Main

Assessment of the development of cerebral white matter in human preterm newborns is extremely important for two major reasons. First, the cerebral white matter in preterm infants is a major locus for tissue injury, due to ischemia or infection, and such injury impairs subsequent white matter development (1,2). Current modes of brain imaging, including conventional MRI, prevent clear assessment of these crucial developmental effects. Second, the vulnerability of cerebral white matter to injury likely relates at least in part to maturation-dependent microstructural changes occurring between 28 and 40 wk after conception. Microstructural changes include fiber development, such as changes in fiber size and orientation, encasement by oligodendroglial cells, and finally myelination (2,3). Relatively little is known about these microstructural changes in living infants, particularly before myelination.

A critical current methodologic need to further define white matter development in living preterm infants is the capacity to make quantitative assessments that provide information about fiber development and myelination. Moreover, of additional importance would be the capacity to use such structural information to provide images that depict microstructural detail in a manner not possible with conventional image techniques. In this study we attempted to address these needs with diffusion-weighted MRI that measure the diffusion of water in the issue (rather than the content of water as measured in conventional MRI) (4–6).

Previous MRI studies of the diffusion of cerebral water molecules have not addressed premature infants and/or have used MRI approaches that do not adequately address the structural complexities of cerebral white matter (7–10). Diffusion of water in the cerebral white matter depends upon the extent to which molecular displacement of water is free or restricted by such structures as fiber tracts. Measurements of water diffusion will vary as a function of the direction in which the measurement is made because water molecules diffuse more readily when moving parallel to fiber direction rather than perpendicular to fiber direction. This characteristic of diffusion (differing according to the medium's relative orientation) is termed anisotropy (11). The estimation of a diffusion tensor, a mathematical description of diffusion, using a vectors, has been proposed as an effective measure of diffusion in anisotropic tissue (12).

Diffusion tensor MRI was used in this study to estimate water diffusion and anisotropy in cerebral white matter of living human premature infants. An overall apparent diffusion coefficient was derived from the measures in at least six gradient directions, and relative diffusion anisotropy was calculated from the magnitude of water mobility along one direction. These measures ultimately depend upon the microstructural architecture of the cerebral white matter (11,13,14).

Thus, in this research, we used a diffusion-weighted line scan sequence (15,16) with diffusion tensor analysis (17) to assess quantitatively the diffusion of water and anisotropy in cerebral white matter of human infants between 28 and 40 wk after conception. The developmental changes of water diffusion were delineated the MRI data were used to generate images of microstructural development. Additionally, we evaluated the effect of prematurity on this microstructural development.

METHODS

Seventeen preterm [gestational age, 30.9 ± 2.3 wk (25-35 wk)] and seven full-term infants [gestational age, 39.4 ± 0.8 wk (38-40)] were studied by MRI within the first 2 wk of life. The patients were selected on the basis of the inclusion criteria given in Table 1. Ten early gestation preterm infants [n = 10; gestational age, 28.8 ± 2.1 wk (25-31 wk)] were studied again at their expected due date [PCA, 39.9 ± 1.3 wk (38-42 wk)]. The study was approved by the hospital Internal Review Board. For the study, no sedation was necessary, cardiorespiratory monitoring was used, and the infant was attended by a neonatologist (P.S.H.).

Table 1 Inclusion criteria for study group
Full size table

Line scan diffusion imaging. The diffusion-weighted line scan sequence was implemented on a 1.5-T conventional whole-body MRI system (SIGNA; General Electric Medical Systems, Milwaukee, WI). This robust diffusion imaging method does not require cardiac gating or special gradient and receiver hardware. Preliminary experience in another study (16) showed that image distortions and incomplete fat-suppression artifacts, sometimes present in diffusion-weighted echo-planar images, are not found with the diffusion-weighted line scan sequence. The maximum gradient strength was 10 mT/m, with a rise time of 600 µs, and images (diffusion-weighted image) were acquired in coronal and axial planes. The effective in-plane resolution was 1.4 × 2.7 mm with an effective section thickness of 7.3 mm (15). Diffusion was measured along seven noncollinear directions, with a diffusion weighting (b factor) of 700 s/mm2. Because the directions of diffusion encoding did not coincide with the gradient coil principal axes, a diffusion gradient strength higher than the maximum gradient strength was selected for a single principal axis alone. The echo time was 105 ms; repetition time and the effective repetition time (15) were 155 ms and 2480 ms, respectively; and the scan time was 120 s/slice.

Data analysis. Diffusion was measured in terms of the apparent diffusion coefficient, according to the Stejskal and Tanner equation (18). Off-diagonal components of the diffusion encoding were neglected, because they are very small with the diffusion-weighted line scan sequence (15). The matrix describing the directional dependence of the apparent diffusion coefficient was estimated for each voxel with nonlinear regression, as suggested by Basser (19). For each estimate the three orthogonal eigenvectors and their related positive eigenvalues were calculated. The eigenvector with the maximum diffusivity described the direction along which maximum diffusion occurs (along the fiber tract axis), whereas the eigenvector with the minimum diffusivity represents the direction with the least diffusion (transverse to the fiber tract axis).

The average apparent diffusion coefficient was calculated as the average of the principal diffusivities for each voxel and averaged for specific areas of interest (diameter = 2.6 mm). In the same area of interest, the RA is the SD of the eigenvalues of the diffusion tensor divided by the apparent diffusion coefficient and is given as a percentage. The magnitude of the percentage corresponds to the degree of preference of water movement along one direction. For an isotropic tissue, in which water diffusion is identical in all directions, regardless of the direction of measurement, the RA is 0%.

The principal direction of diffusion is given by the eigenvector which corresponds to the largest eigenvalue. Therefore, this eigenvector represents the fiber bundle direction, because the principal diffusion during the MRI diffusion encoding time (90 ms) is along the fiber long axis. For even better delineation of fiber tracts, we multiplied the eigenvector, which has the unit length of 1, by the RA. With this method, vectors indicating the direction of principal diffusion would appear longer in areas with high anisotropy (fiber tracts) and shorter in areas without a preferred diffusion direction, as in gray matter) (14,17).

Statistical analysis. To test the relationships between diffusion measures (apparent diffusion coefficients and RA) and PCA, a linear regression model was used to generate coefficients of determination (r2) with p values indicating the correlation of diffusion measures and PCA. Differences in the RA between the two regions of interest were tested in the multiple regression model using the interaction term to assess differences in intercept as well as in slope. For the comparison of apparent diffusion coefficient and RA between preterm infants at term and full-term infants, an unpaired t test was used. A value of p < 0.05 was considered significant.

RESULTS

Quantitative changes of water diffusion in developing cerebral white matter. Two cerebral white matter regions, one actively myelinating and early maturing area (the posterior limb of internal capsule) (Fig. 1) and one nonmyelinating and late maturing (the central white matter) (Fig. 2), were selected for measurement of apparent diffusion coefficients and RA. In the central white matter, apparent diffusion coefficients at 28-30-wk PCA were approximately 1.8 µm2/ms, indicating little restriction to water diffusion, and decreased to 1.2 µm2/ms at term (Fig. 3A). In contrast, in the posterior limb of the internal capsule the apparent diffusion coefficients were approximately 1.2 µm2/ms at 28-30 wk and changed minimally to 1.1 µm2/ms at 40 wk (Fig. 3B).

Figure 1
figure 1

Diffusion weighted images. Apparent diffusion coefficient trace maps in the axial plane. (A) preterm infant at 30-wk PCA and (B) a full-term infant at 40-wk PCA. (The gray scale is set so that an apparent diffusion coefficient of 3.0 µm2/ms appears white). The region of interest for measurements of both the apparent diffusion coefficient and the RA in the corresponding anisotropy map is indicated in the posterior limb of the internal capsule.

Full size image
Figure 2
figure 2

Anisotropy images. RA maps in the coronal plane are displayed. (A) Preterm infant at 30-wk PCA and (B) a full-term infant at 40-wk PCA, respectively. (The gray scale is set so that RA = 40% appears white and RA = 0% appears black). The region of interest for measurements of both the apparent diffusion coefficient and the RA in the corresponding anisotropy map is indicated in the central white matter.

Full size image
Figure 3
figure 3

Apparent diffusion coefficients (ADC). (A) Central white matter in preterm (PT) and full-term (FT) infants vs PCA (nPT + nFT = 24). The linear regression model fits the data with [ADC = -0.055 × PCA (wk) + 3.37] and r2 = 0.8 (p < 0.0001). Ten preterm infants were studied at term (PT at term). ADC measured at term (filled circles). (B) Posterior limb of the internal capsule in preterm (PT) and full-term (FT) infants vs PCA (nPT + nFT = 24). The linear regression model fits the data with [ADC = -0.031 × PCA (wk) + 1.5] and r2 = 0.6 (p < 0.0001). Ten preterm infants were studied at term (PT at term). ADC measured at term (filled circles).

Full size image

RA, the measure of the degree of anisotropy, was low in the central white matter and increased significantly toward term (Fig. 4A). RA in the posterior limb of the internal capsule was higher and increased more rapidly toward term (Fig. 4B) (p value for the slope p < 0.05).

Figure 4
figure 4

RA. (A) Central white matter in preterm (PT) and full-term (FT) infants vs PCA (nPT + nFT = 24). The linear regression model fits the data with [RA (%)= 1.1 × PCA (wk) - 24.9] and r2 = 0.6 (p < 0.0001). Ten preterm infants were studied at term (PT at term). RA measured at term (filled circles). (B) Posterior limb of the internal capsule in preterm (PT) and full-term (FT) infants (nPT + nFT = 24). The linear regression model fits the data with [RA (%)= 1.78 × PCA (wk) - 37.8] and r2 = 0.8 (p < 0.0001). Ten preterm infants were studied at term (PT at term). RA measured at term (filled circles).

Full size image

Preterm infants studied at term showed higher apparent diffusion coefficients in the central white matter than did full-term infants (Fig. 3A, Table 1). Apparent diffusion coefficients in the posterior limb of the internal capsule were similar in the preterm infants at term and in the full-term infants (Fig. 3B, Table 2). RA in both areas was markedly lower in preterm infants at term compared with full-term infants (Fig. 4, Table 2).

Table 2 Mean water diffusion measures in developing white matter
Full size table

Imaging of microstructural development of cerebral white matter by quantitative diffusion tensor MRI. The geometric nature of the diffusion tensor can be used to display the architecture of the developing cerebral white matter. Using the anisotropy measure, RA (13), and its multiplication by the eigenvector, we generated vector maps of fiber orientation (Fig. 5). The drawing of the vectors is done at higher density than the original sampling density, because fiber structures consist of multiple single fibers. The vectors clearly demonstrate the orientation of the in-plane fibers in the genu of the corpus callosum. Fibers perpendicular to the image plane in the posterior limb of the internal capsule can be distinguished clearly by the colors. Although nonmyelinated, at this development stage (28 wk), the fibers of the corpus callosum as well as fibers of the posterior limb of the (myelinating) internal capsule, demonstrate distinct anisotropy. This combination of findings indicates that anisotropy can be caused both by nonmyelinated and myelinated fibers.

Figure 5
figure 5

(A) Anatomical axial T2-weighted MRI. (B) Diffusion vectors for the region indicated by the red box in (A) overlaid on the corresponding anatomical MRI. The blue lines represent the in-plane components of the RA multiplied by the eigenvector. The length of the lines in the image plane is proportional to RA. The out-of-plane components are shown in colors ranging from green through yellow to red, with red indicating the highest value for this component. Dots represent diffusion perpendicular to the image plane. Fiber bundles in the corpus callosum and posterior limb of internal capsule are shown well, whereas on the T2-weighted MRI these bundles are not visible.

Full size image

Diffusion vector images in the coronal plane of a preterm infant at 28-wk postconception (Fig. 6A), a preterm infant at term (Fig. 6B) and a full-term infant (Fig. 6C) illustrate marked differences in vector magnitude (signified by the vector length) and alignment and hence differences in the relative degree of development of the fiber tracts. The corpus callosum in the preterm infant at 28 wk (Fig. 6A) shows a distinct group of directed vectors, the posterior limb of the internal capsule some directed vectors, whereas the central cerebral white matter exhibits short undirected vectors. The preterm infant at term (Fig. 6B) shows distinct groups of vectors in both the corpus callosum and the posterior limb of the internal capsule with some directed vectors in the central white matter. However, in the full-term infant (Fig. 6C), the central cerebral white matter shows multiple parallel aligned vector bundles.

Figure 6
figure 6

Diffusion vector maps overlaid on coronal spoiled-gradient-recalled MRI are presented in (A) for a preterm infant (PCA 28 wk), in (B) for a preterm infant at term (PCA 40 wk) and in (C) for a full-term infant (PCA 40 wk). Vector length and alignment are increased in (B) and (C) compared with (A), both in the area of the internal capsule as well as in the central white matter. Central white matter vectors in (C) compared with vectors of the analogous area in (B), indicate the presence of fiber bundles (vector bundles) in (C), which are missing or less prominent in (B).

Full size image

Comparing the preterm infant at term (Fig. 6B) and the full-term infant (Fig. 6C), differences in vector distribution and orientation are observed in the central white matter, especially in the corona radiata. The vector lengths appear to be longer and more densely packed in the full-term infant (Fig. 6C) than in the preterm infant at term, a finding consistent with the higher RA measured in full-term infants compared with preterm infants at term (Table 2).

DISCUSSION

Quantitative changes in water diffusion in developing cerebral white matter. Our data show striking changes in water diffusion as cerebral white matter developed, whereas diffusion in the posterior limb of the internal capsule remained constant. This combination of findings provides insight into the implications of such changes. Previous studies of infants beyond term have led authors to postulate that overall apparent diffusion coefficients decrease with increasing myelination of the white matter (7,8). Our observation of unchanging apparent diffusion coefficients in the internal capsule, an actively myelinating structure at 28-40 wk, suggests that myelination is not invariably accompanied by diffusion changes. Moreover, the simultaneous changes in apparent diffusion coefficients in central white matter suggest that developmental events other than myelination affect diffusion. Central cerebral white matter does not begin to myelinate until approximately term (20). Our apparent diffusion coefficients at 28-30 wk were relatively high, and then declined markedly over the time to term, when the coefficients were similar to those in the internal capsule. Taken together, these data suggest that declines in apparent diffusion coefficients during brain development can be influenced by factors other than myelination. Our findings suggest that the other factors influencing diffusion in the developing brain are related to anisotropy.

Quantitative measurements of diffusion anisotropy have been shown to be an effective means of differentiating water diffusion changes caused by myelin or from other causes such as fibers (11). Fibers in cerebral white matter have specific but random orientation within a coordinate system. Previously reported measurements of diffusion anisotropy in infant brain (7–9,21) were based on measurements in only three orthogonal directions and, as a result, underestimated diffusion restriction in diagonal and off-diagonal directions. A minimum of six gradient directions are necessary to assess diffusion anisotropy fully (13). The method reported here used seven gradient directions, and thus yielded RA measures that fully represented the three-dimensional characteristic of diffusion anisotropy in the tissue. RA in central cerebral white matter of our patients increased markedly (approximately 2-fold) between 28 and 40 wk postconception (Fig. 4A), whereas overall apparent diffusion coefficients were declining markedly.

Such anisotropic diffusion in the absence of detectable myelin parallels observations in the developing rat brain by Wimberger at al. (22). Because RA measures the directionality of diffusion our findings suggest a change in the development of fiber tracts in cerebral white matter. The nature of this development remains unclear but likely reflects changes that immediately precede the onset of myelination. Thus, this pre-myelination anisotropy and the implied restriction of diffusion perpendicular to the fiber tracts could relate to an increase in fiber diameter, axonal membrane changes, or perhaps most importantly, the early wrapping of axones by oligodendroglial processes, which would be consistent with concurrent reduced extracellular water volume (23–26).

The increased RA in the internal capsule, indicative of increasing directionality of diffusion, also suggests a developmental change in fiber tracts. In this structure, unlike central cerebral white matter, active myelination provides at least a partial explanation for the apparent restriction of diffusion. However, because axonal diameter increases during myelination (27), this diameter change may contribute importantly to the diminished water diffusion perpendicular to the orientation of the fiber and thereby the increase of the RA. Absolute RA values are lower in central white matter than in the posterior limb of the internal capsule, consistent with a multidirectional orientation of fibers within the central cerebral white matter than in the tightly packed unidirectional internal capsule.

Because the measurement of RA alone describes only the presence of preferred diffusion direction but does not permit determination of diffusion direction, further definition of the microstructural architecture of developing cerebral white matter was accomplished by generation of diffusion vector maps. This approach allows the display fiber orientation in three dimensions. These maps allow illustration of premyelinating fibers in the human brain assessed in vivo. For example, in corpus callosum, fiber bundles and orientation were prominently demonstrable on the diffusion vector maps at 28 wk PCA, even though in postmortem studies of human brain myelin deposition in this structure is not apparent until approximately 50 wk PCA (20).

Microstructural development of cerebral white matter in the preterm infant studied at term. The 10 preterm infants studied at term yielded diffusion values markedly different from those at term. Although no abnormalities were apparent on their clinical examination and cranial ultrasonography were apparent, the preterm infants had higher apparent diffusion coefficients in the central cerebral white matter than did the term infants. Concurrently RA was lower than in the infants born at term. These findings suggest that fibers are underdeveloped in the central cerebral white matter of the preterm infants at term. Whether this effect involves axones per se or their encasement by premyelinating oligodendroglia is unknown. A similar conclusion concerning fiber development is suggested by the findings in the posterior limb of the internal capsule.

The apparent diffusion coefficients in the posterior limb of the internal capsule did not differ between the two groups. This finding is consistent with the stable apparent diffusion coefficients from 28 to 40 wk postconception.

The diffusion vector maps confirmed and clarified further architectural differences between the preterm infants studied at term and the infants born at term. Fiber development and orientation, particularly in the white matter but also in internal capsule were more evident in the infants born at term. The central white matter in preterm infants at term exhibited shorter vector lengths, thinner vector bundles, and less organized fibers compared with that of the infants born at term.

Clinical implications. These findings show that diffusion tensor MRI provides the capability to measure the microstructural architecture of the cerebral white matter in living premature infants. Moreover, these quantitative data can be used to generate images that depict fiber development and orientation. The observations show great potential for defining not only normal development events but also the effects of insults, such as ischemic or infectious periventricular white matter injury, or of concomitant perturbations, such as chronic undernutrition. These results agree with delayed gray-white matter differentiation and myelination in preterm infants at term reported in an earlier study (28).