Serial quantitative diffusion tensor MRI of the premature brain: Development in newborns with and without injury




To determine the change over time of the apparent diffusion coefficient (ADC) and relative anisotropy of cerebral water in a cohort of premature newborns serially studied near birth and again near term.

Materials and Methods

Newborns were classified as normal (N = 11), minimal white matter injury (N = 7), or moderate white matter injury (N = 5).


ADC decreased significantly with age in all brain regions in newborns classified as normal and those with minimal white matter injury. ADC increased with age or failed to decline in widespread areas of white matter in newborns with moderate white matter injury. Anisotropy increased with age in all white matter regions in newborns classified as normal. Anisotropy did not increase in frontal white matter in those with minimal white matter injury, and in widespread white matter areas in those with moderate white matter injury.


This study demonstrates that serial diffusion tensor magnetic resonance imaging scans of premature newborns can detect differences in white matter maturation in infants with and without white matter injury. J. Magn. Reson. Imaging 2002;16:621–632. © 2002 Wiley-Liss, Inc.

DURING THE PERIOD OF early post-natal life for most premature newborns (i.e., from 25 to 40 weeks post-conception), the brain undergoes marked structural development (1, 2). This structural brain development can be studied using diffusion tensor imaging (DTI), which can detect significant changes in cerebral water motion with increasing age in premature newborns (3–6). However, these studies have been limited by small numbers and analysis of limited regions (3–6). Additionally, these studies have been primarily cross-sectional with single exams of individual newborns at various ages.

Only two studies have reported changes in both apparent diffusion coefficient (ADC) and anisotropy (3, 4). The ADC decreased and relative anisotropy increased with increasing age in limited white matter regions in 17 preterm newborns, 10 of whom were studied a second time at term (3). Neil et al (4) found that ADC decreased with increasing age in gray matter regions, as well as in white matter regions, while anisotropy increased only in the white matter. On the basis of a single DTI exam of each newborn, Neil et al (4) also found that the increase in anisotropy occurred almost entirely at term, rather than linearly with age. The difference in white matter regions in which anisotropy increased in these two studies underscores the difficulty of quantitatively studying brain development over time with cross-sectional analysis. The current study, however, was designed to obtain serial exams in which we measured the ADC and anisotropy values at given time-points, as well as the change in these values over time for each infant. In this way, the presence or lack of maturational changes can be assessed for each individual patient.

White matter injury (WMI) is the most important type of brain injury in infants born prematurely, and portends cerebral palsy, developmental delay, and visual impairment (1, 7). Histopathologically, WMI is characterized by foci of necrosis that are usually small, multiple, and fairly symmetric (1). Focal WMI is typically in the centrum semiovale and in the white matter surrounding the trigones of the lateral ventricles (1). This is frequently accompanied by a more diffuse injury to white matter characterized by pre-oligodendroglial cell loss and axonal injury (1). One study of diffusion imaging has shown that newborns with early WMI had lower anisotropy at term in the central white matter than newborns without WMI studied at term (8). However, premature newborns with WMI have not previously been serially studied with DTI to determine whether abnormal brain development, measured by ADC and anisotropy, occurs following early brain injury.

The goal of this study was to determine the change over time of ADC and anisotropy in motor, visual, and cognitive brain regions in serially studied premature newborns with and without WMI. A more complete assessment of the change in brain ADC and relative anisotropy over time in premature newborns is important to improve our understanding and identification of both brain development and injury.


The cohort included 23 infants of less than 36 weeks gestational age born in or transferred to our institution's intensive care nursery since September 2000. Gestational age was calculated based on the last menstrual period or early ultrasound (less than 24 weeks); if the difference between the two methods was greater than 7 days, the ultrasound date was used. Infants were excluded if there was: 1) clinical evidence of a congenital malformation or syndrome; 2) congenital infection; or 3) a large parenchymal hemorrhage or infarction evident on ultrasound examination. The protocol was approved by our institution's Committee on Human Research. Infants were studied only after voluntary informed consent was obtained from the parents.

Magnetic Resonance (MR) Imaging

The premature newborns were studied longitudinally as soon after birth as they were stable enough to be transported safely to the MR scanner and imaging time was available, and again near term or just before discharge from the hospital. Most newborns that did not require mechanical ventilatory support were fed before the MR exam and needed no pharmacological sedation. Five newborns required sedation with intravenous Nembutal for both studies, and four newborns required sedation for only one MR imaging study. All studies were performed on a 1.5-Tesla Signa EchoSpeed system (GE Medical Systems) using an MR-compatible isolette, developed for these studies and positioned inside of the standard adult head coil. A neonatologist in the MR suite monitored the newborns during scanning, and hand-ventilated intubated newborns.

The same MR techniques were utilized for the entire cohort. MR imaging of the brain in all newborns included: 1) T1-weighted sagittal and axial spin-echo images (4-mm thickness) using repetition time (TR) = 500 msec, echo time (TE) = 11 msec, theta (flip angle) = 90°, one excitation, and 192 × 256 acquisition matrix; 2) T2-weighted spin-echo (4-mm thickness) with TR = 3000 msec, TE = 60120 msec, theta = 90°, and 192 × 256 acquisition matrix; and 3) coronal spoiled gradient-recalled (SPGR) images (1.5-mm thickness) with TR = 36 msec, TE = 9 msec, theta = 35°, and one excitation.

Two study investigators, both clinical pediatric neuroradiologists, interpreted each of the MR imaging studies blinded to the subjects' clinical condition. Discrepancies were resolved by consensus.

The MR imaging diagnosis of WMI required abnormal T1 hyperintensity in the absence of marked T2 hypointensity in the periventricular white matter, or low intensity on T1-weighted images indicating cavitation (9, 10). These regions were interpreted as representing astrogliosis (9). Areas of T1 hyperintensity and T2 hypointensity were interpreted as foci of hemorrhage and were not included in the study. Newborns were classified as “normal” if there were no periventricular white matter abnormalities, “minimal WMI” if there were three or fewer areas of T1 signal abnormality measuring less than 2 mm, or “moderate WMI” if there were more than three areas of T1 signal abnormality or these areas measured more than 2 mm but less than 5% of the hemisphere was involved. In order to determine the severity of MR imaging abnormalities in the DTI regions of interest (ROIs), these areas were scored as normal (0), hyperintensity on T2-weighted images only (1), or hyperintensity on T2-weighted images and hyperintensity on T1-weighted images (2).


DTI was added to each MR imaging study using a sequence developed by the investigators specifically for this project to assess water diffusion parameters (ADC and anisotropy maps). This sequence used the acquisition of single-shot echo-planar images with seven acquisitions per location with diffusion gradients in six directions (matrix = 101, −101, 011, 01–1, 110, −110) to allow diffusion tensor calculations. The interleaved diffusion echo-planar imaging (EPI) images were acquired in 1 minute in the axial orientation from approximately 10–15 slices. The single-shot EPI images were acquired with a 128 × 256 matrix, field of view (FOV) = 18 × 36 cm, in-plane resolution = 1.4 mm, slice thickness = 4 mm, TR = 5 seconds, TE = 99.5 msec, Δ = 38.7 msec, δ = 32.2 msec, and a b-value of 600 seconds/mm2. The diffusion MR data were analyzed off-line using programs developed by our research group to generate rotationally-invariant ADC and relative anisotropy (RA) images for each slice location. RA was defined as the SD of the eigenvalues divided by the average of the eigenvalues [RA = SD (eigenvalues)/average (eigenvalues)] following the method of Basser et al (11).

To measure ADC and RA values in motor, visual, and cognitive regions of the brain, 0.5-cc ROIs were centered on the ADC images in the following gray matter (1–4) and white matter (5–9) regions bilaterally (Fig. 1) thalamus, 2) basal ganglia, 3) calcarine gray matter, 4) hippocampus (0.15 cc), 5) cortical spinal tracts, 6) parietal white matter, 7) frontal white matter, 8) optic radiations, and 9) visual association area. As WMI is frequently associated with abnormal motor, visual, and cognitive outcome, ROIs were chosen to include regions important in motor, visual, and cognitive function. Use of these regions allowed us to associate the regional DTI measures with motor, visual, and cognitive outcomes of these newborns in the future. The ROIs were defined by a single neuroradiologist according to anatomical landmarks, and were chosen to avoid signal from the cerebrospinal fluid in the ventricular system.

Figure 1.

DTI ROIs defined on the trace ADC maps. A: Hippocampus. B: Calcarine gray matter and optic radiations. C: Basal ganglia, thalamus, and visual association area. D: Frontal white matter and parietal white matter. E: Corticospinal tracts.

Data Analysis

Statistical analyses were performed using Stata (Stata Corporation, College Station, Texas). The outcome variables examined were the ADC and anisotropy in the nine regions examined. The predictor variable of interest was the post-conceptional age at the time of the MR imaging. Linear regression for repeated measures (generalized estimating equation with robust within-subjects correlation estimation) was used in order to account for the fact that each subject had two MR imaging examinations. In order to directly determine the percent change in ADC for each week increase in age, the multivariate linear regression was performed with a log-transformed ADC or anisotropy outcome variable (12). We hypothesized that the relationship of age with ADC and anisotropy would differ in the outcome groups. To explore this, the repeated measures linear regression models for each region were performed separately for each outcome group.


The severity of WMI was the same at both studies for all newborns. Eleven (48%) of the MR images were graded as normal, seven (30%) were graded as minimal WMI (Fig. 2), and five (22%) had moderate WMI at both times of study (Fig. 3).

Figure 2.

Representative coronal T1-weighted SPGR images from a newborn with minimal WMI. A:On the first study at 34.6 weeks post-conception, focal mixed hypo/hyperintensity (arrow) near the right frontal horn of the lateral ventricle is evident. B: On the second study at 43 weeks post-conception, the focal signal abnormality is no longer evident, while focal dilatation of the right frontal horn of the lateral ventricle is noted (arrow), indicating tissue loss in this region.

Figure 3.

Sequential coronal T1-weighted SPGR images from a newborn with moderate WMI. A and B: On the first study at 31.4 weeks post-conception, multi-focal hyperintensities measuring > 2 mm in the periventricular and subcortical white matter (arrows) and bilateral intraventricular hemorrhage are evident. C and D: On the second study at 35.4 weeks post-conception, the multi-focal hyperintensities are less evident (arrows) with some cystic change in these areas. The bilateral intraventricular hemorrhage is still evident but is now more isointense with white matter.

The groups were not significantly different with regards to gestational age at birth, birth weight, age at MR imaging examinations, and time interval between MR examinations (Table 1).

Table 1. Demographic Characteristics Comparing the Three Groups
 NormalMinimal WMIModerate WMIP
  1. Data is presented as median and range.

  2. WMI = white matter injury.

Post-conceptional age29.231300.1
At birth(25–31.8)(26.6–33.8)(26.6–31.8) 
Birth weight (g)1210158010200.1
Age at first MRI31.534.431.40.5
 Weeks post-conception(27.5–38)(30.9–35)(31–36) 
Age at second MRI37.437.3360.2
 Weeks post-conception(35.1–43)(35.7–42.4)(33.2–42.1) 
Time between MRI exams (days)38 (18–68)27 (16–52)25 (15–38)0.1

Two newborns with normal periventricular white matter had focal germinal matrix hemorrhage in the area of the basal ganglia ROI (score = 1–2) bilaterally. In newborns with minimal WMI, the DTI ROIs were normal on conventional MR images except in two newborns with unilateral injury in the basal ganglia ROI (score = 1–2) on the first MR image related to germinal matrix hemorrhage, and in two newborns with injury (score = 2) in the regions of the optic radiations bilaterally. In the moderate WMI group, two newborns had unilateral injury in the basal ganglia ROI (score = 1–2) on the first MR image related to germinal matrix hemorrhage, as well as unilateral injury to the corticospinal tract, and optic radiations or calcarine gray matter, one newborn had unilateral optic radiation injury (score = 2), and two newborns had unilateral injury adjacent to the frontal lobe ROI (score = 1–2). Most of the DTI ROIs were not affected by injury evident on conventional MR images.

Unlike the adult brain, ADC values were substantially higher in white matter regions compared with gray matter regions at both times of study (Figs. 4–6). In gray matter regions of all groups, significant decreases in diffusion were observed with increasing age, indicating less free water motion with increasing age (Table 2 and Figs. 5 and 7). ADC values differed across white matter regions for a given age (Fig. 6). In the white matter regions of newborns classified as normal, each week increase in age revealed a significant decrease in ADC (Table 2 and Figs. 6 and 7). A similar decrease in ADC with age was seen in newborns with minimal WMI (Table 2 and Fig. 7). However, in newborns with moderate WMI, ADC values did not significantly change with increasing age in the posterior white matter, and a significant increase in ADC was seen in the frontal white matter and visual association areas (Table 2 and Fig. 7).

Figure 4.

Representative ADC and anisotropy maps at both times of study in a normal newborn born at 29.2 weeks post-conception. ADC map (A) and anisotropy map (B) acquired at 31.9 weeks post-conception. ADC map (C) and anisotropy map (D) of the same newborn acquired at 37.3 weeks post-conception. Although no visible change in ADC is apparent, quantitative analyses showed definite reduction of ADCs throughout the brain. Increased anisotropy, manifest as increased signal intensity, is present in the internal capsule and corpus callosum.

Figure 5.

Graph of ADC values (10−3 mm2/s) plotted by post-conceptional age in gray matter ROIs in the group of newborns classified as normal.

Figure 6.

Graph of ADC (10−3 mm2/s) values plotted by post-conceptional age in white matter ROIs in the group of newborns classified as normal.

Table 2. Percent Change in ADC With Each Week Increase in Age
RegionNormalPMinimal WMIPModerate WMIP
  1. WMI = white matter injury.

Basal ganglia−1.4<.0001−2<.0001−2.1<.0001
Calcarine gray matter−1.5<.0001−2.5.009−1.4.001
Corticospinal tract−2.2<.0001−3.5<.0001−3.3<.0001
Frontal white matter−1.0.001−1.3.02+1.3<.0001
Posterior white matter−1.0.003−1.6.001−0.70.2
Optic radiations−1.6<.0001−1.2.008−2.4<.0001
Visual association−1.5<.0001−1.6<.0001+1.1<.0001
Figure 7.

Percent change in ADC with each week increase in age. Values are plotted by region for each group. The 95% confidence intervals are indicated.

In newborns classified as normal, anisotropy increased linearly with increasing age in all white matter regions (Table 3, Figs. 8 and 9). In newborns with minimal WMI, the normal development of anisotropy was not observed in the frontal white matter (Table 3 and Fig. 9). In newborns with moderate WMI, anisotropy did not increase with increasing age in all white matter regions except for the corticospinal tracts (Table 3 and Fig. 9).

Table 3. Percent Change in Anisotropy With Each Week Increase in Age
RegionNormalPMinimal WMIPModerate WMIP
  1. WMI = white matter injury.

Corticospinal tract4.4<.00016.1<.00014.0<.0001
Frontal white matter2.9.007−.05.9−1.3.4
Posterior white matter2.1.0083.9.0060.3.9
Optic radiations2.3<.00013.8<.00010.5.3
Visual association2.9.0032.6.0020.1.9
Figure 8.

Graph of relative anisotropy values (×103) plotted by gestational age in white matter ROIs in the group of newborns classified as normal.

Figure 9.

Percent change in relative anisotropy with each week increase in age. Values are plotted by region for each group. The 95% confidence intervals are indicated.


In this longitudinal study of the premature neonatal brain, we observed that the ADC (a measurement of net water motion during a MR sequence) decreased and that the diffusion became more anisotropic from 27 to 42 weeks post-conceptional age. These findings are consistent with previous observations of white matter maturation over this time period (3, 4). The decline of ADC in gray matter regions with increasing age is also consistent with previous cross-sectional observations in premature newborns (4). In premature newborns without WMI, we observed striking increases in diffusion anisotropy in widespread white matter regions. These data are consistent with the report of increase in central white matter anisotropy with increasing age observed in the only other longitudinal study using DTI in premature newborns (3). The coincidence of these diffusion findings with the known profound white matter development during this period suggests that the diffusion changes and the anatomic changes in white matter are related.

ADC values were substantially higher in white matter regions compared with gray matter regions, as others have observed (4). In this study, ADC values in different regions varied with time. At a given age, different areas of the brain were at different stages of development. While the decline in ADC was greatest in the corticospinal tracts and the thalamus, the change in ADC did not significantly differ across brain regions. This may have been due to the relatively small number of newborns studied. Alternatively, brain development as measured by DTI may proceed at a similar rate in different brain regions in premature newborns.

Given previous observations that white matter anisotropy did not significantly change with increasing gestational age, it has been suggested that the decline in white matter ADC is due to a decrease in water content (4). Yet, we and others have found that while relative anisotropy is low early in post-conceptional age, it does change, increasing significantly in white matter towards term (3). In the context of increasing anisotropy, the decline in ADC with brain development likely reflects structural or metabolic changes in the intra- or extra-cellular environment.

In newborns with moderate WMI, ADC failed to decrease and anisotropy failed to increase in widespread regions of the brain that were normal on conventional MR images. In the frontal white matter and visual association areas, ADC even increased with age in this group. Additionally, in newborns with minimal WMI, anisotropy failed to increase in the frontal white matter, a ROI that was normal on conventional MR images. The changes in anisotropy in newborns with WMI were more prominent than the changes in ADC. This is consistent with a prior study of premature newborns studied once at term, in which WMI was associated with abnormal diffusion anisotropy but normal ADC in the internal capsule and central white matter, but not in other white matter regions (8). These abnormalities of water motion may reflect the delay in myelination or axonal damage seen in newborns with WMI; this perturbation of white matter development may underlie some of the decrease in white matter volume, cortical gray matter volume, and cortical complexity identified at term in these newborns (13–16). Additionally, the increase in ADC in the frontal white matter and visual association areas, coupled with the failure of anisotropy to increase in these regions, may reflect selective vulnerability of certain white matter regions. This question of selective vulnerability requires further investigation.

The failure of ADC to decrease and anisotropy to increase normally in individual newborns with WMI may be an early marker of injury in the developing brain. Conventional MR imaging underestimates the extent of WMI, as it poorly detects the diffuse component of this injury (9). DTI detects this alteration in brain development early in post-natal life and in areas that are coincidentally normal by conventional MR imaging. These abnormalities may either reflect a widespread alteration of brain development following early focal WMI or they may represent the “diffuse” component of WMI that is frequently missed with conventional MR imaging. The abnormalities of water motion also suggest the presence of a widespread abnormality of white matter development, detectable with DTI, when conventional MR images are normal.

It has been suggested that detecting abnormal ADC and anisotropy may allow for the earlier and more accurate diagnosis of WMI compared with MR imaging (17, 18). However, determining which premature newborns have early brain injury using regional ADC and anisotropy values is not yet possible, as a large dataset of normative values for these measures in relevant brain regions does not exist. Acquiring this normative data is prohibitive, given the tremendous variability in the post-conceptional age of premature newborns and the logistic and practical issue of studying large numbers of “normal” newborns. Serial quantitative DTI, however, provides a measure of the change in diffusion parameters with age, and thus is a powerful strategy for the study of normal and abnormal brain development in the human newborn that circumvents these limitations.

The serial study design utilized here did not allow for the examination of all newborns at identical ages, making the comparison of the groups at any single time point impossible. As normative data for newborns do not exist, it is difficult to know if ADC and anisotropy values at a single study for a given premature newborn are abnormal. However, serial studies in an individual newborn can distinguish the evolution of ADC and anisotropy during normal or abnormal brain development in an individual newborn. It will be interesting to examine whether abnormalities of evolution might be a sensitive means of determining brain injury.

In summary, we found that the brains of newborns with early injury of the periventricular white matter did not manifest the normal developmental changes in water diffusion. The serial measurement of ADC and anisotropy appears to be a promising technique for the early detection of impaired brain development in premature newborns with this type of injury.


S.P.M. is supported by the Canadian Institutes of Health Research Clinician Scientist Program (Phase 1) and the Glaser Pediatric Research Network Fellowship Award.