Abstract

To date there is little information about brain development during infancy and childhood, although several quantitative studies have shown volume changes in adult brains. We performed three-dimensional magnetic resonance imaging (3D-MRI) in 28 healthy children aged 1 month to 10 years. We examined the volumes of whole brain and frontal and temporal lobes with an advanced method for segmenting images into gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) compartments. Growth spurts of whole brain and frontal and temporal lobes could be seen during the first 2 years after birth. During this period the frontal lobes grew more rapidly than the temporal lobes, the right–left asymmetry was more noticeable in the temporal lobes than in the frontal lobes and the increase in GM was larger than that in WM in the temporal lobes. Subsequently, WM volume increased at a higher rate than GM volume throughout childhood. Quantitative information on normal brain development may play a pivotal role in clarifying brain neurodevelopmental abnormalities.

Introduction

We daily observe the rapidly expanding behavioral repertoire of infants and young children and the corresponding divergence of cognitive, motor and neurological functions. This is thought to reflect normal brain development, yet little is known about neuroanatomical changes in humans that may provide substrates for this process.

MR-based brain volumetric analyses have been established as morphometric techniques, with anatomical and clinical utility (Jernigan and Tallal, 1990; R.C. Gur et al., 1991; Filipek et al., 1994). Quantitative neuroanatomical data in healthy adults have been reported previously, using the volumetric method (Gur et al., 1999), and there have been major recent advances in volumetric analysis of brain changes from childhood to adulthood (Reiss et al., 1996; Giedd et al. 1996, 1999; Lange et al., 1997; Hüppi et al., 1998; Paus et al., 1999; Sowell et al., 1999a,b). Reiss et al. performed volumetric analyses on 85 subjects between the ages of 5 and 17 years and described little change in total cerebral volume after 5 years (Reiss et al., 1996). However, more recent studies examining the age range of 5 to adulthood reported some decrease in gray matter and increase in white matter during the transition from childhood to adolescence and through adulthood (Jernigan and Talla, 1990; Lange et al., 1997; Hüppi et al., 1998). Regional specificity for these changes has been examined using statistical parametric mapping and these studies have indicated that these changes were more pronounced in frontal brain regions (Giedd et al., 1999; Sowell et al., 1999b). Other studies focused on more specific brain regions, such as prefrontal cortex (Kanemura et al., 1999), hippocampal formations and temporal lobes (Utsunomiya et al., 1999).

Only one volumetric MRI study included infants younger than 2 years (Pfefferbaum et al., 1994), although histological studies (Dobbing and Sand, 1973) indicated substantial changes in the early months of life. Pfefferbaum et al. evaluated MRI scans of 88 subjects aged 3 months to 30 years (mean 14 ± 7 years) for gender differences and age-related changes in neuroanatomy (Pfefferbaum et al., 1994). They reported a marked increase in intracranial volume between 3 months and 10 years. However, infants were only a small subgroup and the study presented values for six slices and not the total volumes of the whole brain.

The purpose of this study was to investigate MRI data of normal infants (birth to 2 years of age) and children (ages 2–10 years) using advanced techniques for segmenting brain images into gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) compartments. This method permits the establishment of normative data, which can help evaluate abnormalities in brain development. We focused on infancy since the infantile brain can change dramatically during its maturation process and so far few MRI studies have reported on infancy. We examined whole brain volumes and the frontal and temporal lobes.

Materials and Methods

Subjects

Participants included 28 healthy preadolescent Japanese (14 boys and 14 girls), including infants (aged 1 month to 2 years) and children (aged 2–10 years). They were recruited from University hospital staff and siblings of ambulatory child patients. Their heights and weights were all in the normal range and there were no significant sex differences for height and weight. All subjects had normal neurological development and had no abnormal findings in routine MRI studies. Written informed consent was obtained from parents after the purpose and all procedures of the study were fully explained. Nineteen subjects under 6 years of age were sedated with monosodium trichorethyl phosphate syrup (0.5–1.0 ml/kg). This drug was administered once only and for five children who did not fall asleep after the single dose the study was discontinued.

Because the image analysis software used in our study had been previously used only in adult American samples, six healthy Japanese adult volunteers (three males and three females, mean age 21.3 years, range 19–26 years) were included for comparison with infants and children in this study. They had no history of neurological, psychiatric or medical illness that could potentially affect brain function. Informed consent was obtained from adult subjects and the infants' parents before their participation in the study.

MRI Scan Acquisition

The MRIs were acquired on a 1.5 T Magnetom Vision (Siemens, Erlangen, Germany). Axial images were obtained using a fast low angle shot gradient refocused 3-dimensional sequence with the following parameters: flip angle = 35°, repetition time (TR) = 35 ms, echo time (TE) = 6 ms, nex = 1. The image obtained was T1-weighted with a field of view of 256 mm and a matrix size of 256 × 256 and the entire scan was obtained in ~15 min. The slice thickness was 1.0 mm and between 140 and 170 contiguous slices were obtained in each case.

Image Processing

Each acquisition was transferred to an online UNIX workstation (SPARC20; Sun Microsystems). All image processing was performed using a semi-automated software package (Kohn et al., 1991) (Fig. 1). Brain volume was first extracted by an algorithm (Yan and Karp, 1994a) with optimal thresholding, morphological operations and the Chamfer distance (Borgefors, 1986). This algorithm serves as a pre-segmentation procedure to estimate and extract a whole 3-dimensional brain volume from the MR image before segmentation is performed. An adaptive Bayesian algorithm was then applied to segment 3-dimensional MRIs into three tissue types: GM, WM and CSF (Yan and Karp, 1994b, 1995). The algorithm models MRIs as collections of regions with slowly varying intensity plus white Gaussian noise. Tissue type is modeled by a Markov random field with the 3-dimensional second order neighborhood system, where different potentials are used for the in-plane and axial directions to account for anisotropical images. This model is essential for accurate segmentation because it incorporates spatial integration among adjacent label voxels, which reduces degradation due to a poor signal-to-noise ratio and feature contrast. We introduce a cubic B-spline function to model the slowly varying mean intensity of each tissue type through the least squares fitting technique. The spline is desirable to overcome ‘shading’ effects and reduces bias against small isolated regions, such as sulcal CSF. Combining spline representation and adaptation makes the segmentation more accurate and robust (Gur et al., 1999).

Manual Delineation of Regions of Interest

Manual delineation of cerebral hemispheres and CSF was based on the standard guidelines below (Gur, R.C. et al., 1991; Gur, R.E. et al., 1991; Mozley et al., 1994) and frontal and temporal lobar volumes were calculated using previously described procedures (Cowell et al., 1994; Turetsky et al., 1995).

Cerebral Hemispheres and CSF

Collection of data in the axial plane required neuroanatomical knowledge to separate the supratentorial from infratentorial compartments. All supratentorial slices were analyzed and the infratentorial CSF and tissue were excluded by placing a boundary around the posterior fossa on each slice. The caudal brain stem structures below the level of the cerebral peduncles of the midbrain were excluded. The hypothalamic and chiasmatic cisternae were retained, but the pituitary, the carotid cisterna, the ambient cisterna and the quadrigeminal plates were excluded. Depending on head position, the following was encountered. (i) Occipital lobes often projected onto the same section as the cerebellar hemispheres and brain stem. Because the tentorium slopes, the margins had to be reconciled and CSF in the superior cerebellar and quadrigeminal cisternae had to be subtracted. (ii) The sella turcica was excluded since enlargement of the CSF space in and around the pituitary gland depended on the intactness of the diaphragm sellae. However, CSF in the chiasmatic cisterna was included. (iii) Other anatomical variables included the uppermost portion of the midbrain and the cisternae anterior to it. A line was drawn that connected the two cerebral peduncles with the basilar artery and the brain stem posterior to that was excluded. The most superior portion of the midbrain and the CSF in the chiasmatic cisterna anterior to this, along with structures of the hypothalamus (including the mammillary bodies, tuber cinereum and infundibular stalk and optic chiasma), were included.

Frontal and temporal lobes

In inferior-most slices the temporal lobe did not share common lateral or anterior borders with other structures and was easily outlined. The pons and cerebellum formed the posterio-medial temporal lobe border. At the level of the midbrain the borders of the frontal lobe were along the inter-hemispheric fissure and followed the middle cerebral artery through the suprasellar cisterna. The temporal lobe was separated from adjacent frontal regions by the sylvian fissure. The amygdala and hippocampus were included within the temporal lobe and midbrain structures were excluded. A line extending from the anterior tip of the contralateral cerebral peduncle to the antero-medial tip of the cerebellum (Fig. 21b) defined the posterior temporal lobe. Above the mamillary bodies, a horizontal line that extended from the anteromedial aspect of the sylvian fissure to the midline (Fig. 22b) defined the posterior border of the frontal lobe. The medial borders of the temporal lobe were the sylvian fissure and diencephalon structures. A horizontal line from the posterior tip of the posterior fossa to the lateral cortical perimeter delineated the posterior temporal lobe. The posterior border of the frontal lobe, for remaining superior slices, was delineated on the slice immediately inferior to the splenium of the corpus callosum (Fig. 23b). The horizontal line defined by the anterior aspect of the caudate was projected from this slice onto all superior slices to form the posterior boundary. This slice also denoted the superior extent of the temporal lobes.

We have added a new method of delineation to include more frontal lobar regions, such as the motor area, which were excluded in the original method by Cowell et al. (Cowell et al., 1994). Namely, above all slices where the lateral ventricles were no longer visible, regions anterior to a line connecting the most medial and lateral points of the central sulcus were calculated as frontal area (Fig. 24b). The sagittal slices were adapted to locate the central sulcus (Fig. 3a).

Reliability of Regional Volumetric Measurements

Inter-rater reliability was examined in a sample of 10 randomly selected brain scans (five infants or children and five adult volunteers) analyzed by two raters (J.M. and M.M.). The intra-class correlation for total cerebral volumes ranged from 0.95 to 0.98 and those of frontal and temporal lobar volumes from 0.87 to 0.90. The intra-rater reliability was also examined in the same 10 scans analyzed by one of these raters (J.M.) and the correlation for total cerebral, frontal and temporal lobar volumes ranged from 0.94 to 0.97. The rater (J.M.) then completed the analysis on the remaining scans.

Statistical Analyses

The relationship between the whole brain volume compartment (GM, WM and CSF) and age was modeled using linear regression with fractional polynomials (Royston and Altman, 1994) of age as the covariates. This approach finds the combination of fractional powers of age that best describe each of the relationships. Models with up to three fractional powers were explored. For all three models (GM, WM and CSF) one fractional power of age was sufficient.

Using any curve fitting approach to simultaneously examine sex differences and age effects on both hemispheres and for the frontal and temporal sub-regions, including examination of tissue types and their interactions, would suffer from extremely low power with the present sample size. Therefore, to examine age effects and sex differences on hemispheric whole brain and regional volumes we used age as a grouping factor by dividing the participants into infants (age < 2 years) and children (age > 2 years) and performed a multivariate repeated measures analysis of variance (MANOVA) (Maxwell and Delaney, 1990). Sex was another grouping factor, and GM versus WM and hemisphere were repeated measures (within-group) factors. There were 13 subjects in the infant (seven males, age 9.6 ± 6.0 months, height 70.1 ± 7.8 cm, weight 8.1 ± 1.9 kg; six females, age 11.7 ± 6.1 months, height 71.9 ± 6.2 cm, weight 8.5 ± 1.6 kg; male versus female, age t = 0.62, df = 11, P = 0.545, ns, height t = 0.44, df = 11, P = 0.671, ns, weight t = 0.42, df = 11, P = 0.683, ns) and 15 in the child group (seven boys, age 63.7 ± 33.7 months, height 107.2 ± 17.9 cm, weight 18.5 ± 7.8 kg; eight girls, age 90.9 ± 35 months, height 120.5 ± 17.2 cm, weight 23.8 ± 6.1 kg; boys versus girls, age t = 1.52, df = 13, P =0.151, ns, height t = 1.47, df = 13, P = 0.166, ns, weight t = 1.47, df = 13, P = 0.166, ns). Statistical computations were performed with STATA (STATA Corp., College Station, TX) and SAS software (SAS Institute Inc., Cary, NC).

Results

The distributions by age of the three tissue compartments as well as the best fitting models are shown in Figure 4. For GM the best model was 

\[GM\ {=}\ 653\ {\mbox{--}}\ 33.54\ {\times}\ \mathit{Z}\]
where 
\[\mathit{Z}\ {=}\ 1/{\surd}(age/100)\ {\mbox{--}}\ 1.462\]
Since Z decreases towards 0 with increasing age, the model predicts a maximum GM of 653, which is slightly higher than the average for the adults in this study. Although this is a prediction outside the range of the observed ages, which often yields unstable estimates, the prediction is in line with the observed adult measurements. For WM the best model was 
\[WM\ {=}\ 298.28\ {+}\ 72.74\ {\times}\ \mathit{Z}\]
where 
\[\mathit{Z}\ {=}\ ln(age/100)\ {+}\ 0.7596\]
In this case Z increases with age to a maximum which could not be detected within the present age range. For CSF the best model was 
\[CSF\ {=}\ 149.88\ {+}\ 13.90\ {\times}\ \mathit{Z}\]
with 
\[\mathit{Z}\ {=}\ ln(age/100)\ {+}\ 0.7596\]
Here too Z increases with age. GM and WM increase at a faster rate for the first 2 years than the following years studied. Furthermore, the rate is higher for WM than for GM and WM continues to increase into late childhood. CSF shows a steady volume throughout this period.

The (age group × sex × GM versus WM × hemisphere) MANOVA for the whole brain values, presented in Figure 5a, showed a main effect of age group [F(1,24) = 22.48, P < 0.0001], with children having overall higher volumes than infants. There were also main effects for the within-group (repeated measures) factors. GM volumes were higher than WM [F(1,24) = 1012.76, P < 0.0001] and right hemispheric volumes were overall higher than left [F(1,24) = 84.18, P < 0.0001]. However, these main effects were qualified by higher order interactions. GM versus WM × age [F(1,24) = 9.39, P < 0.0001] indicated that the difference between the age groups was smaller for GM than for WM (Fig. 5a). A GM versus WM × hemisphere interaction [F(1,24) = 28.26, P < 0.0001] indicated that the main effect of higher right hemispheric volumes came entirely from WM, while GM is symmetrical and even slightly larger on the left (Fig. 5a). There were no main effects or interactions with gender and no other main effects or interactions approached significance.

The results of the MANOVA applied to the frontal and temporal volumes (age group × sex × GM versus WM × hemisphere × frontal versus temporal) are summarized in Table 1. As can be seen in Figure 5b,c, the main effect for age group demonstrates that children had overall higher volumes, the GM versus WM effect indicates higher GM values and the main effect for hemisphere reflects higher values in the right than the left hemisphere across regions, groups and compartments. The main effect of frontal versus temporal volumes indicates higher volumes in the frontal than temporal lobes. These main effects were qualified by several higher order interactions. The frontal versus temporal × age interaction indicated greater differences between infants and children in the frontal than temporal lobe. A GM versus WM × frontal versus temporal interaction indicated a larger difference between GM and WM volume for the frontal than for the temporal lobe. Finally, a GM versus WM × hemisphere × frontal versus temporal × age group interaction indicated that the difference between babies and children was relatively larger for right frontal and left temporal GM (Fig. 5b,c). This asymmetry is quite subtle in absolute terms and the statistical significance is attributable to its consistency.

When compared with our limited data on adult brains it seems that GM volumes, for the whole brain and for both frontal and temporal lobes, are as high for children older than 2 years as they are in adults. Indeed, for the whole brain and frontal lobes the values for children over 2 years were higher than for adults, supporting evidence for ‘pruning’ during adolescence (Giedd et al., 1996, Giedd et al., 1999; Paus et al., 1999). In contrast, the volume of WM is lower in both infants and children compared with adults. Examining the lobar values, it appears that by childhood WM has nearly completed its development in the temporal lobe. However, in spite of the greater increase in WM volume from infancy to childhood in the frontal than in the temporal lobe, frontal WM is still lower in children compared with adults.

Discussion

Although many investigators have described changes in cerebral morphology during the adult lifespan, there are relatively few imaging reports describing quantitative in vivo brain development in children and even fewer in healthy infants. Our results show that the volumes of whole brain and frontal and temporal lobes increase rapidly during the first 2 years after birth, followed by a more gradual expansion, predominantly restricted to WM. The increase in brain volume agrees with previous autopsy findings (Dobbing and Sand, 1973). Specifically, the autopsy study observed a transient period of brain growth spurt in early childhood. Dobbing and Sand suggested that the fastest post-natal myelination occurs over the first 2 years after birth and that brain structures drastically change their biochemical composition, such as DNA, cholesterol and water content (Dobbing and Sand, 1973). Moreover, several authors (Barkovich et al., 1988; Hayakawa et al., 1990) have documented changes in MR images corresponding to myelination in the neonate and infant. Thus, the appearance of brain structures is similar to that of adults by the end of 2 years of age and all major fiber tracts can be identified in brains of 3-year-old children.

Several studies have suggested that cortical GM and WM volumes present dynamic changes throughout childhood. Pfefferbaum et al. observed that cortical GM volume peaked at age 4 years and then decreased throughout the lifespan (Pfefferbaum et al., 1994). On the other hand, cortical WM volume increased during the first decade of life and remained stable thereafter. These different patterns of cortical GM and WM volume growth seem to be quite meaningful. In particular, the critical point of the rise and fall in cortical GM volume in their MRI study is consistent with findings such as the change in synaptic counts in post-mortem brains (Huttenlocher et al., 1982), the amplitude of the sleep slow wave on electroenceph-alograms (Feinberg, 1983) and brain metabolism and blood flow studied by positron emission tomography (PET) (Chugani et al., 1987) and single photon emission computed tomography (SPECT) (Chiron et al., 1992). Moreover, this age-related change in GM volume possibly includes the maturational processes of neuronal pruning and cell death (Pfefferbaum et al., 1994; Giedd et al., 1999; Paus et al., 1999; Sowell et al., 1999b). According to previous histological studies (Yakovlev et al., 1967), newborn infant brains showed absence of myelinated fibers, while brains of full-term infants clearly showed myelinated WM. The MRI can visualize this process as known from histology. Several authors have documented the changes in MR corresponding to myelination of the WM in infancy (Barkovich et al., 1988; Hayakawa et al., 1990; Jernigan and Tallal, 1990). Most conclude that a normal adult appearance can be seen at age 2 years and all major fiber tracts can be identified by age 3 years.

In our study the absolute GM and WM volumes showed increases during infancy. Both increased rapidly during the first 2 years, followed by a more gradual expansion. The change in GM volumes seems somewhat different from the previous results of Pfefferbaum et al. (Pfefferbaum et al., 1994). They found that cortical GM volume peaked around age 4 years and decreased thereafter, while our results show an increase of GM volume without a peak. However, Pfefferbaum et al. (Pfefferbaum et al., 1994) examined subjects from 3 months to 30 years of age, while our subjects were all under 10 years and 54% of them were under 2 years. Furthermore, since there have been no published reports that analyzed total (cortical and subcortical) GM and total (myelinated and unmyelinated) WM, we could not directly compare our data with published data of cortical tissue volumes only. The early growth spurt in WM volume was consistent with that of developmental myelination.

Asymmetry

Volume asymmetries have been described in multiple areas of the human brain during fetal life and adulthood (Chui et al., 1980; Bear et al., 1986; R.C. Gur et al., 1991; Cowell et al., 1994; Filipek et al., 1994; Caviness et al., 1996; Giedd et al., 1996). Gur et al. reported higher right hemispheric brain volume for the whole brain in 69 adult volunteers aged 18–80 years (R.C. Gur et al., 1991). Most other studies in adults were consistent with this result. There are also a few studies of volumetric analyses in children, which noted the asymmetry of cerebral volume. Reiss et al. reported that there was no significant lateral difference in hemispheric volumes in 85 children aged 5–17 years (Reiss et al., 1996). However, Giedd et al. reported that the right hemisphere was consistently larger than the left in 104 subjects aged 4–18 years (Giedd et al., 1996). There was no significant effect of age in these two studies. Our results are consistent with these reports in showing slightly higher overall values in the right than the left in children under 10 years of age, although the effect was seen in WM and not in GM.

Several studies have reported asymmetry of the temporal lobes or hippocampal formations, which are associated with seizure disorders and dysfunction of memory and learning. Jack et al. evaluated the right–left asymmetry of temporal lobe volumes in adults and concluded that both the anterior temporal lobe and hippocampal formation were significantly larger in the right hemisphere than in the left (Jack et al., 1989). Such asymmetries were already identifiable in neonates and infants (Utsunomiya et al., 1999), suggesting that the presence of temporal lobe asymmetry exists before the development of speech and language function. Kanemura et al. reported that the prefrontal cortex presented a growth spurt between the ages of 8 and 15 years (Kanemura et al., 1999). However, they did not examine asymmetry of either prefrontal or frontal lobe volume.

Limitations

Our segmentation method has thus far been validated for adult brains more than 17 years old (Gur et al., 1999). We applied this method to infants and children and ascertained the feasibility of this approach. Although our method is useful to analyze the human brain volume in vivo, several problems remain. Firstly, the method is likely to fail in calculating correct brain volume because of the lower density of GM of infants in T1-weighted images compared with adults (Barkovich et al., 1988; Hayakawa et al., 1990). Indeed, the GM volume of our subject aged 1 month was partly calculated as CSF from visual continuity of the regional distribution and we added that volume as GM. Furthermore, the data need to be interpreted carefully for the following reason. As pointed out by Barkovich et al. (Barkovich et al., 1988), in T1-weighted images from infants <4 weeks of age large regions of unmyelinated WM actually have signal values lower than GM. In children between 1 and 6 months of age GM and unmyelinated WM have signal values that become isointense and are indistinguishable from one another. Tissue classification schemes where all voxels in an image must be classified into three distinct categories do not capture all the variability in signal values that occur across the age range studied. The tissue that segments as GM within the centrum semiovale and more peripheral regions where axonal fibers are found in 3- and 6-month-olds is not GM. The composition of tissue that results in a GM signal value in the cortex in children (perhaps more unmyelinated axonal fibers) might be very different from the composition of tissue in the cortical ribbon in young adults. Dynamic and complex cellular events are occurring that account for the changes in signal value between birth and infancy under 6 months old. When interpreting the results we need to consider that GM under 6 months old might include both GM and unmyelinated WM. Secondly, in the automated segmentation of this analysis the brain tissue selected as GM included both cortical and subcortical GM and the tissue as WM also included both myelinated and unmyelinated WM. As for classification of GM (cortical and subcortical), it can be resolved by delineation of subcortical regions such as basal ganglia and analysis of separate volume. Regarding myelination related to WM, we are considering augmentation of the present approach with one that includes MR signal intensity.

The present study analyzed the volumes of the whole brain and frontal and temporal lobes. We did not include other regions, such as the parietal and occipital lobes, because previous studies of volumetry by MRI in adults have concentrated on these regions. However, it would be informative to investigate differences in growth patterns in parietal and occipital lobes in children. Kanemura et al. showed that the prefrontal volume increased rapidly between 8 and 14 years in a 3D-MRI study of 13 cases aged 5 months to 14 years (Kanemura et al., 1999). The volume of the hippocampal formations increased sharply until the age of 2 years (Utsunomiya et al., 1999). This early growth spurt coincides with our findings for the temporal lobe. This is consistent with the findings from brain metabolism and blood flow analyses by PET (Chugani et al., 1987) and SPECT (Chiron et al., 1992). These results suggest that the differences in growth pattern and speed are directly associated with age-related neuro-physiological development. In future, quantitative information on segmented volumes can be correlated with behavioral measures to document the association between neural development and the behavioral repertoire. Such studies can play a pivotal role in clarifying brain abnormalities in infants or children with suspected brain dysfunction.

Our study failed to observe any significant variation between the sexes in infants and children. In post-mortem and imaging studies of adult brains the authors consistently point out gender differences in total brain volume, with adult males having on average a 10% greater volume than females (Dekaban and Sadowsky, 1978; Ho et al., 1980; Filipek et al., 1994; Pfefferbaum et al., 1994; Blatter et al., 1995; Gur et al., 1999), while women have a higher proportion of GM relative to cranial volume (Gur et al., 1982, 1999). Reiss et al. reported that the first appearance of gender differences in brain volume had already occurred in children as young as 5 years of age (Reiss et al., 1996). In our study we detected no significant gender differences, with the testing possibly hampered by low power as a result of the small sample size. Furthermore, the average age of girls was higher than that of boys in both age groups (infants, boys 9.6, girls 11.6 months; children, boys 63.7 and girls 90.9 months). While these differences are not statistically significant (see Data Analysis), even subtle age differences at a time of rapid physical development seem to have obscured sex differences in brain volumes in this small sample. More children of both genders in narrower age bands are needed to establish the first appearance of gender differences and evaluate the possible existence of sex differences in the pattern of growth spurts in brain volumes.

Finally, this study is cross-sectional and thus cannot be used to establish conclusively the time course of volume changes in early infancy through childhood. It is nonetheless the first report including quantitative volumetric analysis of infants and toddlers under 2 years of age. They illustrate the feasibility of this approach to understanding the early stages of neurodevelopment and encourage larger scale and longitudinal studies.

Notes

We acknowledge Professor H. Seto of the Department of Radiology, Toyama Medical and Pharmaceutical University, for his pertinent suggestions and cooperation.

Address correspondence to Dr Mié Matsui, Department of Psychology, School of Medicine, Toyama Medical and Pharmaceutical University, 2630 Sugitani, Toyama 930-0194, Japan. Email: mmatsui@ms.toyama-mpu.ac.jp.

Table 1

Summary table for the MANOVA on regional values

Source Hotelling trace df P value 
GW, gray versus white matter contrast; FT, frontal versus temporal lobe contrast. 
Age group  3.739 7.945 8,17 0.0002 
Sex  0.378 0.803 8,17 0.6086 
Age group × sex  0.370 0.787 8,17 0.6205 
GW 34.731 833.543 1,24 <0.0001 
GW × age group  0.072 1.738 1,24 0.1998 
GW × sex  0.028 0.679 1,24 0.4179 
GW × age group × sex  0.003 0.069 1,24 0.7953 
Hemisphere  1.408  33.802 1,24 <0.0001 
Hemisphere × age group  0.007 0.166 1,24 0.687 
Hemisphere × sex  0.005 0.110 1,24 0.7432 
Hemisphere × age group × sex  0.011 0.275 1,24 0.605 
FT 12.357 296.561 1,24 <0.0001 
FT × age group  0.514  12.328 1,24 0.0018 
FT × sex  0.025 0.596 1,24 0.4475 
FT × age group × sex  0.053 1.280 1,24 0.2692 
GW × hemisphere  0.083 1.986 1,24 0.1716 
GW × hemisphere × age group  0.012 0.291 1,24 0.5944 
GW × hemisphere × sex  0.007 0.177 1,24 0.6779 
GW × hemisphere × age group × sex  0.008 0.181 1,24 0.6744 
GW × FT  4.297 103.133 1,24 <0.0001 
GW × FT × age group  0.124 2.977 1,24 0.0973 
GW × FT × sex  0.018 0.435 1,24 0.5157 
GW × FT × age group × sex  0.038 0.910 1,24 0.3497 
Hemisphere × FT  0.086 2.061 1,24 0.1641 
Hemisphere × FT × age group  0.027 0.638 1,24 0.4322 
Hemisphere × FT × sex  0.029 0.689 1,24 0.4148 
Hemisphere × FT × age group  0.003 0.077 1,24 0.7833 
GW × hemisphere × FT  0.000 0.010 1,24 0.9199 
GW × hemisphere × FT × age group  0.451  10.834 1,24 0.0031 
GW × hemisphere × FT × sex  0.053 1.279 1,24 0.2692 
GW × hemisphere × FT × age group × sex  0.007 0.176 1,24 0.6788 
Source Hotelling trace df P value 
GW, gray versus white matter contrast; FT, frontal versus temporal lobe contrast. 
Age group  3.739 7.945 8,17 0.0002 
Sex  0.378 0.803 8,17 0.6086 
Age group × sex  0.370 0.787 8,17 0.6205 
GW 34.731 833.543 1,24 <0.0001 
GW × age group  0.072 1.738 1,24 0.1998 
GW × sex  0.028 0.679 1,24 0.4179 
GW × age group × sex  0.003 0.069 1,24 0.7953 
Hemisphere  1.408  33.802 1,24 <0.0001 
Hemisphere × age group  0.007 0.166 1,24 0.687 
Hemisphere × sex  0.005 0.110 1,24 0.7432 
Hemisphere × age group × sex  0.011 0.275 1,24 0.605 
FT 12.357 296.561 1,24 <0.0001 
FT × age group  0.514  12.328 1,24 0.0018 
FT × sex  0.025 0.596 1,24 0.4475 
FT × age group × sex  0.053 1.280 1,24 0.2692 
GW × hemisphere  0.083 1.986 1,24 0.1716 
GW × hemisphere × age group  0.012 0.291 1,24 0.5944 
GW × hemisphere × sex  0.007 0.177 1,24 0.6779 
GW × hemisphere × age group × sex  0.008 0.181 1,24 0.6744 
GW × FT  4.297 103.133 1,24 <0.0001 
GW × FT × age group  0.124 2.977 1,24 0.0973 
GW × FT × sex  0.018 0.435 1,24 0.5157 
GW × FT × age group × sex  0.038 0.910 1,24 0.3497 
Hemisphere × FT  0.086 2.061 1,24 0.1641 
Hemisphere × FT × age group  0.027 0.638 1,24 0.4322 
Hemisphere × FT × sex  0.029 0.689 1,24 0.4148 
Hemisphere × FT × age group  0.003 0.077 1,24 0.7833 
GW × hemisphere × FT  0.000 0.010 1,24 0.9199 
GW × hemisphere × FT × age group  0.451  10.834 1,24 0.0031 
GW × hemisphere × FT × sex  0.053 1.279 1,24 0.2692 
GW × hemisphere × FT × age group × sex  0.007 0.176 1,24 0.6788 
Figure 1.

The examples of segmentation (left) and each original image (right). White, gray matter (GM); gray, white matter (WM); black, cerebrospinal fluid (CSF).

Figure 1.

The examples of segmentation (left) and each original image (right). White, gray matter (GM); gray, white matter (WM); black, cerebrospinal fluid (CSF).

Figure 2.

Representative axial slices with tracing. F, frontal region; T, temporal region; c, cerebellum; d, diencephalon; p, pons; cp, cerebral peduncle; sf, sylvian fissure.

Figure 2.

Representative axial slices with tracing. F, frontal region; T, temporal region; c, cerebellum; d, diencephalon; p, pons; cp, cerebral peduncle; sf, sylvian fissure.

Figure 3.

Sagittal and coronal views of frontal and temporal lobe areas traced. a, bottom left; b, top right; c, bottom right. (a,b) Frontal lobe area; (c) temporal lobe area.

Figure 3.

Sagittal and coronal views of frontal and temporal lobe areas traced. a, bottom left; b, top right; c, bottom right. (a,b) Frontal lobe area; (c) temporal lobe area.

Figure 4.

The age-related volumetric changes of three tissue compartments (GM, WM and CSF). The volumes of both GM and WM increase rapidly during the first 2 years after birth. GM, gray matter; WM, white matter; CSF, cerebrospinal fluid.

Figure 4.

The age-related volumetric changes of three tissue compartments (GM, WM and CSF). The volumes of both GM and WM increase rapidly during the first 2 years after birth. GM, gray matter; WM, white matter; CSF, cerebrospinal fluid.

Figure 5.

Means (± SEM) of compartmental volumes (in ml) for infants (<24 months; ▾) and children (24 months; △) for: (a) whole brain; (b) frontal lobe; (c) temporal lobe. Values for the six adults are presented for comparison (♦). GM, gray matter; WM, white matter; CSF: cerebrospinal fluid; L, left hemisphere; R, right hemisphere; MO, months.

Figure 5.

Means (± SEM) of compartmental volumes (in ml) for infants (<24 months; ▾) and children (24 months; △) for: (a) whole brain; (b) frontal lobe; (c) temporal lobe. Values for the six adults are presented for comparison (♦). GM, gray matter; WM, white matter; CSF: cerebrospinal fluid; L, left hemisphere; R, right hemisphere; MO, months.

References

Barkovich AJ, Kjos BO, Jackson DE Jr, Norman D (
1988
) Normal maturation of the neonatal and infant brain: MR imaging at 1.5 T.
Radiology
 
166
:
173
–180.
Bear D, Schiff D, Saver J, Greenberg M, Freeman R (
1986
) Quantitative analysis of cerebral asymmetries: fronto-occipital correlation, sexual dimorphism and association with handedness.
Arch Neurol
 
43
:
598
–603.
Blatter DD, Bigler ED, Gale SD, Johnson SC, Anderson CV, Burnett BM (
1995
) Quantitative volumetric analysis of brain MR: normative database spanning 5 decades of life.
Am J Neuroradiol
 
16
:
241
–251.
Borgefors G (
1986
) Distance transformations in digital images.
Comput Vis Graph Image Process
 
34
:
344
–371.
Caviness VS Jr, Kennedy DN, Richelme C, Rademacher J, Filipek PA (
1996
) The human brain age 7–11 years: a volumetric analysis based on magnetic resonance images.
Cereb Cortex
 
6
:
726
–736.
Chiron C, Raynaud C, Maziere B (
1992
) Changes in regional cerebral blood flow during maturation in children and adolescents.
J Nucl Med
 
33
:
696
–703.
Chugani HT, Phelps ME, Mazziotta JC (
1987
) Positron emission tomography study of human brain functional development.
Ann Neurol
 
22
:
487
–497.
Chui HC, Damasio AR (
1980
) Human cerebral asymmetries evaluated by computed tomography.
J Neurol Psychiat
 
43
:
873
–878.
Cowell PE, Turetsky BI, Gur RC, Grossman RI, Shtasel DL, Gur RE (
1994
) Sex differences in aging of the human frontal and temporal lobes.
J Neurosci
 
14
:
4748
–4755.
Dekaban A, Sadowsky D (
1978
) Changes in brain weights during the span of human life: relation of brain weights to body heights and body weights.
Ann Neurol
 
4
:
345
–356.
Dobbing J, Sand J (
1973
) Quantitative growth and development of human brain.
Arch Dis Child
 
48
:
757
–767.
Feinberg I (
1983
) Schizophrenia caused by a fault in programmed synaptic elimination during adolescence?
J Psychiat Res
 
17
:
319
–334.
Filipek PA, Richelme C, Kennedy DN, Caviness VS Jr (
1994
) The young adult human brain: an MRI-based morphometric analysis.
Cereb Cortex
 
4
:
334
–360.
Giedd JN, Snell JW, Lange N, Rajapakse JC, Casey BJ, Kozuch PL, Vaituzis AC, Vauss YC, Hamburger SD, Kaysen D, Rapoport JL (
1996
) Quantitative magnetic resonance imaging of human brain development: ages 4–18.
Cereb Cortex
 
6
:
551
–560.
Giedd JN, Blumenthal J, Jeffries NO, Castellanos FX, Liu H, Zijdenbos A, Paus T, Evans AC, Rapoport JL (
1999
) Brain development during childhood and adolescence: a longitudinal MRI study.
Nature Neurosci
 
2
:
861
–863.
Gur RC, Gur RE, Obrist WD, Hungerbuhler JP, Younkin D, Rosen AD, Skolnick BE, Reivich M (
1982
) Sex and handedness differences in cerebral blood flow during rest and cognitive activity.
Science
 
217
:
659
–661.
Gur RC, Mozley PD, Resnick SM, Gottlieb GL, Kohn M, Zimmermann R, Herman G, Atlas S, Grossman R, Berretta D, Erwin R, Gur RE (
1991
) Gender differences in age effect on brain atrophy measured by magnetic resonance imaging.
Proc Natl Acad Sci USA
 
88
:
2845
–2849.
Gur RC, Turetsky BI, Matsui M, Yan M, Bilker W, Hughett P, Gur RE (
1999
) Sex differences in brain gray and white matter in adults: correlation with cognitive performance.
J Neurosci
 
19
:
4065
–4072.
Gur RE, Mozley PD, Resnick SM, Shtasel D, Kohn M, Zimmerman R, Herman G, Atlas S, Grossman R, Erwin R, Gur RC (
1991
) Magnetic resonance imaging in schizophrenia.
Arch Gen Psychiat
 
48
:
407
–412.
Hayakawa K, Konishi Y, Kuriyama M, Konishi K, Matsuda T (
1990
) Normal brain maturation in MRI.
Eur J Radiol
 
12
:
208
–215.
Ho KC, Roessmann U, Straumfjord JV, Monroe G (
1980
) Analysis of brain weight. 1. Adult brain weight in relation to sex, race, and age.
Arch Pathol Lab Med
 
104
:
635
–639.
Hüppi PS, Warfield S, Kikinis R, Barnes PD, Zientara GP, Jolesz FA, Tsuji MK, Volpe JJ (
1998
) Quantitative magnetic resonance imaging of brain development in premature and mature newborns.
Ann Neurol
 
43
:
224
–235.
Huttenlocher PR, de Courten C, Garey LJ, Van Der Loos H (
1982
) Synaptogenesis in human visual cortex: evidence for synapse elimination during normal development.
Neurosci Lett
 
33
:
247
–252.
Jack CR Jr, Twomey CK, Zinsmeister AR, Sharbrough FW, Petersen RC, Cascino GD (
1989
) Anterior temporal lobes and hippocampal formations: normative volumetric measurements from MR imaging in young adults.
Radiology
 
172
:
549
–554.
Jernigan TL, Tallal P (
1990
) Late childhood changes in brain morphology observable with MRI.
Dev Med Child Neurol
 
32
:
379
–385.
Kanemura H, Aihara M, Aoki S, Hatayama K, Kamiya Y, Ono C, Sada Y, Nakazawa S (
1999
) Quantitative measurement of prefrontal lobe volume on three dimensional magnetic resonance imaging scan.
No To Hattatsu
 
31
:
519
–524 (in Japanese).
Kohn MI, Tanna NK, Herman GT, Resnick SM, Mozley PD, Gur RE, Alavi A, Zimmerman RA, Gur RC (
1991
) Analysis of brain and cerebrospinal fluid volumes with MR imaging.
Radiology
 
178
:
115
–22.
Lange N, Giedd JN, Castellanos FX, Vaituzis AC, Rapaport JL (
1997
) Variability of human brain structure size: aged 4–20 years.
Psychiat Res
 
74
:
1
–12.
Maxwell WE, Delaney HD (1990) Designing experiments and analyzing data: a model comparisons approach. Belmont, CA: Wadsworth.
Mozley PD, Gur RE, Resnick SM, Shtasel DL, Richards J, Kohn M, Grossman R, Herman G, Gur RC (
1994
) Magnetic resonance imaging in schizophrenia: relationship with clinical measures.
Schizophr Res
 
12
:
195
–203.
Paus T, Zijdenbos A, Worsley K, Collins DL, Blumenthal J, Giedd JN, Rapoport JL, Evans AC (
1999
) Structural maturation of neural pathway in children and adolescents: in vivo study.
Science
 
283
:
1908
–1911
Pfefferbaum A, Mathalon DH, Sullivan EV, Rawles JM, Zipursky RB, Lim KO (
1994
) A quantitative magnetic resonance imaging study of changes in brain morphology from infancy to late adulthood.
Arch Neurol
 
51
:
874
–887.
Reiss AL, Abrams MT, Singer HS, Ross JL, Denckla MB (
1996
) Brain development, gender and IQ in children: a volumetric imaging study.
Brain
 
119
:
1763
–1774.
Royston P, Altman DG (
1994
) Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling (with discussion).
Appl Statist
 
43
:
429
–467.
Sowell ER, Thompson PM, Holmes CJ, Batth R, Jernigen TL, Toga AW (
1999
) Localizing age-related changes in brain structure between childhood and adolescence using statistical parametric mapping.
NeuroImage
 
9
:
587
–597.
Sowell ER, Thompson PM, Holmes CJ, Jernigen TL, Toga AW (
1999
) In vivo evidence for post-adolescent brain maturation in frontal and striatal regions.
Nature Neurosci
 
2
:
859
–861
Turetsky BI, Cowell PE, Gur RC, Grossman RI, Shtasel DL, Gur RE (
1995
) Frontal and temporal lobe brain volume in schizophrenia: relation to symptomatology and clinical subtype.
Arch Gen Psychiat
 
52
:
1061
–1070.
Utsunomiya H, Takano K, Okazaki M, Mitsudome A (
1999
) Development of the temporal lobe in infants and children: analysis by MR-based volumetry.
Am J Neuroradiol
 
20
:
717
–723.
Yan MXH, Karp JS (
1994
) Image registration of MR and PET based on surface matching and principal axes fitting.
Proc IEEE Med Imaging Conf
 
4
:
1677
–1681.
Yan MXH, Karp JS (
1994
) Segmentation of 3D brain MR using an adaptive K-means clustering algorithm.
Proc IEEE Med Imaging Conf
 
4
:
1529
–1533.
Yan MXH, Karp JS (1995) An adaptive bayesian approach to three-dimensional MR brain segmentation. In: Information processing in medical imaging (Bizais Y, Barillot C, Di Paola R, eds), pp. 201–213. Dordrecht: Kluwer Academic Publishers.
Yakovlev PL, Lecours AR (1967) The myelogenetic cycles of regional maturation of the brain. In: Regional development of the brain in early life (Minkowski A, ed.), pp. 3–70. Oxford: Blackwell.