Quantifying Cerebral Changes in Adolescence With MRI and Deformation Based Morphometry

Transcription

1 JOURNAL OF MAGNETIC RESONANCE IMAGING 28: (2008) Original Research Quantifying Cerebral Changes in Adolescence With MRI and Deformation Based Morphometry William R. Riddle, PhD, 1 * Susan C. DonLevy, MSN, MPH, 2,3 Curtis A. Wushensky, MD, 1 Benoit M. Dawant, PhD, 4 J. Michael Fitzpatrick, PhD, 4 and Ronald R. Price, PhD 1 Purpose: To identify and quantify structural changes in the maturing brain between childhood and adolescence. Materials and Methods: Two three-dimensional T1- weighted MR volumes of the brain were acquired from eight subjects, 6 to 7 years apart. The subjects were 9 to 12 years old on the first scan and 15 to 19 years old on the second scan. The MR scans were converted to one millimeter isotropic volumes, globally aligned with a rigid transform, inhomogeneity corrected, and nonrigid deformation fields between the aligned volumes were calculated. Masks for brain regions were automatically warped with the deformation fields and volumes of brain regions calculated. Color overlays based on the nonrigid deformation fields were generated to identify local volume changes. Results: Gray matter decreased as much as 60% and white matter increased as much as 250%. The biggest gray matter changes were in the head of the caudates, areas of the putamens, and areas of the thalamus. Some of the biggest white matter changes were in the forceps minor, forceps major, and internal capsule. Conclusion: Deformation based morphometry with serial scans provides a method to study regional structural changes with brain growth and maturation. Key Words: adolescence; brain growth; deformation based morphometry J. Magn. Reson. Imaging 2008;28: Wiley-Liss, Inc. 1 Department of Radiology and Radiological Sciences, Vanderbilt University, Nashville, Tennessee. 2 St. Thomas Research Institute, Nashville, Tennessee. 3 Department of Pediatrics, Vanderbilt University, Nashville, Tennessee. 4 Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, Tennessee. *Address reprint requests to: W.R.R., R-1311, Medical Center North, Vanderbilt University Medical Center, Nashville, TN bill.riddle@vanderbilt.edu Received November 13, 2007; Accepted April 7, DOI /jmri Published online in Wiley InterScience ( WHILE IT IS WELL KNOWN that there are cognitive, behavioral, and emotional differences when children become teenrs, little is known about structural brain changes during this time. Autopsy studies have shown that, by 10, the weight of the brain is approximately 95% of adult values and that male brains are 10% heavier than female brains (1). Cross-sectional magnetic resonance (MR) imaging studies (2 5) have demonstrated that gray matter decreases and white matter increases as the brain matures, but these studies do not identify the specific areas that are changing. In another cross-sectional study, Sowell et al (6) showed gray matter volumes in the frontal, parietal, and temporal s decreasing and volumes of white matter in the frontal, parietal, and occipital s increasing from 6 to 18. Im registration is the process of establishing spatial correspondence between two ims or im volumes, and can be achieved with rigid and/or nonrigid transforms (7). A rigid transform accounts for differences in positioning and a nonrigid transform describes shape changes due to differences between individuals or shape changes associated with growth and atrophy (8). A transformation is called rigid when only translations and rotations are allowed. If the transformation also includes scaling and shear factors that maps parallel lines onto parallel lines, it is called affine (7). Rigid and affine transforms are global transforms because they are applied uniformly to all the voxels. With three-dimensional volumes, a rigid transform uses 6 parameters (three translation vectors and three rotation angles) and an affine transform uses 12 parameters (3 translation vectors, 3 rotation angles, 3 scaling factors, and 3 shearing factors). Several different types of nonrigid transforms have been described in the literature, including matching intensities without using an explicit physical model for deformations (9), matching the spatial coordinates of one volume to those of another volume with discrete cosine basis functions (10), elastic matching (11), viscous fluid registration algorithms (12), curvature-based registration (13), free-form deformation based on B- splines (14), deformation using B-splines with constraint terms (15), deformations with radial basis functions (16), and demons registration (17,18). Demons registration, which uses optical flow, is available in the Insight Toolkit (ITK) (19). With serial scans in a fixed coordinate system, the deformation at time t is s 3 s U(s,t), where s is the position (x, y, z) in the first volume and U is the displacement vector (20). The local volume change of the deformation in the neighborhood of position s at time t 2008 Wiley-Liss, Inc. 320

2 Cerebral Changes in Adolescence 321 is determined by the Jacobian, which is defined as J(s,t) det I U/ s) where I denotes an identity matrix and U/ s is calculated from the 3 3 displacement gradient matrix of U, and is given by: U s s,t Ux x Uy x Uz x Ux y Uy y Uz y Ux z Uy z Uz z The nine components of U/ s form scalar fields used to measure the second-order morphological variability. When the first im volume is warped to the second im volume, the Jacobian (J) of the deformation field at position s is equal to the ratio of the volume in the first im volume (V1) divided by the corresponding volume in the second im volume (V2), or J V1 / V2. This volume ratio can be expressed as the percent volume change with respect to the first im (change 1/J 1). Deformation based morphometry, also called tensor based morphometry, deformation tensor morphometry, and voxel compression mapping, has been used to study brain changes. Reports include brain atrophy in HIV/ AIDS (21), brain atrophy in semantic dementia (22,23), brain atrophy in frontotemporal dementia (24,25), brain atrophy in Alzheimer s disease (26 28), growth patterns in developing brains (20,27,29), mapping brain volumes and brain differences in premature infants (30,31), brain abnormalities in Fragile X syndrome (32), brain abnormalities in Williams syndrome (33), brain changes in alcoholism (34), brain changes in schizophrenia (27), and cerebral differences between 7-year-old twins that had discordant growth in utero (35). However, results from deformation based morphometry can vary depending on the nonrigid deformation method (15). In this report, demons based registration (17 19) was used first to warp masks for brain regions from a reference scan to the other scans, then the Jacobian of the deformation field between longitudinal scans was used to quantify 1 mm 3 volume changes in the cerebrum. Figure 1. Illustrates the first six processing steps for a midsagittal slice from subject #2. Step 1 converts the sampled im volumes to isotropic volumes with 1-mm resolution. Step 2 globally registers the im volumes. Step 3 applies the N3 inhomogeneity correction. Step 4 calculates the nonrigid deformation field between the reference volume and the subject s first volume, then automatically warps the masks from the reference subject to the subject s first volume with the deformation field. Step 5 calculates the nonrigid deformation field between the subject s first volume and the subject s second volume, then automatically warps the masks from the subject s first volume to the subject s second volume. Step 6 segments the subject s whole brain volumes into CSF, gray matter, and white matter. METHODS Subjects MR ims used for this investigation were obtained from children who agreed to participate in a study to evaluate structural brain anatomy in premature children. Two longitudinal scans were obtained from eight subjects, 6 to 7 years apart. Seven sets of ims we analyzed were obtained from full-term controls and one set of ims was obtained from a child born extremely premature (subject #5). The subjects, five female and three male, were 9 to 12 years old on the first scan (aver 11.0 years) and 15 to 19 years old on the second scan (aver 17.5 years). All of the children were healthy without history of head trauma or meningitis. The full-term control subjects were recruited from university families and all of the parents had college education. The subjects had normal IQ sores, were academically successful, and have pursued college educations. Written informed consent was obtained from the subjects and parents (if under 18) before each scan. MR Im Acquisition Three-dimensional MR volumes were acquired on 1.5 Tesla (T) GE Signa scanners (GE Medical Systems, Mil-

3 322 Riddle et al. Figure 2. Five views of a transverse slice through the genu and splenium of the corpus callosum for subject #2. A: A midsagittal slice from the second scan with a horizontal line indicating the location of the transverse slice. B,C: A slice from the first scan (B) and a slice from the second scan (C). D: The transverse slice from the second scan and color overlays for gray matter changes. The color overlays show gray matter decreases from 18% to 60% in the caudates, medial thalamus, frontal, limbic, occipital, and temporal. E: The transverse slice from the second scan and color overlays for white matter changes. The color overlays show white matter increases from 21% to over 250% in the genu and splenium of the corpus callosum, the anterior region of the corona radiata (forceps minor), the internal capsule, and the posterior region of the corona radiata (forceps major). waukee, WI) with a 3D T1-weighted sequence (SPoiled Gradient Recalled, sagittal slices with no gaps, inversion time 400 ms, echo time 3.6 ms, repetition time 21 ms, one excitation, flip angle 20, matrix ). The first im volumes were acquired with fields of view (FOV) of mm. The second im volumes (6 to 7 years later) were acquired with FOV mm. Ims were transferred to a SUN Blade 1000 workstation (Sun Microsystems, Palo Alto, CA) for processing. Defining Brain Regions The first scan of subject #8 (from control group) was selected as the reference because he had good head placement in the scanner. Masks for frontal, limbic (cingular gyrus), occipital, parietal, temporal, sub lobar region (corpus callosum, caudate, putamen, globus pallidus, thalamus), cerebrum, cerebellum, medulla, midbrain, pons, and whole brain were created from manual tracings with custom software. These masks include cerebral spinal fluid (CSF), gray matter, and white matter. Processing of MR Volumes The two MR volumes for each subject were processed with seven steps. Figure 1 illustrates the first six processing steps for a midsagittal slice from subject #2. Figure 3. Transverse slices through the genu and splenium of the corpus callosum for all eight subjects showing volume increases/decreases in the white matter. Seven subjects showed volume increases in the forceps minor (ellipse) while subject #1 showed decreases. All subjects showed volume increases in the internal capsule (rectangle).

4 Cerebral Changes in Adolescence 323 Figure 5. Superior, anterior, and left lateral views of surface plots of gray matter decreases in the sub lobar region of subject #2, generated with voxels containing contractions from 34% to 60%. These voxels are located in the head of the caudates, areas of the putamens, and areas of the thalamus. As a reference, 1-mm sagittal and coronal slices of the white matter are shown in each view. Figure 4. Gray matter volume changes of cerebral regions for subject #2. The subject was 9.4 years old on the first scan and 15.9 years old on the second scan. 1. Convert the im volumes to 1-mm isotropic volumes with trilinear interpolation. 2. Globally register the subject s first im volume to the reference (first scan of subject #8) with a rigid transform (three translation vectors and three rotation angles). Globally register the subject s second im volume to the subject s first globally registered im volume with a rigid transform. 3. Apply N3 inhomogeneity correction (36) from MIPAV (37) to the globally aligned im volumes. 4. Calculate the nonrigid deformation field between the reference s N3 corrected volume and the subject s first N3 corrected volume with demons registration (17 19). Automatically warp the masks (frontal, limbic, occipital, parietal, temporal, sub lobar region, cerebrum, cerebellum, medulla, midbrain, pons, and whole brain) from the reference subject to the subject s first N3 corrected volume. 5. Calculate the nonrigid deformation field between the subject s first N3 corrected volume and the subject s second N3 corrected volume with demons registration. Automatically warp the masks (frontal, limbic, occipital, parietal, temporal, sub lobar region, cerebrum, cerebellum, medulla, midbrain, pons, and whole brain) from the subject s first N3 corrected volume to the subject s second N3 corrected volume. 6. Segment the N3 corrected whole brain volumes into CSF-Gray Matter-White Matter with Fuzzy C- means from MIPAV (37). While Fuzzy C-means worked well for separating gray matter from white matter, it did not do well separating CSF from gray matter. Thresholds between CSF and gray matter were manually selected by viewing a sagittal slice through the lateral ventricles with IMAGEJ (38). Create segmented volumes based on these thresholds. 7. Calculate the Jacobian of the nonrigid deformation field between the subject s first N3 corrected im volume and the subject s second N3 corrected im volume (from step 5). Map the Jacobian into 13 different bins and display bins 1 6 and 8 13 as color overlays on the MR slices. After these seven processing steps, each 1 mm 3 voxel in the brain is assigned a tissue type, a brain region, a Jacobian value, and an expansion/contraction color overlay. Color Overlays The same colors that were previously reported (35) were used in this study, but the order of the colors was Table 1 Cerebral Gray Matter Volume Changes by Region Subject Gender 1st 2nd Limbic Sub lobar region Occipital Parietal Temporal Frontal Cerebrum 1 F % 10.6% 7.7% 9.8% 2.1% 4.5% 5.5% 2 F % 19.8% 5.5% 10.0% 5.0% 5.4% 7.0% 3 F % 11.4% 1.7% 7.0% 1.0% 3.8% 3.9% 4 F % 8.9% 7.3% 14.4% 4.5% 8.5% 8.6% 5 F % 1.2% 4.1% 7.8% 1.5% 6.2% 5.1% 6 M % 10.3% 13.2% 17.3% 8.5% 12.7% 12.3% 7 M % 7.3% 2.3% 10.5% 3.4% 9.0% 6.8% 8 M % 0.8% 2.7% 8.4% 0.5% 5.4% 3.9% Aver % 8.6% 5.6% 10.7% 3.3% 6.9% 6.6% SD % 6.4% 3.8% 3.5% 2.7% 3.0% 2.8%

5 324 Riddle et al. Figure 6. White matter volume changes of cerebral regions for subject #2. The subject was 9.4 years old on the first scan and 15.9 years old on the second scan. changed. This reordering produced contrasting colors between adjacent bins, making the volume increases/ decreases easier to quantify. Colors C01 C06 indicate volume decreases of 18% to greater than 71%. Colors C07 C12 indicate volume increases of 21% to over 250%. Volume changes between 17% and 20% were displayed as the gray scale im. The volume ranges are symmetrical for expansion and contraction, but not for percent change. A volume change of 200% (Jacobian 3) means that the volume from the first im has increased by a factor of three in the second im. A volume change of 68% (Jacobian 1/3) means that the volume from the first im has decreased by a factor of three in the second im. RESULTS Figure 2 shows five views of a transverse slice through the genu and splenium of the corpus callosum for subject #2 with color overlays showing volume increases/ decreases in the gray and white matter. Figure 3 shows the same transverse slice through the genu and splenium of the corpus callosum for all eight subjects with volume increases/decreases in the white matter. Seven subjects showed volume increases in the forceps minor while subject #1 showed decreases. It is not known why this subject showed decreases. All subjects showed volume increases in the internal capsule. While Figures 2 and 3 shows the specific locations of contractions and expansions in the gray and white matter, it requires 125 to 140 one-millimeter transverse slices to cover the brain. It is desirable to summarize the results. As a representation of the typical findings, Figure 4 shows the gray matter volumes of cerebral regions in subject #2, with gray matter decreasing in all regions from 9.4 to Table 1 lists the percent change of gray matter in the cerebral regions for all eight subjects. Figure 5 shows superior, anterior, and left lateral views of gray matter decreases in the sub lobar region for subject #2, generated with voxels that decreased from 34% to 60%. These voxels are located in the caudates, putamens, and thalamus. Figure 6 shows the white matter volumes of cerebral regions in subject #2. White matter volumes increased in all regions except for the limbic. Table 2 lists the percent change of white matter volume in each region for all eight subjects. Figure 7 shows superior, anterior, and left lateral views of white matter expansions in the cerebrum of subject #2, generated with voxels that increased from 101% to greater than 250%. These voxels are located in the anterior region of the corona radiata (forceps minor), the internal capsule, the superior region of internal capsule, the inferior longitudinal fasciculus, and the posterior region of the corona radiata (forceps major). To assess the validity of the deformation based morphometry processing, two three-dimensional im volumes from a different study, taken 92 minutes apart, were analyzed. The maximum value of the Jacobian of the deformation field was ( 6.0% change) and the minimum value was (7.8% change). Figure 8 shows five views of a transverse slice through the genu and splenium of the corpus callosum for this subject. Because the contractions were less than 18% and the expansions were less than 21%, there are no color overlays for the gray or white matter changes. DISCUSSION Rohlfing (15) evaluated warping of an electronic phantom with several nonrigid registration algorithms and Table 2 Cerebral White Matter Volume Changes by Region Subject Gender 1st 2nd Limbic Sub lobar region Occipital Parietal Temporal Frontal Cerebrum 1 F % 21.0% 17.9% 11.2% 2.4% 10.6% 11.5% 2 F % 28.7% 5.2% 3.2% 21.0% 17.6% 15.2% 3 F % 19.3% 4.5% 6.7% 12.6% 12.3% 11.2% 4 F % 5.2% 4.2% 4.9% 0.1% 7.3% 3.5% 5 F % 3.0% 2.0% 3.1% 1.6% 8.4% 4.8% 6 M % 10.0% 25.9% 22.2% 27.9% 26.8% 23.9% 7 M % 3.3% 10.4% 15.0% 12.2% 13.8% 12.1% 8 M % 1.8% 18.5% 17.5% 11.8% 25.1% 16.6% Aver: % 11.1% 10.0% 10.5% 11.2% 15.2% 12.4% SD % 10.7% 10.0% 7.2% 9.8% 7.4% 6.5%

6 Cerebral Changes in Adolescence 325 Figure 7. Superior, anterior, and left lateral views of surface plots of white matter increases in subject #2, generated with voxels containing increases from 101% to greater than 250%. These voxels are located in the anterior region of the corona radiata (forceps minor), the internal capsule, the superior region of internal capsule, the inferior longitudinal fasciculus, and the posterior region of the corona radiata (forceps major). As a reference, 1-mm sagittal and coronal slices of the white matter are shown in each view. noted that free-form deformation based on B-splines (14) produced a noticeable geometric structure in the deformation field. Some of the warped cerebral volumes shown by Boardman et al (30), using free-form deformation based on B-splines (14), extend outside the skull. This suggests that there may be artifacts when warping with free-form deformation based on B-splines. In the present study, masks for brain regions in the reference subject were generated as adjacent regions without overlaps. After warping the reference masks to each of the other im volumes with demons registration, there were no overlaps or regions outside the skull. Midsagittal outlines of the masks (frontal, parietal, occipital, limbic, sub lobar, cerebellum, midbrain, pons, and medulla) for the reference subject and the warped masks for subject #2 are shown in Figure 1. Using demons registration, we have had success warping brain masks between T1-weighted im volumes acquired with different imaging parameters, acquired on magnets from different manufacturers, and acquired on magnets with different field strengths (1.5T and 3.0T). However, for deformation based morphometry, the two imaging volumes must be acquired with the same type coil, the same imaging sequence, and the same imaging parameters. While most studies using deformation based morphometry illustrate significance maps, some have also quantified the volume changes. In adult brains, Rohlfing et al (34) showed volume changes of 22% to 28% in alcoholics, Thompson et al (27) showed volume changes of 0% to 15% in Alzheimer s disease, and Leow et al (39) showed volume changes of 30% to 40% in a semantic dementia patient. In developing brains, Thompson et al (29) showed volume changes of 20% to 80% in a midsagittal slice of the corpus callosum. As shown in Figures 2 and 3, the biggest white matter changes were not in the midsagittal corpus callosum. The surface plots in Figure 7, generated with white matter voxels that increased from 101% to greater than 250%, did not show the corpus callosum increasing by this amount. Decreases in gray matter may reflect the dendrite pruning process and increases in white matter volumes could be due to myelination, increases in axonal size, and/or glial cell proliferation (5). However, these mechanisms are at the cellular level and cannot be determined with 1-mm resolution. While this study had only eight subjects, the techniques we have presented can be applied to other data sets, such as the National Institutes of Health Pediatric Database (40), which will have three longitudinal MR scans for each subject spaced at 2-year intervals. The first MR scans became available in the summer of 2007 with 409 3D T1-weighted scans from subjects ranging from 4.7 to 18.3 years of. The second and third Figure 8. Five views of a transverse slice through the genu and splenium of the corpus callosum of a 32-year-old subject. A: A midsagittal slice from the second scan with a horizontal line indicating the location of the transverse slice. B,C: A slice from the first scan (B) and a slice from the second scan (C). The two scans were acquired 92 min apart. Because the contractions were less than 18% and the expansions were less than 21%, there are no color overlays for the gray matter (8D) or white matter (8E) changes.

7 326 Riddle et al. scans will be available in Using deformation based morphometry with this data set will help to establish when volumes of brain regions, gray matter, and white matter change. In conclusion, we have presented a technique to quantify grey matter and white matter changes with longitudinal scans. Understanding the normal maturational changes of the brain may lead to a better understanding of the interruption of brain growth by pathological processes. REFERENCES 1. Dekaban AS. Changes in brain weights during the span of human life: relation of brain weights to body heights and body weights. Ann Neurol 1978;4: Jernigan TL, Tallal P. Late childhood changes in brain morphology observable with MRI. Dev Med Child Neurol 1990;32: Jernigan TL, Trauner DA, Hesselink JR, Tallal PA. Maturation of human cerebrum observed in vivo during adolescence. Brain 1991; 114(Pt 5): Reiss AL, Abrams MT, Singer HS, Ross JL, Denckla MB. Brain development, gender and IQ in children. A volumetric imaging study. Brain 1996;119(Pt 5): De Bellis MD, Keshavan MS, Beers SR, et al. Sex differences in brain maturation during childhood and adolescence. Cereb Cortex 2001;11: Sowell ER, Trauner DA, Gamst A, Jernigan TL. Development of cortical and subcortical brain structures in childhood and adolescence: a structural MRI study. Dev Med Child Neurol 2002;44: Maintz JB, Viergever MA. A survey of medical im registration. Med Im Anal 1998;2: Schnabel JA, Tanner C, Castellano-Smith AD, et al. Validation of nonrigid im registration using finite-element methods: application to breast MR ims. IEEE Trans Med Imaging 2003;22: Collins DL, Holmes CJ, Peters TM, Evans AC. Automatic 3-D model-based neuroanatomical segmentation. Hum Brain Mapp 1995; 3: Ashburner J, Friston KJ. Nonlinear spatial normalization using basis functions. Hum Brain Mapp 1999;7: Bajcsy R, Kovacic S. Multiresolution elastic matching. Comput Vision Graph Imag Processing 1989;46: Christensen GE, Rabbitt RD, Miller MI. Deformable templates using large deformation kinematics. IEEE Trans Im Processing 1996;5: Fischer B, Modersitzki J. Curvature based im registration. J Math Imaging Vision 2003;18: Rueckert D, Sonoda LI, Hayes C, Hill DL, Leach MO, Hawkes DJ. Nonrigid registration using free-form deformations: applications to breast MR ims. IEEE Trans Med Imaging 1999;18: Rohlfing T. Transformation model and constraints cause bias in statistics on deformation fields. Med Im Comput Comput Assist Interv Int Conf Med Im Comput Comput Assist Interv 2006;9(Pt 1): Rohde GK, Aldroubi A, Dawant BM. The adaptive bases algorithm for intensity based nonrigid im registration. IEEE Trans Med Imaging 2003;22: Thirion JP. Im matching as a diffusion process: an analogy with Maxwell s demons. Med Im Anal 1998;2: Dawant BM, Hartmann SL, Thirion J-P, Vandermeulen D, Demaerei P. Automatic 3D segmentation of internal structures of the head in MR ims using a combination of similarity and free-form transformations: I. Methodology and validation on normal subjects. IEEE Trans Med Imaging 1999;18: Insight Segmentation and registration toolkit, sponsored by the National Library of Medicine. Accessed 29 August Chung MK, Worsley KJ, Paus T, et al. A unified statistical approach to deformation-based morphometry. Neuroim 2001;14: Chiang MC, Dutton RA, Hayashi KM, et al. 3D pattern of brain atrophy in HIV/AIDS visualized using tensor-based morphometry. Neuroim 2007;34: Studholme C, Cardenas V, Blumenfeld R, et al. Deformation tensor morphometry of semantic dementia with quantitative validation. Neuroim 2004;21: Brambati SM, Rankin KP, Narvid J, et al. Atrophy progression in semantic dementia with asymmetric temporal involvement: a tensor-based morphometry study. Neurobiol Aging [Epub ahead of print]. 24. Brambati SM, Renda NC, Rankin KP, et al. A tensor based morphometry study of longitudinal gray matter contraction in FTD. Neuroim 2007;35: Cardenas VA, Boxer AL, Chao LL, et al. Deformation-based morphometry reveals brain atrophy in frontotemporal dementia. Arch Neurol 2007;64: Janke AL, de Zubicaray G, Rose SE, Griffin M, Chalk JB, Galloway GJ. 4D deformation modeling of cortical disease progression in Alzheimer s dementia. Magn Reson Med 2001;46: Thompson PM, Hayashi KM, Sowell ER, et al. Mapping cortical change in Alzheimer s disease, brain development, and schizophrenia. Neuroim 2004;23(Suppl 1):S2 S Fox NC, Crum WR, Scahill RI, Stevens JM, Janssen JC, Rossor MN. Imaging of onset and progression of Alzheimer s disease with voxelcompression mapping of serial magnetic resonance ims. Lancet 2001;358: Thompson PM, Giedd JN, Woods RP, MacDonald D, Evans AC, Toga AW. Growth patterns in the developing brain detected by using continuum mechanical tensor maps. Nature 2000;404: Boardman JP, Counsell SJ, Rueckert D, et al. Early growth in brain volume is preserved in the majority of preterm infants. Ann Neurol 2007;62: Boardman JP, Counsell SJ, Rueckert D, et al. Abnormal deep grey matter development following preterm birth detected using deformation-based morphometry. Neuroim 2006;32: Lee AD, Leow AD, Lu A, et al. 3D pattern of brain abnormalities in Fragile X syndrome visualized using tensor-based morphometry. Neuroim 2007;34: Chiang MC, Reiss AL, Lee AD, et al. 3D pattern of brain abnormalities in Williams syndrome visualized using tensor-based morphometry. Neuroim 2007;36: Rohlfing T, Sullivan EV, Pfefferbaum A. Deformation-based brain morphometry to track the course of alcoholism: differences between intra-subject and inter-subject analysis. Psychiatry Res 2006;146: Riddle WR, Li R, Fitzpatrick JM, DonLevy SC, Dawant BM, Price RR. Characterizing changes in MR ims with color-coded Jacobians. Magn Reson Imaging 2004;22: Sled JG, Zijdenbos AP, Evans AC. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans Med Imaging 1998;17: MIPAV (Medical im processing, analysis, and visualization), Center for Information Technology, U.S. National Institutes of Health. Accessed 29 August IMAGEJ (Im processing and analysis in Java), National Institutes of Health. Accessed 29 August Leow AD, Yanovsky I, Chiang MC, et al. Statistical properties of Jacobian maps and the realization of unbiased large-deformation nonlinear im registration. IEEE Trans Med Imaging 2007;26: Evans AC. The NIH MRI study of normal brain development. Neuroim 2006;30: