Eric Courchesne, PhD Heather J. Chisum, BA Jeanne Townsend, PhD Angilene Cowles, BA James Covington, MA Brian Egaas, MS Mark Harwood, BA Stuart Hinds, MD Gary A. Press, MD Index terms: Aging Brain, growth and development, 10.91 Brain, MR, 10.121411, 10.12142 Brain, volume, 10.91 Magnetic resonance (MR), volume measurement, 10.121411, 10.12142 Radiology 2000; 216:672–682 Abbreviations: CSF 5 cerebrospinal fluid GM 5 gray matter IQ 5 intelligence quotient WM 5 white matter 1 From the Laboratory for Research on the Neuroscience of Autism, Chil- dren’s Hospital Research Center, 8110 La Jolla Shores Dr, Suite 201, La Jolla, CA 92037 (E.C., H.J.C., J.T., A.C., J.C., B.E., M.H., S.H.); the Department of Neurosciences, School of Medicine, University of California at San Diego, La Jolla (E.C., J.T.); and the Kaiser Per- manente Hospital, San Diego, Calif (G.A.P.). Received April 5, 1999; revi- sion requested June 7; final revision received December 7; accepted De- cember 21. This work is supported by funds from NINDS (2-RO1-NS-19855) awarded to Eric Courchesne and NINDS (5RO1-NS-34155) award to Jeanne Townsend. Address corre- spondence to E.C. (e-mail: ecourchesne @ucsd.edu). © RSNA, 2000 Author contributions: Guarantor of integrity of entire study, E.C.; study concepts, E.C., G.A.P.; study design, E.C.; definition of intellectual content, E.C.; literature research, E.C., H.J.C.; clinical studies, B.E., S.H.; data acquisition, J.T., J.C., E.C., A.C.; data analysis, H.J.C., A.C., B.E.; statistical analysis, J.T., M.H.; manuscript prepara- tion, E.C., H.J.C., G.A.P., J.T.; manu- script editing, E.C., G.A.P., H.J.C.; manuscript review, E.C., G.A.P. Normal Brain Development and Aging: Quantitative Analysis at in Vivo MR Imaging in Healthy Volunteers1 PURPOSE: To quantitate neuroanatomic parameters in healthy volunteers and to compare the values with normative values from postmortem studies. MATERIALS AND METHODS: Magnetic resonance (MR) images of 116 volunteers aged 19 months to 80 years were analyzed with semiautomated procedures vali- dated by means of comparison with manual tracings. Volumes measured included intracranial space, whole brain, gray matter (GM), white matter (WM), and cere- brospinal fluid (CSF). Results were compared with values from previous postmortem studies. RESULTS: Whole brain and intracranial space grew by 25%–27% between early childhood (mean age, 26 months; age range, 19–33 months) and adolescence (mean age, 14 years; age range, 12–15 years); thereafter, whole-brain volume decreased such that volunteers (age range, 71–80 years) had volumes similar to those of young children. GM increased 13% from early to later (6–9 years) child- hood. Thereafter, GM increased more slowly and reached a plateau in the 4th decade; it decreased by 13% in the oldest volunteers. The GM-WM ratio decreased exponentially from early childhood through the 4th decade; thereafter, it gradually declined. In vivo patterns of change in the intracranial space, whole brain, and GM-WM ratio agreed with published postmortem data. CONCLUSION: MR images accurately depict normal patterns of age-related change in intracranial space, whole brain, GM, WM, and CSF. These quantitative MR imaging data can be used in research studies and clinical settings for the detection of abnormalities in fundamental neuroanatomic parameters. Knowledge of the course of normal brain development and aging provides the foundation for the recognition of pathologic brain development and decline. Quantitative magnetic resonance (MR) imaging studies have provided such information for a number of specific brain structures (eg, cerebrum, corpus callosum, cerebellum, brainstem, hippocampus) (1–18) but only for limited age windows and never for the entire life span from early childhood to late adulthood. Moreover, several fundamental indicators of normal brain development—volumes of intracranial space, whole brain, gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF)—have yet to be thoroughly documented by using contemporary, quantitative, and radiologic methods in a large group of healthy young children. Even in the postmortem literature, there is a paucity of information on the parameters of brain development. For example, one study of postmortem specimens (19) provided measures of intracranial volume from birth through the age of 20 years, but for only males (see table 115 in the article by Blinkov and Glezer [19]), while another (20) provided such measures separately for both boys and girls, but for only the narrower age group of 2.5–8 years. In another study (21), intracranial volume was estimated from measures of the Neuroradiology 672 external height, width, and circumfer- ence of the skull. Since the routine method for measuring brain weight at autopsy includes the weight of CSF within the leptomeninges and the weight of parenchyma, in the postmortem liter- ature, age-related changes in CSF may confound estimates of developmental changes in parenchyma. Furthermore, postmortem swelling and/or shrinkage due to fixation may introduce artifacts into measures of fresh or fixed postmor- tem brains (19). Both postmortem and in vivo imaging studies (13,22–28) have demonstrated that, with advancing age, brain weight or volume, GM, and GM-WM ratios de- crease and CSF volume increases. How- ever, the timing of such age-related de- creases and increases is uncertain. Transitions from growth-related in- creases to age-related decreases in several fundamental neuroanatomic parameters have been incompletely documented with in vivo MR imaging because devel- opmental studies have commonly in- cluded samples of only children and ad- olescents or young adults, while aging studies have commonly included sam- ples of only adults. Comparison of the neuroanatomic volume changes in devel- opmental studies with those in aging studies would not help to resolve the tim- ing of developmental and aging transi- tions because different imaging, mea- surement, and quantitative procedures were used. Also, while many aging stud- ies included healthy volunteers, most studies of development included pa- tients. To our knowledge, there exist no pub- lished, quantitative, in vivo MR imaging studies of normal brain changes in the life span from early childhood (1 or 2 years of age) to late adulthood. We used currently available MR imaging proce- dures to accurately and objectively docu- ment the normal course of development and aging in whole brain, GM, WM, CSF, and intracranial volumes in male and fe- male healthy volunteers aged 19 months through 80 years. Quantitation of these parameters in healthy volunteers allowed us to compare these values with pub- lished values in the postmortem litera- ture. MATERIALS AND METHODS Volunteers The institutional review board of the Children’s Hospital of San Diego, Calif, approved the neuropsychological testing procedures and imaging protocols used in our study. One hundred sixteen healthy volunteers (79 male, 37 female) were aged 19 months to 80 years (mean age, 21.4 years 6 20 [SD]); 50 of these volunteers were aged 19 months to 12 years. The ethnic distribution of the pa- tients was as follows: 99 (85%) white, 10 (9%) Hispanic American, two and a half (2%) Asian American, three and a half (3%) African American, and one (1%) un- known, where volunteers of mixed eth- nicity were counted as half in each group. Intelligence quotient (IQ) data were available for 86 of the 116 volunteers and revealed that IQ was comparable across the age range of the sample. Composite IQ scores obtained with the Stanford-Bi- net test (29) in 25 volunteers younger than 6 years was 95–141 (116.3 6 11.0). Wechsler Intelligence Scale for Children- Revised or Wechsler Intelligence Scale for Children III full-scale IQ scores (30,31) for 33 volunteers aged 6 to 16 years were 90–133 (112.7 6 11.0). Wechsler Adult Intelligence Scale-Revised or Wechsler Adult Intelligence Scale III full-scale IQ scores (32,33) for 36 volunteers older than 16 years were 90–137 (115.6 6 14.0). To screen for signs of memory loss, volunteers older than 50 years under- went the California Verbal Learning Test (34). All volunteers performed this test in the normal or well-above-normal range. (Standardized scores based on age norms, or T scores, for overall performance were 42–66 [55.1 6 7.0].) Volunteers were recruited by means of advertisements distributed in the com- munity. On the basis of their responses to questionnaires administered at screening before MR imaging about medical, fam- ily, and educational history, volunteers showed no evidence of developmental, educational, medical, psychological, or psychiatric abnormalities or deficiencies. Prior to testing, the nature of the study and the procedures were explained to each volunteer, and informed consent was ob- tained. For each volunteer younger than 18 years, informed consent was obtained from a parent. Imaging MR images were acquired by using a 1.5-T Signa MR unit (GE Medical Sys- tems, Milwaukee, Wis). To provide im- ages for the computation of CSF, GM, and WM volumes, a dual spin-echo intermediate- and T2-weighted sequence was performed in the transverse plane (3,000/30 and 80 [repetition time msec/ echo time msec], one signal acquired, 20-cm field of view, 256 3 256 matrix, 3-mm interleaved [no gaps]). No medica- tion or sedation was used during the im- aging procedure. To help ensure that younger children remained still for imag- ing, scanning was performed during nat- ural sleep. Occasionally, the staff would pacify younger children who were not sleeping by singing to them during the MR imaging examination. The MR imag- ing data were transferred in digital for- mat to workstations (Silicon Graphics; Mountain View, Calif) for analysis. All images were coded so that the operators (H.J.C., A.C., J.C.) were blinded to each volunteer’s identity, sex, and age. Automated Correction for Signal Falloff Images were preprocessed (by B.E.) by using SEGMENT (Egaas B and Courchesne E, San Diego, Calif) to correct for decreased signal intensity toward the inferior sec- tions by using a fully automated algo- rithm. In this algorithm, pixels in the lowest signal intensity range (ie, bone of the skull base, air in the sinuses) were automatically excluded prior to calcula- tion of the difference in the mean signal intensity of the remaining pixels be- tween pairs of consecutive section loca- tions. Then, on the basis of a higher- order polynomial best fit of those difference values, a separate gain coeffi- cient for each section was derived and applied to each section so that the mean signal intensity of adjacent sections dif- fered by less than 1%. The gain coeffi- cient was constant within a section, and the same coefficient was applied to corre- sponding intermediate- and T2-weighted images. Next, semiautomated and fully automated steps were used to remove skull and extracranial structures and to classify GM, WM, and CSF as follows. Semiautomated Removal of Skull and Extracranial Structures In all volunteers, skull and extracranial structures were removed from the T2- weighted images with a semiautomated procedure that used a computer algo- rithm called “brain removal” (S.H.) and with decision making by experienced neuroanatomists (H.J.C., A.C., J.C.) in a section-by-section process. One of the operators set the initial threshold level on each T2-weighted image to separate the higher signal intensity of the brain and CSF from the lower signal intensity Volume 216 z Number 3 Brain Development and Aging: Quantitative Analysis at MR Imaging z 673 of the skull and air. Next, manual tracing was performed to separate contiguous nonbrain pixels from brain or CSF pixels with similar signal intensity values. The remaining intracranial pixels (ie, all brain and CSF pixels) were selected by using an automated region-growing algorithm that retained the original signal intensity val- ues of all intracranial pixels (Fig 1a). The final T2-weighted images that depicted only brain and CSF were used as a mask on the original intermediate-weighted im- ages to automatically obtain the corre- sponding intermediate-weighted images of only brain and CSF (Fig 1b). The entire process was completed in less than 45 min- utes per brain. The lowest section that matched the external morphology of the inferior me- dulla rather than the cervical spinal cord was chosen as the inferior boundary of the brain; external landmarks, such as the vertebral arteries, were not consistent among volunteers and did not provide reliable landmarks for the identification of the inferior boundary of the brain. All brain tissue and CSF spaces at or above the level of the lowest section, including the pituitary and infundibulum, were se- lected. Excluded from the intracranial measure were the skull, subcutaneous and orbital fat, mastoid and nasal sinuses, du- ral venous sinuses, larger blood vessels beyond the surface of the brain, bony protuberances (ie, dorsum sellae), and cranial nerve roots, since they extend be- yond the surface of the brain. Interopera- tor reliability was R . .99 (P , .001), and the mean percentage difference be- tween operators was 0.22%, with a max- imum difference in any brain of 1.47%. Anatomically, then, whole brain vol- ume was defined as all brainstem, infun- dibular and pituitary, cerebellar, subcor- tical, and cerebral parenchyma. Total intracranial CSF volume was defined as all CSF in ventricular and subarachnoid spaces, and intracranial space volume was defined as the sum of whole brain and CSF volumes. Automatic Classification of GM, WM, and CSF In the next step of the SEGMENT process with each brain data set, all pixels (in- cluding all skull and extracranial pixels) were used to form a global histogram in a intermediate-weighted (y axis) versus T2- weighted (x axis) feature space (Fig 1c). Each cluster (Fig 1c) was mathematically mapped as a joint Gaussian distribution in this intermediate- and T2-weighted feature space; the location and orienta- tion of each cluster may be visualized as an ellipse that has major axis points ori- ented toward the origin (where the signal intensity values of intermediate- and T2- weighted images both equal 0). By using a maximum likelihood criteria algorithm based on previous approaches (35), all pixels were then automatically classified (by B.E.) as parenchyma (WM and GM clusters in Fig 1c), CSF (CSF cluster in Fig 1c), CSF partially volumed with skull (cluster not shown), or other (ie, all re- maining nonparenchymal and non-CSF pixels [eg, nonbrain clusters in Fig 1c]). Figure 1. Images and graphs illustrate how SEGMENT, a fully automated algorithm, uses (a) T2- and (b) intermediate-weighted images to plot all pixels as (c) a global histogram in a T2-weighted (x axis) versus intermediate-weighted (PD; y axis) feature space. SEGMENT then uses an algorithm with maximum likelihood criteria to classify pixel clusters as brain, CSF, CSF partially volumed with skull (cluster not shown in figure), or other (nonbrain). Next, for all pixels classified as brain, SEGMENT uses a three-dimensional local-contrast algorithm to separate GM and WM pixels, which creates (d) final GM-, WM-, and CSF-segmented images. For qualitative comparison, manual tracings of T2- and intermediate-weighted pairs of images (eg, see tracings in a and b) were overlaid onto SEGMENT-created images (tracing in d). (e) Validation of SEGMENT was accomplished by means of statistical comparison with results from manual tracing; analyses showed that GM and WM measures obtained by using SEGMENT and manual-tracing methods correlated to 98.2% and 96.1%, respectively. 674 z Radiology z September 2000 Courchesne et al All parenchymal pixels were then au- tomatically separated into GM and WM pixels (by B.E.) with a three-dimensional local-contrast algorithm SEGMENT. With this algorithm, the local threshold for GM and WM pixels was computed from pixel statistics within a 2 3 2 3 2-cm cube surrounding the pixel being classi- fied. (The cube was 29 pixels 3 29 pix- els 3 7 sections, or 5,887 surrounding voxels.) The use of this local contrast makes segmentation relatively insensi- tive to the signal intensity inhomogene- ities intrinsic to imaging that plague sim- pler methods of segmentation, such as uniform thresholding over an entire sec- tion. In particular, this method success- fully handles the signal intensity reduc- tions often found at the occipital pole and lateral extent of the temporal lobes. Pixels that the three-dimensional local contrast algorithm determined to be mostly WM were classified as WM, and pixels determined to be mostly GM were classified as GM. This classification pre- served a sharp outline between the corti- cal ribbon and adjacent axonal material (Fig 1d). The approach is analogous to the previous approaches that use feature space in the semiautomated segmenta- tion of pixels on images with nearly iden- tical intermediate- and T2-weighted pro- tocols (28,36); the principal difference is that the present algorithm is fully auto- mated. Validation of the Automated Classification Algorithm The automated classification was vali- dated by using randomly selected images of eight brains from those of the study group without any foreknowledge of im- age quality; they included images in five male and three female young children and adults. In each brain, all GM, WM, and CSF spaces within the right hemi- brain were manually traced (by H.J.C.) from the T2- and intermediate-weighted images at the following three section lev- els—the centrum semiovale, the thala- mus and basal ganglia, and the pons (see Fig 1a, 1b for tracings in one section of one brain). The area was calculated for pixels designated as GM, WM, or CSF within each manual tracing of each sec- tion. Manually traced and automatically segmented tissue classifications were compared by means of qualitative visual examination and quantitative statistical analysis. Visual examination was performed (by E.C., G.A.P., H.J.C., and B.E.) by overlay- ing the results of the manual tracings onto the automatically classified images (Fig 1d). Correspondence between auto- mated and manual approaches was eval- uated in the cerebrum, cerebellum, glo- bus pallidus, lateral thalamus, substantia nigra, and red nucleus. Since the com- bined volume of these subcortical struc- tures accounts for approximately 1% of total brain volume (12), a modest mea- surement error in these structures will result in an error of much less than 1% in the whole brain, GM, and WM volumes calculated in our study. In an additional validation step, the experienced neuroanatomist (H.J.C.) manually traced cortical GM on 20 sec- tions in a ninth brain using the interme- diate- and T2-weighted images; measures from these manual tracings and those of the automated classification were statisti- cally compared. That these fully automated (SEGMENT) and semiautomated steps resulted in valid brain measures is further supported by the results described later that showed comparable age-related changes in intra- cranial space volume in MR imaging and postmortem studies, comparable changes in the mean intracranial space volume in adolescents and adults, comparable esti- mates of whole brain weight, and compa- rable age-related changes in the GM-WM ratio. Calculation of Intracranial, Whole Brain, CSF, GM, and WM Volumes Each two-dimensional pixel on the im- ages represents a three-dimensional vol- ume (voxel), and the intracranial space volume was determined by summing the volume of all voxels designated as GM, WM, and CSF and half of the volume of all voxels designated as CSF partially vol- umed with skull. Whole brain volume was determined in a similar fashion by using only GM and WM voxels. Total CSF volume was constructed from the full volume of CSF voxels and half of the volume of all voxels designated as CSF partially volumed with skull. Estimation of Brain Weight at MR Imaging At autopsy, it is common practice to weigh brains with the leptomeninges and ventricles intact, trapping CSF within and around the brain. Therefore, to com- pare brain size at in vivo MR imaging with brain size in postmortem studies, it is necessary to obtain an estimation of brain weights from MR imaging-derived measures that are comparable to norma- tive brain weights published in the post- mortem literature. This estimation can be performed by converting MR imaging– derived volumes of parenchyma (ie, whole brain) and CSF into grams per mil- liliter and then by summing these values as follows: brain weight in grams 5 (whole brain volume in milliliters 3 1.0365 g/mL) 1 (CSF volume in millili- ters 3 1.00 g/mL) (37,38). Statistical Analyses of Volume Measurement Data One author (J.T.) ensured the use of the appropriate statistical tests. Statistical analysis was performed (by J.T.) with the Biomedical Statistical Software Package, or BMDP (SPSS; Berkeley, Calif) (39). Sta- tistical probabilities reported for regres- sion analyses were tests of significance for the orthogonal polynomial regression coefficients. Comparison of the change with age in our in vivo brain volume estimates and the brain weight measures obtained from published postmortem data involved the use of standard regression analyses with effect coding in an evaluation of both linear and quadratic differences with age between the in vivo MR imaging data and postmortem data (40). Independent variables in these analyses were study type (postmortem or in vivo MR imag- ing), linear and quadratic age factors, in- teractions of study type, and linear and quadratic age. Measures from the 11 pub- lished postmortem studies (see tables 111, 113, 115, 117, and 119 in reference 19) (41–46) cited here for comparison were available as only means (for various age groups) and not as individual data points. In our analyses, we assumed that these means were robust and that they represented the data of an individual at the given age. A P value of .05 indicated a statistically significant difference. RESULTS Intracranial Space Age-related change in intracranial space is shown in Figure 2a. Intracranial space volume grew exponentially by 27% between early childhood (mean age, 26 months; age range, 19–33 months) and early adolescence (mean age, 14.0 years; age range, 12–15 years), when the devel- opmental maximum was reached (linear component, R 5 0.58, R2 5 0.34, t58 5 5.48, P , .001; quadratic component, change in R2 5 0.06, t57 5 2.37, P , .05) (Fig 2a). In the age range of 16–80 years, there was no statistically signifi- Volume 216 z Number 3 Brain Development and Aging: Quantitative Analysis at MR Imaging z 675 cant change in intracranial space vol- ume. Across all ages, intracranial space volumes were about 10% smaller in fe- male volunteers than in males volunteers (1,305.7 mL 6 106 vs 1,450.5 mL 6 137, t114 5 5.68, P , .001). Figure 2c shows growth in intracranial space based on our data obtained with in vivo quantitative MR imaging procedures plotted with postmortem intracranial space volumes reported in two previous developmental studies (see table 115 in reference 19) (20). Age-related changes in intracranial space volume in our in vivo study and in these previous postmortem studies closely match (Fig 2c). Whole Brain Volume Age-related change in whole brain vol- ume is shown in Figure 2b. Whole brain volume grew exponentially by 25% be- tween early childhood (19–33 months) and adolescence (12–15 years), when the developmental maximum volume was reached (linear component, R 5 0.53, R2 5 0.28, t58 5 4.72, P , .001; qua- dratic component, change in R2 5 0.06, t57 5 2.22, P , .05) (Fig 2b). In the age range of 16–80 years, whole brain vol- ume slowly but steadily declined so that in those aged 71–80 years, it had dropped by 26% and was smaller than that of healthy 2- or 3-year-old children (linear component, R 5 0.57, R2 5 0.32, t54 5 5.09, P , .001; quadratic compo- nent, change in R2 5 0.06, t53 5 2.28, P , .05). Across all ages, whole brain volumes were about 12% smaller in fe- male volunteers than in male volunteers (1,137.8 mL 6 109 vs 1,286.4 mL 6 133, t114 5 5.92, P , .001). Figure 2d depicts a graph of age-related changes in postmortem brain weight from 11 studies (see tables 111, 113, 115, 117, and 119 in reference 19) (41–46), age-related changes in our in vivo whole brain volumes (brain only), and our esti- mated in vivo whole brain weights (esti- mated brain weight 1 CSF weight; see Materials and Methods). For each study, weight or volume at each age was plotted as a function of a percentage of the max- imum weight or volume for that study (eg, for our study, the maximum was achieved at 12–15 years of age). The graph and statistical analyses showed no difference in change with age between postmortem data and our in vivo estimated whole brain weight. Re- gression analysis with effect coding for study type (see Materials and Methods, Statistical Analyses of Volume Measure- ment Data) yielded a model in which there were significant linear and qua- dratic components in the association of age with brain weight (R 5 0.40, F2,224 5 Figure 2. Graphs depict age-related volume changes in (a) intracranial space and (b) whole brain in our 116 healthy volunteers. Volume intercepts at birth are based on quantitative MR imaging findings of Huppi et al (55). (c) Graph shows that our in vivo age-related intracranial space growth rates are similar to those of previous postmortem studies (table 115 in the study by Blinkov and Glezer [19]; Lichtenberg [20]). (d) Graph shows similar growth rates in our in vivo whole brain volume (red symbols), estimated in vivo whole brain weight (black symbols; estimated whole brain weight 5 brain weight 1 CSF weight), and brain weight (open symbols) from 11 previously published postmortem studies (tables 111, 113, 115, 117, 119 in reference 19) (41–46). In the present study and in each published study, volume or weight at each age is plotted as a function of a percentage of the maximum volume or weight for that study. Postmortem weight includes CSF, but our in vivo MR imaging whole brain volume does not. Therefore, during development, CSF weight makes a small contribution to the reported brain weight, but, with aging, its weight makes a successively greater contribution. (On x axes, age scales change after 20-year point.) 676 z Radiology z September 2000 Courchesne et al 21.85, P , .001, linear coefficient t 5 5.44, P , .001; quadratic coefficient t 5 6.56, P , .001 [Fig 2d]). There was no association between study type (post- mortem weight, in vivo estimated weight) and brain weight, and there were no significant interactions between study type and age. The correlations of brain weight with study type was 0.16, and the interaction of study type with linear age and quadratic age was 0.003 and 20.27, respectively. A regression model with study type, linear age with study type interaction, and quadratic age with study type interaction as independent variables showed no relationship between these factors and brain weight (R2 5 0.005). The graph and statistical analyses showed similar changes in growth but not aging in the postmortem data and in our in vivo brain volume. A comparison of postmortem brain weight and whole brain volume (only GM and WM) at MR imaging showed significant differences in the rate of change with age as a func- tion of the study type (postmortem weight, in vivo whole brain volume). This difference was due to the increase in CSF and the complementary decrease in whole brain volume with aging seen in the MR imaging measures but not in the postmortem measures of weight, which included the weight of CSF. By using the same set of independent variables as those used in the previous analysis, the model of best fit selected study type, the quadratic age component, and the inter- action between study type and linear age (R 5 0.56, F3,223 5 34.53, P , .001; quadratic age coefficient, t 5 8.25, P , .001; study-type coefficient, t 5 4.45, P , .001; study-type and linear-age in- teraction coefficient, t 5 6.28, P , .001) (Fig 2d). In addition, in the present study, the mean estimated brain weight from in vivo MR imaging data was 1,527 g 6 120 for male adolescents and adults (age range, 16–43 years) and 1,363 g 6 99 for female adolescents and adults (age range, 17–49 years). These MR imaging–derived estimates of brain weight were about 4% greater than the mean brain weights re- ported for male and female participants in six different postmortem normative studies involving more than 6,000 partic- ipants aged 17–44 years (table 118 in ref- erence 19) (41–43,47,48). In those six studies, the median of the mean brain weight was 1,460 g 6 113 in male partici- pants (range of study means, 1,443–1,504 g) and 1,305 g 6 106 in female participants (range of study means, 1,288–1,398 g). The small 4% difference between our in vivo estimated brain weight and the postmor- tem weight may be due to sampling differ- ences or to secular increases in brain size in the past several decades. Alternatively, it may be that our slightly greater estimate in the living brain reflects the fact that, in living brain, 2%–3% of its volume is blood most of which is in GM capillaries (19). GM Volume Age-related change in GM is shown in Figure 3a. GM increased 13% from early childhood (19–33 months) to later child- hood (6–9 years) (linear component, R 5 0.37, R2 5 0.14, t41 5 2.64, P , .05). Thereafter, it decreased linearly by ap- proximately 5% per decade throughout life (linear component, R 5 0.75, R2 5 0.57, t71 5 9.63, P , .001) (Fig 3a). The percentage of GM contributing to intra- cranial space volume was not different in male and female volunteers (female vol- unteers, 58.5% 6 0.08%; male volun- teers, 59.9% 6 0.07%). WM Volume Age-related change in WM volume is shown in Figure 3b. WM volume in- creased 74% from early childhood (19–33 months) to adolescence (12–15 years) (linear component, R 5 0.81, R2 5 0.65, t58 5 10.28, P , .001; quadratic component, R 5 0.82, R2 5 0.68, t57 5 2.59, P , .01). Thereafter, it increased at a slower rate and reached a plateau by the 4th decade of life; relative to this maximum level, it decreased by 13% in the oldest volunteers (aged 70–80 years) (significant quadratic component only, R 5 0.39, R2 5 0.15, t53 5 2.61, P , .01) (Fig 3b). The percentage of WM contributing to intracranial space vol- ume was not different in male and female volunteers (female volunteers, 28.7% 6 0.04%; male volunteers, 28.9% 6 0.03%). GM-WM Ratio As shown in Figure 3c, the GM-WM ratio decreased rapidly from early child- hood (19–33 months) to later adulthood (about 50–60 years). Thereafter, it slowly but steadily declined (linear component, R 5 0.69, R2 5 0.47, t114 5 10.10, P , .001; quadratic component, change in R2 5 0.24, t113 5 9.63, P , .001). GM-WM ratios were not different in male and female volunteers (female volun- teers, 2.11 6 0.6; male volunteers, 2.12 6 0.5). The present MR imaging–based age-re- lated decline in the GM-WM ratio is gen- erally consistent with postmortem study– derived age-related differences in this ratio, as originally reported nearly a century ago (49,50), and with the more contemporary postmortem data (25) and recent MR im- aging–derived values (Fig 3d) (27,51,52). Total Intracranial CSF Volume Total ventricular and extracerebral CSF volume increased during the life span (Fig 4a). The total volume of CSF (ven- tricular and subarachnoid) more than doubled between early childhood (19–33 months) (approximately 96 mL) and middle adulthood (40–55 years) (approx- imately 250 mL). This total CSF volume increased linearly with age in the entire sample of both male and female volun- teers (male volunteers, R 5 0.84, R2 5 0.71, t77 5 13.84, P , .001; female volunteers, R 5 0.88, R2 5 0.78, t35 5 11.19, P , .001). On the other hand, as a percentage of intracranial space volume (Fig 4b), total CSF remained steady at 7–9% during the period of rapid intracranial space growth and whole brain growth from early child- hood to adolescence and then increased after the 2nd decade of life (R 5 0.81, R2 5 0.65, t54 5 10.03, P , .001). In the oldest volunteers (aged 71–80 years), CSF ranged from 20%–33% of the intra- cranial volume. The percentage of CSF contributing to intracranial space vol- ume was not different in male and female volunteers across the sample (female vol- unteers, 12.8% 6 0.01%; male volun- teers, 11.2 % 6 0.06%) (Fig 4b). Correlation Between Manual and Automatic Classification Methods Statistical analyses were used to con- firm that our automated method of mea- surement provided classifications with high fidelity to the results of the manual tracings performed on intermediate- and T2-weighted images of eight randomly selected volunteers. Correlations be- tween manual and automatic classifica- tion of three different section locations within the right hemibrain in these eight brains were 0.982 for GM, 0.961 for WM, and 0.887 for CSF (Fig 1e). Visual examination conducted by overlaying the results of the manual trac- ings on the images of the eight brains onto the automatically classified images (eg, Fig 1d) showed that correspondence between the automated and manual ap- proaches was best and equally good in the cerebrum and cerebellum and was worst in the globus pallidus, lateral thal- amus, substantia nigra, and red nucleus. Volume 216 z Number 3 Brain Development and Aging: Quantitative Analysis at MR Imaging z 677 We also studied the correlation be- tween the manual tracings and the auto- matic classification of the cortical GM (alone) on 20 intermediate- and T2- weighted images from the ninth randomly selected brain. Measures from those man- ual tracings and from the automated clas- sification correlated to a level of more than 99%. DISCUSSION Intracranial Volume and Brain Size Our study of parameters of normal brain development and aging with MR images of 116 healthy volunteers shows that the brain reaches maximum volume in early adolescence then steadily and linearly declines from early through mid- dle adulthood, with an accelerated rate of decline after 55 years of age. Intracranial volume increases with brain volume but does not, thereafter, decline with aging. Between early childhood and early ado- lescence, we found that the healthy brain (Fig 2b) and intracranial space (Fig 2a) grew exponentially by about 25%–27%; however, by 71–80 years of age, brain volume was less than that of our 2- or 3-year-old children. These brain growth and aging effects were similar in male and female volunteers (with a hint of a sharper decline in brain volume in the oldest male volunteers [Fig 2]). This pattern of life-span change is con- sistent with findings of previous studies of either development or aging. Studies of development.—Postmortem study findings concur that the brain of neonates weighs 370 g on average and that its weight increases nearly threefold by 2 years of age (19,42,44–46). The post- mortem data (42) show that, from in- fancy to adolescence, brain weight in- creases by 29%, an amount similar to the 25% increase in brain volume noted in our study (Fig 2d). In postmortem studies (19,46) that focused on the life span from birth to adolescence or young adulthood, brain weight increased by about 32% from early childhood to adolescence. As with brain growth (Fig 2d), growth of intracranial volume is also similar be- tween postmortem studies and our in vivo study (Figs 2c). To our knowledge, no previous in vivo MR imaging study has quantified growth in whole brain and intracranial volumes in healthy young children; only groups of clinically-referred pediatric patients Figure 3. Graphs depict growth and aging changes in (a) GM and (b) WM volumes in 116 living healthy volunteers. Volume intercepts at birth are based on the findings of Huppi et al (55). (On x axes, age scales change after 20-year point.) GM volume reached a maximum by 6–9 years of age and thereafter declined linearly. WM volume rapidly increased until 12–15 years of age, and thereafter increased at a slower rate to a plateau at about the 4th decade of life. For both (c) the present in vivo and (d) the published MR imaging and postmortem data, the GW-WM ratio drops sharply from infancy to about the end of adolescence after which it declines at a slower rate. Symbols in d represent data from the postmortem studies of Anton (49) (¹), Jaeger (50) ({), and Miller et al (25) (w) and the MR imaging studies of Filipek et al (51) (■ 5 data in males, h 5 data in females), Pfefferbaum et al (27) (ƒ), and Narayana and Borthakur (52) (v). 678 z Radiology z September 2000 Courchesne et al who are termed “neurologically normal” or “radiologically normal” have been studied. For example, in one group of such patients, there was “a rapid increase by 2 years of age, and thereafter a gradual increase until about 12 years” (10). Their growth curves show that, from early childhood to adolescence, cerebral vol- umes grew exponentially by about 22%, which is a magnitude of change compa- rable with that of the whole brain vol- ume (ie, 25%) in the present study. In another study (27), pediatric patients had roughly a 30% increase in intracranial volume from infancy and early child- hood to about 10 years of age, when maximum volumes were reached. In rare studies (53,54), cerebral volumes were measured in healthy children and adoles- cents. Neither study reported statistically significant age-related changes in cere- bral volume, possibly because few young volunteers were included. Studies of aging.—A large number of study findings agree that brain size de- creases with advancing age (22), but the magnitude of decrease differs substan- tially between postmortem brain weight data on the one hand and in vivo whole brain volumes on the other (Fig 2d). Ac- cording to findings of postmortem stud- ies, brain weight decreases by only about 9% in males and females alike from a maximum in adolescence to late adult- hood (71–80 years) (19,41–43,48). When we converted our brain and CSF volumes to brain weight (autopsy brain weight in grams 5 brain volume in mil- liliter 3 1.0365 g/mL 1 CSF volume in milliliter 3 1.00 g/mL), our results showed a similar small decrease (9%) in weight from adolescence to late adult- hood (71–80 years; compare open and black symbols in Fig 2d). However, this is a far smaller decrease than the 26% de- crease found in brain (GM and WM) vol- umes in our study (compare open and red symbols in Fig 2d). The discrepancy between age-related changes based on brain weight determined at autopsy and those based on in vivo brain volume is likely caused by the inclusion of the weight of CSF trapped in the ventricles and leptomeninges at autopsy as brain weight. In healthy older individuals, CSF volume and its weight can be consider- able. As noted previously for the present study, CSF accounts for a small percent- age (7%–12%) of the total intracranial volume in a healthy young person (and, therefore, would account for a small per- cent of brain weight in a standard mea- surement at autopsy); however, in 71– 80-year-old adults, CSF in the ventricles and leptomeninges can account for 16%– 25% of the intracranial volume (present study) (22,24,26,28). Whole Brain GM Volume In vivo MR imaging data show that cerebral GM volume increases linearly from 29 to 41 weeks after conception (55), continues to increase to 4–6 years of age (27), and decreases thereafter (27). Our in vivo study revealed a similar age- dependent profile for whole brain GM (compare our Fig 3a to figure 3a in the article by Pfefferbaum et al [27]). GM vol- ume increased 13% from early childhood (mean age, 26 months) to later child- hood (6–9 years), when maximum vol- ume was reached (which is several years before peak whole brain volume is reached). From late childhood, GM de- creased linearly by about 5% per decade throughout life. One recent study (13) revealed that most sampled regions showed linear age-related decreases in ce- rebral GM from 20 to 80 years; prefrontal and superior parietal cortex, respectively, decreased by 4.9% and 4.3% per decade, but some cortical areas, such as the ante- rior cingulate and primary motor cortex, decreased little. The particular biologic mechanisms underlying this striking age-dependent profile of GM volume increase and de- crease are not known with certainty. There are no comparable age-related in- creases and decreases in the number of neurons in the healthy brain that can account for these changes in GM volume. However, for the period from 29 weeks after conception through 6–9 years of age (during which the combined in vivo MR imaging data from our study and oth- ers [27,55] demonstrate increasing GM volume), prior postmortem studies report corresponding increases in the size of neuronal cells (19), increases in the ex- tent of neuronal arbors (56), and in- creases in the number of synapses (57) and neurons (58). The rapid growth of GM (including each of its neuronal components and the enhanced functional responses that char- acterize the period from 19 months to 9 years of age) is followed by a period of elimination of cerebral synapses (57) slowed the growth in cell and arbor size slows (19,56) and decreases in cerebral metabolic and neurophysiologic indices (59–61). While these changes might be related to the onset of the decline in GM volume (which follows the achievement Figure 4. (a) Graph shows that total intracranial CSF volume increased resolutely across the life span included in the present study. (b) None- theless, graph shows that the percentage of CSF occupying intracranial remained at about 7%–9% during the period of rapid intracranial space and whole brain growth (Fig 1a, 1b) from early childhood to early adolescence. Thereafter, the percentage increased as whole brain volume declined, and intracranial space volume changed little. (On x axis, age scales change after 20-year point.) Volume 216 z Number 3 Brain Development and Aging: Quantitative Analysis at MR Imaging z 679 of peak volumes at about 6–9 years of age and which might also continue to con- tribute to the decline in late childhood and adolescence), they are largely com- pleted by adolescence, and, therefore, they cannot explain the continued linear decline throughout early, middle, and late adulthood. Conversely, the various hypotheses that have been advanced to explain why GM declines during aging (eg, age-related atrophy in subcortical monoaminergic structures that supply cortical neurons and decline in the abil- ity to renew synapses and repair of neu- rons) cannot explain the GM decline from late childhood through adoles- cence. It seems unlikely, therefore, that a single factor can account for the decline in GM that lasts nearly throughout the life-span; perhaps at different ages differ- ent factors operate to cause GM volume changes. Whole Brain WM Volume Prior in vivo MR imaging data show a sudden increase in whole-brain myelin- ated WM between 29 and 41 weeks after conception (55). From infancy through early adulthood, in vivo studies of cere- bral WM reveal large increases in volume (10,27), as do studies of older children and adolescents (54). In our study, whole brain WM volume increased in a manner parallel to these developmental changes by about 74% from early childhood to adolescence; thereafter, it increased at a slower rate and finally reached a plateau by the 4th decade of life (Fig 3b). We detected a decrease in WM volume of 13% in the oldest volunteers (aged 71–80 years). Others (13) found age-re- lated volume decreases to be slight or not detectable in several cortical regions, and only prefrontal and superior parietal re- gions had statistically significant declines in volume with advancing age. GM-WM Ratio The opposing age-related patterns of volume change in GM and WM combine to make a striking curve of change in the GM-WM ratio throughout the life span (Fig 3c, 3d). In our study and in that of others (27), the steady rise in WM and decline in GM led to a sharply declining, curvilinear change in the GM-WM ratio from early childhood through adoles- cence. Apparently, beginning at about 20 years of age, the much slower rate of in- crease in WM combined with the steady decrease in GM substantially slowed the rate of decline in the GM-WM ratio in our study population. This life-span pat- tern of change is consistent with post- mortem (49,50) and in vivo MR imaging data (27,51,52) that show that the GM-WM ratio in the very young brain is far larger than that of the adult brain and that, from 20 to 80 years of age, this ratio changes only slightly (25). CSF Volumes In our volunteer population, the abso- lute volume of ventricular and extracere- bral CSF increased substantially and nearly linearly throughout the life span (from approximately 96 mL in toddlers to approximately 350 mL in 80-year-old adults) (Fig 4a). Therefore, increases in the absolute volume of intracranial CSF are a phenomenon of not only aging but the entire life span. Nevertheless, the percentage of CSF that occupied intracranial space (Fig 4b) remained nearly constant at 7%–9% dur- ing the entire period of rapid intracranial and brain growth from early childhood to early adolescence. Thereafter, with lit- tle further change in intracranial space, but with decreasing brain volume with advancing age, the percentage of CSF that occupied intracranial space, as well as the absolute volume of CSF, increased from 14% during young and middle adulthood to 25% in our healthy older adults. It appears, therefore, that from adolescence through late adulthood, the percentage of CSF that occupies intracra- nial space gradually increases. There are no developmental, postmor- tem, or in vivo MR imaging studies of changes in total intracranial CSF vol- umes. However, among a large number of in vivo studies of aging (22,26,27), there is uniform agreement that intracra- nial CSF increases with advancing age. For instance, a recent study (24) reported that CSF that occupied intracranial space increased from 12% to 17% between 26 and 70 years of age. The absolute amounts of intracranial CSF reported in older adults varies some- what from study to study (typically from 250 to 400 mL) most likely because dif- ferent studies used different definitions of intracranial volume or CSF spaces. Our study included all brain, CSF, and intra- cranial space from the vertex to the level of the bottom of the medulla, and, so, we gave a “true” total intracranial CSF vol- ume that was larger than, for instance, the volume calculated by others (22) who excluded CSF in the posterior fossa. Cross-sectional and Longitudinal Study Designs Cohort effects are unavoidable in any cross-sectional study of age-related changes during development and ag- ing. Longitudinal studies are also vul- nerable to cohort effects; for instance, longitudinal changes observed in a sin- gle cohort may not be representative of the population at large. Nonetheless, longitudinal MR imaging studies can have greater statistical power than that of cross-sectional studies in the detec- tion of small effects, and the power of the longitudinal MR imaging design has been convincingly demonstrated in healthy individuals (62) and in those with different disorders (63,64). However, the most powerful design for studying development and aging changes and for ruling out potential cohort effects is the cross-sequential design that com- bines cross-sectional and longitudinal pro- cedures. (An example of such a longitudi- nal MR imaging study of children and adolescents if that of Giedd et al [65], which was published while the present ar- ticle was in press.) Quantitative MR imag- ing is ideally suited for this, and the present article represents the cross-sec- tional phase of our cross-sequential design; our group has nearly completed data col- lection in a long-term, longitudinal MR imaging phase; results from this combined, cross-sequential, quantitative MR imaging study will be reported in the future. Clinical and Research Utility of Quantitative MR imaging Our quantitative MR imaging proce- dures yielded indices of normal brain de- velopment and aging throughout the life span that faithfully followed the age-re- lated time courses predicted from quan- titative postmortem procedures. Our in vivo normative life-span MR imaging data may be useful in clinical settings and in research studies for detecting and monitoring pathologic alterations in fun- damental neuroanatomic indices in pa- tients with suspected congenital, devel- opmental, or acquired disorders. For example, using the normative values and quantitative MR imaging procedures in the present study, we detected macren- cephaly in young children suspected of being autistic (Courchesne E, unpublished data, 1991–1996). In spinocerebellar ataxia type 2 (66), a statistically significant corre- lation exists between supratentorial atro- phy and disease duration (whereas the de- gree of infratentorial atrophy does not 680 z Radiology z September 2000 Courchesne et al correlate with disease duration). The com- bination of normative age-related changes with results from serial quantitative MR imaging of intracranial compartments in spinocerebellar ataxia type 2 may help in the detection of the progressive steps of this disease. Moreover, individual research studies of a number of disorders have shown that volumetric MR imaging analy- ses of basic brain parameters can be used to distinguish disorders from normal states and to distinguish diseases from each other (67,68), but the routine clinical use of such information via quantitative MR imaging assays has remained limited. Our study adds evidence indicating that quantitative MR imaging assays of several fundamental neuroanatomic indices in a routine clinical setting is feasible. Until re- cently, quantitative MR imaging examina- tions of the neuroanatomy were not a practical option in routine clinical neuro- radiologic examinations primarily because they were (a) dependent on specialized im- aging protocols, (b) labor intensive, (c) de- pendent on a high level of detailed ana- tomic knowledge and manual tracing, (d) dependent on specialized computer hardware, or (e) based on homegrown soft- ware that required specialized computer or biomedical knowledge to operate. Our study, along with a number of other recent studies (26,28,36), involved the use of routine intermediate- and T2-weighted imaging protocols and semiautomated ap- proaches that circumvent these obstacles. For instance, in our study, classification of GM, WM, and CSF pixels was accom- plished by using a fully automated algo- rithm that (a) was based on known and tested principles (28,35,36); (b) was vali- dated against neuroanatomic tracings per- formed by an experienced neuroanatomist on images in young children, adolescents, and adults; (c) provided results with mea- surement errors of only 1%–2% or less; and (d) clinical MR imaging technologists could easily operate. Moreover, this algorithm can be used on computers that are commonly avail- able at clinical imaging facilities. The pri- mary drawback with our algorithm is a single semiautomated “skull removal” step that involves manual tracing and that requires about 45 minutes to com- plete. However, adaptation of this step to meet time and personnel needs in a clin- ical setting is practical, and is, in fact, fast (approximately 5 minutes), although easy- to-use, less precise methods that imple- ment that single step are already commer- cially available. With current procedures, therefore, reliable and accurate normative brain indices, such as those of the present report, can be used in any clinical facility as quantitative clinical standards against which possible congenital, acquired, or treatment-related deviations from normal development across the life span can be judged. Acknowledgments: This work is supported by funds from NINDS (5P50-NS-22343) and NIDCD (5P50-DC-021289) awarded to Elizabeth Bates. Funds from her grants paid for recruit- ment and imaging in a subsection of the control subjects whose data are reported in the article. The authors thank Lynn K. Lord, MA, and Beth M. Hanenburg, AF, for MR imaging technical assistance, Rachel Yeung Courchesne, BS, for project coordination and helpful comments on the manuscript, and Christina M. Karns, BS, for technical assistance. References 1. Andreasen NC, Flashman L, Flaum M, et al. Regional brain abnormalities in schizo- phrenia measured with magnetic resonance imaging. JAMA 1994; 272:1723–1769. 2. Baumgardner TL, Singer HS, Denckla MB, et al. Corpus callosum morphology in children with Tourette syndrome and at- tention deficit hyperactivity disorder. Neurology 1996; 47:477–482. 3. Berquin PC, Giedd JN, Jacobsen LK, et al. Cerebellum in attention-deficit hyperac- tivity disorder: a morphometric MRI study. Neurology 1998; 50:1087–1093. 4. Courchesne E, Yeung-Courchesne R, Press GA, Hesselink JR, Jernigan TL. Hy- poplasia of cerebellar vermal lobules VI and VII in autism. N Engl J Med 1988; 318:1349–1354. 5. Courchesne E, Saitoh O, Yeung- Courchesne R, et al. Abnormality of cere- bellar vermian lobules VI and VII in pa- tients with infantile autism: identification of hypoplastic and hyperplastic subgroups with MR imaging. AJR Am J Roentgenol 1994; 162:123–130. 6. Egaas B, Courchesne E, Saitoh O. Reduced size of corpus callosum in autism. Arch Neurol 1995; 52:794–801. 7. Gosain AK, McCarthy JG, Glatt P, Staffen- berg D, Hoffmann RG. A study of intra- cranial volume in Apert syndrome. Plast Reconstr Surg 1995; 95:284–295. 8. Hashimoto T, Tayama M, Murakawa K, et al. Development of the brainstem and cerebellum in autistic patients. J Autism Dev Disord 1995; 25:1—18. 9. Hynd GW, Semrud-Clikeman M, Lorys AR, Novey ES, Eliopulos D, Lyytinen H. Corpus callosum morphology in atten- tion deficit-hyperactivity disorder: mor- phometric analysis of MRI. J Learn Disabil 1991; 24:141–146. 10. Iwasaki N, Hamano K, Okada Y, et al. Volumetric quantification of brain devel- opment using MRI. Neuroradiology 1997; 39:841–846. 11. Jacobsen LK, Giedd JN, Berquin PC, et al. Quantitative morphology of the cerebel- lum and fourth ventricle in childhood- onset schizophrenia. Am J Psychiatry 1997; 154:1663–1669. 12. Raz N, Torres IJ, Briggs SD, et al. Selective neuroanatomic abnormalities in Down’s syndrome and their cognitive correlates: evidence from MRI morphology. Neurol- ogy 1995; 45:356–366. 13. Raz N, Gunning FM, Head D, et al. Selec- tive aging of the human cerebral cortex observed in vivo: differential vulnerabil- ity of the prefrontal gray matter. Cereb Cortex 1997; 7:268–282. 14. Reiss AL, Faruque F, Naidu S, et al. Neu- roanatomy of Rett syndrome: a volumet- ric imaging study. Ann Neurol 1993; 34: 227–234. 15. Rumsey J. Thatcher RW, Lyon GR, Rum- sey J, Krasnegor N, eds. Developmental neuroimaging: mapping the develop- ment of brain and behavior. New York, NY: Academic Press, 1997. 16. Rumsey J. Ernst M, Rumsey JM, eds. Func- tional neuroimaging in child psychiatry. New York, NY: Cambridge University Press, 2000. 17. Schlaepfer TE, Harris GJ, Tien AY, et al. Decreased regional cortical gray matter volume in schizophrenia. Am J Psychia- try 1994; 151:842–848. 18. Toft PB, Leth H, Ring PB, Peitersen B, Lou HC, Henriksen O. Volumetric analysis of the normal infant brain and in intrauter- ine growth retardation. Early Hum Dev 1995; 43:15–29. 19. Blinkov SM, Glezer II. The human brain in figures and tables: a quantitative hand- book. New York, NY: Plenum Press and Basic Books, 1968; 334–338. 20. Lichtenberg R. Radiographie du crane de 226 enfants normaux de la naissance a 8 ans: impressions digitiformes, capacite, angles et indicies. Paris, France: Univer- sity of Paris, 1960. 21. Dekaban AS. Tables of cranial and orbital measurements, cranial volume, and de- rived indexes in males and females from 7 days to 20 years of age. Ann Neurol 1977; 2:485–491. 22. Coffey CE, Lucke JF, Saxton JA, et al. Sex differences in brain aging: a quantitative magnetic resonance imaging study. Arch Neurol 1998; 55:169–179. 23. Jernigan TL, Trauner DA, Hesselink JR, Tallal PA. Maturation of human cere- brum observed in vivo during adoles- cence. Brain 1991; 114:2037–2049. 24. Murphy DG, DeCarli C, McIntosh AR, et al. Sex differences in human brain mor- phometry and metabolism: an in vivo quantitative magnetic resonance imaging and positron emission tomography study on the effect of aging. Arch Gen Psychi- atry 1996; 53:585–594. 25. Miller AK, Alston RL, Corsellis JA. Varia- tion with age in the volumes of grey and white matter in the cerebral hemispheres of man: measurements with an image analyser. Neuropathol Appl Neurobiol 1980; 6:119–132. 26. Mueller EA, Moore MM, Kerr DCR, et al. Brain volume preserved in healthy el- derly through the eleventh decade. Neu- rology 1998; 51:1555–1562. 27. Pfefferbaum A, Mathalon DH, Sullivan EV, Rawles JM, Zipursky RB, Lim KO. A quantitative magnetic resonance imaging study of changes in brain morphology from infancy to late adulthood. Arch Neurol 1994; 51:874–887. 28. Matsumae M, Kikinis R, Maorocz IA, et al. Age-related changes in intracranial com- partment volumes in normal adults as- sessed by magnetic resonance imaging. J Neurosurg 1996; 84:982–991. 29. Thorndike RL, Hagen EP, Sattler JM. The Volume 216 z Number 3 Brain Development and Aging: Quantitative Analysis at MR Imaging z 681 Stanford-Binet intelligence scale. 4th ed. Chicago, Ill: Riverside, 1986. 30. Wechsler D. Wechsler Intelligence Scale for Children. 3rd ed. San Antonio, Tex: Psychological Corporation, 1991. 31. Wechsler D. Wechsler Intelligence Scale for Children-Revised (WISC-R). San Anto- nio, Tex: Psychological Corporation, 1974. 32. Wechsler D. Manual for the Wechsler Adult Intelligence Scale. 3rd ed. San An- tonio, Tex: Psychological Corporation, 1997. 33. Wechsler D. Wechsler Adult Intelligence Scale-Revised. San Antonio, Tex: Psycho- logical Corporation, 1981. 34. Delis D, Kramer JH, Kaplan E, Ober BA. California Verbal Learning Test: adult version. San Antonio, Tex: Psychological Corporation, 1987. 35. Vannier MW, Butterfield RL, Jordan D, Murphy WA, Levitt RG, Gado M. Multi- spectral analysis of magnetic resonance images. Radiology 1985; 154:221–224. 36. Jackson EF, Narayana PA, Falconer JC. Re- producibility of nonparametric feature map segmentation for determination of normal human intracranial volumes with MR imaging data. J Magn Reson Imaging 1994; 4:692–700. 37. Frontera JG. Evaluation of the immediate effects of some fixatives upon the mea- surements of the brains of macaques. J Comp Neurol 1958; 109:417–438. 38. Stephan H. Methodische studien uber den quantitativen vergleich architekto- nischer struktureinheiten des gehirns. Z Wiss Zool 1960; 164:1–2. 39. Dixon WJ, Brown MB, Engelman L, Jennrich R. BMDP statistical software manual. Berkeley, Calif: University of California Press, 1990. 40. Cohen J, Cohen P. Applied multiple re- gression/correlation analysis for the be- havioral sciences. 2nd ed. Hillsdale, NJ: Lawrence Erlbaum, 1983. 41. Chrzanowska G, Beben A. Weight of the brain and body height in man between the ages of 20 and 89 years. Folia Mor- phol (Warsz) 1973; 32:391–406. 42. Dekaban AS, Sadowsky D. Changes in brain weights during the span of human life: relation of brain weights to body heights and body weights. Ann Neurol 1978; 4:345–356. 43. Pakkenberg H, Voight J. Brain weight of the Danes. Acta Anat (Basel) 1964; 56: 297–307. 44. Craig CF. Laboratory methods of the U.S. Army. Philadelphia, Pa: Lea and Febiger, 1929. 45. Feer E. Textbook of pediatrics. Philadel- phia, Pa: Lippincott, 1922. 46. Coppoletta JM, Wolbach SB. Body length and organ weights of infants and chil- dren: a study of the body length and nor- mal weights of the more important vital organs of the body between birth and twelve years of age. Am J Pathol 1933; 9:55–70. 47. Dam AM. The density of neurons in the human hippocampus. Neuropathol Appl Neurobiol 1979; 5:249–264. 48. Ho K, Roessman U, Straumfjord JV, Mon- roe G. Analysis of brain weight. I. Adult brain weight in relation to sex, race and age. Arch Pathol Lab Med 1980; 104:635– 639. 49. Anton G. Gehirnvermessung mittelst des Kompensations-Polar-Planimeters. Wien Klin Wochenschr 1903; 16:1263–1267. 50. Jaeger R. Inhaltsberechnungen der Rin- den-und Mark-substanz des Grosshirns durch planimetrische Messungen. Arch Psychiatr Nervenkr 1914; 54:261–272. 51. Filipek PA, Richelme C, Kennedy DN, Caviness VS Jr. The young adult human brain: an MRI-based morphometric anal- ysis. Cereb Cortex 1994; 4:344–360. 52. Narayana PA, Borthakur A. Effect of radio frequency inhomogeneity correction on the reproducibility of intra-cranial vol- umes using MR image data. Magn Reson Med 1995; 33:396–400. 53. Giedd JN, Snell JW, Lange N, et al. Quan- titative magnetic resonance imaging of human brain development: ages 4–18. Cereb Cortex 1996; 6:551–560. 54. Reiss AL, Abrams MT, Singer HS, Ross JL, Denckla MB. Brain development, gender and IQ in children: a volumetric imaging study. Brain 1996; 119:1763–1774. 55. Huppi PS, Warfield S, Kikinis R, et al. Quantitative magnetic resonance imag- ing of brain development in premature and mature newborns. Ann Neurol 1998; 43:224–235. 56. Schade JP, van Groenigen WB. Structural organization of the human cerebral cor- tex. I. Maturation of the middle frontal gyrus. Acta Anat 1961; 47:72–111. 57. Huttenlocher PR, Dabholkar AS. Regional differences in synaptogenesis in human cerebral cortex. J Comp Neurol 1997; 387:167–178. 58. Shankle WR, Romney AK, Landing BH, Hara J. Developmental patterns in the cy- toarchitecture of the human cerebral cor- tex from birth to 6 years examined by correspondence analysis. Proc Natl Acad Sci USA 1998; 95:4023–4028. 59. Chugani HT, Phelps ME, Mazziotta JC. Positron emission tomography study of human brain functional development. Ann Neurol 1987; 22:487–497. 60. Courchesne E. Event-related brain poten- tials: comparison between children and adults. Science 1977; 197:589–592. 61. Courchesne E. Chronology of postnatal human brain development: event-related potential, positron emission tomogra- phy, myelino-genesis, and synaptogen- esis studies. In: Rohrbaugh JW, Parasura- man R, Johnson R, eds. Event-related brain potentials: basic issues and applica- tions. New York, NY: Oxford Press, 1990; 210–241. 62. Pujol J, Vendrell P, Junquae C, Martai- Vilalta JL, Capdevila A. When does hu- man brain development end? evidence of corpus callosum growth up to adulthood. Ann Neurol 1993; 34:71–75. 63. de Leon MJ, Convit A, DeSanti S, et al. Contribution of structural neuroimaging to the early diagnosis of Alzheimer’s dis- ease. Int Psychogeriatr 1997; 9(suppl 1): 183–190. 64. Pfefferbaum A, Sullivan EV, Mathalon PK, Shear PK, Rosenbloom MJ, Lim KO. Lon- gitudinal changes in magnetic resonance imaging brain volumes in abstinent and relapsed alcoholics. Alcohol Clin Exp Res 1995; 19:1177–1191. 65. Giedd JN, Blumenthal J, Jeffries NO, et al. Brain development during childhood and adolescence: a longitudinal MRI study. Nature Neurosci 1999; 2:861–863. 66. Giuffrida S, Saponara R, Restivo DA, et al. Supratentorial atrophy in spinocerebellar ataxia type 2: MRI study of 20 patients. J Neurol 1999; 246:383–388. 67. Milhorat TH, Chou MW, Trinidad EM, et al. Chiarai 1 malformation redefined: clinical and radiographic findings for 364 symptomatic patients. Neurosurgery 1999; 44:1005–1017. 68. Matsumae M, Kikinis R, Morocz I, et al. Intracranial compartment volumes in pa- tients with enlarged ventricles assessed by magnetic resonance-based image pro- cessing. J Neurosurg 1996; 84:972–981. 682 z Radiology z September 2000 Courchesne et al