Abstract
Background
A framework for understanding rapid diffusion changes from 0 to 6 years of age is important in the detection of neurodevelopmental disorders.
Purpose
To quantify patterns of normal apparent diffusion coefficient (ADC) development from 0 to 6 years of age.
Materials and Methods
Previously constructed age-specific ADC atlases from 201 healthy full-term children (108 male; age range, 0–6 years) with MRI scans acquired from 2006 to 2013 at one large academic hospital were analyzed to quantify four patterns: ADC trajectory, rate of ADC change, age of ADC maturation, and hemispheric asymmetries of maturation ages. Patterns were quantified in whole-brain, segmented regional, and voxelwise levels by fitting a two-term exponential model. Hemispheric asymmetries in ADC maturation ages were assessed using t tests with Bonferroni correction.
Results
The posterior limb of the internal capsule (mean ADC: left hemisphere, 1.18 ×103μm2/sec; right hemisphere, 1.17 ×103μm2/sec), anterior limb of the internal capsule (left, 1.11 ×103μm2/sec; right, 1.09 ×103μm2/sec), vermis (1.26 ×103μm2/sec), thalami (left, 1.17 ×103μm2/sec; right, 1.15 ×103μm2/sec), and basal ganglia (left, 1.26 ×103μm2/sec; right, 1.23 ×103μm2/sec) demonstrate low initial ADC values, indicating an earlier prenatal time course of development. ADC maturation was completed between 1.3 and 2.4 years of age, depending on the region. The vermis and left thalamus matured earliest (1.3 years). The frontolateral gray matter matured latest (right, 2.3 years; left, 2.4 years). ADC maturation occurred earlier in the left hemisphere (P < .001) in several regions, including the frontal (mean ± standard deviation) (left, 2.16 years ± 0.29; right, 2.19 years ± 0.31), temporal (left, 1.93 years ± 0.22; right, 1.99 years ± 0.22), and parietal (left, 1.92 years ± 0.30; right, 2.03 years ± 0.28) white matter. Maturation occurred earlier in the right hemisphere (P < .001) in several regions, including the thalami (left, 1.63 years ± 0.32; right, 1.45 years ± 0.33), basal ganglia (left, 1.79 years ± 0.31; right, 1.70 years ± 0.37), and hippocampi (left, 1.93 years ± 0.34; right, 1.78 years ± 0.33).
Conclusion
Normative apparent diffusion coefficient developmental patterns on diffusion-weighted MRI scans were quantified in children aged 0 to 6 years. This work provides knowledge about early brain development and may guide the detection of abnormal patterns of maturation.
© RSNA, 2020
Online supplemental material is available for this article.
See also the editorial by Rollins in this issue.
Summary
Analysis of apparent diffusion coefficient values on diffusion-weighted MRI can be used to detect regional and voxelwise brain developmental patterns, as well as hemispheric asymmetries, from birth to 6 years of age.
Key Results
■ Age-based normative values of apparent diffusion coefficient (ADC) development patterns from 201 healthy full-term children aged 0–6 years are provided (108 male).
■ ADC maturation was completed between 1.3 and 2.4 years of age; the vermis and left thalamus matured earliest (1.3 years), and the frontolateral gray matter matured latest (right hemisphere, 2.3 years; left hemisphere, 2.4 years).
■ The left hemisphere matured earlier (P < .001) in the frontal, temporal, and parietal white matter; the right hemisphere matured earlier (P < .001) in the thalami, basal ganglia, and hippocampi.
Introduction
There is limited information about normal diffusion-weighted imaging patterns in the early childhood developmental periods. Specifically, there is an absence of normative data to describe the changes in apparent diffusion coefficient (ADC) values from birth to early childhood (1–7). These gaps in knowledge occur during the most critical time points when the brain changes most rapidly. Changes in brain ADC values are important for understanding the normal regional variations in brain maturation (8). For many diseases, this developmental period is a key time point for the early diagnosis of brain disorders (1,9,10). Since clinical radiologists rely heavily on diffusion-weighted imaging to discriminate between abnormalities, it is important to understand and quantify the patterns of normal development. Thus, there is a need to quantify ADC development patterns with spatial and temporal granularity.
Multiple reasons contribute to the current lack of knowledge about early changes in brain diffusion parameters. Previously, the understanding of normal ADC development was limited to two-dimensional image analyses with limited regions of interest in few and sparsely sampled age groups (1–7,11). MRI scans of term-born and normal-developing children are scarce, particularly longitudinal studies for children aged 2 years or younger. Also, it is a technical challenge to parcellate three-dimensional (3D) regions with accuracy throughout the developing brain and to model changes over time.
In our prior publication, normative ADC atlases spanning 10 densely sampled age groups from 0 to 6 years of age were constructed and are publicly available (12). The method of 3D atlas construction has been detailed in this prior work. In the current study, these atlases are analyzed using a novel approach for quantification of ADC changes. The purpose of this work was to quantify whole-brain voxelwise and regional variations in ADC maps across the 10 age groups. This article aims to describe key aspects of ADC development patterns, including trajectory of change, rate of change, maturation age, and hemispheric asymmetry. The work investigates the spatial and temporal heterogeneity of ADC values and the differences in brain hemispheric development.
Materials and Methods
Participants
This study was compliant with the Health Insurance Portability and Accountability Act and was approved by the institutional review board, with waiver of informed consent. As detailed previously (12), cross-sectional ADC maps from 201 term-born (>37 weeks of gestation) children between 0 and 6 years of age were acquired without sedation at a large academic hospital from 2006 to 2013 and were then retrospectively analyzed to construct 10 age-specific atlases. Please refer to the prior publication on atlas construction for additional details on demographics, data collection, and confirmation of normal brain development (12). Briefly, exclusion criteria were as follows: history of preterm birth, incomplete or motion-degraded ADC images, and reported brain abnormalities, either by clinical chart or radiology report.
MRI Protocol
Scans were obtained with a Siemens Trio 3-T MRI scanner (Siemens Healthineers) and a clinical diffusion protocol (repetition time msec/echo time msec, 7500–9500/80–115; b value, 1000 sec/mm2; matrix, 128 × 128 × 60; voxel size, 2 × 2 × 2 mm; 32 diffusion directions). ADC maps were automatically generated.
Constructing Age-specific Atlases
Atlas construction was detailed previously (12). Atlases span 10 age groups: birth to 2 weeks, 2 weeks to 3 months, 4–6 months, 6–9 months, 9–12 months, 1–2 years, 2–3 years, 3–4 years, 4–5 years, and 5–6 years (Table 1) (12). The atlases represent average neuroanatomy and average ADC at each voxel in each age group (12–14). Sample size was justified previously (12), as voxelwise ADC values remained relatively stable, with a change of less than 3% (<1% in most voxels) when moving from eight or 20 subjects in each age group, but changed more than 5% when moving from two to eight subjects in each age group. Atlases are publicly available at https://www.nitrc.org/projects/mgh_adcatlases.
Table 1:
Demographic Information
Spatiotemporal Registration and Structural Segmentation
Atlases were registered spatiotemporally with four-dimensional (3D + time) and time-deformable registration (15) and were extensively validated in children of various ages, with MRI scans acquired at multiple sites and with different imaging sequences (16–18). A multiatlas segmentation algorithm was used to segment the SRI24 anatomic T1-weighted atlas (19), with a subsequent nonrigid registration transferring the segmentation into our 5–6-year-old atlas leading to parcellation of 62 brain structures (12,20). Structural segmentation was propagated from the 5–6-year-old atlas to the other nine ADC atlases in the order of descending ages with spatiotemporal registration (12,20). Structural segmentations in the atlases were verified by two expert pediatric neuroradiologists (P.E.G., S.S.; >20 years and >5 years of experience in pediatric neuroradiology, respectively).
Statistical Analysis
Regional trajectories.—The median ADC value for each brain region was calculated, with 25th–75th percentiles. Next, the trajectory of the median ADC in each brain structure was modeled by fitting a two-term exponential model. This model was chosen since diffusion tensor imaging parameters have been shown to closely approximate this nonlinear model from early life to maturation (21). By using SciPy (Python, version 3.2), parameters a, b, c, and d were fit to minimize overall prediction error between the predicted value, medianADCPredicted, and the observed median ADC values in the atlases, medianADCObserved, over the median ages of the 10 age groups (t1–t10) (t1 = 0.02, t2 = 0.13, t3 = 0.375, t4 = 0.625, t5 = 0.875, t6 = 1.5, t7 = 2.5, t8 = 3.5, t9 = 4.5, t10 = 5.5 years of age):
 |
The two-term exponential model is derived in each of the 62 segmented regions. Eight regions were selected to represent anterior and posterior, superior and inferior, and cortical and deep brain structures, including the corpus callosum, left occipital white matter (WM), temporal WM, parietal WM, frontal WM, thalamus, cerebellum, and posterior limb of internal capsule (PLIC).
Histograms for ADC values were calculated in the whole brain and in each of the 62 regions. These histograms show the percentage of voxels at each ADC value across each age group.
Rate of ADC change.—Rate of change in ADC values was analyzed using a fitted two-term exponential model, which derives the rate of decrease from median ADC values in each segmented region and at every voxel. The rate of decrease is represented as the first-order derivative of the fitted model. These methods were repeated at the voxel level, where spatiotemporal voxel correspondences were computed by the four-dimensional deformable registration, as described previously.
The goodness of fit for the two-term exponential model in both the regional level and the voxelwise level was quantified with the root mean square error (RMSE) between predicted and observed ADC values averaged across all 10 age groups with a leave-one-out (age group) cross-validation method. Since different regions and voxels had different ADC values at different ages, percentage RMSE, or root mean square of the percentage error (sqrt (average{[(fitted–actual)/actual]2})) was also quantified. Regions were ranked by RMSE and percentage RMSE in ascending order. Voxelwise RMSE and percentage RMSE were quantified. The model fit ADC changes with less than 0.4% error in all ROIs (Table E3 [online]) and less than 1% in most voxels (Fig E1 [online]), providing high accuracy for rate of decrease and maturation age.
Maturation age.—Regional median ADC values and voxelwise mean ADC values were calculated. ADC maturation age was defined as the time when the regional median ADC or the voxelwise mean ADC values decreased less than 35 μm2/sec and the rate of decrease was less than 2% in two consecutive age groups (sampling interval, 0.1 years). To encourage neighboring voxels sharing similar maturation ages and to account for minor registration errors, the voxelwise ADC maturation age map was smoothed with an isotropic 2-mm 3D Gaussian kernel.
Hemispheric asymmetries.—A two-tailed Student t test of voxelwise maturation ages was performed in the same segmented structure between the left and right hemispheres, using Bonferroni correction to control for false-positive findings. Hemispheric asymmetry for the ADC maturation age was determined using a Pcorrected < .05 threshold.
Results
Participant Characteristics
ADC maps from 201 children (age range, 0–6 years; 108 male children) were collected (Table 1, Fig 1). A more detailed summary of demographics was previously published (12).
Figure 1:
Diagram shows flow of participants through the study. ADC = apparent diffusion coefficient.
Whole-Brain ADC Trajectory
Ten age-specific ADC atlases and corresponding regional segmentation were evaluated (Fig 2). ADC histograms for six age-groups over the first 2 years show the pattern of whole-brain ADC change (Fig 3, A). With increasing age, whole-brain ADC values decrease, and the distribution of ADC values becomes narrower in range, indicating progression toward an adult morphology. Whole-brain ADC histogram animation across age groups is available in Movie 1 (online).
Figure 2:
Age-specific three-dimensional (3D) apparent diffusion coefficient (ADC) atlases and structural segmentations. Top: Age-specific 3D ADC atlases (reported in reference 12). Bottom: Region of interest segmentation results. Different colors denote different regions.
Figure 3:
Histogram analysis of apparent diffusion coefficient (ADC) values. A, Changes in ADC histograms in the whole brain, showing the percentage of voxels by ADC value, in six age groups over the first 2 years of life. B, Changes in ADC histograms in eight representative regions of interest at the regional level analysis, demonstrating the percentage of voxels in each region of interest (ROI) by ADC values and by age. PLIC = posterior limb of the internal capsule, WM = white matter.
Movie 1:
Change of Whole-Brain Apparent Diffusion Coefficient (ADC) Histograms in the First 3 Years of Life. This animated movie shows the change of whole brain ADC histograms, at 5 age groups over the first 3 years of life. Over time, whole-brain ADC values decrease (ie, the curves move to the left), and the distribution of ADC values becomes more limited in range (ie, the curves grow taller and narrower). While the whole brain histograms demonstrate a dominant peak in the high 900 to low 1000 um2/sec range, a second smaller peak is noted in the 800 range, which is more conspicuous at later ages and is being driven by the relatively lower ADC values of the cerebellum, basal ganglia, PLIC and ALIC.
Rate of ADC Decrease at the Regional and Voxel Levels
Regional and voxelwise analyses show patterns of ADC change from birth through the early developmental time periods. Initial ADC values from birth to 2 weeks differ by region. Deep gray matter regions, such as the thalami (mean: left hemisphere, 1.17 ×103μm2/sec; right hemisphere, 1.15 ×103μm2/sec) and basal ganglia (left hemisphere, 1.26 ×103μm2/sec; right hemisphere, 1.23 ×103μm2/sec), as well as the PLIC (left hemisphere, 1.18 ×103μm2/sec; right hemisphere, 1.17 ×103μm2/sec), anterior limb of the internal capsule (left hemisphere, 1.11 ×103μm2/sec; right hemisphere, 1.09 × 103μm2/sec) and vermis (1.26 ×103μm2/sec), demonstrate relatively low (mature) initial ADC values, indicating an earlier prenatal time course of development for these structures.
Brain ADC values change most rapidly during the 1st year of life and gradually slow down during the 2nd year. Regional ADC histograms demonstrate a decrease in ADC values over the first 2 years and a narrow distribution of ADC values, indicating progression toward an adult morphology (Fig 3b). Regional ADC analysis in the six age groups shows predominantly two patterns in the rate of ADC decrease (Fig 4): the first pattern is a dramatic decrease over the course of the 1st year and the second is a more gradual decrease over the 1st year. The left cerebellum (Fig 4, region 7) shows the rapid pattern over the 1st year, whereas the corpus callosum (Fig 4, region 1) demonstrates the gradual pattern. After the 2nd year of life, the curves are all relatively stable, with respect to the rate of change.
Figure 4:
Median apparent diffusion coefficient (ADC) values and rate of ADC decrease by regional analyses. Median ADC values (in square micrometers per second) and rate of decrease (in square micrometers per second per year) of the median ADC values in eight representative regions of interest (ROIs) in the regional level analysis. In the upper graph of each region, the median ADC values are plotted with 25th–75th percentiles at each age group. For each ROI, the red line shows a fitted continuous curve of the changes in median ADC in that region, and the blue line shows the rate of decrease in the median ADC values of that region across time, which was derived from the fitted continuous red curve of the same ROI. Age of maturation is shown in years. Red and blue curves in the same box (for the same region) share the same x-axis label. PLIC = posterior limb of the internal capsule, WM = white matter.
The relative rates of ADC decrease between regions change over time. Initially, at 0–2-weeks, the cerebellum and occipital WM had rapid rates of decrease, relative to other regions. However, these regions showed the slowest rates of decrease in ADC values during the 2-week to 3-month old and 3–6-month-old periods. In the first 2 weeks of life, the areas with the most rapid rates of decrease were temporal inferior gray matter (GM), occipital inferior GM, temporal lateral GM (left), temporal WM, frontal WM, cerebellum, occipital WM, frontal lateral GM (left), and PLIC (right). The slowest rate of ADC decrease in the first 2 weeks of life occurred in the parietal medial GM (left), occipital medial GM (left), amygdala, anterior limb of the internal capsule (left), ventral diencephalon, thalamus (left), brainstem, parietal lateral GM (right), limbic medial temporal GM (left), and frontal inferior GM.
With voxelwise analysis, rates of ADC decrease within the 1st year of life were analyzed (Fig 5). During the first 2 weeks, a diffusely high rate of ADC decrease was observed. This rate was dramatically different from the rate in later age groups and is shown with a different scale (0–400 μm2/sec/year). Within the first 2 weeks of life, the highest rate of ADC decrease was seen within the periventricular WM, inferior temporal lobes, and cerebellum, concordant with regional analysis. The perirolandic and calcarine regions (dark blue areas in Fig 5) demonstrated the slowest rates of ADC decrease, indicating an earlier maturation of these structures. During the 2-week to 3-month-old period, ADC decreases most rapidly within the genu of the corpus callosum, cingulate gyrus, and frontal WM and operculum. The rate of decrease of ADC slows dramatically in the 3–6 month-old period, but a high rate of ADC decrease in the genu of the corpus callosum and frontal WM is seen from 3 to 6 months of age. The 6–9-month-old and 9–12-month-old periods demonstrate overall relatively slow rates of decrease, suggesting stabilization of ADC values. Voxel-level animations of ADC changes are available in Movie 2 (online).
Figure 5:
Rate of decrease in apparent diffusion coefficient (ADC) values according to voxelwise analysis is shown in the axial (left column), sagittal (middle column), and coronal (right column) planes. The 0–2-week-old age group is shown with a scale of 0–400 μm2/sec/y. The other four age groups are shown with a scale of 0–150 μm2/sec/y.
Movie 2:
Spatiotemporal Apparent Diffusion Coefficient (ADC) Changes and Rates of ADC Changes in the First 6 Years of Life at the Voxel Level. Top image: Atlas in this age; Middle image: Atlas in this age normalized into the 0–2 week-old atlas; Bottom image: Rate of ADC decrease at this age. This animated movie demonstrates ADC changes and the rate of ADC decrease at the voxel level in three planes, over the first 6 years of life. During the first two week age-group, a diffusely high rate of ADC decrease was observed, with the highest rate of ADC decrease within the periventricular white matter, inferior temporal lobes and cerebellum. This rate was dramatically different from later age-groups; therefore, this time period is shown with a different scale (0–400 (µm2/s)/year). In the 0–2 week age-group, the peri-Rolandic and calcarine regions (in dark blue) demonstrate the slowest rates of ADC decrease. During the 2 week to 3 month-old age-groups, the ADC decreases most rapidly within the genu of the corpus callosum and frontal white matter and the frontal operculum and cingulate gyrus. The rate of decrease of ADC slows down dramatically in the 3–6-month-old period. However, a high rate of ADC decrease in the genu of the corpus callosum and frontal white matter is seen from 3–6 months. The 6–9 and 9–12-month-old periods demonstrate overall relatively slow rates of decrease, suggesting ADC values have reached maturation.
Maturation Age of ADC at the Regional and Voxel Levels
Regional and voxelwise analyses show the pattern of ADC value maturation. Supratentorially, maturation of ADC progresses in a posterior-to-anterior distribution (Fig 6, A and B). The deep GM structures mature earlier relative to frontal and temporal GM, demonstrating a central-to-peripheral pattern of maturation.
Figure 6:
Age of apparent diffusion coefficient (ADC) maturation by voxelwise analysis. A, Age (in years) of ADC maturation by each voxel using the voxel-based analysis and shown in the axial (left), sagittal (middle), and coronal (right) planes. The color scale is shown in years, where birth is displayed in dark blue and 3 years of age is displayed in red. All ADC maturation occurred over the first 3 years of life. B, Each voxel displayed at the time of maturation (in years), in the sagittal, coronal, and axial planes, using the same color scale as A. Hemispheric asymmetry can be appreciated in the age of maturation.
All regional rates of median ADC decrease and reach a steady-state plateau, indicating a maturated state, after 1.3 to 2.4 years of age, depending on the region. The earliest regions to reach their respective plateaus were the vermis and left thalamus, at around 1.3 years of age. Regions in the bilateral amygdala and right thalamus also plateaued relatively early in their maturation, reaching a plateau by 1.4 years of age. The regions in which ADC matured latest were the right and left frontal lateral GM, occurring at 2.3 and 2.4 years of age, respectively.
With voxelwise analysis, a relatively early age of maturation was observed in the posterior regions, including the occipital and parietal lobes, cerebellum, and deep gray nuclei, (Fig 6, Table 2), concordant with regional analysis. The perirolandic region was one of the earliest areas to mature (dark green areas in Fig 6). The age of maturation was higher (red areas in Fig 6) for the frontal WM and operculum. By 3.0–3.5 years of age, most of the voxels have already matured, except for a few voxels predominantly within the frontal lobe cortex. Animated movies of ADC maturation ages in the 3D brain are available as supplements (Movies 3–5 [online] from the axial, coronal, and sagittal views, respectively).
Table 2:
Maturation Age across Regions by Hemisphere
Movie 3:
Apparent Diffusion Coefficient (ADC) Maturation Ages over the First 3 Years of Life (axial view, displayed from inferior to superior). The color-coded map displays the age at which the rate of ADC decrease reaches its plateau, where dark blue indicates maturation at birth (0 years-old) and red indicates a maturation age of 3 years-old. This analysis was performed using the voxel-based approach and mapped onto the 0–2-week-old brain atlas. The maturation age was earlier (shown in dark blue and green), in the posterior regions including the occipital and parietal lobes and the cerebellum and deep gray nuclei, consistent with prior literature. The peri-Rolandic region was one of the earliest areas to mature (dark green). The age of maturation occurred later (shown in red) in the frontal white matter and operculum.
Movie 4:
Apparent Diffusion Coefficient (ADC) Maturation Ages over the First 3 Years of Life (coronal view, displayed from posterior to anterior). The color-coded map displays the age at which the rate of ADC decrease reaches its plateau, where dark blue indicates maturation at birth (0 years-old) and red indicates a maturation age of 3 years-old. This analysis was performed using the voxel-based approach and mapped onto the 0–2-week-old brain atlas. The maturation age was earlier (shown in dark blue and green), in the posterior regions including the occipital and parietal lobes and the cerebellum and deep gray nuclei, consistent with prior literature. The peri-Rolandic region was one of the earliest areas to mature (dark green). The age of maturation occurred later (shown in red) in the frontal white matter and operculum.
Movie 5:
Apparent Diffusion Coefficient (ADC) Maturation Ages over the First 3 Years of Life (sagittal view, displayed from left to right). The color-coded map displays the age at which the rate of ADC decrease reaches its plateau, where dark blue indicates maturation at birth (0 years-old) and red indicates a maturation age of 3 years-old. This analysis was performed using the voxel-based approach and mapped onto the 0–2-week-old brain atlas. The maturation age was earlier (shown in dark blue and green), in the posterior regions including the occipital and parietal lobes and the cerebellum and deep gray nuclei, consistent with prior literature. The peri-Rolandic region was one of the earliest areas to mature (dark green). The age of maturation occurred later (shown in red) in the frontal white matter and operculum.
Hemispheric Asymmetry in ADC Development
Brain ADC value maturation differs between the right and left hemispheres. The mean maturation ages are shown by right and left hemispheric regions (Table 2). Regional hemispheric asymmetries in the age of ADC maturation were earlier in the left hemisphere (P < .001) for the lateral occipital lobe (left, 1.53 years ± 0.16; right, 1.68 years ± 0.17); parietal GM (left, 1.92 years ± 0.30; right, 2.03 years ± 0.28); frontal (left, 2.16 years ± 0.29; right, 2.19 years ± 0.31), temporal (left, 1.93 years ± 0.22; right, 1.99 years ± 0.22), and parietal (left, 1.92 years ± 0.30; right, 2.03 years ± 0.28) WM; and the medial frontal (left, 2.37 years ± 0.49; right, 2.51 years ± 0.57) and lateral (left, 2.45 years ± 0.39; right, 2.53 years ± 0.44) GM. Maturation occurred earlier in the right hemisphere (P < .001) for the medial occipital GM (left, 1.61 years ± 0.21; right, 1.56 years ± 0.30), thalami (left, 1.63 years ± 0.32; right, 1.45 years ± 0.33), medial limbic temporal GM (left, 1.76 years ± 0.37; right, 1.51 years ± 0.42), basal ganglia (left, 1.79 years ± 0.31; right, 1.70 years ± 0.37), and hippocampi (left, 1.93 years ± 0.34; right, 1.78 years ± 0.33).
Discussion
Our study provides a comprehensive four-dimensional (three-dimensional plus time) atlas-based evaluation of early childhood brain development using apparent diffusion coefficient (ADC) maps from diffusion-weighted MRI. This research builds on our prior method on atlas development (12). In this current work, parameters for ADC change are extensively characterized, providing a more comprehensive understanding of developmental patterns. Our results detail the spatiotemporal changes in absolute and relative ADC values. ADC maturation progresses in a posterior-to-anterior and central-to-peripheral distribution. Maturation was completed between 1.3 and 2.4 years of age, depending on the region. Hemispheric asymmetries in maturation age were earlier in the left hemisphere for several regions, including frontal, temporal, and parietal white matter (P < .001). Maturation occurred earlier in the right hemisphere for several regions, including the thalami, basal ganglia, and hippocampi (P < .001). Therefore, these parameters of ADC development, trajectory of change, rate of change, maturation age, and hemispheric asymmetry provide fundamental information for the understanding and interpretation of changes in ADC values over the early developmental time periods.
This work addresses substantial gaps in knowledge of diffusion-weighted parameters and patterns of change. Previously, our understanding of ADC changes in the early development period was limited to whole-brain and regional two-dimensional image analyses (4,22–25). Mean ADC values from this 3D analysis are similar to those reported in the literature (Table E1 [online]) (1–7,11,23,24). However, using 3D atlases, this work surpasses prior studies by integrating comprehensive whole-brain, regional, and voxelwise analyses, with densely sampled age groups across the critical early developmental periods.
As the field of neuroscience moves toward a more integrated and comprehensive approach to regional brain variations, this work establishes the role of ADC maps in assessing brain development. Specification of absolute and relative changes in ADC values is a critical step toward defining normal ranges for reference charts (Table E2 [online]) to allow for more accurate image analysis in clinical and research settings. Each brain region can be understood by both its absolute ADC change and its maturation relative to surrounding structures. Relative regional maturation reflects the underlying sequence of neurocognitive development and points to the early formation of structural and functional connections. Regional hemispheric asymmetries may provide a basis for adult patterns of functional lateralization (26–28) and predict the onset of disorders (29–32). Areas with a high rate of ADC change and potentially higher metabolic demands may be vulnerable to injury depending on the age of the child and the mechanism of injury (33–35).
In the absence of normal reference values, clinical radiologists underuse routine ADC maps in the detection of abnormal brain development during infancy and early childhood. Emphasis is placed on abnormally low ADC values to identify brain abnormalities. However, areas with increased ADC values are largely ignored. The lack of knowledge about variations in ADC values by region and age results in up to 20%–40% intra- and interreader variability in interpretations (36,37). Recently, our group showed these atlases to reduce interrater variability in neonatal hypoxic ischemic encephalopathy (38) and Sturge-Weber syndrome (39). Therefore, this work may serve as a valuable quantitative reference to assist radiologists in recognizing normal brain ADC patterns in clinical practice, as well as for future artificial intelligence image analysis.
There were several limitations to our study. Variability in imaging protocols and scanners may alter ADC values (40). In this study, all children were scanned with the same scanner type, the same head coil, the same diffusion-weighted sequence, and relatively stable parameters. In this context, ADC values may differ by 1%–2% (39), which is relatively small compared with the ADC differences across age groups, (approximately, 10%–20% depending on regions and age groups) (12). Nevertheless, generalizability of these normative ADC values must be verified in future multisite studies using different imaging protocols and scanning parameters. Future studies will also need to stratify participants by sex, gestational age, race, and ethnicity. Our study had a cross-sectional design; these findings need to be further verified in a cohort of patients with longitudinal follow-up examinations. However, we expect subtle variations between subjects within the same age group will be relatively suppressed by averaging across subjects during atlas construction. Lastly, anatomic boundaries can be challenging to delineate during segmentation, particularly when the regions are small, as in the neonatal period. However, this 3D approach is more accurate and consistent in accounting for errors in segmentation than manual two-dimensional slice annotation. By averaging across the region, these 3D atlases are more robust to minor segmentation errors, particularly in large cohorts. Additionally, voxelwise results do not depend on regional boundaries and represent more objective confirmation of segmented findings.
This comprehensive three-dimensional atlas-based analysis of changes in diffusion-weighted imaging addresses the prior lack of knowledge regarding regional and age-related apparent diffusion coefficient variations in the early brain developmental periods. These atlases may serve as valuable quantitative references for distinguishing normal and abnormal diffusion-weighted imaging patterns among brain regions. Future studies are needed to construct an atlas-based machine learning framework to quantitatively detect abnormalities and assist in image interpretation for both two- and three-dimensional regions of interest. With these approaches, the relative vulnerability of brain regions will need to be validated with pathology-proven MRI data in larger multisite studies.
SUPPLEMENTAL TABLES
SUPPLEMENTAL FIGURES
Acknowledgments
Acknowledgments
We thank Kallirroi Retzepi, Nathaniel Reynolds, Victor Castro, Christopher Herrick, Joseph Chou, and Yanbing (Bill) Wang for helping access and organize the clinical data. We also thank Dr Rudolph Pienaar, Dr Steve Pieper, and Dr Lilla Zöllei for early discussions of some of the presented work.
Disclosures of Conflicts of Interest: S.S. disclosed no relevant relationships. R.L.G. disclosed no relevant relationships. S.V.B. disclosed no relevant relationships. R.W. disclosed no relevant relationships. S.N.M. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: received research funds from AstraZeneca, institution received funds from Amgen. Other relationships: disclosed no relevant relationships. P.E.G. disclosed no relevant relationships. Y.O. disclosed no relevant relationships.
Supported by the National Institutes of Health (R01EB014947) (R.L.G., S.N.M., P.E.G.), Thrasher Research Fund Early Career Award (THF13411) (Y.O.), Boston Children’s Hospital/Harvard Medical School Faculty Career Development Fellowship (Y.O.), and St. Baldrick's Scholar Award with support from the Grace for Good Fund (Y.O.).
Abbreviations:
- ADC
- apparent diffusion coefficient
- GM
- gray matter
- PLIC
- posterior limb of the internal capsule
- RMSE
- root mean square error
- 3D
- three-dimensional
- WM
- white matter
References
- 1.Engelbrecht V, Scherer A, Rassek M, Witsack HJ, Mödder U. Diffusion-weighted MR imaging in the brain in children: findings in the normal brain and in the brain with white matter diseases. Radiology 2002;222(2):410–418. [DOI] [PubMed] [Google Scholar]
- 2.Kwan S, Boudes E, Benseler A, et al. Evolution of Apparent Diffusion Coefficient and Fractional Anisotropy in the Cerebrum of Asphyxiated Newborns Treated with Hypothermia over the First Month of Life. Neural Plast 2015;2015:653727. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Morriss MC, Zimmerman RA, Bilaniuk LT, Hunter JV, Haselgrove JC. Changes in brain water diffusion during childhood. Neuroradiology 1999;41(12):929–934. [DOI] [PubMed] [Google Scholar]
- 4.Bültmann E, Mußgnug HJ, Zapf A, Hartmann H, Nägele T, Lanfermann H. Changes in brain microstructure during infancy and childhood using clinical feasible ADC-maps. Childs Nerv Syst 2017;33(5):735–745. [DOI] [PubMed] [Google Scholar]
- 5.Klimas A, Drzazga Z, Kluczewska E, Hartel M. Regional ADC measurements during normal brain aging in the clinical range of b values: a DWI study. Clin Imaging 2013;37(4):637–644. [DOI] [PubMed] [Google Scholar]
- 6.Bartha AI, Yap KR, Miller SP, et al. The normal neonatal brain: MR imaging, diffusion tensor imaging, and 3D MR spectroscopy in healthy term neonates. AJNR Am J Neuroradiol 2007;28(6):1015–1021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Neil JJ, Shiran SI, McKinstry RC, et al. Normal brain in human newborns: apparent diffusion coefficient and diffusion anisotropy measured by using diffusion tensor MR imaging. Radiology 1998;209(1):57–66. [DOI] [PubMed] [Google Scholar]
- 8.Dubois J, Dehaene-Lambertz G, Kulikova S, Poupon C, Hüppi PS, Hertz-Pannier L. The early development of brain white matter: a review of imaging studies in fetuses, newborns and infants. Neuroscience 2014;276:48–71. [DOI] [PubMed] [Google Scholar]
- 9.Izbudak I, Grant PE. MR imaging of the term and preterm neonate with diffuse brain injury. Magn Reson Imaging Clin N Am 2011;19(4):709–731, vii. [DOI] [PubMed] [Google Scholar]
- 10.Dennis EL, Babikian T, Giza CC, Thompson PM, Asarnow RF. Diffusion MRI in pediatric brain injury. Childs Nerv Syst 2017;33(10):1683–1692. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Galloway NR, Tong KA, Ashwal S, Oyoyo U, Obenaus A. Diffusion-weighted imaging improves outcome prediction in pediatric traumatic brain injury. J Neurotrauma 2008;25(10):1153–1162. [DOI] [PubMed] [Google Scholar]
- 12.Ou Y, Zöllei L, Retzepi K, et al. Using clinically acquired MRI to construct age-specific ADC atlases: Quantifying spatiotemporal ADC changes from birth to 6-year old. Hum Brain Mapp 2017;38(6):3052–3068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Guimond A, Meunier J, Thirion JP. Average Brain Models: A Convergence Study. Comput Vis Image Underst 2000;77(2):192–210. [Google Scholar]
- 14.Fonov V, Evans AC, Botteron K, et al. Unbiased average age-appropriate atlases for pediatric studies. Neuroimage 2011;54(1):313–327. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Ou Y, Sotiras A, Paragios N, Davatzikos C. DRAMMS: Deformable registration via attribute matching and mutual-saliency weighting. Med Image Anal 2011;15(4):622–639. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Ou Y, Akbari H, Bilello M, Da X, Davatzikos C. Comparative evaluation of registration algorithms in different brain databases with varying difficulty: results and insights. IEEE Trans Med Imaging 2014;33(10):2039–2065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ou Y, Gollub RL, Retzepi K, et al. Brain extraction in pediatric ADC maps, toward characterizing neuro-development in multi-platform and multi-institution clinical images. Neuroimage 2015;122:246–261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ou Y, Zöllei L, Da X, et al. Field of View Normalization in Multi-Site Brain MRI. Neuroinformatics 2018;16(3-4):431–444. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Rohlfing T, Zahr NM, Sullivan EV, Pfefferbaum A. The SRI24 multichannel atlas of normal adult human brain structure. Hum Brain Mapp 2010;31(5):798–819. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Ou Y, Weinstein SP, Conant EF, et al. Deformable registration for quantifying longitudinal tumor changes during neoadjuvant chemotherapy. Magn Reson Med 2015;73(6):2343–2356. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Mukherjee P, McKinstry RC. Diffusion tensor imaging and tractography of human brain development. Neuroimaging Clin N Am 2006;16(1):19–43, vii. [DOI] [PubMed] [Google Scholar]
- 22.Watanabe M, Sakai O, Ozonoff A, Kussman S, Jara H. Age-related apparent diffusion coefficient changes in the normal brain. Radiology 2013;266(2):575–582. [DOI] [PubMed] [Google Scholar]
- 23.Boyer AC, Gonçalves LF, Lee W, et al. Magnetic resonance diffusion-weighted imaging: reproducibility of regional apparent diffusion coefficients for the normal fetal brain. Ultrasound Obstet Gynecol 2013;41(2):190–197. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Cannie M, De Keyzer F, Meersschaert J, et al. A diffusion-weighted template for gestational age-related apparent diffusion coefficient values in the developing fetal brain. Ultrasound Obstet Gynecol 2007;30(3):318–324. [DOI] [PubMed] [Google Scholar]
- 25.Klose U, Batra M, Nägele T. Age-dependent changes in the histogram of apparent diffusion coefficients values in magnetic resonance imaging. Front Aging Neurosci 2013;5:78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Ratnarajah N, Rifkin-Graboi A, Fortier MV, et al. Structural connectivity asymmetry in the neonatal brain. Neuroimage 2013;75:187–194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Liu Y, Balériaux D, Kavec M, et al. Structural asymmetries in motor and language networks in a population of healthy preterm neonates at term equivalent age: a diffusion tensor imaging and probabilistic tractography study. Neuroimage 2010;51(2):783–788. [DOI] [PubMed] [Google Scholar]
- 28.Dubois J, Hertz-Pannier L, Dehaene-Lambertz G, Cointepas Y, Le Bihan D. 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 2006;30(4):1121–1132. [DOI] [PubMed] [Google Scholar]
- 29.Liu W, Mao Y, Wei D, et al. Structural Asymmetry of Dorsolateral Prefrontal Cortex Correlates with Depressive Symptoms: Evidence from Healthy Individuals and Patients with Major Depressive Disorder. Neurosci Bull 2016;32(3):217–226. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Shaw P, Lalonde F, Lepage C, et al. Development of cortical asymmetry in typically developing children and its disruption in attention-deficit/hyperactivity disorder. Arch Gen Psychiatry 2009;66(8):888–896. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Yücel M, Stuart GW, Maruff P, et al. Paracingulate morphologic differences in males with established schizophrenia: a magnetic resonance imaging morphometric study. Biol Psychiatry 2002;52(1):15–23. [DOI] [PubMed] [Google Scholar]
- 32.Eyler LT, Pierce K, Courchesne E. A failure of left temporal cortex to specialize for language is an early emerging and fundamental property of autism. Brain 2012;135(Pt 3):949–960. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Rice D, Barone S Jr. Critical periods of vulnerability for the developing nervous system: evidence from humans and animal models. Environ Health Perspect 2000;108(Suppl 3):511–533. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Franceschini MA, Thaker S, Themelis G, et al. Assessment of infant brain development with frequency-domain near-infrared spectroscopy. Pediatr Res 2007;61(5 Pt 1):546–551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Choe MS, Ortiz-Mantilla S, Makris N, et al. Regional infant brain development: an MRI-based morphometric analysis in 3 to 13 month olds. Cereb Cortex 2013;23(9):2100–2117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Goergen SK, Ang H, Wong F, et al. Early MRI in term infants with perinatal hypoxic-ischaemic brain injury: interobserver agreement and MRI predictors of outcome at 2 years. Clin Radiol 2014;69(1):72–81. [DOI] [PubMed] [Google Scholar]
- 37.Ozturk A, Sasson AD, Farrell JA, et al. Regional differences in diffusion tensor imaging measurements: assessment of intrarater and interrater variability. AJNR Am J Neuroradiol 2008;29(6):1124–1127. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Jaimes C, Ou Y, Shih J et al. Apparent Diffusion Coefficient Z-score Maps Compared to Normative Atlas in Hypoxic Ischemic Encephalopathy. ASNR 53rd Annual Meeting & The Foundation of the ASNR Symposium; 2015; Chicago, IL, USA. 2015. [Google Scholar]
- 39.Pinto ALR, Ou Y, Sahin M, Grant PE. Quantitative Apparent Diffusion Coefficient Mapping May Predict Seizure Onset in Children With Sturge-Weber Syndrome. Pediatr Neurol 2018;84:32–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Kıvrak AS, Paksoy Y, Erol C, Koplay M, Özbek S, Kara F. Comparison of apparent diffusion coefficient values among different MRI platforms: a multicenter phantom study. Diagn Interv Radiol 2013;19(6):433–437. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.