www.elsevier.com/locate/ynimg

NeuroImage 41 (2008) 223–232

Diffusion tensor imaging of normal white matter maturation from latechildhood to young adulthood: Voxel-wise evaluation of meandiffusivity, fractional anisotropy, radial and axial diffusivities, andcorrelation with reading development

Deqiang Qiu,a Li-Hai Tan,b Ke Zhou,c and Pek-Lan Khonga,⁎

aDepartment of Diagnostic Radiology, State Key Laboratory of Brain and Cognitive Sciences, The University of Hong Kong, Hong KongbDepartment of Linguistics, State Key Laboratory of Brain and Cognitive Sciences, The University of Hong Kong, Hong KongcInstitute of Biophysics, Chinese Academy of Sciences, Beijing, China

Received 16 October 2007; revised 13 February 2008; accepted 13 February 2008Available online 29 February 2008

Using diffusion tensor MR imaging (DTI) and advanced voxel-wiseanalysis tools, we study diffusivity and anisotropy changes of whitematter from late childhood to young adulthood, and correlate quan-titative diffusion indices with Chinese and English reading performancescores. Seventy-five normal healthy school going ethnic Chinese stu-dents and young adults of three age groups were recruited (group 1,n=24, mean±SD=7.4±0.3 years; group 2, n=27, mean±SD=10.3±0.5 years; group 3, n=24, mean±SD=22.8±2.3 years). DTI wasperformed with 3 mm isotropic resolution to cover the entire brain.Voxel-wise analysis was performed using Tract-Based Spatial Statistics(TBSS) to localize regions of whitematter showing significant changes offractional anisotropy (FA), mean diffusivity (MD), and axial and radialdiffusivities between groups.We found increased FA and decreasedMDwith increasing age in regions of cerebellar white matter, right temporalwhite matter, and a large portion of the superior frontal and parietalwhite matter driven by both the reduction of radial diffusivity and axialdiffusivity with the former to a greater extent. Changes were continualfrom late childhood to young adulthood. Findings were confirmed byregion-of-interest analysis in specific white matter tracts. After con-trolling for the effect of age, significant correlation was found betweendiffusion indices of the anterior limb of the left internal capsule andChinese reading score (p=0.05), and of the corona radiata and Englishreading score (p=0.026 and p=0.029 for left and right, respectively).These DTI indices likely reflect the multiple biological processes thatoccur during brain development which provide the neural substrate forongoing functional connectivity including for reading development.

© 2008 Elsevier Inc. All rights reserved.

⁎ Corresponding author. Department of Diagnostic Radiology, Blk. K, Rm406, Queen Mary Hospital, The University of Hong Kong, 102 PokfulamRoad, Hong Kong. Fax: +852 28551652.

E-mail address: plkhong@hkucc.hku.hk (P.-L. Khong).Available online on ScienceDirect (www.sciencedirect.com).

1053-8119/$ - see front matter © 2008 Elsevier Inc. All rights reserved.doi:10.1016/j.neuroimage.2008.02.023

Introduction

The human brain continues to undergo complex and prolongeddevelopment after birth. The study of this process can help identifykey maturational milestones and provide a valuable reference withwhich various pathologies can be compared to and characterized(Huppi and Dubois, 2006; Toga et al., 2006). Multiple MR imagingtechniques have been applied to evaluate noninvasively the variousaspects of normal brain development. Volumetric studies have de-monstrated that gray matter volume increases dramatically until latechildhood (6–9 years) and subsequently declines gradually, whilewhite matter volume increases dramatically up to adolescence (12–15 years) and continues to increase at a slower rate until as late as the4th decade of life (Courchesne et al., 2000). It is also known thatbrain maturation follows a spatially and temporally ordered pattern;with cortical maturation generally following a posterior to anteriorgradient (Gogtay et al., 2004; Sowell et al., 2002; Toga et al., 2006),and white matter tract myelination also following distinct temporalpatterns (Barkovich et al., 1988; Kinney et al., 1988).

Diffusion tensor MR imaging (DTI) can estimate water diffu-sivities in tissues along the three orthogonal principle diffusiondirections (Basser et al., 1994). It is unique in its sensitivity toalterations of tissue microstructure and the ability to illustrate neuralpathways (Wakana et al., 2004). In structured tissue such as whitematter, water diffusion is less restricted along the fiber than perpen-dicular to it due to barriers such as axon membrane and myelin. Theoverall diffusivity and the degree of directionality of diffusion in atissue can be quantified by mean diffusivity (MD) and fractionalanisotropy (FA), respectively (Beaulieu, 2002). Due to its superbsensitivities to whiter matter changes, DTI has been used to studybrain maturation from infancy to childhood and until adulthood(Ben et al., 2005; Huang et al., 2006; Huppi et al., 1998; Mukherjeeet al., 2001; Schneider et al., 2004). It has been found that white

224 D. Qiu et al. / NeuroImage 41 (2008) 223–232

matter maturation is manifested as nonlinearly increased FA anddecreased mean diffusivity (MD) with the most dramatic changestaking place during the first 2–3 years of life. Several studies havebeen carried out to investigate brain maturation specifically fromlate childhood to adolescence and young adulthood (Barnea-Goralyet al., 2005; Bonekamp et al., 2007; Giorgio et al., 2008; Klingberget al., 1999; Schmithorst et al., 2002; Snook et al., 2005, 2007). Thisperiod is critical for emotional and cognitive development and theongoing maturation of neural pathways is important in thedevelopment of functional connectivity. It is generally agreed thatwhite matter maturation during this period is characterized bycontinual widespread changes of decreased MD and increased FA.Apart from quantification of FA andMD, recent studies have shownthat directional diffusivities such as axial diffusivity (presumed tobe the diffusivity along the axon) and radial diffusivity (presumed tobe the diffusivity perpendicular to the axon) are more specific tounderlying biological processes, such as myelin and axonal changes(Song et al., 2002). In general, the relationship between thesediffusivities, FA and MD, is such that FA increases when radialdiffusivity decreases and/or axial diffusivity increases while MDincreases when axial diffusivity and/or radial diffusivity increases,and vice versa.

The previously used conventional method of voxel-wise analysisto evaluate DTI indices, including FA, is limited by the use ofregistration methods with low dimensional transformations whichrestricts the accuracy in establishing correspondence of anatomicalpoints between individuals, largely due to the heterogeneous sig-nal intensity of the FA map. Furthermore, the requirement of imagesmoothing and the arbitrary selections of smoothing kernel andstatistical thresholds may lead to different, and even contradictoryresults (Jones et al., 2005). Tract-Based Spatial Statistics (TBSS)(Smith et al., 2006) is a recent advancement in voxel-wise analysis ofDTI data, which can overcome the limitations encountered withconventional methods. This is achieved by using a nonlinear re-gistration method and also, by projecting the center of fiber tracts ofindividual subjects into a common white matter skeleton, residualmisregistration issues can be greatly circumvented. Furthermore, themethod does not require image smoothing and uses nonparametricstatistical test with a correction of multiple comparisons, instead of

Fig. 1. The manually drawn ROIs in the superior frontal and parietal white matterinternal capsule (D), genu of corpus callosum (E), rostral and caudal parts of the p

assuming unjustified Gaussian distribution of DTI indices asconventional methods do, thereby delivering better localizationability and providing a solid ground for the validity of the results.

The structural changes found by MR imaging during braindevelopment from childhood to young adulthood may providethe neural substrate for the development of functional connectiv-ity, including for higher cognitive functions and reading de-velopment. From an on-going study of reading development inChinese children from late childhood to young adulthood, weevaluate DTI parameters FA, MD, and axial and radial dif-fusivities by TBSS analysis of three age groups at different stagesof reading development, and determine the relationships withreading performance.

Method

Subjects

Seventy-five normal healthy ethnic Chinese right-handed schoolgoing students and young adults were recruited according to three agegroups. Informed consent was obtained from the subjects and/or theirparent as appropriate, and the study was approved by the institutionalreview board. All subjects were native Chinese (Mandarin) speakerswho learned English as a second language. Group 1 comprised sub-jects in late childhood (n=24, male=13, age range=6.8 to 7.9 years,mean±SD=7.4±0.3 years). These were first grade students, typicallyhaving just started to learn to read. Group 2 comprised subjects inearly adolescence (n=27, male=16, age range=9.4 to 11.5 years,mean±SD=10.3±0.5 years), whowere fourth grade students and hadacquired substantial reading skills. Group 3 comprised young adults(n=24, male=11, age range=18.6 to 26.1 years, mean±SD=22.8±2.3 years), who were undergraduate students in a Beijing universityand highly fluent readers.

Reading scores

We evaluated subjects’ reading ability in Chinese and English byasking subjects to read aloud 200 Chinese characters and 200

(A), corona radiata (B), splenium of corpus callosum (C), anterior limb ofosterior limb of internal capsule (F), and anterior temporal white matter (G).

225D. Qiu et al. / NeuroImage 41 (2008) 223–232

English words (giving a maximum possible score of 200 each forChinese reading and English reading). These characters/words wereselected from textbooks used in primary schools in Beijing for firstto sixth graders, 20 from each. The remaining 160 were selectedfrom low-frequency characters/words in a linguistic corpus whichwere not covered in primary school textbooks. Characters and wordswere arranged in a sequence of increasing difficulty (as determinedby grade level and visual complexity or stroke number). Subjectswere asked to read the characters/words aloud as quickly and ac-curately as possible within 3 minutes.

Among the 75 subjects, Chinese and English reading scores wereobtained from 23, 26, and 17 subjects from groups 1, 2, and 3, re-spectively. The mean±SD of Chinese reading scores was 56.6±26.6,97.38±13.8, and 113.5±16.4 for the three groups, respectively, andthe mean±SD of English reading score was 8±9.6, 43.7±35.9, and94.5±13.4 for the three groups, respectively.

Data acquisition

MRI was performed at the Beijing MRI Center for Brain Researchof the Chinese Academy of Sciences using a 3 Tesla imager (Siemens,Erlangen, Germany)with a standard head coil. DTI data were acquiredusing single-shot spin-echo echo-planar imaging with TR=6000 ms,TE=84 ms, acquisition matrix=64×64, field of view=192 mm (in-plane resolution=3×3 mm), and slice thickness=3 mm with no gap.Diffusion-sensitizing gradient encoding was applied in 6 directionswith a diffusion-weighted factor b=1000 s/mm2, and one image

Fig. 2. Regions of significantly decreased MD with increasing age, superimposed omultiple comparisons by suprathreshold test with tN3). Green color indicates whindicates significant decreased MD in group 2 compared to group 1, the red color inyellow color indicates overlap of the above two contrasts. All images are in radiolimages.

(b0 image)was acquiredwithout use of a diffusion gradient, i.e. b=0 s/mm2. For each encoding direction, around 48 axial images parallel tothe anterior commisure-posterior commisure plane were acquired tocover the entire brain. DTI sequence was repeated four times toimprove signal to noise ratio.

Data analysis

Images were processed using the FSL (FMRIB Software Library,FMRIB, Oxford, UK) (Smith et al., 2004) software package. Foreach subject, all images including diffusion weighted and b0 imageswere affinely coregistered to the b0 image of the first repetition usingFLIRT (FMRIB’s Linear Image Registration Tool) (Jenkinson andSmith, 2001) so as to correct for eddy current induced distortion andsubject motion effect. Brain mask was created from the first b0image using BET (Brain extraction Tool) (Smith, 2002) and FDT(FMRIB’s Diffusion Toolbox) (Behrens et al., 2003) was used to fitthe tensor model and to compute the FA, MD, axial diffusivity andradial diffusivity maps. Voxel-wise analysis was performed usingTBSS (Smith et al., 2006) with the following steps; first, an FAimage was arbitrarily selected as the template and all FA imageswere nonlinearly normalized to it. The template was affinely re-gistered to the T1 template in MNI152 (Montreal NeurologicalInstitute, McGill, USA) standard space. The two transformationsthus obtained were combined to transform all the FA images to theMNI152 standard space. Then, the mean FA image was created andfiltered to retain only the center of the white matter tracts so as to

n the mean FA image (pb0.05, nonparametric permutation test, corrected forite matter skeleton where no significant results were found, the blue colordicates significantly decreased MD in group 3 compared to group 2, and theogical convention, i.e. the right side of the subjects is on the left side of the

226 D. Qiu et al. / NeuroImage 41 (2008) 223–232

create the mean FA skeleton. A threshold of FAN0.2 was applied tothe skeleton to include only major fiber bundles. Each subject’s FAdata was subsequently projected to the FA skeleton by searching forthe local center of the relevant fiber tract. This step was undertakenso as to account for residual local misregistration uncorrected for bythe aforementioned nonlinear registration. Statistical analysis wasperformed voxel by voxel to detect regions of significant differencesof FA among three groups by using nonparametric permutation testwith a correction for multiple comparisons (pN0.05, suprathresholdcluster test, tN3) (Nichols and Holmes, 2002). TBSS analysis wasrepeated for MD, axial diffusivity, and radial diffusivity maps.Results were overlaid on the mean FA image for better anatomicalidentification.

Region-of-interest (ROI) analysis was performed to obtain thequantitative DTI indices of significant regions and to confirm ourresults of voxel-wise analysis. We manually outlined ROIs on themean FA image according to the voxel-wise analysis findings. Themean FA, MD, axial diffusivity, and radial diffusivity were sub-sequently quantified from the same ROIs on nonlinearly normal-ized images before being mapped onto the skeleton. Voxels withFAb0.2 were excluded from the analysis to avoid partial volumeeffect from neighboring gray matter. ROIs were drawn bilaterallyand symmetrically over the anterior temporal white matter, rostraland caudal parts of the posterior limb of internal capsule, anteriorlimb of internal capsule, corona radiata, and superior frontal andparietal white matter. Since we found discrepancy in the result ofvoxel-wise analysis in the corpus callosum between this and pre-vious studies, we also included the genu and splenium of thecorpus callosum in ROI analysis to evaluate if this is related to the

Fig. 3. Regions of significantly increased FAwith increasing age, superimposed ongreen indicates the white matter skeleton, blue group 2Ngroup 1, red group 3Ngr

use of TBSS. Posterior limb of the internal capsule was separatedinto two parts because there was evidence of discrepancy betweenthem from voxel-wise analysis (Fig. 1). ANOVA analysis wasperformed to compare differences in DTI indices among the threeage groups, and, if significant, Tukey’s test was followed to detectgroup-wise difference. Chinese and English reading scores werecorrelated with DTI indices of the above measured ROIs. This wasfollowed by a partial correlation analysis to detect significant cor-relations between DTI indices and reading performance scoresafter controlling for the effect of age. A p value of b0.05 wasconsidered significant.

Results

Voxel-wise analysis

Figs. 2–6 show the statistical result superimposed on the meanFA image of MD, FA, radial diffusivity and axial diffusivity (bothincreased and decreased), respectively. Green color indicates thewhite matter skeleton where no significant differences were foundwhile blue color indicates significant increase/decrease in group 2 ascompared with group 1, red color indicates significant increase/decrease in group 3 as compared with group 2, and yellow colorindicates overlap of the above two contrasts.

Mean diffusivity (MD)DecreasedMDwas found in group 2 compared to group 1 in small

clusters in the external capsule, superior corona radiata, cingulum,temporal, occipital, and the superior frontal lobes. DecreasedMDwas

the mean FA image (see Fig. 2 for details on the meaning of colors; briefly,oup 2, and yellow the overlap of the latter two).

Fig. 4. Regions of significantly decreased radial diffusivity with increasing age, superimposed on the mean FA image (see Fig. 2 for details on the meaning ofcolors; briefly, green indicates the white matter skeleton, blue group 2bgroup 1, red group 3bgroup 2, and yellow the overlap of the latter two).

227D. Qiu et al. / NeuroImage 41 (2008) 223–232

much more extensive in group 3 compared to group 2 in the frontallobes, and in the right anterior temporal lobe white matter, cerebralpeduncles, anterior and posterior limbs of the internal capsule, andsmall clusters in the cerebellar white matter and the splenium of thecorpus callosum. No region was found to have significant increase inMD.

Fractional anisotropy (FA)Increased FAwas found in group 2 compared to group 1 only in

two small clusters, the genu of the left internal capsule and leftcerebral peduncle. Increased FAwas found in group 3 compared togroup 2 in multiple regions including cerebellum, cerebral pedun-cles, mid and posterior right temporal lobe, anterior limb, genu androstral part of the posterior limb of the internal capsules, anteriorpart of the external capsule, superior corona radiata, cingulum, andparts of the frontal and parietal lobes, more widespread in thefrontal lobe. Corpus callosum was not found to have age-relatedchanges in FA. No region was shown to have significant decreasein FA.

Radial diffusivityThe areas of decreased radial diffusivity closely mirrored the

areas of decreased MD. Compared to FA, in addition to the sig-nificant areas found on FA, much more extensive areas of sig-nificant differences were found, most evident in the frontal lobeand right anterior temporal lobe. Specifically, decreased radial dif-fusivity in group 2 compared to group 1 was found in the genu ofthe left internal capsule and left cerebral peduncle, and the frontallobes. Decreased radial diffusivity in group 3 compared to group 2

was found in much more widespread areas including cerebellum,cerebral peduncles, anterior, mid and posterior right temporal lobe,anterior limb, genu and rostral part of the posterior limb of theinternal capsules, anterior part of the external capsule, superiorcorona radiata, cingulum, and much more widespread in the frontallobe than on the FA map. On further inspection of the superiorfrontal and parietal lobes, a gradient of change is evident in thatsignificant regions between group 2 and group 1 were posteriorwhile significant regions between group 3 and group 2 weredistinctly in the anterior, i.e. frontal regions. No region was foundto have significant elevation of radial diffusivity.

Axial diffusivityBoth regions of significant increase and decrease in axial dif-

fusivity were found, although changes were predominantly ofdecreased axial diffusivity and between the two older age groups.Decreased axial diffusivity was found in group 2 compared togroup 1 in small clusters in the corona radiata and frontal lobes.Decreased axial diffusivity was found in group 3 compared togroup 2 in much more widespread areas including anterior andmid right temporal lobe, posterior part of the bilateral posteriorlimb of internal capsule, a small cluster in the splenium of corpuscallosum, and widespread areas in the frontal lobes. There wereno regions of increased axial diffusivity in group 2 compared togroup 1, while areas of increased axial diffusivity were found ingroup 3 compared to group 2 in small clusters in the left cerebellarhemisphere, bilateral occipital lobes, and a small area in the leftcorona radiata (which may include fibers of the superior longi-tudinal fasciculus).

Fig. 5. Regions of significantly increased axial diffusivity with increasing age, superimposed on the mean FA image (see Fig. 2 for details on the meaning ofcolors; briefly, green indicates the white matter skeleton, blue group 2bgroup 1, red group 3bgroup 2 and yellow the overlap of the latter two).

228 D. Qiu et al. / NeuroImage 41 (2008) 223–232

In summary, areas of significant change were more widespreadbetween group 3 and group 2 than group 2 and group 1 for all indices.Also, among the four indices, significant change was much morewidespread in mean diffusivity and radial diffusivity than FA or axialdiffusivity. As expected, areas of decreased mean diffusivity andradial diffusivity closely mirrored each other. The areas of signi-ficance overlapped between FA and radial diffusivity, but there wereadditional significant areas of reduced radial diffusivity in the frontaland right anterior temporal lobes in particular. These latter regionsoverlapped with significant areas of decrease in axial diffusivity.There was little overlap in areas of significance between FA and axialdiffusivity.

ROI analysis

Supplementary Tables 1–4 (see Supplementary Figs. 1–4 for barplots) show the normative values of DTI indices of specific regionsamong the three age groups. Statistical results of comparisonsamong the age groups confirmed our voxel-wise analysis findings.

Structure–function correlationTable 1 shows correlations between Chinese and English reading

scores and DTI indices of the ROIs studied. Significant correlationswere found between DTI indices of most of the ROIs and readingscores. After controlling for the effect of age, partial correlationanalysis revealed a significant correlation between FA and radialdiffusivity of the anterior limb of the left internal capsule (ALIC) andChinese reading score (r=0.245, p=0.05 and r=−0.245, p=0.05respectively), and mean diffusivity of left and right corona radiata

and English reading score (r=−0.277, p=0.026 and r=−0.271,p=0.029, respectively).

Discussion

We have performed TBSS analysis, a fully automated voxel-wisemethod, to investigate changes of DTI indices among three agegroups from late childhood to early adolescence and young adult-hood. We found increased FA and decreased MD in widespreadareas of the cerebral and cerebellar white matter, consistent withprevious studies (Bonekamp et al., 2007; Giorgio et al., 2008;Mukherjee et al., 2002; Schmithorst et al., 2002). Moreover, wefound that changes of FA and MD are primarily driven by reductionin both radial diffusivity and axial diffusivity, but the latter to a muchlesser extent. These findings of the changes in directional dif-fusivities are in keeping with the results of a previous study ofeigenvalues between early development and young adulthood in asmaller cohort (Suzuki et al., 2003). Recently, Giorgio et al. reportedthe DTI findings in normal brain maturation in an older cohort fromadolescence to adulthood (13.5 years–42 years). The study alsofound that FA increase was primarily driven by decreases in radialdiffusivity. However, the changes in radial diffusivity found byTBSS analysis were much less widespread with significant cor-relations found only in a white matter cluster in the body of thecorpus callosum. This may be due to age differences between thetwo cohorts andmethodological differences including the fact that inour study, we intentionally acquired subjects in three age groupswith relatively small age span in each group and therefore makingthe results more sensitive to nonlinear age effect. We have further

Fig. 6. Regions of significantly decreased axial diffusivity with increasing age, superimposed on the mean FA image (see Fig. 2 for details on the meaning ofcolors; briefly, green indicates the white matter skeleton, blue group 2bgroup 1, red group 3bgroup 2, and yellow the overlap of the latter two).

229D. Qiu et al. / NeuroImage 41 (2008) 223–232

demonstrated the correlation of structure and function, and shownthat changes in DTI parameters concur with development of readingabilities.

Reduction of radial diffusivity during maturation is suggested tobe attributed tomyelination as this modulates the diffusion anisotropy

Table 1Pearson's correlation between reading scores and DTI data

Regions Mean diffusivity FA

Chinese reading English reading Chinese reading English read

Left FWM −0.453⁎⁎⁎ −0.573⁎⁎⁎ 0.347⁎⁎ 0.206Right FWM −0.442⁎⁎⁎ −0.595⁎⁎⁎ 0.254⁎ 0.299⁎

Left PWM −0.303⁎ −0.273⁎ 0.115 0.084Right PWM −0.256⁎ −0.338⁎⁎ 0.333⁎⁎ 0.317⁎⁎

Left ALIC −0.454⁎⁎⁎ −0.545⁎⁎⁎ 0.497⁎⁎⁎ 0.446⁎⁎⁎

Right ALIC −0.330⁎⁎ −0.252⁎ 0.431⁎⁎⁎ 0.340⁎⁎

Left RPLIC −0.524⁎⁎⁎ −0.473⁎⁎⁎ 0.362⁎⁎ 0.347⁎⁎

Left CPLIC −0.559⁎⁎⁎ −0.580⁎⁎⁎ −0.046 −0.175Right RPLIC −0.534⁎⁎⁎ −0.586⁎⁎⁎ 0.413⁎⁎⁎ 0.441⁎⁎⁎

Right CPLIC −0.499⁎⁎⁎ −0.574⁎⁎⁎ 0.087 0.020Left ATWM −0.236 −0.335⁎⁎ 0.208 0.222Right ATWM −0.322⁎⁎ −0.413⁎⁎⁎ 0.393⁎⁎ 0.426⁎⁎⁎

Left CR −0.481⁎⁎⁎ −0.610⁎⁎⁎ 0.284⁎ 0.450⁎⁎⁎

Right CR −0.547⁎⁎⁎ −0.683⁎⁎⁎ 0.199 0.324⁎⁎

GCC −0.101 −0.329⁎⁎ −0.143 −0.065SCC −0.232 −0.239 0.039 0.040

⁎pb0.05, ⁎⁎pb0.01, and ⁎⁎⁎pb0.001. Abbreviations: FWM (frontal white matteRPLIC (rostral part of posterior limb of internal capsule), CPLIC (caudal part of pos(corona radiata), GCC (genu of corpus callosum), and SCC (splenium of corpus c

originating from cell membranes by creating an additional barrier(Beaulieu, 2002). Furthermore, the lack/disruption of myelin sheathhas been found to increase radial diffusivity without affecting axialdiffusivity (Song et al., 2002). These findings, along with the hi-stological observation of continuing white matter myelination into

Radial diffusivity Axial diffusivity

ing Chinese reading English reading Chinese reading English reading

−0.452⁎⁎⁎ −0.398⁎⁎⁎ −0.088 −0.319⁎⁎−0.378⁎⁎ −0.473⁎⁎⁎ −0.235 −0.359⁎⁎−0.224 −0.187 −0.097 −0.107−0.391⁎⁎ −0.412⁎⁎⁎ 0.065 −0.024−0.575⁎⁎⁎ −0.562⁎⁎⁎ 0.149 0.030−0.437⁎⁎⁎ −0.339⁎⁎ 0.205 0.164−0.445⁎⁎⁎ −0.411⁎⁎⁎ −0.432⁎⁎⁎ −0.376⁎⁎−0.237 −0.148 −0.547⁎⁎⁎ −0.636⁎⁎⁎−0.491⁎⁎⁎ −0.522⁎⁎⁎ −0.303⁎ −0.362⁎⁎−0.327⁎⁎ −0.307⁎ −0.377⁎⁎ −0.481⁎⁎⁎−0.264⁎ −0.344⁎⁎ −0.121 −0.204−0.408⁎⁎⁎ −0.476⁎⁎⁎ −0.001 −0.072−0.439⁎⁎⁎ −0.581⁎⁎⁎ −0.263⁎ −0.294⁎−0.397⁎⁎⁎ −0.552⁎⁎⁎ −0.267⁎ −0.248⁎0.085 −0.045 −0.216 −0.411⁎⁎⁎

−0.093 −0.094 −0.330⁎⁎ −0.341⁎⁎

r), PWM (parietal white matter), ALIC (anterior limb of internal capsule),terior limb of internal capsule), ATWM (anterior temporal white matter), CRallosum).

230 D. Qiu et al. / NeuroImage 41 (2008) 223–232

adulthood support the fact that decrease in radial diffusivity found inour study reflects primarily the myelination process (Bonekamp et al.,2007). The decrease in axial diffusivitymaybe related to the growth ofneurofibrils, such asmicrotubules and neurofilaments, and the growthof glial cells during brain development which would increase thetortuosity of the extra-axonal space. Conversely, the destruction ofneurofibrils has been found to increase axial diffusivity (Kinoshita etal., 1999). Other factors which have been proposed to affect axialdiffusivity include fiber coherence (Dubois et al., 2008), and axonalinjury, which in animal studies is associated with reduction in axialdiffusivity (Kim et al., 2006). Therefore, it is likely that changes inaxial diffusivity reflect complex interactions of multiple biologicalfactors that drive it in different directions.

Among the four DTI indices, significant differences were muchmore widespread in MD and radial diffusivity, than FA and axialdiffusivity. Significant regions of FA increase largely overlappedwith that of radial diffusivity, highlighting decreased radial diffu-sivity as the primary factor driving FA elevation. However, therewere additional significant regions of reduced radial diffusivitywhich was associated with decreased axial diffusivity, thereforenegating their effect on FA. In contrast, reduction in radial andaxial diffusivity boosted the reduction in MD in multiple regions.

Changes of DTI indices were found to continue from latechildhood to young adulthood, and were more widespread betweenearly adolescence and young adulthood than between late child-hood and early adolescence. Although this difference may be partlyaccounted for by the smaller age difference between group 1 (latechildhood) and group 2 (early adolescence), than group 2 andgroup 3 (young adulthood), our finding is consistent with a pre-vious report which also showed more widespread FA increase be-tween childhood and young adulthood groups than within thechildhood group (Snook et al., 2005). Structural imaging studieshave similarly demonstrated prominent brain growth during thisperiod rather than earlier in childhood (Sowell et al., 2001, 1999).These findings during the period when complex cognitive func-tions such as executive function and attention are maturing suggestthat these developments may be the neural substrate of maturingcognitive functions.

We found prominent changes in the frontal lobe and the as-sociational fibers that connect it to other parts of the brain, includingthe external capsule, cingulum, and probably the superior long-itudinal fasciculus. The frontal lobe is associated with high-ordercognitive functions such as executive, attention, and coordinationand this region undergoes late and protracted process of myelination(Yakovlev and Lecours, 1967). Structural image studies havesimilarly shown overt and widespread gray matter change localizedprimarily to the frontal lobe during adolescence to adulthood(Sowell et al., 1999). Significant regions found in our study closelyresemble the findings of structural imaging studies on gray matter,suggesting that there is concurrent development of neural functionunit and neural pathways. Superior longitudinal fasciculus connectsfrontal lobe to temporal and parietal lobes and is important for theinformation exchange between Broca and Wernicke language re-gions (Paus et al., 1999). There is evidence that integrity of leftarcuate fasciculus (part of superior longitudinal fasciculus) is cor-related with reading abilities (Beaulieu et al., 2005; Klingberg et al.,2000) and a lesion to this fiber would cause conduction aphasia.Interestingly, we found asymmetrical change in axial diffusivityin the corona radiata, with only the left side showing increase. Partof this area may correspond to the superior longitudinal fascicu-lus and may reflect asymmetrical re-organization of the fronto-

parieto-temporal fiber system to be more coherent during languagedevelopment. We also found asymmetrical changes in the temporalwhite matter with significant changes of DTI indices in the right sideonly. This area likely encompasses the right inferior longitudinalfasciculus (ILF) and inferior fronto-occipital fasciculus (IFO).The ILF connects the visual cortex in the occipital lobe to thetemporal pole and is believed to be important for visual memory(Catani et al., 2003), which is further supported by a recent casereport that damage to the right ILF was associated with impairmentof visual memory (Shinoura et al., 2007). Our findings of the de-velopment of ILF beyond childhood may provide a neural substratefor the protracted maturation of the ventral (or occipitotemporal)visual-stream function (Kovacs et al., 1999). These observationsconverge in supporting the notion that maturation of specific neuralpathways is essential to language and other high-order cognitivefunction development.

Internal capsule was found in TBSS analysis to show increase inFA and decrease in both axial and radial diffusivities, which isconsistent with previous studies reporting continuous developmentof internal capsule into adulthood (Paus et al., 1999). Furthermore,we found segregation of developmental patterns in the rostral andcaudal parts of posterior limb, whereby the caudal part did not showa significant change in FA and radial diffusivity while remarkablechanges were found in its rostral part. This segregation within theposterior limb of internal capsule has never been previously re-ported, and is now appreciated due possibly to the better localizationability achieved with TBSS in this study. It has been shown by DTItractography studies that the rostral part of internal capsule includesmainly connections from/to the motor and prefrontal cortices whilethe caudal part includes mainly connections from/to the sensory andpost-parietal cortices (Zarei et al., 2007). Also, the motor system hasbeen found to myelinate later than the sensory system (Kinney et al.,1988). Therefore, the segregation may be due to the fact that therostral and caudal parts of the internal capsule contain fibers fromdifferent parts of brain with distinct developmental patterns. Wehave found a significant correlation between FA and radial dif-fusivity of the anterior limb of the left internal capsule and Chinesereading ability. The maturation of this fiber bundle may reflectconnectivity with the left frontal lobe, which has been shown to beimportant for Chinese reading (Siok et al., 2004). For Englishreading ability, our structure–function correlation analysis hasshown that mean diffusivity of the superior corona radiata correlatedsignificantly with English reading scores after controlling for theeffect of age. It is noteworthy that our subjects learned English astheir second language. The fact that different regions were identifiedto correlate with English and Chinese reading leads us to postulatethat brain networks underlying Chinese and English (as a secondlanguage) may be language-related. Further studies are warranted toconfirm these findings.

We found significant reduction in axial diffusivity in a smallarea of the splenium of corpus callosum, which is confirmed by theROI analysis. Previous DTI studies on similar subject have foundmore widespread changes of FA and MD in the corpus callosum(Barnea-Goraly et al., 2005; Snook et al., 2007). Such discrepancymay be partially accounted for by the fact that TBSS specificallydetects changes in DTI indices while conventional VBM method isalso sensitive to morphological differences, such as fiber width.Moreover, variations in subject samples between the studies mayalso play a role.

There are several limitations to our study; the study is notlongitudinal, thus limiting the detection of subtle changes due to

231D. Qiu et al. / NeuroImage 41 (2008) 223–232

inter-subject variation. In the analysis of TBSS, we applied athreshold FAN0.2 to the mean FA image so as to obtain the meanFA skeleton. Such an operation may result in a conservative es-timate of group differences, as white matter with low FA is ex-cluded from the analysis. However, regions with FAb0.2 areusually located close to the vicinity of gray matter, rendering themsubject to partial volume effect.

In conclusion,we found continual development-related changes inDTI parameters, which provide insights complementary to otherstructural and functional studies of brain development. In addition, wedemonstrated correlations between structure and function in thedevelopment of reading abilities from childhood to young adulthood.

Acknowledgments

This research was supported by a 973 grant from the Ministry ofScience and Technology of China (2005CB522802) and the Knowl-edge Innovation Program of the Chinese Academy of Sciences.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.neuroimage.2008.02.023.

References

Barkovich, A.J., Kjos, B.O., Jackson Jr., D.E., Norman, D., 1988. Normalmaturation of the neonatal and infant brain:MR imaging at 1.5 T.Radiology166, 173–180.

Barnea-Goraly, N., Menon, V., Eckert, M., Tamm, L., Bammer, R., Kar-chemskiy, A., Dant, C.C., Reiss, A.L., 2005. White matter developmentduring childhood and adolescence: a cross-sectional diffusion tensorimaging study. Cereb. Cortex 15, 1848–1854.

Basser, P.J., Mattiello, J., LeBihan, D., 1994. MR diffusion tensor spec-troscopy and imaging. Biophys. J. 66, 259–267.

Beaulieu, C., 2002. The basis of anisotropic water diffusion in the nervoussystem—a technical review. NMR Biomed. 15, 435–455.

Beaulieu, C., Plewes, C., Paulson, L.A., Roy, D., Snook, L., Concha, L.,Phillips, L., 2005. Imaging brain connectivity in children with diversereading ability. Neuroimage 25, 1266–1271.

Behrens, T.E., Woolrich, M.W., Jenkinson, M., Johansen-Berg, H., Nunes,R.G., Clare, S., Matthews, P.M., Brady, J.M., Smith, S.M., 2003. Char-acterization and propagation of uncertainty in diffusion-weighted MRimaging. Magn. Reson. Med. 50, 1077–1088.

Ben, B.D., Ben, S.L., Graif, M., Pianka, P., Hendler, T., Cohen, Y., Assaf, Y.,2005. Normal white matter development from infancy to adulthood:comparing diffusion tensor and high b value diffusion weighted MRimages. J. Magn. Reson. Imaging 21, 503–511.

Bonekamp, D., Nagae, L.M., Degaonkar, M., Matson, M., Abdalla, W.M.,Barker, P.B., Mori, S., Horska, A., 2007. Diffusion tensor imaging inchildren and adolescents: reproducibility, hemispheric, and age-relateddifferences. Neuroimage 34, 733–742.

Catani, M., Jones, D.K., Donato, R., Ffytche, D.H., 2003. Occipito-temporalconnections in the human brain. Brain 126, 2093–2107.

Courchesne, E., Chisum, H.J., Townsend, J., Cowles, A., Covington, J.,Egaas, B., Harwood, M., Hinds, S., Press, G.A., 2000. Normal braindevelopment and aging: quantitative analysis at in vivo MR imaging inhealthy volunteers. Radiology 216, 672–682.

Dubois, J., haene-Lambertz, G., Perrin, M., Mangin, J.F., Cointepas, Y.,Duchesnay, E., Le Bihan, D., Hertz-Pannier, L., 2008. Asynchrony ofthe early maturation of white matter bundles in healthy infants:Quantitative landmarks revealed noninvasively by diffusion tensorimaging. Hum. Brain Mapp. 29, 14–27.

Giorgio, A., Watkins, K.E., Douaud, G., James, A.C., James, S., De, S.N.,Matthews, P.M., Smith, S.M., Johansen-Berg, H., 2008. Changes inwhite matter microstructure during adolescence. Neuroimage 39, 52–61.

Gogtay,N.,Giedd, J.N., Lusk, L., Hayashi, K.M.,Greenstein,D.,Vaituzis, A.C.,Nugent III, T.F., Herman, D.H., Clasen, L.S., Toga, A.W., Rapoport, J.L.,Thompson, P.M., 2004. Dynamic mapping of human cortical developmentduring childhood through early adulthood. Proc. Natl. Acad. Sci. U. S. A.101, 8174–8179.

Huang, H., Zhang, J., Wakana, S., Zhang, W., Ren, T., Richards, L.J.,Yarowsky, P., Donohue, P., Graham, E., van Zijl, P.C., Mori, S., 2006.White and gray matter development in human fetal, newborn and pediatricbrains. Neuroimage 33, 27–38.

Huppi, P.S., Dubois, J., 2006. Diffusion tensor imaging of brain develop-ment. Semin. Fetal Neonatal Med. 11, 489–497.

Huppi, P.S., Maier, S.E., Peled, S., Zientara, G.P., Barnes, P.D., Jolesz, F.A.,Volpe, J.J., 1998.Microstructural development of human newborn cerebralwhite matter assessed in vivo by diffusion tensor magnetic resonanceimaging. Pediatr. Res. 44, 584–590.

Jenkinson, M., Smith, S., 2001. A global optimisation method for robustaffine registration of brain images. Med. Image Anal. 5, 143–156.

Jones, D.K., Symms, M.R., Cercignani, M., Howard, R.J., 2005. The effectof filter size on VBM analyses of DT-MRI data. Neuroimage 26,546–554.

Kim, J.H., Budde, M.D., Liang, H.F., Klein, R.S., Russell, J.H., Cross, A.H.,Song, S.K., 2006. Detecting axon damage in spinal cord from a mousemodel of multiple sclerosis. Neurobiol. Dis. 21, 626–632.

Kinney, H.C., Brody, B.A., Kloman, A.S., Gilles, F.H., 1988. Sequence ofcentral nervous system myelination in human infancy: II. Patterns ofmyelination in autopsied infants. J. Neuropathol. Exp. Neurol. 47,217–234.

Kinoshita, Y., Ohnishi, A., Kohshi, K., Yokota, A., 1999. Apparent diffusioncoefficient on rat brain and nerves intoxicated with methylmercury.Environ. Res. 80, 348–354.

Klingberg, T., Vaidya, C.J., Gabrieli, J.D., Moseley, M.E., Hedehus, M.,1999. Myelination and organization of the frontal white matter in child-ren: a diffusion tensor MRI study. Neuroreport 10, 2817–2821.

Klingberg, T., Hedehus,M., Temple, E., Salz, T., Gabrieli, J.D.,Moseley,M.E.,Poldrack, R.A., 2000. Microstructure of temporo-parietal white matter as abasis for reading ability: evidence from diffusion tensor magnetic reso-nance imaging. Neuron 25, 493–500.

Kovacs, I., Kozma, P., Feher, A., Benedek, G., 1999. Late maturation ofvisual spatial integration in humans. Proc. Natl. Acad. Sci. U. S. A. 96,12204–12209.

Mukherjee, P., Miller, J.H., Shimony, J.S., Conturo, T.E., Lee, B.C., Almli,C.R., McKinstry, R.C., 2001. Normal brain maturation during child-hood: developmental trends characterized with diffusion-tensor MRimaging. Radiology 221, 349–358.

Mukherjee, P.,Miller, J.H., Shimony, J.S., Philip, J.V., Nehra, D., Snyder, A.Z.,Conturo, T.E., Neil, J.J., McKinstry, R.C., 2002. Diffusion-tensor MRimaging of gray andwhite matter development during normal human brainmaturation. AJNR Am. J. Neuroradiol. 23, 1445–1456.

Nichols, T.E., Holmes, A.P., 2002. Nonparametric permutation tests forfunctional neuroimaging: a primer with examples. Hum. Brain Mapp.15, 1–25.

Paus, T., Zijdenbos, A.,Worsley, K., Collins, D.L., Blumenthal, J., Giedd, J.N.,Rapoport, J.L., Evans, A.C., 1999. Structural maturation of neural path-ways in children and adolescents: in vivo study. Science 283, 1908–1911.

Schmithorst, V.J., Wilke, M., Dardzinski, B.J., Holland, S.K., 2002. Cor-relation of white matter diffusivity and anisotropy with age during child-hood and adolescence: a cross-sectional diffusion-tensor MR imagingstudy. Radiology 222, 212–218.

Schneider, J.F., Il'yasov, K.A., Hennig, J., Martin, E., 2004. Fast quantitativediffusion-tensor imaging of cerebral white matter from the neonatalperiod to adolescence. Neuroradiology 46, 258–266.

Shinoura, N., Suzuki, Y., Tsukada, M., Katsuki, S., Yamada, R., Tabei, Y.,Saito, K., Yagi, K., 2007. Impairment of inferior longitudinal fasciculusplays a role in visual memory disturbance. Neurocase 13, 127–130.

232 D. Qiu et al. / NeuroImage 41 (2008) 223–232

Siok, W.T., Perfetti, C.A., Jin, Z., Tan, L.H., 2004. Biological abnormality ofimpaired reading is constrained by culture. Nature 431, 71–76.

Smith, S.M., 2002. Fast robust automated brain extraction. Hum. BrainMapp. 17, 143–155.

Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens,T.E., Johansen-Berg, H., Bannister, P.R., De, L.M., Drobnjak, I.,Flitney, D.E., Niazy, R.K., Saunders, J., Vickers, J., Zhang, Y., De, S.N.,Brady, J.M., Matthews, P.M., 2004. Advances in functional and struc-tural MR image analysis and implementation as FSL. Neuroimage 23(Suppl 1), S208–S219.

Smith, S.M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T.E.,Mackay, C.E., Watkins, K.E., Ciccarelli, O., Cader, M.Z., Matthews,P.M., Behrens, T.E., 2006. Tract-based spatial statistics: voxelwiseanalysis of multi-subject diffusion data. Neuroimage 31, 1487–1505.

Snook, L., Paulson, L.A., Roy, D., Phillips, L., Beaulieu, C., 2005. Diffusiontensor imaging of neurodevelopment in children and young adults.Neuroimage 26, 1164–1173.

Snook, L., Plewes, C., Beaulieu, C., 2007.Voxel based versus region of interestanalysis in diffusion tensor imaging of neurodevelopment. Neuroimage 34,243–252.

Song, S.K., Sun, S.W., Ramsbottom, M.J., Chang, C., Russell, J., Cross,A.H., 2002. Dysmyelination revealed through MRI as increasedradial (but unchanged axial) diffusion of water. Neuroimage 17,1429–1436.

Sowell, E.R., Thompson, P.M., Holmes, C.J., Jernigan, T.L., Toga, A.W.,1999. In vivo evidence for post-adolescent brain maturation in frontaland striatal regions. Nat. Neurosci. 2, 859–861.

Sowell, E.R., Thompson, P.M., Tessner, K.D., Toga, A.W., 2001. Mappingcontinued brain growth and gray matter density reduction in dorsalfrontal cortex: inverse relationships during postadolescent brainmaturation. J. Neurosci. 21, 8819–8829.

Sowell, E.R., Trauner, D.A., Gamst, A., Jernigan, T.L., 2002. Developmentof cortical and subcortical brain structures in childhood and adolescence:a structural MRI study. Dev. Med. Child Neurol. 44, 4–16.

Suzuki, Y., Matsuzawa, H., Kwee, I.L., Nakada, T., 2003. Absoluteeigenvalue diffusion tensor analysis for human brain maturation. NMRBiomed. 16, 257–260.

Toga, A.W., Thompson, P.M., Sowell, E.R., 2006. Mapping brain matu-ration. Trends Neurosci. 29, 148–159.

Wakana, S., Jiang,H.,Nagae-Poetscher, L.M., vanZijl, P.C.,Mori, S., 2004. Fibertract-based atlas of human white matter anatomy. Radiology 230, 77–87.

Yakovlev, P.I., Lecours, A.R., 1967. The myelogenetic cycles of regionalmaturation of the brain. In: Minkowski, A. (Ed.), Regional developmentof the brain in early life. Blackwell Scientific, Oxford(UK), pp. 3–70.

Zarei, M., Johansen-Berg, H., Jenkinson,M., Ciccarelli, O., Thompson, A.J.,Matthews, P.M., 2007. Two-dimensional population map of corticalconnections in the human internal capsule. J. Magn. Reson. Imaging 25,48–54.