Diffusion tensor fractional anisotropy of the normal-appearing seven segments of the corpus callosum in healthy adults and relapsing-remitting multiple sclerosis patients


  • Khader M. Hasan PhD,

    Corresponding author
    1. Department of Interventional and Diagnostic Imaging, University of Texas Medical School at Houston, Houston, Texas, USA
    • Department of Interventional and Diagnostic Imaging, University of Texas Medical School at Houston, 6431 Fannin Street, MSB 2.100, Houston, TX 77030
    Search for more papers by this author
  • Rakesh K. Gupta MD,

    1. Department of Interventional and Diagnostic Imaging, University of Texas Medical School at Houston, Houston, Texas, USA
    Search for more papers by this author
  • Rafael M. Santos MD, FRCS,

    1. Department of Interventional and Diagnostic Imaging, University of Texas Medical School at Houston, Houston, Texas, USA
    Search for more papers by this author
  • Jerry S. Wolinsky MD,

    1. Department of Neurology, University of Texas Medical School at Houston, Houston, Texas, USA
    Search for more papers by this author
  • Ponnada A. Narayana PhD

    1. Department of Interventional and Diagnostic Imaging, University of Texas Medical School at Houston, Houston, Texas, USA
    Search for more papers by this author

  • Presented in part at the 12th Annual Meeting of ISMRM, Kyoto, Japan, 2004 (abstract 1498). The acquisition, processing, and quantitative analysis methodologies were also described in abstracts 338 and 1350 presented at the same meeting.



To investigate the utility of whole-brain diffusion tensor imaging (DTI) in elucidating the pathogenesis of multiple sclerosis (MS) using the normal-appearing white matter (NAWM) of the corpus callosum (CC) as a marker of occult disease activity.

Materials and Methods

A high signal-to-noise ratio (SNR) and optimized entire brain DTI data were acquired in 26 clinically-definite relapsing and remitting multiple sclerosis (RRMS) patients and 32 age-matched healthy adult controls. The fractional anisotropy (FA) values of seven functionally distinct regions in the normal-appearing CC were compared between patients and controls.


This study indicates that 1) there was a gender-independent FA heterogeneity of the functionally specialized CC segments in normal volunteers; 2) FA in the MS group was significantly decreased in the anterior (P = 0.0039) and posterior (P = 0.0018) midbody subdivisions of the CC, possibly due to a reduction of small-caliber axons; and 3) the FA of the genu of the CC was relatively intact in the MS patients compared to the healthy age-matched controls (P = 0.644), while the splenium showed an insignificant trend of reduced FA values (P = 0.248). The decrease in FA in any of the CC subdivisions did not correlate with disease duration (DD) or the expanded disability status scale (EDSS) score.


The preliminary results are consistent with published histopathology and clinical studies on MS, but not with some published DTI reports. This study provides insights into the pathogenesis of MS, and the role played by compromised axonal integrity in this disease. J. Magn. Reson. Imaging 2005;21:735–743. © 2005 Wiley-Liss, Inc.

MULTIPLE SCLEROSIS (MS) is the most common white matter (WM) disease in humans, and carries significant social, economic, and emotional burdens. MS involves a complex pathophysiology and unpredictable disease course that complicate objective evaluation of the disease state (1–5).

The confounding pathological factors in MS include axonal loss, gliosis, demyelination, remyelination, edema, and inflammation (1–3). Magnetic resonance imaging (MRI) has significantly improved our understanding of MS (4). Unfortunately, conventional MRI measures, such as lesion load (LL), show relatively poor correlation with clinical measures (1–5). In addition, there is considerable evidence, based on both MR spectroscopy (5) (MRS) and histology (6, 7), that axonal pathology appears early in the course of the disease. Thus, noninvasive methods to localize and evaluate disease activity early on are essential for implementing therapeutic approaches to retard and/or reverse the course of the disease (1). The identification of a noninvasive and robust marker of the disease would be very useful in multicenter clinical trials and to customize patient management (1, 2, 5).

Diffusion tensor imaging (DTI) is a promising technique for probing the microstructural organization of WM (8). DTI offers orientation (connectivity) and scalar (anisotropy) maps on a voxel-by-voxel basis that directly reflect the microstructural organization of the tissue (8–10). Rotationally invariant DTI indices, such as the mean diffusivity (Dav) and fractional anisotropy (FA) (8–11), have the potential to detect compromised fiber tract integrity early on, and thus could be used to detect pathology in normal-appearing WM (NAWM) (12–15). Specifically, DTI measures promise to provide noninvasive and more specific pathological markers that reflect the state of myelin and axonal integrity, and may exhibit better correlation with clinical disability (12–19). WM demyelination of compact and highly directional fiber bundles may increase the diffusivity along the perpendicular direction without a concomitant alteration along the direction parallel to the fiber (20). Acute breakdown of myelin would result in a loss of organization, and may offer a transient hindrance to water motion, resulting in a decrease in the bulk mean diffusivity Dav and a loss of anisotropy (12, 14). DTI measures have been shown to be useful in studying the NAWM of different MS phenotypes. For example, a recent study by Guo et al (15) showed that tensor anisotropy in connected WM decreased distally from eloquent focal MS lesions and provided a sensitive probe for microstructural tissue injury.

Unfortunately, the published FA values in both healthy control and MS populations have been inconsistent and often contradictory, even on compact WM structures such as the corpus callosum (CC). The contradictions in the published literature may reflect variation in ages, technical and instrumentation limitations (such as the use of suboptimal and rotationally variant tensor encoding schemes), poor signal-to-noise ratios (SNRs), large slice thickness, and variable region-of-interest (ROI) placement.

The CC is the largest compact fiber bundle that connects cortical and subcortical regions of the brain, and is involved in interhemispheric transfers of auditory, sensory, and motor information. The CC has been implicated in many aspects of human behavior and cognition (21), normal aging (22), and various pathologies (23), including traumatic brain injuries (24), chronic alcoholism (25), schizophrenia (26, 27), and MS (6, 7, 13, 16, 28). The CC is the most discernable structure of the central nervous system (CNS) on DTI (29). The compact fiber bundles run right to left, with high anisotropy and directional coherence (see Fig. 1). The CC fiber distribution is known from postmortem studies to be heterogeneous (21, 27, 30, 31) with larger myelinated fibers located at the body segment and the posterior portion, whereas the genu is populated by a large number of smaller fibers (30). To rationally interpret DTI measures of CC, it is important to recognize the spatial heterogeneity of the fibers in both normal and pathological states (29). Most published studies appear to have ignored this spatial heterogeneity.

Figure 1.

An illustration of the complementary role of fused DTI and conventional MRI in localizing lesions and providing clues on the pathogenesis and evolution of MS lesions. The figure shows the internal capsule, and anterior and posterior callosal tracks in the vicinity of the FLAIR-highlighted lesion (right front). The figure is also a demonstration that the genu of the CC in MS patients was not significantly different from that of healthy controls (see Results and Discussion). Note that the right frontal MS lesion did not disrupt the structural connectivity of the frontal and posterior callosal fibers. The DTI orientation color table is also shown (red: right-to-left; green: anterior-to-posterior; and blue: inferior-to-superior). The human face is an imaginative and artistic rendering (by R.M.S.) that illustrates the potential of DTI for building a digital atlas of MS lesion interactions with NAWM, and future applications of DTI in therapy and lesion staging.

The unique location, heterogeneity, and functional specialization of the CC in the human brain (29), and the availability of postmortem axonal morphology data on this structure (27, 30, 31) make it an ideal subject for the study of NAWM pathology, and could lead to insights that enhance our understanding and modeling of MS pathogenesis (3, 13, 16, 23, 28).

To increase the accuracy of DTI metric estimation, we measured regional FA values using optimized DTI protocols at high SNR from seven functionally distinct regions within the NAWM of the CC from the mid-sagittal section. The CC subdivision is commonly used in psychometric and volumetric studies to provide an objective means of estimating the axonal number and area of the different callosal segments. In this study, FA was used as a probe of the CC regional morphology (9–11), in contradistinction to volumetric studies that used the entire midsagittal area of the CC (23, 28).

We believe this is the first study to compare the measured FA values in the NAWM of the seven subdivisions of the CC in both healthy adults and age-matched MS patients, and to relate these changes to results from published histopathology studies.



Twenty-six patients with clinically definite relapsing-remitting multiple sclerosis (RRMS; 22 females and four males; mean age ± SD = 37.9 ± 11.7) and 32 age-matched healthy adult controls (16 females and 16 males; age = 38.1 ± 12.1 years) were included in these studies. The median expanded disability status scale (EDSS) score for the 26 RRMS patients was 1.25 (range = 0.0, 6.0) and the median disease duration (DD) to imaging was 7.6 years (range = 0.3, 17.6). All of the RRMS patients had been receiving medication for various periods of time when the MRI data were acquired. These studies were approved by the Committee for the Protection of Human Subjects, University of Texas Health Science Center at Houston, and written consent was obtained from all subjects.

MRI Protocol

All MR scans were performed on a GE 1.5 T neurovascular optimized MR scanner (LX echo speed, version 8.4-9.1) using a quadrature transmit/receive head coil and a gradient system with a slew rate of 150 T/m/second and a maximum gradient amplitude of 40 mT/m on all three channels. To minimize motion during the scans, foam wedges were placed on both sides of the subject's head inside the RF coil and secured with a head strap, with care taken to ensure the patient's comfort. We have used this procedure successfully over the last several years.

Conventional MR brain images, covering the brain from the vertex to the foramen magnum, were acquired using both dual fast spin echo (FSE; TE1/TE2/TR = 10/80/7000 msec) and fast fluid-attenuated inversion recovery (FLAIR; TE/TI/TR = 85/2200/10000 msec) sequences. A total of 42 contiguous, 3-mm-thick axial sections were acquired with a field of view (FOV) of 240 mm × 240 mm and an image matrix of 256 × 256. In addition, pre- and postcontrast T1-weighted images were also collected from the MS patients using the same graphical prescription as the FLAIR and FSE sequences.

DT-encoded data from the same region as the FLAIR images were acquired with a dual SE-prepared and diffusion-sensitized, single-shot, echo-planar imaging (EPI) sequence with spectral selective pulses for fat suppression. The dual SE EPI sequence minimizes the geometric distortions introduced by the eddy currents that are induced when large diffusion gradients are switched (32). A multifaceted, rotationally invariant and balanced Icosa21 DT encoding scheme with a diffusion-weighting b-factor of 1400 seconds mm−2 was used (33). A T2-weighted or b = 0 reference image was also acquired for DT analysis. The diffusion-weighted (DW) acquisition parameters were: slice thickness = 3 mm with no gap, total number of slices = 42, FOV = 240 mm × 240 mm, number of ky lines = 80, and echo spacing = ∼640 μs. The image matrix after homodyne construction and zero-filling was 256 × 256, TE = 80 msec, and TR = 6800 msec. The sequence utilized ramp sampling to further reduce geometric distortions. The number of excitations (NEX) was 4 per slice and per diffusion-encoding direction. To reduce the phase fluctuations and image data overload, the images were constructed and magnitude-averaged by the scanner. The SNR in the b = 0 images (SNR0) was roughly 50 in the brain parenchyma. The DTI full-brain examination generated 42 * (21 + 1) or 924 images in approximately 10 min.

Data Processing and Analysis

The DW data were transferred to a workstation via ftp. The data were then spatially filtered using a median 3 × 3 filter. Acquiring the DTI data with the dual-echo EPI sequence and ramp sampling considerably reduced but not eliminate the geometric distortions. We corrected the residual distortions by registering the DW images with the reference image using the two-dimensional module of the automated image registration (AIR) package (eight-parameter with perspective) (34). The DW image for each slice, with the least distortion based on a water phantom calibration examination acquired with identical image parameters, was used as the reference image. Furthermore, since the CC is located close to the center of the image, the susceptibility-induced distortions in this region are expected to be fairly minimal. Based on our observations, the distortion was minimal along the frequency-encoding or x-direction, and largest along the phase-encoding or y-direction (11). We performed the DT decoding, cubic voxel interpolation, diagonalization, and Dav and FA computation using an in-house-developed DTI design and analysis toolbox based on the generalized decoding (33) and analytical diagonalization (35) methods, written under IDL (Research Systems, Boulder, CO, USA).

ROI Selection and DTI Data Overlay Method

To increase the utility of DTI for neurological disorders, we implemented procedures to allow the visualization of three-dimensional maps of fiber orientation along any direction, and fused the DTI maps with any other data set, including conventional FLAIR and dual SE images. Since the DTI maps are self-registered, the FA modulated principal vector (e1) map fused with Dav, T2w, FLAIR, exp(−b*Dav) maps or any coregistered map are useful for placing the ROI and visualizing compact fiber bundles in the vicinity of MS lesions. An example of this image fusion is shown in Fig. 1. This method helped us considerably to identify and avoid partial averaging from adjacent gray matter or CSF in the ROI placement.

Witelson Subdivision of the CC and ROI Placements

A semiautomated subregional division of the CC, as described by Witelson (21) and used in other non-DTI studies (6, 7, 24, 27, 30, 31), was implemented in which the CC was divided into seven segments: the CC1-rostrum, CC2-genu, CC3-rostral midbody, CC4-anterior midbody, CC5-posterior midbody, CC6-isthmus, and CC7-splenium. These subdivisions of the CC correspond to the projection areas delineated by the parcellation system described above: the midbody, isthmus, and splenium are associated with their respective parcellation units (sensorimotor, midtemporal, and occipital), and the rostrum and genu are associated with regions comprising frontal lobe structures (21). The DTI-guided Witelson subdivisions described above were identified on the midsagittal section from the reformatted axial DTI data after isotropic voxel interpolation, and the column of the fornix and the interthalamic mass were used as anatomical landmarks. This procedure was supervised by a neurosurgeon (R.M.S.). The mean and SD of the DTI metrics were obtained from rectangular ROIs covering a homogenous tissue region of the CC subdivisions, as shown in Fig. 2. The CC was considered to be normal-appearing on the basis of its normal signal intensity on proton- and T2-weighted FSE and FLAIR images with no contrast enhancement on the post-Gd images. The NAWM in the CC was evaluated by an experienced neuroradiologist (R.K.G.).

Figure 2.

An illustration of the DTI-guided implementation of the Witelson CC subdivision procedure. The figure shows a mid-sagittal section of the principal vector modulated by the FA fused with the mean diffusivity map.

Compact Fiber Bundle Tracking

To demonstrate the feasibility of connecting adjacent and directionally similar voxels in the NAWM of the CC, we used the fiber assignment by continuous tracking (FACT) approach (36). The initial seeds for fiber tracking were placed in the genu and splenium of the CC with a tracking threshold of FA ≥ 0.2. Additional seeds were also placed in the corticospinal tract and the posterior limb of the internal capsule. While we performed fiber tracking in a number of subjects, we did not attempt to quantitatively analyze these data because the choice of the threshold would have had a major impact on the quantitative analysis. At this stage, we do not know the optimum threshold value.

DTI Data Collection and Processing Quality Control

Since the data were collected over a period of one year and involved ROI placement by trained raters (K.M.H. and R.M.S.), the results of the ROI analyses for the same subjects from two independent sessions were compared. For b-factor calibration and residual eddy current distortion correction, we also collected the DTI data on a spherical water phantom. The large diffusion-encoding gradients induce mechanical motion that could compromise the estimated DTI measures. To ensure that mechanical vibrations would have a minimal effect on the DTI measures, we performed temporal and spatial stability and bootstrap analyses on water phantoms, as described elsewhere (11, 33).

Statistical Analysis

We used a Student's t-test and a χ2 test to examine group age and gender differences, respectively. The confidence level was set at 95% (or α = 0.05) unless otherwise stated. The statistical significance of the intracallosal comparisons of the seven segments of the CC was evaluated by analysis of variance (ANOVA). To investigate the effect of gender on the intracallosal heterogeneity of the CC segments on the FA, we performed a two-way ANOVA followed by post-hoc multiple paired t-tests. The Bonferroni procedure was adopted to adjust the significant α whenever multiple comparisons were done (37). For intracallosal comparisons, spatial covariances between adjacent CC segments were also incorporated, and the smallest P-value provided by four methods (Tukey-Kramer, Bonferroni, Scheffe, and Dunn-Sidak) was used for significance. In general, the Bonferroni method is the most conservative, and the significance P for the 21 (7 * 6/2) comparisons was set at ≤0.002 (α < 0.05/21). The statistical significance of the FA and Dav values for comparisons between different regions of the CC was determined by means of a one-way ANOVA. The differences between group means were computed with the use of a two-tailed t-test for unpaired comparisons. Note that the t-test in this case is equivalent to an ANOVA for two groups. For intercallosal comparisons between groups, the Bonferroni adjustment of the significant α was accounted for by setting the significance level at ≤0.05/7 ∼ 0.007 (seven group comparisons or subdivisions of the CC used in this work). The DTI measures (FA and Dav) were also correlated with the EDSS and DD using Spearman's rank correlation coefficient. For the repeatability analysis, we used the Bland-Altman analysis (37).


Demographics of the MS Patients and Healthy Adults

Female RRMS patients (22/26) were disproportionately represented in the MS group compared to the normal controls (16/32), but the ages of the two groups were well balanced (P = 0.566, t-test). The ages of males and females were quite similar in the control group (P = 0.998). Across the MS group, the Spearman correlation coefficient (r) between EDSS and age was (r = 0.388; P = 0.05), as were correlations between DD and age (r = 0.083, P = 0.69) and between EDSS and DD (r = 0.363; P = 0.069). Since the control and MS groups were age-matched, we did not include age as a covariate in the data analysis.

Quality Control and Repeatability of FA

The results from a Bland-Altman bias analysis of the FA results obtained from 34 randomly selected subjects and three representative structures of the CC (CC2, CC4, and CC7) are shown in Fig. 3. This analysis was done to determine the scanner stability, confidence in the ROI placement, and rater bias, and was performed by two independent raters who were blind to the subject's group and identity. The abscissa in Fig. 3 shows the mean values of the two sessions (considered to be the true values), and the ordinate shows the difference (bias) in the values obtained in the two sessions. The linear regression between the bias and mean shown in this figure demonstrates no significant trend between the two sessions, indicating minimal bias (r = 0.16, P = 0.096) and excellent repeatability.

Figure 3.

A Bland-Altman bias analysis of the FA results (×1000) obtained from 34 randomly selected subjects and three representative structures of the CC (CC2, CC4, and CC7). The horizontal axis shows the mean of two rating sessions, and the vertical axis shows the difference between the two sessions. The linear regression least-squares fit curve shows no significant trend between the two rating sessions.

Our spatial and temporal analyses on the water phantom and normal data failed to detect any effect of gradient-induced mechanical vibrations on the DTI measures.

Inter- and Intracallosal Comparisons in Age-Matched Male and Female Healthy Adults

Figure 4 shows the results of the post-hoc analysis of the FA values of different segments of the CC for the healthy male and female controls. The stratification by gender was done to ensure that the results would not reflect the bias in the gender distribution in our MS patients. The spatial heterogeneity of FA in CC was gender-independent in the healthy controls, as evaluated by the two-way ANOVA analysis. The corresponding follow-up paired t-test comparisons are shown in Fig. 4. For intracallosal comparisons, the FA values of the anterior, middle, and posterior CC subdivisions were significantly different (P < 0.002), and FA (splenium) > FA (genu) > FA (body) in both the normal controls and MS patients. This trend was observed for both genders. Thus, both male and female data were pooled to reduce redundant comparisons and possible loss of statistical power due to the small populations.

Figure 4.

A bar plot showing the comparison of FA group means and SD (μ ± σ) of the seven subdivisions of the CC between the age-matched healthy males and females. The P-value was computed using a two-tailed t-test, and significance accounted for the seven comparisons.

Normative vs. Pathological FA and Dav CC Spatial Distribution and Intergroup Comparisons

Figure 5 graphically summarizes the results of the comparison of FA between the 32 healthy adults and the 26 RRMS patients. The figure shows group means, standard error of the mean, and the corresponding P-values of the group differences. The FA values of the posterior, anterior, and midbody were significantly different (P < 0.002) in normal controls. This regional heterogeneity was also observed in MS patients: FA (posterior CC) > FA (anterior CC) > FA (middlle CC) for both controls and patients. The FA values were significantly decreased in the MS patients compared to the controls in the CC4 (anterior, P = 0.0039) and CC5 (posterior, P = 0.0018) midbody segments, and did not reach statistical significance in the CC1, CC6, and CC7 segments.

Figure 5.

A bar plot of the mean and SD FA values of the seven segments of the CC in the 26 RRMS group and the 32 age-matched healthy adult controls. The P-values were computed using the two-tailed t-test for unequal samples. The significance level was taken as 0.007. FA (MS) < FA (Controls): FA (splenium) > FA (genu) > FA (body of CC) and CC4-CC5 differ significantly between the two groups.

Genu and Splenium Anisotropy, and Feasibility of Fiber Tracking

As indicated in Fig. 5, the FA values in the anterior (CC2-CC3; P = 0.644, 0.434), and posterior (CC7; P = 0.249) subdivisions did not significantly differ between RRMS patients and the age-matched healthy controls. As a demonstration of this quantitative observation, Fig. 1 illustrates the ability to perform compact fiber tracking of the frontocallosal fibers traversing through the genu of the CC in one RRMS patient with a right frontal MS lesion as indicated by the FLAIR images. In addition, Fig. 6 shows four representative cases and demonstrates the feasibility of performing compact fiber tracking in the entire CCs of two RRMS patients (38-year-old females: EDSS = 0, DD = 11 years, and EDSS = 2.0, DD = 2.0, respectively) and two healthy adults (a 38-year-old male and a 33-year-old female).

Figure 6.

Illustration of the feasibility of compact fiber tracking using whole-brain, high-angular-resolution, icosahedral DTI data acquisition in both RRMS patients and healthy controls (from left-to-right, up-to-down). a: A 38-year-old female with RRMS (DD = 14 years, EDSS = 0.0). b: A 38-year-old female with RRMS (DD = 11 years, EDSS = 2.0) and two healthy adult controls. c: A 38-year-old healthy male. d: A 33-year-old healthy female. The figure shows commisural fibers (red: CC), long association fibers (green) and projection fibers (blue: tracks coursing through the posterior limb of the internal capsule, corticospinal (CST) and corona radiata). The directional color table and encoding scheme are also shown on the unit sphere. The three-dimensional view is set so that most of the fibers are shown. The grayscale background maps (axial-sagittal and coronal) were obtained from the b = 0 interpolated and DTI coregistered volumes. Notice the reduced number of tracks in b compared to a, c, and d, reflecting the lesion activity, connectivity, and volume in the periventricular areas (FLAIR-based segmentation not shown). The DTI orientation color table and the tessellated tensor encoding scheme are also shown.


We compared the FA values of seven functionally distinct regions in the CC between 26 RRMS patients and 32 healthy adult controls. Our findings on the spatial heterogeneity in the anisotropy of CC trends and gender independence in both healthy adults and MS patients are consistent with recent quantitative ROI reports that included only two or three representative CC segments (16, 22, 29).

It is noteworthy to compare our regional FA values of the CC in both healthy adults and RRMS with results from the published reports. In our studies, the following trend was observed in both healthy adult males and age-matched females, and RRMS patients: FA (splenium) > FA (genu) > FA (body). Using only normal CC measurements in three age groups, Chepuri et al (29) found in both adult males and females that tensor anisotropy was significantly larger in the splenium compared to the genu of the CC. The trend of FA (splenium) > FA (genu) has also been observed in other studies with large populations of healthy adult males and females (22, 25). On the other hand, Foong et al (26) reported very similar FA values for the splenium and genu of the CC. In contrast, Ciccarelli et al (16) reported the following values for the CC of healthy adults: FA (genu) > FA (splenium) > FA (body). They also found that the splenium FA significantly decreased, while the genu was normal in MS patients. Cercignani et al (17) reported values of FA (splenium) > FA (genu) in healthy adults, and observed that FA was significantly reduced in both the genu and splenium of MS patients compared to controls. Bammer et al (18) reported that FA (splenium) > FA (body), and found that while FA (body) of the normal-appearing CC was not significantly different from that of healthy controls, the FA (splenium) of MS patients showed an insignificant decrease in the splenium. Griffin et al (19) reported that FA (genu) > FA (splenium) in healthy controls, and failed to detect any differences in Dav or FA in major NAWM structures, including the NAWM CC of the MS patients studied.

Our observed nonsignificant FA decreasing trend in the splenium of MS patients compared with age-matched healthy controls is generally consistent with some reports (18). Our results on the nonsignificant FA differences in the genu of age-matched healthy controls and MS patients are consistent with some studies (16, 18), but contradict other reports (13, 16) in which the genu anisotropy significantly decreased. As shown in Fig. 1, in one RRMS patient we were able to track frontocallosal fibers traversing through the genu, even in the close vicinity of a non-enhancing and inflammatory frontal lesion. The ability to track compact fiber bundles in the vicinity of frontal lesions (Fig. 1) and the entire CC (Fig. 6) may indicate certain patterns with regard to the fiber size, function, selectivity, location, severity, and connectivity of MS lesions (1–4, 15, 38). It is clear that longitudinal follow-up of the evolution of these FLAIR-localized lesions using DTI would be helpful in assessing Wallerian degeneration. Such investigations are currently under way.

Our findings that the FA (CC body) is significantly decreased in the NAWM of the CC of RRMS patients are in agreement with previous findings by Ge et al (13), who observed a significant reduction in FA of the genu and splenium in 15 MS patients and 12 healthy adults. The contradictions between our findings in the body of the CC and published works on the RRMS patients may be attributed to the use of different encoding strategies (in the present study, a multifaceted rotationally invariant strategy), acquisition with dual SEs with minimal image distortion due to eddy currents, SNR, processing, ROI selection methods, and possibly the patients studied (2). Note that the selection of a high SNR (88 T2w and DW images per slice), balanced encoding scheme, and slice thickness increased the confidence and repeatability of the method, and suggest that our results are robust (11, 35). Poor selection of the working SNR would elevate the estimated FA, since noise would lead to overestimation of the intrinsic anisotropy (11). The trend measured in our studies was for an FA decrease in the NAWM of the CC of the MS subjects relative to the healthy age-matched controls. This observed decrease may not be explained by an effect due to SNR alone. A larger slice thickness should reduce the estimated FA and decrease the specificity of DTI, resulting in reduced heterogeneity of the CC structure as reported in other studies. At higher magnetic fields, the spatial resolution and hence specificity may be enhanced without compromising the SNR. We plan to explore and extend these studies at 3.0 T in the future.

A comparison of the histopathological studies of CC and spinal cord in MS patients might provide clues about the relationships among the DT anisotropy FA, axonal integrity, and lesion spatial distribution (38). Evangelou et al (6, 7) described spatially-dependent and DD-independent axonal density loss in the NAWM of the CC in MS subjects. They showed a significant decrease in the callosal axonal density in the body of the NAWM of the CC, and correlated the loss in axonal density and regional lesion load (6, 7). The authors attributed the callosal axonal loss to Wallerian degeneration. These histopathologic studies suggest that the reduced FA value observed in the body of the CC in the current studies could result from the loss of small fibers (1–4) (see also Fig. 6b). Our studies suggest that the reduced FA value observed in the midbody (CC4-CC5) may reflect reduced axonal density in MS (6, 7, 10, 27, 30, 38).

The poor correlation of the EDSS with regional callosal FA values observed in the current study is not surprising given that EDSS measures motor disability and the CC is not involved directly in ambulation (Table 1). The poor correlation of regional callosal FA with DD may reflect a general characteristic of MS. It is worth noting that in histological studies, Evangelou et al (6, 7) reported that axonal density loss did not correlate with DD, and that fiber morphology, proximity, and connectivity to lesions would provide important clues for understanding the pathogenesis of MS.

Table 1. A Summary of the Spearman Correlation Coefficient r and P values (rs(p)) Between CC1–CC7 Regional FA and DDV Values, Disease Duration (DD) and EDSS for the 26 RRMS Patients Included*
 FA vs. DD rs(p)FA vs. EDSS rs(p)Dav vs. DD rs(p)Dav vs. EDSS rs(p)
  • *

    The linear regression statistical significance was considered at P* · 0.05.

CC10.0607 (0.7684)0.0701 (0.7337)0.0115 (0.9557)−0.0176 (0.9320)
CC2−0.0253 (0.9024)−0.3031 (0.1323)0.1752 (0.3919)0.1024 (0.6187)
CC30.2267 (0.2655)−0.0969 (0.6376)−0.2544 (0.2099)0.3323 (0.0972)
CC40.1701 (0.4061)−0.0174 (0.9326)−0.1174 (0.5678)0.1573 (0.4429)
CC50.1516 (0.4597)0.0256 (0.9011)−0.1679 (0.4124)−0.0003 (0.9987)
CC60.1385 (0.5000)−0.0870 (0.6726)−0.2036 (0.3185)0.1352 (0.5102)
CC70.0937 (0.6490)0.0285 (0.8899)−0.1385 (0.5000)0.2126 (0.2970)

In our studies we observed that the CC regional anisotropy is heterogeneous: FA (splenium) > FA (genu) > FA (body of healthy adults (age-matched males and females). The same trend was also maintained in the CC of RRMS patients (Figs. 4 and 5). It is clear that histopathological knowledge of the normal adult human CC is required for proper interpretation of FA and its possible association with axonal density. A complete association between FA in the human brain compact WM and fiber morphology remains to be established (10, 39), and only a preliminary correlative study has been performed on live excised sea lamprey spinal cords (40). A close look at Fig. 1 in Aboitiz et al (30) (after correction for the shrinkage factor) and Fig. 8 in Lemantia and Rakic (31) reveals that the density of myelinated fibers with diameter larger than 1.5 μm correlates with the regional trend we observed in our studies. The splenium and genu myelination levels are about 97% and 72%, respectively, and thus the interfiber distance in the splenium is smaller than that in the genu. The body of the CC region contains larger-diameter fibers, and hence it has a greater interfiber distance compared to the genu (30, 31). Using histology, Aboitz et al (30) also found that the axonal packing regional heterogeneity trends in the CC were gender-independent. This may explain our preliminary results that there was little difference in the CC regional anisotropy values between the 16 healthy adult age-matched males and females (Fig. 4). It is important to recognize that the interpretation provided in this preliminary study is only rudimentary and speculative, and combined DTI and histopathological analyses on the CC, in addition to systematic computational modeling and simulations (10, 27, 30, 39), are required.

In this study we qualitatively observed a causal relationship between the periventricular lesion load and location and the CC connectivity. However, a rigorous and unbiased quantitative analysis of the relationship between the spatial location of the lesions and the regional FA values observed in the CC requires an automatic analysis of the regional distribution of the lesions. At the present time we do not have this capability.

In conclusion, we used optimized whole-brain DTI to investigate the involvement of the seven functionally different CC segments in MS patients relative to a cohort of age-matched healthy adults. This study 1) demonstrates a relatively gender-independent spatial heterogeneity in the DTI metrics in the NAWM of CC in both controls and MS patients, 2) implicates both the anterior and posterior midbody of the CC (CC4-CC5) as being vulnerable in MS, 3) reveals no significant difference in the FA of the genu between the MS cohorts and age-matched normal controls, and 4) indicates a lack of correlation between regional CC FA values and both EDSS and DD.


We thank all of the volunteers and MS patients for participating in these studies. The editorial assistance of Marci A. Harris is appreciated. They thank Vimala Ahobila for assisting with the MRI data collection and Bhavik P. Knabar for technical assistance. K.M.H. wishes to thank Drs. M. Esiri, R. Highley, and F. Aboitiz for providing useful details on their cited publications. This work was funded in part by the NIH (grants R01 NS31499 and R01 EB02095). K.M.H. is funded in part by the Dunn Research Foundation and the Department of Diagnostic and Interventional Imaging, University of Texas at Houston.