing axonal pathways (Kostovic and Jovanov-Milosevic, 2006). Using tractography, recent diffusion tensor imaging (DTI) postmortem studies of the human fetal brain have successfully identified immature axonal pathways as early as the beginning of the mid-fetal period (Huang et al., 2006, 2009; Takahashi et al., 2012; Vasung et al., 2010). Furthermore, using these methods, predominant radial coherent pathways crossing the cerebral wall, originating possibly from radial glial fascicles, columns of migrating neurons and/or afferent immature axons, have been described in the mid/late fetal period (17–25 gestational weeks) (Takahashi et al., 2012; Xu et al., 2014). Nevertheless, application of these methods, in particular high angular resolution diffusion imaging (HARDI), to in vivo fetal and preterm imaging is limited by subject motion, small brain size and the long duration of scan (above 2 h) required for achieving high-resolution and high signal-to-noise-ratio anatomical images of the developing cerebral architecture. Furthermore, the information gained is related to tissue organization only. Imaging early preterm neonates at birth gives a unique opportunity to investigate cerebral development and monitor dynamic maturational events in vivo. DTI is a powerful method for investigating maturation of WM tracts in the developing brain (Berman et al., 2005; Dubois et al., 2006, 2008; Hermoye et al., 2006; Huppi et al., 1998; Miller et al., 2002; Neil et al., 1998; Schneider et al., 2004). Using tractography or region-of-interest approaches, these studies demonstrated, based on a linear model, significant age-related changes in diffusion indices, where fractional anisotropy (FA) increased and diffusivity values decreased, in most WM structures measured during the preterm and term periods as well as in the first few months of life. Nevertheless, a recent in utero DTI tractography study in normal fetuses between 23 and 38 gestational weeks depicted structure-specific non-linear curves of normalized FA and mean, axial and radial diffusivities (MD, AD and RD, respectively) as a function of age in different WM tracts (Zanin et al., 2011). Polynomial curve fitting with respect to age demonstrated three different phases, related by the authors to axonal organization, myelination gliosis, and myelination. Magnetization transfer imaging (MTI) is another method used to assess brain development. It is sensitive to the concentration of semisolids in the tissue and therefore can be used for investigating white as well as gray matter maturation (Engelbrecht et al., 1998). A handful of studies have demonstrated positive linear evolution of magnetization transfer ratio (MTR) values with age in different selected WM and gray matter (GM) regions during early development (Engelbrecht et al., 1998; Nossin-Manor et al., 2012; van Buchem et al., 2001; Xydis et al., 2006a,b). In a previous work we demonstrated the use of MTI and DTI to create group-wise average parametric maps characterizing tissue microstructure of the neonatal brain in very preterm infants (24–32 gestational weeks) scanned at preterm and term equivalent age (Nossin-Manor et al., 2013). Using a region-of-interest and voxel-based approach, we showed that these parametric maps present distinct contrasts whose interrelations varied across brain regions and between the preterm and term period, corresponding to various aspects of brain maturation

31

such as tissue organization and myelination. MTR values showed a marked change in the pattern of regional variation at term equivalent age compared to the preterm period, such that the ordinal ranking of regions by signal contrast changed, corroborating myelination, for example, in the posterior limb of the internal capsule (PLIC). This was unlike DTI parameters where the regional ranking was similar at the two time points. Interpreting the data in terms of myelination and structural organization, we reported on the concordance with previous histological findings and demonstrated the value of multi-contrast MRI for tracking various aspects of brain maturation over the neonatal period. In the present study we use MTI and DTI along with structural imaging, group-wise image registration, semiautomatic segmentation and voxel-based linear regression analysis for obtaining volumetric regression coefficient maps that describe population based developmental trajectories in the very preterm brain between 28 and 32 gestational weeks. Looking at the evolution of MTR and DTI indices with age, we follow spatiotemporal variations in cerebral maturation using in vivo data. Our goal was to determine distinct region-specific cerebral developmental trajectories over the whole brain and corroborate them with known cellular events occurring during the preterm period such as cell migration, the gradual regression of radial glial fibers, the development of axonal circuitry and the disappearance of the radial coherence, the gradual disappearance of the subplate and pre-/early myelination events. Materials and methods Subjects The study included 18 preterm neonates, nine male, born between 27 and 31 gestational weeks (mean ± SD, 29.4 ± 1.2 weeks) and scanned within 2 weeks after birth (age range, 28–32 weeks; mean ± SD, 30.8 ± 1.4 weeks) without sedation. Neonates presented with normal findings on conventional MR images (T1-, T2-, T2⁎- and diffusionweighted) (n = 17, including one with non-specific minor globi pallidi T1 hyperintensity, probably due to total parenteral nutrition (TPN) administration with Manganese), and grade II intraventricular hemorrhage with no extension to the brain parenchyma or evidence of ventricular dilatation (n = 1). None had evidence of genetic, metabolic or viral infection disorders. MRIs were acquired between March 2008 and April 2010 as a part of a broader cohort of a prospective longitudinal study of 105 preterm neonates. Exclusion criteria for the present analyses included white matter and deep gray matter injuries, grade III intraventricular and grade IV intraperenchymal hemorrhages, and ventriculomegaly. Furthermore, only neonates who had a complete and successful (no susceptibility artifacts or severe motion) multicontrast MRI session were included. All data sets were visually inspected; scans rejected for severe motion had rotation N 0.12 rad and displacements on the order of 3–5 mm (2D scans) but also included scans showing strong artifacts or signal loss (2D and 3D scans). DTI data was rated by two of the authors (RNM and DC) for severe, moderate and mild (rotational/translational) motion. Adequate data sets were chosen upon agreement between the raters. Out of 22 neonatal data sets passing these criteria, four data sets were excluded as MTR volumes were acquired after a scanner upgrade when the pulse type and angle of the MT pulse changed. Radiological assessments were completed by a neuroradiologist with N10 years of experience in neonatal imaging. The study was approved by the hospital's research ethics board, and informed, written consent was given by the infants' parents. MR acquisition All scans were completed on a 1.5 T GE Signa Excite HD scanner (GE, Milwaukee, WI) using an MR-compatible incubator and neonatal head coil (AIR, Inc., Cleveland, OH) according to a published imaging protocol (Nossin-Manor et al., 2013). To summarize the sequences analyzed here

32

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

Table 1 MRI protocol of the present study. MRI scan

Sequencea

TR (ms)

TE (ms)

Flip angle

BW (kHz)

FOV (mm)

Resolution (mm)

Time (min:sec)

T1w T2w DTI MTI

3D-SPGR 2D-FRFSE SE-EPI (twice-refocused) 3D-SPGRb

23 4000 15000 27

4 145 85 4

19° 90° 90° 10°

15.63 25 250 15.63

128 × 128 128 × 128 205 × 205 128 × 128

1×1×1 1×1×1 1.6 × 1.6 × 1.6 1 × 1 × 1.5

5:39 4:16 5:06 9:00

a

Images were acquired in an axial orientation. SPGR = spoiled gradient recalled; FRFSE = fast recovery fast spin-echo; EPI = echo planner imaging. Total scan time was 24 min. With and then without an off-resonance preparatory pulse: excitation pulse angle = 726°, duration = 8.192 ms, off-set frequency = 1.6 kHz; composed by multiplying a Gaussian envelope with standard deviation of 1.5 ms with a Hanning window. For the given magnetization transfer sequence, with TR of 27 ms, this is equivalent to continuous wave irradiation with a nutation rate of 107 Hz (Nossin-Manor et al., 2012). b

(Table 1): high resolution axial T1- and T2-weighted (T1w and T2w) volumes were acquired using 3D SPGR (TR/TE/FA = 23 ms/4 ms/19°, BW = 15.63 kHz, FOV = 12.8 cm, matrix = 128 × 128, 110 slices of 1 mm) and multi-slice 2D FRFSE (TR/TE/ETL = 4000 ms/145 ms/19, BW = 25 kHz, FOV = 12.8 cm, matrix = 128 × 128, 90 slices of 1 mm). MT images were obtained using a proton-density (PD) weighted 3D SPGR sequence by acquiring the sequence twice—once with an offresonance MT saturation pulse (TR/TE/FA = 27 ms/4 ms/10°, BW = 15.63 kHz, FOV = 12.8 cm, matrix = 128 × 128, 74 slices of 1.5 mm, excitation pulse angle = 726°, Duration = 8.192 ms, offset frequency = 1.6 kHz) and once without; MTR maps were obtained by calculating the percent difference of these two images. Twice refocused spin echo planar DTI was acquired with three non-diffusion and 15 noncollinear diffusion weighted volumes and b = 700 s/mm2 using: 2D axial oblique slices, FOV = 205 mm, with 1.6 mm cubic voxels, TR/TE/ FA = 15 s/85 ms/90°. Eddy currents and distortion corrections, motion correction and outlier rejection were then performed on DTI data, and volumetric FA, MD, AD and RD maps were obtained using the DROP-R algorithm (Morris et al., 2011). Image processing Registration The brain extraction tool (BET) was used to segment T2w volumes into brain and non-brain (Smith, 2002). Images were reviewed on a case-by-case basis and an inter-slice motion correction algorithm based on the MNI AutoReg software package (Collins et al., 1994) was applied where needed. To allow for accurate intra-subject registration and in light of the difficulties in producing accurate brain masks using BET with T1w volumes in very preterm neonates the following steps were completed: 1) Structural images (T1w and T2w) were corrected for intensity non-uniformity using the MNI N3 algorithm (Sled et al., 1998) with a mask for T2w images and without a mask for T1w images. 2) Brain masks were refined using BET and T2w images after MNI N3 correction. 3) All volumes within each scan were aligned using rigidbody registration and the masks produced for the T2w volumes (Collins et al., 1994). To align all scans the following steps were completed (Spring et al., 2007): 1) Rigid body registration of T2w volumes was accomplished using a target model, a 30 week neonate (Collins et al., 1994; Kovacevic et al., 2005). 2) All volumes were then coregistered using all possible pair-wise affine registrations to create a linear average of the entire data set. 3) All images were subsequently nonlinearly aligned in an iterative manner towards the 12-parameter average. The resulting non-linear transformations were applied to all individual scans to create average structural and quantitative volumes of the whole cohort (Collins et al., 1994; Kovacevic et al., 2005). Segmentation T1w, T2w and PD (MT without an off-resonance saturation pulse) images were used to create group WM and GM masks using the following steps (Card et al., 2011): 1) Training data were manually selected for each subject on T2w volume for GM, WM, and cerebral spinal fluid (CSF) and used as inputs to a non-parametric neural network classifier to create initial individual classifications (Zijdenbos et al., 1998). 2) The

trimmed minimum covariance determinant method (Tohka et al., 2004) with modified Markov smoothing was used to identify boundaries between WM/GM and GM/CSF as partial volumes. 3) A maximum a posteriori classifier was used to convert partial volumes classifications into GM/WM/CSF classifications (Tohka et al., 2004). 4) Individual segmentations were co-registered using the non-linear transformations described above; the resulting average classification was refined manually to fit the corresponding average anatomy, and then incorporated as a spatial prior to re-classify all subjects using the above procedure. 5) A threshold of N 50% representation across subjects was applied to the segmented averaged volumes to construct the corresponding group average masks for gray and white matter. Statistical analysis Voxel-based linear regressions of 1) MTR values against FA, MD and RD, 2) AD against RD and 3) each of the above parametric values against age at scan across subjects were calculated to produce volumetric maps of the regression coefficient. All volumes were blurred using a 3 mm Gaussian kernel prior to regression. At each voxel these analyses included only the subset of subjects where the tissue classification (GM or WM) matched the classification of the group average masks. According to the segmentation algorithm described above, at least 10 out of 18 subjects were included in the analysis for each voxel. That said, 18 subjects were typically included in the analysis for voxels in deep GM and WM structures (15–18 subjects in cortical GM voxels and 14–17 subjects for WM voxels at the borders between GM and WM structures). False discovery rate (FDR) of 20% was used to threshold the regression coefficient maps using R statistical software (www.r-project.org). Results To investigate the temporal variations in cerebral maturation and the behavior of the different contrast mechanisms over the early preterm period, we conducted a whole-brain voxel-based analysis where each parametric contrast was regressed against age at scan. Fig. 1 presents representative axial and sagittal slices of the resulting regression coefficient volumetric maps, standard error maps and tvalue (t-statistic) maps overlaid on the corresponding average T2w volume. A distinct regional pattern of regression coefficient values is observed for the different quantitative MRI-derived parameters regressed against age. These maps show different positive as well as negative trends throughout the brain in MTR values and DTI-derived parameters vs. age. Fiber tracts of the somatosensory and primary motor axonal pathways are seen vividly on MTR (sagittal view) and FA (axial and sagittal views) regression maps. FA was found to be the most robust parameter of all measured indices, showing the highest absolute t-values (Fig. 1l, Tables 2a and 2b) and lowest FDR values for the regression against age. Fig. 2 presents representative slices of FDR values ≤20% for the regression of FA values against age overlaid on the average T2w volume in the axial, sagittal and coronal planes throughout the preterm brain. Patches of significant age dependence (q ≤ 0.2) were found in various anatomical locations such as in the primary sensory and

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

33

Fig. 1. Regression coefficient maps—representative axial and sagittal slices of the volumetric regression coefficient maps for the regression of MRI-derived parameters against age (upper panel), standard error maps (middle panel) and t-value (t-statistic) maps (lower panel) overlaid on the corresponding study-specific average T2w volume. A distinct regional pattern of regression coefficient values is observed across the brain: In regression coefficient maps and t-value maps, warm colors represent an increase in regression coefficient values whereas cold colors represent a decrease in regression coefficient values; in the t-value maps, green and cyan indicate the increase/decrease in regression coefficient values measured for this cohort is not statistically significant (|t| ≤ 1.75; p N 0.05).

motor pathways, association fibers, commissural fibers and in various other regions, detailed in Fig. 2. Tables 2a and 2b present the mean values of the measured MRI indices along with regression coefficients values, standard errors and tvalues of representative voxels in subcortical GM structures and the SP, showing the highest absolute t-values in Fig. 1k, and in significant WM voxels, showing the lowest FDR values, chosen according to Fig. 2. Eighteen subjects were included for each voxel except the one representing the occipital WM (n = 14; see Statistical Analysis). Fig. 3 depicts voxel plots of MTR and DTI-derived parameters vs. age at scan for various GM and WM voxels representing structures detailed in Tables 2a and 2b. As demonstrated with Fig. 3 and Tables 2a and 2b, anatomical regions differed in terms of their mean quantitative parametric

values, trajectory (the relation between cross-sectional parametric data and age) and interrelations between measured MRI indices, suggesting structure-specific developmental changes occur at this stage. Table 3 summarizes the trends in MTR and DTI-derived parameters observed across the very preterm brain, dividing the structures presented in Tables 2a and 2b into nine major subgroups according to the change observed in MTR and FA values. Voxels in the frontal IZ, such as the crossroads between the genu of the corpus callosum (gCC) and the anterior corona radiata (ACR), and in the WM of the occipital lobe, corresponding to the primary visual area, (group 1 in Table 3) showed relatively low MTR (16.2 ± 0.4% and 18.0 ± 0.5%, respectively) and FA (0.134 ± 0.006 and 0.159± 0.009, respectively) values and a negative trend in these values when

34 Table 2a Mean valuesa of the measured quantitative MR parameters in representative voxels chosen according to Fig. 2 and the corresponding regression coefficientsb and t-values for the regression of these indices with age (n = 18 for all voxels but the Occipital WM (n = 14)). Location

MTR (%)

Caudate Putamen GP VLN Pulvinar Frontal SP Crossroads (IZ) Temporal (auditory) Occipital (visual) b c d

FA

Regression (t-value)

MD × 10−3 (mm2/s)

Regressionc (t-value)

AD × 10−3 (mm2/s)

Regressionc (t-value)

RD × 10−3 (mm2/s)

Regressionc (t-value)

0.63 ± 0.22 (2.88) 0.47 ± 0.18 (2.59) 0.23 ± 0.16 (1.50) 0.40 ± 0.13 (3.18) 0.32 ± 0.16 (2.00) −0.63 ± 0.40 (−1.58) −0.69 ± 0.22 (−3.18) 0.70 ± 0.28 (3.19) −0.67 ± 0.27 (−2.45)

0.151 ± 0.006 0.119 ± 0.005 0.216 ± 0.006 0.248 ± 0.006 0.209 ± 0.006 0.121 ± 0.007 0.134 ± 0.006 0.161 ± 0.007 0.159 ± 0.009

0.004 ± 0.004 (0.97) 0.002 ± 0.003 (0.60) 0.000 ± 0.004 (−0.07) 0.002 ± 0.004 (0.58) 0.002 ± 0.004 (0.52) −0.001 ± 0.005 (−0.28) −0.012 ± 0.003 (−3.77) 0.014 ± 0.004 (3.67) −0.013 ± 0.003 (−3.70)

1.48 ± 0.02 1.46 ± 0.02 1.30 ± 0.02 1.18 ± 0.01 1.30 ± 0.01 1.79 ± 0.04 1.83 ± 0.03 1.63 ± 0.03 1.59 ± 0.04

−0.008 ± 0.014 (−0.61) −0.021 ± 0.013 (−1.67) −0.016 ± 0.011 (−1.41) −0.014 ± 0.010 (−1.46) −0.008 ± 0.011 (−0.74) 0.029 ± 0.028 (1.05) 0.036 ± 0.018 (1.97) −0.053 ± 0.019 (−2.76) 0.044 ± 0.025 (1.73)

1.70 ± 0.02 1.63 ± 0.02 1.60 ± 0.02 1.51 ± 0.02 1.58 ± 0.02 2.01 ± 0.04 2.07 ± 0.03 1.88 ± 0.03 1.84 ± 0.05

−0.001 ± 0.018 (−0.07) −0.019 ± 0.016 (−1.19) −0.022 ± 0.012 (−1.88) −0.015 ± 0.014 (−1.05) −0.006 ± 0.017 (−0.36) 0.029 ± 0.028 (1.02) 0.018 ± 0.020 (0.90) −0.039 ± 0.020 (−1.93) 0.029 ± 0.029 (0.98)

1.36 ± 0.02 1.37 ± 0.02 1.15 ± 0.02 1.02 ± 0.01 1.16 ± 0.01 1.68 ± 0.04 1.71 ± 0.03 1.51 ± 0.03 1.46 ± 0.04

−0.012 ± 0.013 (−0.93) −0.022 ± 0.011 (−1.90) −0.013 ± 0.012 (−1.06) −0.014 ± 0.009 (−1.57) −0.009 ± 0.009 (−1.03) 0.029 ± 0.028 (1.01) 0.045 ± 0.018 (2.50) −0.060 ± 0.019 (−3.12) 0.048 ± 0.024 (2.16)

Values are given in mean ± standard error (SE). Values are given in (index units) per week as coefficient ± SE. The given values ×10−3. The difference between the MTR values of the frontal SP and IZ is statistically significant (p ≤ 0.05; post-hoc two-tailed Student's t-test).

Table 2b Mean valuesa of the measured quantitative MR parameters in representative voxels chosen according to Fig. 2 and the corresponding regression coefficientsb and t-values for the regression of these indices with age (n = 18 for all voxels). Location ALIC PLIC OR gCC sCC SFO SLF ILF U-fiber (central) a b c

MTR (%) 20.8 ± 0.3 21.6 ± 0.2 19.3 ± 0.3 24.9 ± 0.6 26.3 ± 0.5 17.4 ± 0.3 19.0 ± 0.4 17.8 ± 0.3 19.6 ± 0.4

Regression (t-value)

FA

Regression (t-value)

MD × 10−3 (mm2/s)

Regressionc (t-value)

AD × 10−3 (mm2/s)

Regressionc (t-value)

RD × 10−3 (mm2/s)

Regressionc (t-value)

−0.10 ± 0.22 (−0.44) 0.30 ± 0.14 (2.17) 0.42 ± 0.23 (1.83) −0.25 ± 0.16 (−1.60) 0.36 ± 0.26 (1.40) −0.24 ± 0.25 (−0.94) 0.55 ± 0.32 (1.72) 0.15 ± 0.23 (0.65) 0.48 ± 0.28 (1.71)

0.272 ± 0.005 0.347 ± 0.006 0.321 ± 0.011 0.435 ± 0.011 0.411 ± 0.016 0.161 ± 0.006 0.181 ± 0.008 0.180 ± 0.010 0.165 ± 0.006

−0.008 ± 0.003 (−2.97) 0.011 ± 0.003 (3.77) 0.016 ± 0.007 (2.41) 0.015 ± 0.007 (2.20) 0.016 ± 0.007 (4.90) −0.010 ± 0.004 (−2.87) 0.014 ± 0.005 (3.15) 0.016 ± 0.005 (3.80) 0.013 ± 0.003 (4.98)

1.36 ± 0.02 1.23 ± 0.02 1.51 ± 0.03 1.46 ± 0.03 1.69 ± 0.04 1.71 ± 0.03 1.59 ± 0.03 1.62 ± 0.02 1.55 ± 0.03

0.006 ± 0.014 (0.47) −0.021 ± 0.011 (−1.97) −0.002 ± 0.026 (−0.07) 0.042 ± 0.022 (1.90) −0.046 ± 0.028 (−1.61) 0.042 ± 0.019 (2.20) −0.039 ± 0.019 (−2.04) −0.003 ± 0.015 (−0.22) −0.029 ± 0.018 (−1.58)

1.77 ± 0.03 1.74 ± 0.02 2.05 ± 0.04 2.21 ± 0.06 2.46 ± 0.04 2.01 ± 0.03 1.85 ± 0.03 1.91 ± 0.03 1.79 ± 0.03

−0.005 ± 0.020 (−0.25) −0.012 ± 0.013 (−0.87) 0.018 ± 0.025 (0.70) 0.053 ± 0.039 (2.42) 0.017 ± 0.034 (0.50) 0.027 ± 0.020 (1.37) −0.024 ± 0.021 (−1.11) 0.028 ± 0.018 (1.58) −0.010 ± 0.021 (−0.50)

1.15 ± 0.02 0.98 ± 0.02 1.23 ± 0.03 1.08 ± 0.02 1.30 ± 0.05 1.57 ± 0.03 1.45 ± 0.03 1.47 ± 0.02 1.43 ± 0.03

0.012 ± 0.011 (1.05) −0.026 ± 0.010 (−2.58) −0.030 ± 0.031 (−0.96) 0.016 ± 0.017 (0.90) −0.077 ± 0.028 (−2.71) 0.048 ± 0.020 (2.52) −0.046 ± 0.018 (−2.52) −0.019 ± 0.015 (−1.25) −0.038 ± 0.018 (−2.19)

Values are given in mean ± standard error (SE). Values are given in (index units) per week as coefficient ± SE. The given values ×10−3.

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

a

19.2 ± 0.3 19.9 ± 0.3 21.7 ± 0.2 23.4 ± 0.2 21.4 ± 0.2 17.5 ± 0.6d 16.2 ± 0.4d 17.9 ± 0.5 18.0 ± 0.5

Regression (t-value)

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

35

Fig. 2. FDR map—q-values ≤ 0.2 (FDR ≤ 20%) for the regression of fractional anisotropy (FA) against age at scan overlaid on study-specific average T2w image of six representative slices in the axial (upper panel: inferior → superior), sagittal (middle panel: left → right) and coronal (lower panel: posterior → anterior) planes. On the top left corner of each slice is the corresponding study-specific average FA-RGB map (red, left↔right; green, anterior↔posterior; blue, superior↔inferior). Abbreviations: ACR: anterior corona radiata; ALIC: anterior limb of the internal capsule; CP: cortical plate; EC: external capsule; ILF: inferior longitudinal fasciculus; IZ: intermediate zone; OR; optic radiation, PLIC: posterior limb of the internal capsule; gCC: genu of the corpus callosum; sCC: splenium of the corpus callosum; SFO: superior fronto-occipital fasciculus; SLF: superior longitudinal fasciculus; SS: sagittal stratum (OR/ILF); WM: white matter.

measured against age, with the decrease in FA driven primarily by an increase in RD (Table 2a). The frontal subplate zone (group 6 in Table 3), on the other hand, demonstrated higher MTR values compared to the IZ, a trend of MTR decrease with age and no change in FA, resulting from a concomitant slow increase in AD and RD values (Table 2a). Similar to the subplate, subcortical GM structures (groups 7a and 7b in Table 3) showed no significant change in FA. However, in contrast to a similar increase in diffusivity values measured in the SP, the globus pallidus (GP), ventolateral thalamic nucleus (VLN) and putamen showed similar decreases in all diffusivity values, while the caudate and pulvinar showed no significant change. Voxels in anatomical locations such as the PLIC, optic radiation (OR), splenium of the corpus callosum (sCC) and U-fibers (group 2a in Table 3), the superior longitudinal fasciculus (SLF) and the WM in the middle temporal gyrus, corresponding to the auditory area, (group 2b in Table 3) showed a pronounced trend of MTR and FA increase with age, with the increase in FA (and the concomitant decrease in MD) driven primarily by the decrease in RD. The sCC (group 2a in Table 3) and the gCC (group 3 in Table 3) share the highest MTR, FA and AD values (Table 2b). However, unlike the sCC, the gCC showed a trend of MTR decrease and a significant FA increase with age, with the increase in FA (and the concomitant increase in MD) driven by the increase in AD. Association fibers such as the inferior longitudinal fasciculus (ILF) and structures containing projection fibers such as the anterior limb of the internal capsule (ALIC) show no change in MTR values along with a significant change in FA. Nevertheless, each of those structures presents distinct interrelations between DTI-derived indices resulting in an opposite change in FA. In the ILF (group 4 in Table 3), an increase in AD in parallel with a slower decrease in RD values produces an increase in

FA and no significant change in MD, while the ALIC (group 5 in Table 3) showed a significant decrease in FA values reflecting the increase in RD values. These developmental trajectories of the different quantitative MR indices are correlated. Fig. 4 shows voxel-based regression coefficient maps of MTR values against FA, MD and RD and AD against RD overlaid on average T2 volume, showing coefficient values for FDR ≤ 20%. Regression of MTR against FA shows positive coefficient values across WM regions in the frontal, parietal, temporal and occipital lobes, highlighting motor, somatosensory and visual pathways, marking mostly structures included in groups 1 and 2 in Table 3. Regression of MTR against MD and RD values, on the other hand, revealed significant negative correlations for all the above-mentioned anatomical regions in addition to the frontal subplate zone, group 6 in Table 3. Regression of AD against RD yielded significant coefficient values around 1 across the entire brain. Regression coefficient values higher than 1 suggest the change in AD values is larger than RD, while values lower than 1 suggest the opposite. Accordingly, note the posterior to anterior gradual increase in regression coefficient values across the very preterm brain, in particular across the internal capsule as observed on the axial view. Discussion This study follows spatiotemporal variation in developmental trajectories across the very preterm brain. Distinct regional variations were observed for regression coefficient maps obtained for the different quantitative MRI-derived parameters. Using a whole-brain approach, we were able to demonstrate region-specific positive as well as negative trends of linear evolution over the very preterm period (28–32 weeks)

36

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

Fig. 3. Voxel plots of MTR (upper panel), FA (middle panel) and diffusivity values (lower panel*) vs. age at scan for representative subcortical GM structures, the SP zone, the IZ and different WM tracts, chosen according to Tables 2a and 2b, show linear regression lines fitted to the data and the 95% confidence bands. *At the lower panel: green = AD, red = MD, blue = RD.

in all MRI-derived parameters measured. In accordance with our previous region-of-interest results for the periventricular WM (NossinManor et al., 2013), the frontal WM showed that these changes in

MTR and FA were positively correlated. Interestingly, however, here we were able to show this was a result of an associated decrease, rather than an increase, in both indices as a function of age. The crossroads area

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42 Table 3 Trends of developmental trajectoriesa showing the direction of the linear evolution of MTR and DTI-derived parameters with age, according to the results presented in Tables 2a and 2b b. Structures are divided into nine major groups according to the change observed in MTR and FA values. Group 1 2a 2b 3 4 5 6 7a 7b a b

MTR ↓ ↑ ↑ ↓ (small) ~ ~ ↓ ↑ ↑

FA

MD

AD

↓ ↑ ↑ ↑ ↑ ↓ ~ ~ ~

↑ ↓ ↓ ↑ ~ ~ ↑ ↓ ~

↑ ~ ↓ ↑ ↑ ~ ↑ ↓ ~

≈ ≈

b b N ≈ ≈

RD

Structures

↑ ↓ ↓ ~ ↓ ↑ ↑ ↓ ~

Crossroads, occipital WM, SFO PLIC, OR, sCC, U-fiber (central) Temporal WM, SLF gCC ILF ALIC Frontal SP Putamen, GP, VLN Caudate, pulvinar

↑ = increase, ↓ = decrease, ~ = no significant change. Trajectories of representative voxels are shown in Fig. 3.

in the frontal IZ, for example, demonstrated a linear decrease in MTR and FA with age, with the decrease in FA (and the increase in MD) driven primarily by an increase in RD values. These findings could be considered unexpected in light of previous in vivo findings in the preterm brain obtained using region-of-interest approach, studying MTR of deep GM structures (Nossin-Manor et al., 2012) and the trajectory of DTI-derived parameters in various WM tracts (Berman et al., 2005; Dudink et al., 2007; Gilmore et al., 2007; Miller et al., 2002; Partridge et al., 2004). These studies demonstrate a linear increase in MTR and FA and a decrease in MD, governed by the decrease in RD values with age. The trends describe the accepted behavior of MTR and DTI-derived parameters in very early stages of brain development documented to date in the literature (Engelbrecht et al., 1998; Lebel et al., 2008; Mukherjee and McKinstry, 2006; Neil et al., 2002; Schneider et al., 2004; Xydis et al., 2006b). Nevertheless, recent DTI

37

papers in the adult (Hasan et al., 2010) and fetal (Trivedi et al., 2009; Zanin et al., 2011) brain suggest that the polarity of the change in FA values, for example, is region-specific and age-dependant and does not follow a positive trend across the whole developmental trajectory, supporting the results presented here for the first time for the very preterm brain.

MRI-derived parameters and region-specific variation in maturation Cerebral maturation is a cascade of well characterized timely organized cellular and axonal events (Volpe, 2008). Previous models that explain the change in DTI-derived parameters with age in early maturation of major WM bundles have described three distinct maturation processes, namely, axonal organization, pre-myelination (myelination gliosis) and myelination (Dubois et al., 2008; Zanin et al., 2011). These models, however, do not explain our findings observed in areas such as the frontal IZ and SP, showing a decrease and no change in FA values, respectively, as they do not take into account processes such as the development of crossing axonal pathways (Vasung et al., 2010) and the gradual regression of radial organization of axons and glial fibers (Xu et al., 2014), that occur in the early preterm period and that can further influence the temporal evolution of DTI-derived parameters. Combining MTR and DTI-derived parameters over the whole brain, we interpret our in vivo data in terms of cerebral maturation events occurring between 28 and 32 gestational weeks. With this aim, we extended the three-phase model previously suggested for WM development by Dubois et al. (2008). The proposed model in Fig. 5 considers additional maturation processes, such as the reduction in cellular density, the regression of radial glial fibers and the development of crossing fibers, and is based on recent postmortem fetal studies obtained using diffusion tractography and histochemical tissue characterization (Takahashi et al., 2012; Trivedi et al., 2009; Vasung et al., 2010; Xu et al., 2014).

Fig. 4. Regression coefficient maps—representative axial (upper panel) and sagittal (lower panel) slices of the volumetric regression coefficient maps for the regression of MTR vs. FA (a), MTR vs. MD (b), MTR vs. RD (c), and AD vs. RD (d) overlaid on the corresponding study-specific average T2w volume. *The internal capsule in (d)—note the posterior to anterior gradual increase in regression coefficient values.

38

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

radial glial fibres established

afferent thalamocortical axon growth

afferent axons penetrate cortical plate / radial glial fibres regress toward cortical plate

Subplate toward intermediate zone MTR↑ FA↑ MD↓ AD↑ RD↓

MTR↓ FA− MD↑ AD↑ RD↑ very preterm period

radial glial fibres established

afferent thalamo- cross roads develop / cortical growth radial glial fibres regress

premyelination

Intermediate Zone

MTR↑ FA↑ MD↓ AD↑ RD↓

MTR↓ FA↓ MD↑ AD↑ RD↑

MTR↑ FA− MD↓ AD↓ RD↓

very preterm period axon growth

axonal organization / ECM decreased

premyelination

myelination

Fibre Tracts

MTR– FA↑ MD↑ AD↑ RD−

MTR↑ FA↑ MD↓ AD↓ RD↓

MTR↑ FA↑ MD↓ AD− RD↓

very preterm PLIC very preterm genu of CC very preterm splenium of CC Fig. 5. Biophysical model—a model for the biophysical interpretation of the MRI-derived parameters and brain development in the fetal, preterm (blue bar) and term periods. Three representative tissues are depicted at different stages of development, namely, the subplate (SP) (upper panel), the intermediate zone (IZ) (middle panel) and WM structures (lower panel). In the preterm period, designated by the blue bar, structures differ in the stage of development. For example, the posterior limb of the internal capsule (PLIC) is already starting the myelination process, while the genu of the corpus callosum (gCC) is still undergoing organizational processes. ECM: extracellular matrix.

Each anatomical location presented distinct interrelations between DTI-derived parameters and between these indices and MTR values, reflecting structure-specific maturational changes taking place in the pre-myelinating and early myelinating tissue.

Axonal growth Histological studies of developing human brain have demonstrated the radial organization (perpendicular to the pial surface) of the radial glial fibers and the active axonal growth in the cerebral WM to the SP

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

zone before 20 gestational weeks (Kostovic and Jovanov-Milosevic, 2006; Kostovic and Rakic, 1990; Marin-Padilla, 2011; Volpe, 2008). By 20 weeks, thalamocortical fibers, the corpus callosum and even some association pathways like ILF can be identified using diffusion tractography (Huang et al., 2006, 2009; Takahashi et al., 2012; Vasung et al., 2010). Therefore, this process (Fig. 5, axonal growth) occurs prior to the preterm period investigated in the present study and is unlikely to contribute to the changes observed in MRI indices measured; nevertheless, it supports the observation of commissural, projection and association axonal pathways detailed here. Further major events taking place between 21 and 25 gestational weeks are the radial migration of late neurons (Rakic, 2003) and oligodendrocyte progenitor cells (Jakovcevski and Zecevic, 2005) in the direction of the CP and the accumulation of the thalamocortical fibers in the SP (Fig. 5, upper panel, SP, afferent thalamocortical axonal growth) (Kostovic and Judas, 2002). Jakovcevski and Zecevic (2005) showed pre-myelinating oligodendrocytes accumulate mainly in the SP zone. These events likely explain the gradual spatial increase in MTR values from the SVZ to the CP, captured by our average MTR map obtained in the early preterm period, and reflecting the radial increase in cellular (neuronal and glial) density towards the CP (Nossin-Manor et al., 2013). A postmortem DTI study demonstrated an increase in FA values with gestational age in the IZ until 28 weeks, supporting the increase in FA values for axonal growth in the proposed model (Fig. 5, middle panel, IZ, afferent thalamocortical axonal growth) (Trivedi et al., 2009). Interestingly, until 28 weeks, the latter study reported a parallel decrease in FA in the SP corroborating the observed temporal decrease in the columnar migration of neuronal cells along the radial glial fibers by histology. The development of crossing fibers/regression of radial glial fibers Diffusion tractography images and histochemical sections acquired in the early preterm period show the development of the corona radiata and the growing interactions (crossings) of projection, callosal and associative processes in periventricular territories in the IZ adjacent to the internal capsule (Judas et al., 2005; Takahashi et al., 2012; Vasung et al., 2010). In parallel, migration events cease and the radial glial fibers begin to transform into WM astrocytes and oligodendrocytes and gradually disappear (Rakic, 2003; Sidman and Rakic, 1973). The combination of these events was suggested as an explanation for the gradual decrease in radial coherence observed in the fetal interstitial WM (Xu et al., 2014). These postmortem findings support our results (Fig. 5, middle panel, IZ, cross roads develop/radial glial fibers regress): The anterior caps and the crossroads between gCC and the ACR in the frontal IZ showed relatively low MTR and FA values, reflecting low tissue density and directionality, respectively. The further decrease in MTR values between 28 and 32 gestational weeks, seen at the same time at the frontal SP (Fig. 5, upper panel, SP, afferent axons penetrate cortical plate/radial glial fibers regress), can be explained by the temporary decrease in tissue density resulting from the reduction in migrating radial glial cells and SP neurons and the regression of radial glial fibers towards the frontal CP. The associated decrease in FA values (and the increase in MD), driven primarily by an increase in RD, can be explained by the concomitant increase in the extra cellular space in parallel with the development of crossing fibers, causing the same effect on DTI-derived parameters. The growth of thalamocortical axons into the CP The major events in axonal pathway development observed in the early preterm period by histology is the penetration of “waiting” (at the SP) afferent thalamocortical and basal forebrain fibers into the CP (Kostovic and Jovanov-Milosevic, 2006). In parallel, the SP zone is thickening due to the addition of non-myelinated callosal and other corticocortical fibers (Innocenti and Price, 2005) and there is a decrease in the extracellular matrix-rich neuropil component, possibly with the growing front of thalamocortical fibers in the superficial SP (Vasung et al.,

39

2010). The growth of thalamocortical fibers into the CP has not been directly demonstrated by diffusion tractography. Nevertheless, previous studies have attributed the decrease in anisotropy measures of the CP (seen also in this paper) to the ingrowth of these afferent fibers and the formation of synapses resulting in the loss of its columnar organization (McKinstry et al., 2002; Trivedi et al., 2009). In accordance, our results provide indirect complementary evidence by demonstrating a decreasing trend in MTR with age in the frontal SP zone along with no change in FA (and an increase in MD), due to concomitant similar increases in AD and RD values. These results corroborate the decrease in extracellular matrix and the reduction in overall tissue density (neuronal, glial), as described above, accompanying the penetration of thalamocortical fibers into the CP (Fig. 5, upper panel, SP, afferent axons penetrate the CP/radial glial fibers regress). The radial decrease in the change in tissue density towards the CP (from the IZ to the SP zone), inter alia due to the gradual reduction in migrating glial cells and the distribution of oligodendrocyte progenitor cells (Jakovcevski and Zecevic, 2005), is corroborated by MTR regression coefficient values in the frontal WM (faster decrease in MTR values with age in the IZ; Fig. 1a) and is well delineated by the radial decrease of MD and RD regression values (faster increase in MD and RD values in the IZ; Figs. 1c and e, respectively). Given the negative biophysical correlation previously observed between MTR and T1 values (Nossin-Manor et al., 2012, 2013), these findings may explain the gradual blurring of distinct T1w characteristics of the IZ and the SP, starting at 28 gestational weeks (Rados et al., 2006), by a faster increase in T1 values in the IZ compared to the SP causing MRI signal intensities to become approximately the same. This may provide the explanation for the discrepancy between the time the SP zone attains its developmental peak (i.e. maximal thickness; 29–32 gestational weeks) and the time its structural MRI transient features reach their prominence peak (i.e. lowest T1w intensity; 24–28 gestational weeks) (Kostovic and Judas, 2002; Kostovic et al., 2002). While at 28 gestational weeks onward the clear lamination pattern in the cerebral wall gradually disappears on structural MR images (Rados et al., 2006), the SP and the IZ continue to show distinct quantitative MRI characteristics in the early preterm period (Huang et al., 2006; Kostovic and Judas, 2002; Maas et al., 2004). The present study shows these two adjacent cellular zones have different MTR values. Furthermore, in accordance with previous results, we report low anisotropy in the SP (Neil et al., 1998). Nevertheless, as opposed to the decrease described above for the IZ, no change was found in tissue directionality (no change in FA) in the SP zone reflecting an isotropic increase in the extra cellular space with the penetration of thalamocortical fibers into the CP. These in vivo observations are in agreement with postmortem DTI findings showing a plateau at the edge of a graph describing a decrease in FA values in the SP as a function of gestational age until 28 weeks (Trivedi et al., 2009). Axonal organization In the early preterm period the corpus callosum is already the most bulky and highly organized fiber system in the brain (Kostovic and Judas, 2002). This is the time of exuberant development of callosal fibers and thickening of the fiber-rich subventricular zone (Innocenti and Price, 2005). In accordance, corpus callosum structures are characterized by the highest MTR and AD values (Nossin-Manor et al., 2013; Partridge et al., 2004). The process of organization itself, however, is characterized by a dominant increase in AD values with age that govern the increase in FA values, as demonstrated here for the gCC and ILF. This distinction separates the gCC and sCC in terms of maturational changes taking place in the tissue, namely, axonal organization (AD↑, RD-) and pre-myelination (AD↓, RD↓), respectively (Fig. 5, lower panel, fiber tracts). Furthermore, gCC showed low (unchanged) RD values, indicating restricted diffusion perpendicular to the long axis of the fibers, possibly due to the low extra-cellular space between exuberant axons. These results are inconsistent with previous DTI findings in the fetal brain measured between 28 and 32 gestational weeks and interpreted

40

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

without exception as myelination gliosis (pre-myelination events) in the gCC, sCC, OR and cortico-spinal tract (Zanin et al., 2011). The paradoxical trend of a small MTR decrease in the case of the gCC may be explained by competing processes, namely, the thickening of the unmyelinated callosal fiber-rich zone, corresponding to the anterior associative cortices, and the decrease in extracellular matrix content. Pre-myelination Pre-myelination events include the oligodendroglia lineage and the maturation of oligodendroglia progenitors into immature and then mature oligodendrocytes along with the increase in membrane density and a decrease in water content in tissue (Volpe, 2008). In agreement with previous literature (Drobyshevsky et al., 2005; Dubois et al., 2008; Neil et al., 2002; Nossin-Manor et al., 2013), pre-myelinating structures such as the PLIC demonstrated an increase in MTR and FA values with age (Fig. 5, lower panel, fiber tracts, pre-myelination), with the increase in FA (and decrease in MD), driven by a dominant decrease in RD, corresponding to myelination gliosis, the accumulation of premyelinating oligodendrocytes and the increase in water restriction in tissue. The OR, sCC, U-fibers (central), SLF and the temporal WM showed a similar behavior in MTR and DTI-derived parameters. Nevertheless, each of those structures, including the PLIC, demonstrated a specific combination of parametric values reflecting a distinct sequential order of maturation. The PLIC, presenting the highest measure for restriction, namely, lowest MD and RD values, is the most progressed pre-myelinating structure of the WM structures measured here and the one that myelinates first, followed by the OR (2nd lowest RD) and the highly organized, closely packed sCC (2nd lowest RD, highest MTR, FA and AD values). That AD changed less in these structures reflects their earlier maturation–myelination compared to the SLF and temporal WM. This interpretation corroborates the sequence of cerebral myelination in human infancy (Brody et al., 1987; Yakovlev and Lecours, 1967). Myelination Myelin has been identified in the VLN using MRI by 28 gestational weeks (Counsell et al., 2002). The lowest MD value combined with significant change in MTR was associated with myelination in our previous paper (Nossin-Manor et al., 2013). In accordance, high (increasing) MTR values, low (unchanged) FA and lowest (decreasing) diffusivities values, in particular MD, distinguish this deep gray matter nucleus from a pre-myelinating WM structure such as the PLIC. The GP and putamen show similar trends in the temporal evolution of MTR and DTIderived parameters. Nevertheless, lower MTR and higher diffusivity values indicate later myelination in accordance with histological findings (Brody et al., 1987; Yakovlev and Lecours, 1967). Quantification of tissue maturation To demonstrate the importance of using whole-brain analysis to quantify cerebral maturation, it is important to explain the discrepancy described above between developmental trends presented here for very preterm neonates and trends reported previously for fetuses of the same age (Zanin et al., 2011) (see Axonal organization). A physiological factor that can cause differences between fetuses and preterm neonates is the increase in brain perfusion following birth (Kehrer et al., 2005; Meek et al., 1998). In principle, increased cerebral blood flow may attenuate the diffusion signal measured by an EPI sequence. Nevertheless, while measured values may alter slightly following birth, neurodevelopmental trends should follow similar cellular maturation events, unless pathology is involved. Therefore, it seems reasonable to postulate that increased cerebral blood flow is probably not responsible for the discrepancy between developmental trends measured in fetuses and preterm neonates. As suggested previously by our group (NossinManor et al., 2013), regional quantitative values reported in the MRI literature are greatly influenced by the acquisition and analysis methods

used to obtain these values. Therefore, assuming that our biophysical model holds for both fetuses and preterm neonates, a methodical reason that can explain these differences is the fact that our results were obtained using voxel-based analysis, enabling us to measure values in each and every voxel across the brain, while the results reported by Zanin et al. (2011) were obtained using region-of-interest approach, looking at specific structures as complete and homogenous developing units. Previous histological findings demonstrate that this is not how the brain matures (Volpe, 2008). For example, the spatiotemporal developmental trends have been shown to increase progressively with age in a posterior-to-anterior direction, as demonstrated here for the internal capsule using coefficient values for the regression of AD against RD. The unique aspect of the present study is in providing in vivo MRI data that can explain macro- and micro-structural changes in the neuroanatomy of the developing human brain. Our findings, obtained on “normal” preterm neonates, parallel results attained recently by postmortem morphological analyses using histological and diffusion tractography techniques (Takahashi et al., 2012; Trivedi et al., 2009; Vasung et al., 2010; Xu et al., 2014). The novel whole-brain multicontrast approach we present for very preterm neonates is obtained by high-resolution DTI and MTI sequences that are applicable on clinical MRI scanners. No more than 40 min were needed to complete the MRI protocol on a 1.5 scanner (without the use of parallel or multi-band imaging; image acquisition techniques that were unavailable using the neonatal head coil (receive only) at the time the protocol was first designed (2008)). Axonal pathways were demonstrated on MTR and FA regression maps. In contrast to most tractography methods, however, we used a low number of 15 non-collinear diffusion directions, and could avoid the need to define structure-of-interest a priori and to apply the conventional FA mask of FA ≥ 0.2. The change in FA with age was the most robust effect found and was detected in anatomical locations with very low FA. For example, we could identify long association fibers having FA values around 0.18, an observation previously provided only by postmortem studies of the fetal and very preterm period (Huang et al., 2006, 2009; Takahashi et al., 2012), and we demonstrated statistically significant changes with age in the order of 10−2 per week in these structures. While FA was found most sensitive to changes in WM micro-structure, low MD values were associated with maturation–myelination. This may explain previous experimental results in 1–4-month-old infants describing a delay in MD compared to anisotropy in the corpus callosum, external capsule and uncinate fasciculus (Dubois et al., 2008). Taken alone, DTI-derived parameters extracted from the Gaussian diffusion model suffer from a lack of specificity to change in microstructure, e.g., the coupling between cellular/axonal density, axonal coherent alignment and myelination (Avram et al., 2013). Using a multi-contrast approach and obtaining both MTR and DTI indices as part of the same protocol enabled us to further dissociate these changes. For example, using the interrelations between MTR and FA values, we demonstrated different maturational events in all association fibers identified (SFO, SLF, ILF) and distinguished processes taking place in corpus callosum structures. Furthermore, using a voxel-based approach we were able to demonstrate the heterogeneous pattern of tissue maturation across the whole brain by showing, for example, the gradual posterior-toanterior increase in coefficient values for the regression of AD against RD across the internal capsule, similar to recent results obtained using multi-shell diffusion MRI and the biophysical compartment diffusion models CHARMED-light and NODDI (Kunz et al., 2014). The objective of this work was to characterize whole-brain 3D spatiotemporal maturational changes in the tissue using study-specific template. Therefore, a limitation to our study is the low number of subjects in the cohort (n = 18 out of 105). To avoid confounding the comparison with available literature data on autopsy specimens, the inclusion criteria was restricted to preterm neonates presenting with normal findings on conventional MR images and having no severe

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

motion artifacts on the series of high-resolution structural and quantitative volumes. As scans were acquired without sedation, motion was a major obstacle and up to 20% of the images of the different modalities were lost for that reason, in particular DTI. With imaging small brains, another limitation is the partial volume effect cause by the large voxel size compared to the bundle size which was exacerbated by the use of a Gaussian smoothing kernel prior to voxel-based regressions. Conclusions Previous work has demonstrated that dissociating MTR and DTIderived parameters aids in following the spatiotemporal pattern of early brain maturation, taking into account both organization and myelination processes (Nossin-Manor et al., 2013). Here we analyzed MTR and DTI-derived parameters using a voxel-based approach to characterize spatiotemporal cerebral maturation across the whole brain between 28 and 32 gestational weeks, a period of tremendous brain development. Fiber tracts of major axonal pathways such as the somatosensory and primary motor pathways were evident on MTR and FA regression coefficient maps. Moreover, developmental trajectories of MTR and DTI-derived parameters differed between regions dissociating the coupling between changes in cellular/axonal density, axonal coherent alignment and myelination, and elucidating distinct developmental processes in the very preterm brain. The relations between changes in quantitative MRI parameters and region-specific maturational events, including anomalous decreases in MTR and FA, could be explained by a biophysical model taking into account cell migration and the reduction of overall tissue density, the gradual regression of radial glial fibers, the development of axonal circuitry and the disappearance of the radial coherence, and pre-/early myelination events. Looking at quantitative MRI indices and the interrelations among these measures, we observed the lamination pattern in the cerebral wall that is no longer evident on structural MR images after 28 gestational weeks. These findings support the use of multiple MRI contrasts in assessing brain development and suggest the value of future studies examining a broader age range and the neurodevelopmental effects of specific pathologies. Acknowledgments This research was supported by the Canadian Institute of Health Research (CIHR MOP-84399). We would like to thank the time and effort spent by neuroradiologist Dr. Manohar M. Shroff for assessing the conventional MR images. We thank neonatologists Drs. Hilary E. Whyte and Aideen M. Moore who were responsible for patient recruitment. We thank the Mouse Imaging Centre at the Hospital for Sick Children in Toronto for supplying the mouse-build-model algorithm that we adapted to neonatal brain registration. References Avram, A.V., Ozarslan, E., Sarlls, J.E., Basser, P.J., 2013. In vivo detection of microscopic anisotropy using quadruple pulsed-field gradient (qPFG) diffusion MRI on a clinical scanner. NeuroImage 64, 229–239. 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. Berman, J.I., Mukherjee, P., Partridge, S.C., Miller, S.P., Ferriero, D.M., Barkovich, A.J., Vigneron, D.B., Henry, R.G., 2005. Quantitative diffusion tensor MRI fiber tractography of sensorimotor white matter development in premature infants. NeuroImage 27, 862–871. Brisse, H., Fallet, C., Sebag, G., Nessmann, C., Blot, P., Hassan, M., 1997. Supratentorial parenchyma in the developing fetal brain: in vitro MR study with histologic comparison. AJNR Am. J. Neuroradiol. 18, 1491–1497. Brody, B.A., Kinney, H.C., Kloman, A.S., Gilles, F.H., 1987. Sequence of central nervous system myelination in human infancy. I. An autopsy study of myelination. J. Neuropathol. Exp. Neurol. 46, 283–301. Card, D., Nossin-Manor, R., Taylor, M.J., Sled, J.G., 2011. Automated Partial Volume Tissue Classification in Preterm Infants (abstr). The Nineteenth International Society for

41

Magnetic Resonance in Medicine. Proc Intl Soc Mag Reson Med, Montréal, Canada, p. 4321. Childs, A.M., Ramenghi, L.A., Evans, D.J., Ridgeway, J., Saysell, M., Martinez, D., Arthur, R., Tanner, S., Levene, M.I., 1998. MR features of developing periventricular white matter in preterm infants: evidence of glial cell migration. AJNR Am. J. Neuroradiol. 19, 971–976. Collins, D.L., Neelin, P., Peters, T.M., Evans, A.C., 1994. Automatic 3D intersubject registration of MR volumetric data in standardized Talairach space. J. Comput. Assist. Tomogr. 18, 192–205. Counsell, S.J., Maalouf, E.F., Fletcher, A.M., Duggan, P., Battin, M., Lewis, H.J., Herlihy, A.H., Edwards, A.D., Bydder, G.M., Rutherford, M.A., 2002. MR imaging assessment of myelination in the very preterm brain. AJNR Am. J. Neuroradiol. 23, 872–881. Drobyshevsky, A., Song, S.K., Gamkrelidze, G., Wyrwicz, A.M., Derrick, M., Meng, F., Li, L., Ji, X., Trommer, B., Beardsley, D.J., Luo, N.L., Back, S.A., Tan, S., 2005. Developmental changes in diffusion anisotropy coincide with immature oligodendrocyte progression and maturation of compound action potential. J. Neurosci. 25, 5988–5997. Dubois, J., Hertz-Pannier, L., Dehaene-Lambertz, G., Cointepas, Y., Le Bihan, D., 2006. Assessment of the early organization and maturation of infants' cerebral white matter fiber bundles: a feasibility study using quantitative diffusion tensor imaging and tractography. NeuroImage 30, 1121–1132. Dubois, J., Dehaene-Lambertz, G., Perrin, M., Mangin, J.F., Cointepas, Y., Duchesnay, E., Le Bihan, D., Hertz-Pannier, L., 2008. Asynchrony of the early maturation of white matter bundles in healthy infants: quantitative landmarks revealed noninvasively by diffusion tensor imaging. Hum. Brain Mapp. 29, 14–27. Dudink, J., Lequin, M., van Pul, C., Buijs, J., Conneman, N., van Goudoever, J., Govaert, P., 2007. Fractional anisotropy in white matter tracts of very-low-birth-weight infants. Pediatr. Radiol. 37, 1216–1223. Engelbrecht, V., Rassek, M., Preiss, S., Wald, C., Modder, U., 1998. Age-dependent changes in magnetization transfer contrast of white matter in the pediatric brain. AJNR Am. J. Neuroradiol. 19, 1923–1929. Gadisseux, J.F., Evrard, P., Misson, J.P., Caviness, V.S., 1989. Dynamic structure of the radial glial fiber system of the developing murine cerebral wall. An immunocytochemical analysis. Brain Res. Dev. Brain Res. 50, 55–67. Gilmore, J.H., Lin, W., Corouge, I., Vetsa, Y.S., Smith, J.K., Kang, C., Gu, H., Hamer, R.M., Lieberman, J.A., Gerig, G., 2007. Early postnatal development of corpus callosum and corticospinal white matter assessed with quantitative tractography. AJNR Am. J. Neuroradiol. 28, 1789–1795. Girard, N., Raybaud, C., Poncet, M., 1995. In vivo MR study of brain maturation in normal fetuses. AJNR Am. J. Neuroradiol. 16, 407–413. Hasan, K.M., Kamali, A., Abid, H., Kramer, L.A., Fletcher, J.M., Ewing-Cobbs, L., 2010. Quantification of the spatiotemporal microstructural organization of the human brain association, projection and commissural pathways across the lifespan using diffusion tensor tractography. Brain Struct. Funct. 214, 361–373. Haynes, R.L., Borenstein, N.S., Desilva, T.M., Folkerth, R.D., Liu, L.G., Volpe, J.J., Kinney, H.C., 2005. Axonal development in the cerebral white matter of the human fetus and infant. J. Comp. Neurol. 484, 156–167. Hermoye, L., Saint-Martin, C., Cosnard, G., Lee, S.K., Kim, J., Nassogne, M.C., Menten, R., Clapuyt, P., Donohue, P.K., Hua, K., Wakana, S., Jiang, H., van Zijl, P.C., Mori, S., 2006. Pediatric diffusion tensor imaging: normal database and observation of the white matter maturation in early childhood. NeuroImage 29, 493–504. Huang, H., Zhang, J., Wakana, S., Zhang, W., Ren, T., Richards, L.J., Yarowsky, P., Donohue, P., Graham, E., van Zijl, P.C., Mori, S., 2006. White and gray matter development in human fetal, newborn and pediatric brains. NeuroImage 33, 27–38. Huang, H., Xue, R., Zhang, J., Ren, T., Richards, L.J., Yarowsky, P., Miller, M.I., Mori, S., 2009. Anatomical characterization of human fetal brain development with diffusion tensor magnetic resonance imaging. J. Neurosci. 29, 4263–4273. Huppi, P.S., Maier, S.E., Peled, S., Zientara, G.P., Barnes, P.D., Jolesz, F.A., Volpe, J.J., 1998. Microstructural development of human newborn cerebral white matter assessed in vivo by diffusion tensor magnetic resonance imaging. Pediatr. Res. 44, 584–590. Innocenti, G.M., Price, D.J., 2005. Exuberance in the development of cortical networks. Nat. Rev. Neurosci. 6, 955–965. Jakovcevski, I., Zecevic, N., 2005. Sequence of oligodendrocyte development in the human fetal telencephalon. Glia 49, 480–491. Judas, M., Rados, M., Jovanov-Milosevic, N., Hrabac, P., Stern-Padovan, R., Kostovic, I., 2005. Structural, immunocytochemical, and mr imaging properties of periventricular crossroads of growing cortical pathways in preterm infants. AJNR Am. J. Neuroradiol. 26, 2671–2684. Kehrer, M., Blumenstock, G., Ehehalt, S., Goelz, R., Poets, C., Schoning, M., 2005. Development of cerebral blood flow volume in preterm neonates during the first two weeks of life. Pediatr. Res. 58, 927–930. Kostovic, I., Jovanov-Milosevic, N., 2006. The development of cerebral connections during the first 20-45 weeks' gestation. Semin. Fetal Neonatal Med. 11, 415–422. Kostovic, I., Judas, M., 2002. Correlation between the sequential ingrowth of afferents and transient patterns of cortical lamination in preterm infants. Anat. Rec. 267, 1–6. Kostovic, I., Rakic, P., 1990. Developmental history of the transient subplate zone in the visual and somatosensory cortex of the macaque monkey and human brain. J. Comp. Neurol. 297, 441–470. Kostovic, I., Judas, M., Rados, M., Hrabac, P., 2002. Laminar organization of the human fetal cerebrum revealed by histochemical markers and magnetic resonance imaging. Cereb. Cortex 12, 536–544. Kovacevic, N., Henderson, J.T., Chan, E., Lifshitz, N., Bishop, J., Evans, A.C., Henkelman, R.M., Chen, X.J., 2005. A three-dimensional MRI atlas of the mouse brain with estimates of the average and variability. Cereb. Cortex 15, 639–645. Kunz, N., Zhang, H., Vasung, L., O'Brien, K.R., Assaf, Y., Lazeyras, F., Alexander, D.C., Huppi, P.S., 2014. Assessing white matter microstructure of the newborn with multi-shell diffusion MRI and biophysical compartment models. NeuroImage 96, 288–299.

42

R. Nossin-Manor et al. / NeuroImage 112 (2015) 30–42

Lebel, C., Walker, L., Leemans, A., Phillips, L., Beaulieu, C., 2008. Microstructural maturation of the human brain from childhood to adulthood. NeuroImage 40, 1044–1055. 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., 2011. The Human Brain: Prenatal Development and Structure. Springer, Berlin, Heidelberg. 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. Meek, J.H., Tyszczuk, L., Elwell, C.E., Wyatt, J.S., 1998. Cerebral blood flow increases over the first three days of life in extremely preterm neonates. Arch. Dis. Child Fetal Neonatal Ed. 78, F33–F37. 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. Morris, D., Nossin-Manor, R., Taylor, M.J., Sled, J.G., 2011. Preterm neonatal diffusion processing using detection and replacement of outliers prior to resampling. Magn. Reson. Med. 66, 92–101. Mukherjee, P., McKinstry, R.C., 2006. Diffusion tensor imaging and tractography of human brain development. Neuroimaging Clin. N. Am. 16, 19–43 (vii). Neil, J.J., Shiran, S.I., McKinstry, R.C., Schefft, G.L., Snyder, A.Z., Almli, C.R., Akbudak, E., Aronovitz, J.A., Miller, J.P., Lee, B.C., Conturo, T.E., 1998. Normal brain in human newborns: apparent diffusion coefficient and diffusion anisotropy measured by using diffusion tensor MR imaging. Radiology 209, 57–66. Neil, J., Miller, J., Mukherjee, P., Huppi, P.S., 2002. Diffusion tensor imaging of normal and injured developing human brain - a technical review. NMR Biomed. 15, 543–552. Nossin-Manor, R., Chung, A.D., Whyte, H.E., Shroff, M.M., Taylor, M.J., Sled, J.G., 2012. Deep gray matter maturation in very preterm neonates: regional variations and pathologyrelated age-dependent changes in magnetization transfer ratio. Radiology 263, 510–517. Nossin-Manor, R., Card, D., Morris, D., Noormohamed, S., Shroff, M.M., Whyte, H.E., Taylor, M.J., Sled, J.G., 2013. Quantitative MRI in the very preterm brain: assessing tissue organization and myelination using magnetization transfer, diffusion tensor and T(1) imaging. NeuroImage 64, 505–516. 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. Rados, M., Judas, M., Kostovic, I., 2006. In vitro MRI of brain development. Eur. J. Radiol. 57, 187–198. Rakic, P., 2003. Developmental and evolutionary adaptations of cortical radial glia. Cereb. Cortex 13, 541–549.

Schneider, J.F., Il'yasov, K.A., Hennig, J., Martin, E., 2004. Fast quantitative diffusion-tensor imaging of cerebral white matter from the neonatal period to adolescence. Neuroradiology 46, 258–266. Sidman, R.L., Rakic, P., 1973. Neuronal migration, with special reference to developing human brain: a review. Brain Res. 62, 1–35. Sled, J.G., Zijdenbos, A.P., Evans, A.C., 1998. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans. Med. Imaging 17, 87–97. Smith, S.M., 2002. Fast robust automated brain extraction. Hum. Brain Mapp. 17, 143–155. Spring, S., Lerch, J.P., Henkelman, R.M., 2007. Sexual dimorphism revealed in the structure of the mouse brain using three-dimensional magnetic resonance imaging. NeuroImage 35, 1424–1433. Takahashi, E., Folkerth, R.D., Galaburda, A.M., Grant, P.E., 2012. Emerging cerebral connectivity in the human fetal brain: an MR tractography study. Cereb. Cortex 22, 455–464. Tohka, J., Zijdenbos, A., Evans, A., 2004. Fast and robust parameter estimation for statistical partial volume models in brain MRI. NeuroImage 23, 84–97. Trivedi, R., Husain, N., Rathore, R.K., Saksena, S., Srivastava, S., Malik, G.K., Das, V., Pradhan, M., Pandey, C.M., Gupta, R.K., 2009. Correlation of diffusion tensor imaging with histology in the developing human frontal cerebrum. Dev. Neurosci. 31, 487–496. van Buchem, M.A., Steens, S.C., Vrooman, H.A., Zwinderman, A.H., McGowan, J.C., Rassek, M., Engelbrecht, V., 2001. Global estimation of myelination in the developing brain on the basis of magnetization transfer imaging: a preliminary study. AJNR Am. J. Neuroradiol. 22, 762–766. Vasung, L., Huang, H., Jovanov-Milosevic, N., Pletikos, M., Mori, S., Kostovic, I., 2010. Development of axonal pathways in the human fetal fronto-limbic brain: histochemical characterization and diffusion tensor imaging. J. Anat. 217, 400–417. Volpe, J.J., 2008. Neuronal Proliferation, Migration, Organization, and Myelination. Neurology of the Newborn. Saunders, Elsevier, Inc., Philadelphia, pp. 51–118. Xu, G., Takahashi, E., Folkerth, R.D., Haynes, R.L., Volpe, J.J., Grant, P.E., Kinney, H.C., 2014. Radial coherence of diffusion tractography in the cerebral white matter of the human fetus: neuroanatomic insights. Cereb. Cortex 24, 579–592. Xydis, V., Astrakas, L., Drougia, A., Gassias, D., Andronikou, S., Argyropoulou, M., 2006a. Myelination process in preterm subjects with periventricular leucomalacia assessed by magnetization transfer ratio. Pediatr. Radiol. 36, 934–939. Xydis, V., Astrakas, L., Zikou, A., Pantou, K., Andronikou, S., Argyropoulou, M.I., 2006b. Magnetization transfer ratio in the brain of preterm subjects: age-related changes during the first 2 years of life. Eur. Radiol. 16, 215–220. Yakovlev, P.I., Lecours, A.R., 1967. The Myelogenetic Cycles of Regional Maturation of the Brain. In: Minkowski, A. (Ed.), Regional Development of the Brain in Early Life. Blackwell, Oxford, pp. 3–70. Zanin, E., Ranjeva, J.P., Confort-Gouny, S., Guye, M., Denis, D., Cozzone, P.J., Girard, N., 2011. White matter maturation of normal human fetal brain. An in vivo diffusion tensor tractography study. Brain Behav. 1, 95–108. Zijdenbos, A., Forghani, R., Evans, A., 1998. Automatic quantification of MS lesions in 3D MRI brain data sets: validation of INSECT. Medical Image Computing and Computer-Assisted Intervention–MICCAI'98, pp. 439–448.