Diffusion tensor imaging reveals regional differences in the cervical spinal cord in amyotrophic lateral sclerosis

NeuroImage 53 (2010) 576–583

Contents lists available at ScienceDirect

NeuroImage j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / y n i m g

Diffusion tensor imaging reveals regional differences in the cervical spinal cord in amyotrophic lateral sclerosis Govind Nair a, John D. Carew b,c, Sharon Usher d, Debbie Lu d, Xiaoping P. Hu a,⁎, Michael Benatar c,d,e,⁎ a

Biomedical Imaging Technology Center, Department of Biomedical Engineering, Emory University and Georgia Institute of Technology, Atlanta, GA, USA Epidemiology and Biostatistics, Carolinas Medical Center, Dickson Institute for Health Sciences, Charlotte, NC, USA School of Public Health, Emory University, Atlanta, GA, USA d Department of Neurology, Emory University School of Medicine, Atlanta, GA, USA e Department of Epidemiology, Emory University School of Medicine, Atlanta, GA, USA b c

a r t i c l e

i n f o

Article history: Received 31 March 2010 Revised 15 June 2010 Accepted 23 June 2010 Available online 28 June 2010 Keywords: Diffusion tensor imaging Amyotrophic lateral sclerosis Cervical spinal cord Imaging biomarker of disease severity Dying-back hypothesis Neurodegeneration

a b s t r a c t Therapeutic development in amyotrophic lateral sclerosis (ALS) is hampered by the lack of suitable biomarkers that might be sensitive to spatial and temporal patterns of neurodegeneration. Diffusion tensor imaging is a useful non-invasive tool that permits detection of microstructural tissue changes due, for example, to neurodegeneration. Even though the spinal cord bears the brunt of the disease process, diffusion tensor imaging has mainly been used to study white matter changes in the brain. The aim of this study was to examine the diffusion tensor imaging parameters of the cervical spinal cord (C1 through C6 segments) and brainstem (corticospinal tracts in the pyramids and pons) among ALS patients, to compare these to findings in age-matched healthy controls, and to correlate these differences with clinical measures of disease severity. Fractional anisotropy in the white matter of the cervical cord was 12% lower (p b 0.01) in ALS patients (n = 14) compared to age-matched healthy control subjects (n = 15), and showed significant positive correlation with the average finger and foot tapping speed (r = 0.61, p b 0.05) in ALS patients. Radial diffusivity in the cervical cord was 15% higher (p b 0.05) in ALS patients compared to healthy control subjects. Radial diffusivity in the white matter of the cervical cord was significantly correlated with clinical measures of disease severity such as forced vital capacity (FVC % predicted, r = −0.69, p b 0.01), average finger and foot tapping speed from all four limbs (r = −0.59, p b 0.05), and ALSFRS-R (r = −0.55, p b 0.05) in ALS patients. There were no significant differences in mean diffusivity or axial diffusivity in the cervical spinal cord, or in any diffusion tensor imaging parameters measured in the brainstem. Analysis of diffusion tensor imaging parameters from individual cervical segments as well as profile plots along the length of the cervical cord showed larger differences in fractional anisotropy and radial diffusivity at more distal cervical segments, providing evidence that supports the “dying-back” hypothesis of neurodegeneration in ALS. © 2010 Elsevier Inc. All rights reserved.

Introduction Progress in developing therapies for amyotrophic lateral sclerosis (ALS) has been slow in part because of a paucity of objectively measurable indicators of the disease process. Such indicators, often described as biomarkers, have been sought using a wide variety of techniques including serology, physiology and imaging. Efforts to develop imaging biomarkers for ALS have focused largely on the brain with relative neglect of the spinal cord. This is perhaps surprising given Abbreviations: ALS, Amyotrophic lateral sclerosis; CST, Corticospinal tract; ALSFRS-R, ALS Functional Rating Score (Revised); FVC, Forced vital capacity. ⁎ Corresponding authors. M. Benatar is to be contacted at Woodruff Memorial Research Building, Suite 6000, 101 Woodruff Circle NE, Atlanta, GA 30322, USA. Fax: + 1 404 778 3767. X.P. Hu, Woodruff Memorial Research Building, Suite 4000, 101 Woodruff Circle NE, Atlanta, GA 30322, USA. Fax: + 1 404 712 2707. E-mail addresses: [email protected] (X.P. Hu), [email protected] (M. Benatar). 1053-8119/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2010.06.060

that the spinal cord bears the brunt of the disease process in ALS, but understandable given the technical challenges inherent to applying sophisticated imaging techniques to the study of this anatomical region. Diffusion tensor imaging (DTI) is a non-invasive quantitative MRI method that provides information about the microstructural properties of tissue. Reduced axonal integrity, a hallmark of many neurological diseases including ALS, alters tissue microstructure, which in-turn alters the diffusion of water molecules in the tissue and is measurable with DTI. As such DTI can be used to demonstrate disease-correlated tissue changes that are not evident on conventional MRI. Sensitivity to microstructure and tissue organization makes DTI an important tool for studying neurodegenerative disease (Gulani and Sundgren, 2006; Le Bihan et al., 2001; Mori and Zhang, 2006). The preponderance of DTI studies in ALS have focused on the brain, revealing a decrease in fractional anisotropy (FA), and an increase in mean diffusivity (MD) (Abe et al., 2004; Cosottini et al., 2005; Ellis et al., 1999; Sach et al., 2004; Sage et al., 2007; Segawa et al., 1994;

G. Nair et al. / NeuroImage 53 (2010) 576–583

Toosy et al., 2003); these changes in FA and MD may correlate with clinical measures of disease severity (Abe et al., 2004; Agosta et al., 2009; Cosottini et al., 2005; Ellis et al., 1999; Sach et al., 2004; Sage et al., 2007; Segawa et al., 1994; Toosy et al., 2003). More recently, similar findings have been reported in the spinal cord of ALS patients, but without any clear correlation with clinical measures of disease severity (Agosta et al., 2009). The development of DTI as a biomarker of disease progression, or potentially as a marker of therapeutic response, requires a better understanding of the nature of the spinal cord changes that are seen in ALS. One possibility is that other DTI parameters, such as axial diffusivity (AD) and radial diffusivity (RD), might be more sensitive biomarkers with better correlation with clinical measures of disease severity. Alternatively, it is possible that a better correlation between DTI and clinical measures of disease severity may become apparent through an evaluation of individuals cervical spinal cord segments. This study evaluates differences in DTI parameters in the cervical cord and brainstem in patients with ALS and age-matched healthy controls. FA, MD, RD, and AD are examined in regions of interest (ROIs) that either encompasses the entire cervical spinal cord or brainstem as well as in ROIs at individual cervical segments; correlations between these DTI metrics and clinical measures of ALS severity are also examined. Materials and methods Fourteen patients with ALS and 15 healthy control subjects were enrolled in this study. The severity of disease in the ALS patients was evaluated using the revised ALS functional rating scale (ALSFRS-R) and the forced vital capacity (FVC). The study protocol required that participants had an FVC of at least 40% predicted and were able to lay recumbent for an hour. Finger and foot tapping speeds were measured individually from each of the four limbs over 10 s using an electronic tapping device (Western Psychological Services, CA, USA). Average tapping speed from all four limbs was calculated and used as a clinical measure of upper motor neuron dysfunction in ALS (although recognizing that coexisting lower motor neuron dysfunction may also impact tapping speed). MRI was performed on a 3T Siemens TIM™ TRIO whole-body MR scanner (Siemens Medical Solutions, Malvern, PA, USA). Images were acquired with body RF coil in transmit mode and a 12-element head matrix coil as well as 2-channel neck coil in receive mode. Coronal slices for DTI were positioned on sagittal T2-weighted scans with the inferior aspect of the thalamus on the mid-sagittal images forming the superior boundary of the DTI slices. DTI was performed using standard single-shot EPI sequence with TR = 3200 ms, TE = 105 ms, FOV = 160 mm2, 128 × 128 imaging matrix, acquisition bandwidth of 908 Hz/pixel, in-plane resolution of 1.25 mm2, 19 coronal slices of 2.5 mm thickness and 2 averages. Parallel imaging was used with a GRAPPA factor of 2, and the readout time for susceptibility correction was determined to be 47.19 ms. Diffusion gradients were applied evenly along 30 directions with a b-value of 1000 s/mm2. The DTI data acquisition was repeated (for a total of 4 averages) with the phase encode direction rotated by 180°, to remove susceptibility induced artifacts (Andersson et al., 2003). The total acquisition time was approximately 7 min for 4 averages. Data processing was performed using scripts written in Matlab, SPM toolbox, and UNIX using functions in FSL (Analysis Group, FMRIB, Oxford, UK). The diffusion weighted images and the B0 image were first co-registered with each other using affine registration to eliminate eddy current and movement artifacts. Image distortions due to susceptibility changes were then corrected by method described elsewhere (Andersson et al., 2003), and combined to yield the raw images for DTI processing. The phase swap method of distortion correction was preferred to others such as the phase map technique (Jezzard and Balaban, 1995) as it maximized the number of signal averages in the DTI scan for the available scan time. FA, MD, and

577

principle eigenvalue maps were calculated using DTI toolbox in FSL (Analysis Group, FMRIB, Oxford, UK). The principal eigenvalue map was assigned as AD and an average of the second and third eigenvalue maps were assigned as RD. Brainstem and spinal cord (to the inferior aspect of the C6 vertebral body) were extracted from the FA, RD, MD, and AD maps using ROIs drawn manually for each subject. Voxelwise statistical analyses of these DTI metric maps were performed using Tract-Based Spatial Statistics (TBSS v 1.2) tools in FSL (Smith et al., 2006, 2004). The FA maps of all subjects were co-registered to each other, first using linear registration with 6 and 12 degrees-of-freedom, and then using nonlinear registration. The registration matrices were then used to transform RD, MD and AD images into the common space. The FA skeleton mask was created from the mean FA map of all subjects, at a threshold of 0.2. Individual aligned FA maps were then projected onto the skeleton. Voxelwise comparison between ALS and control groups were performed with nonparametric 2-sample Student's t-test corrected for multiple comparisons across space, using the randomize function in FSL with threshold free cluster comparison option (Bullmore et al., 1999; Nichols and Holmes, 2002). A p-value b 0.1 corrected for multiple comparisons was taken to be statistically significant, since the ALS group was heterogeneous with MRI scans done between 56 and 2366 days of symptom onset. Despite using a relatively high p-value for the group comparison, experiment-wise type I error rates were controlled by computing the permutation distribution of the maximal voxel statistics over the volume of interest, by using the corrected-p values in TBSS (Holmes et al., 1996). The TBSS and voxelwise statistics were then repeated for the RD, MD and AD maps using the skeleton mask derived from the FA map. Profile plots of DTI parameters were constructed along the cervical cord and brainstem using FSL tools and Matlab scripts. First, the mean FA skeleton was cropped to preserve only the corticospinal tract (CST) along the brainstem, and the white matter in the cervical cord using anatomical landmarks and atlases as guides. The lateral column was not separately analyzed in the cervical cord due to the limitation of its size, compared to the resolution of the acquired DTI. This croppedskeleton was then used as a mask for DTI maps from each subject. The non-zero DTI values along the slice and L-R directions were averaged to obtain a plot of average DTI parameter along the length of the cord and brainstem. A two-tailed t-test was performed between the ALS and control groups at each point along the profile plot and a p b 0.05 was considered to be statistically significant. ROIs were drawn on the cropped WM skeleton (obtained to perform the profile plot) along various sections of the cord (C1 through to C6) and brainstem (pons and pyramidal tracts), again using anatomical landmarks from transformed B0 images, mean FA image, as well as atlas as guides. Separate ROIs were also drawn for comparison of DTI parameters within the entire cervical cord from C1 through C6 (designated as the cervical cord ROI) and the pons and pyramid combined (designated as the brainstem ROI). Percent difference in mean of DTI parameters in ALS patients compared to healthy control subjects were determined at each CST segment as well as from the cervical cord and brainstem ROIs. DTI parameters from individual ROIs were correlated with clinical measures of ALS severity such as ALSFRS-R, FVC, tapping speed, and duration of disease. All values reported are mean ± SD and a p-value less than 0.05 was considered to be statistically significant. This study was approved by Emory University Institutional Review Board (IRB) and all participants provided written informed consent. Results Study participants The ALS group included 14 patients (4 females; 50 ± 13 years age; median disease duration of 2 years, with IQR of 1.1–3.8 years; ALSFRS-

578

G. Nair et al. / NeuroImage 53 (2010) 576–583

R of 38 ± 5) with El Escorial definite (n = 3), possible (n = 6) and familial ALS (n = 5). Limb onset disease was most common (n = 13) with some limb involvement in all. Bulbar onset disease occurred in a single patient and 3 of the 14 ALS patients reported some bulbar symptoms. The control group comprised 15 healthy individuals (8 females, 49 ± 14 years age). The average age was comparable between the two groups (p N 0.85), but there was a preponderance of males in the ALS group (71% vs. 49%). Average finger and foot tapping speed in the ALS group was 37 ± 14/10 s, which was significantly lower (p bb 0.01) than that in the control group (53 ± 8/10 s), as expected. Eight of the 14 ALS patients (2 females) were taking Riluzole at the time of evaluation; 2 additional patients (1 female) had used the drug within the past year. Mean ALSFRS-R score was significantly higher (p b 0.05) among Riluzole users (41 ± 4) compared to patients not on Riluzole (35 ± 4), and disease duration was significantly shorter (1.4 ± 1.3 years) compared to patients not

taking Riluzole (3.8 ± 2 years) (p b 0.05). Patients on or off Riluzole were otherwise comparable. Regional differences in DTI parameters of the cervical cord and the brainstem B0 and diffusion weighted images were successfully acquired in all ALS patients and healthy control subjects. The distortions seen in cord B0 and DWI images were successfully corrected (Fig. 1A and B) using a previously described phase swap method (Andersson et al., 2003), and FA, RD, MD (representative maps shown in Fig. 1C–E) and AD maps were reconstructed using the corrected B0 and DWIs. The FA maps from individual subjects as well as mean FA map from all subjects obtained after affine and nonlinear registrations (Fig. 1F) did not show any misregistration based on visual inspection. The skeleton derived in FSL with a threshold of 0.2 in FA (Fig. 1F, green overlay)

Fig. 1. DTI of cervical spinal cord. (A) A single coronal B0 slice (2 averages) from a representative healthy control subject depicting susceptibility distortion, which causes artificial curvature of the cord in the LR plane and stretching near intervertebral discs. The cord moves out of the imaging slice at approximately the C6 vertebral body. (B) Distortion corrected B0 slice from the same subject showing a straight cord. (C) Representative FA (range 0–0.9), (D) MD (range 0–2.3 × 10−3 mm2/s), and (E) RD (range 0–2.1 × 10−3 mm2/s) maps of the brainstem and spinal cord extracted shows sufficient signal-to-noise ratio within the cord for accurate DTI quantification. (F) White matter skeleton (in green), derived using a threshold of 0.2 and FSL tools, overlaid on corresponding mean fractional anisotropy (FA) slice from all subjects after nonlinear registration. The cord and skeleton moves out of the slice plane between approximately the C1 though C3 levels and is not seen in this image. Note: the cord was truncated at the level of C6–C7 disk and levels below were not included in the analysis so as to remove variations in the size of cervical spinal cord among subjects. Volume rendered image (slightly magnified) showing regions with (G) a significant reduction in FA (in red, p b 0.1), and (H) a significant increase in radial diffusivity (in green, p b 0.1) in ALS patients compared to age-matched healthy controls. Differences in FA are more extensive along the cervical cord than differences in RD. (I) Volume rendered image showing ROIs drawn along the corticospinal tract starting superiorly at the region of the pons (yellow), pyramids (cyan), C1 and C2 region (red), C3 (blue), C4 (green), C5 (violet), and C6 (yellow). The ROIs were drawn with B0 image as reference. (n = 14 ALS and n = 15 age-matched healthy controls).

G. Nair et al. / NeuroImage 53 (2010) 576–583

corresponded well to the expected locations of WM fiber tracts in the cord. There were no significant differences in DTI metrics between male and female subjects in the healthy control group (data not shown) at the given sample size. However, RD, MD and AD were higher, and the FA lower, in the cervical cord than in the brainstem of healthy control subjects (p b 0.01, data not shown). Larger variations in FA were seen along the cervical cord than the brainstem, and there was a prominent dip in FA at approximately the level of medulla oblongata. Prominent DTI differences seen in the cervical cord but not in the brainstem of ALS FA is reduced and RD is increased in the spinal cord of ALS patients relative to controls. The region of decreased FA spanned roughly from C1 to C6 cervical segments (p b 0.1 corrected for multiple comparisons across space, Fig. 1G, regions in red) and the region of increased RD was observed at approximately the C3, C4 and C5 levels (p b 0.1, corrected for multiple comparisons across space, Fig. 1H, regions in blue). There were no regions that showed differences between ALS patients and healthy controls in the MD and AD maps. Group-average FA was lower in patients with ALS (0.45 ± 0.06) compared to healthy controls (0.51 ± 0.03) within the cervical ROI, a reduction of 11% (p = 0.003, Fig. 2A). However, FA values in the brainstem ROI were no different between ALS patients (0.48 ± 0.04) and healthy controls (0.48± 0.04) (a difference of 1.0%, p = 0.74). Groupaveraged RD in the cervical cord ROI was higher among patients with ALS (0.72± 0.13 × 10−3 mm2/s) compared to healthy controls (0.63 ± 0.05× 10−3 mm2/s), a difference of 14.6% (p = 0.027, Fig. 2B). As with

579

FA, there were no differences in RD within the brainstem ROI between the ALS (0.43 ± 0.05 × 10−3 mm2/s) and control group (0.43 ± 0.04 × 10−3 mm2/s, p = 0.9, a difference of 0.5%). Differences in mean MD values between ALS (0.96 ± 0.13 × 10−3 mm2/s) and control (0.89 ± 0.06 × 10−3 mm2/s) groups approached significance (p = 0.098, a difference of 7.7%, Fig. 2C) in the cervical ROI, but were not different in the brainstem ROI (p = 0.81, a difference of 0.8%). In contrast, mean AD values of ALS and control groups were not significantly different in either the cervical cord (p = 0.67, difference of 1.4%, Fig. 2D) or the brainstem ROIs (p = 0.74, difference of 1.1%, Table 1). There were also no significant differences in any DTI parameter between patients on Riluzole and those not on Riluzole in the ALS group, but these groups were not matched for disease duration or ALSFRS-R scores. To better understand the differences in DTI that might occur with ALS progression, the DTI parameters were correlated with clinical measures of disease severity among the patients with ALS. FA values from the cervical cord ROI correlated with the average finger and foot tapping speed (Fig. 2E, r = 0.61, p = 0.019), but not with any other clinical measure of disease severity. There were significant negative correlations between RD of the cervical cord ROI and average finger and foot tapping speed (Fig. 2F, r = −0.59, p = 0.027), FVC (data not shown, r = −0.69, p = 0.006), and ALSFRS-R (data not shown, r = −0.55, p = 0.04). Similarly, MD values obtained from the cervical ROI were significantly negatively correlated with FVC (data not shown, r = −0.67, p = 0.008) and ALSFRS-R (data not shown, r = −0.56, p = 0.036). AD values correlated only with FVC (data not shown, r = −0.54, p = 0.046). In the brainstem, only ALSFRS-R showed any correlation with RD (data not shown, r = −0.61, p = 0.020), MD (data not shown, r = −0.70,

Fig. 2. The decreased diffusion anisotropy seen in the cervical cord of ALS correlates with disease severity. ROI drawn over the WM in cervical cord (C1 through C6) reveals (A) decreased fractional anisotropy (FA, p = 0.003), and (B) increased radial diffusivity (RD, p = 0.027) in the cervical cord of ALS patients compared to age-matched healthy controls. (C) Mean diffusivity (MD) approaches statistical significance (p = 0.089), but (D) axial diffusivity (AD) was no different between the two groups. There is significant correlation between DTI measures in ALS patients and clinical measures of disease progression, such as (E) RD and (F) FA with average tapping speed of limbs, among others (* p b 0.05; ** p b 0.01; n = 14 for ALS and n = 15 for agematched healthy controls).

580

G. Nair et al. / NeuroImage 53 (2010) 576–583

Table 1 Group-averaged percentage differences in DTI parameters in ALS patients relative to age-matched healthy controls are calculated from the brainstem (corticospinal tracts of the pyramids and pons), cervical cord ROIs (all segments C1 through C6) and individual cervical segments C1 through C6. The numbers in parenthesis indicate standardized differences (* p b 0.05; ** p b 0.01; n = 14 for ALS group and n = 15 for age-matched healthy control group). % Difference in ALS vs. HC

FA

Brainstem ROI Cervical ROI C1 and C2 C3 C4 C5 C6

1.0 −11.8 −11.0 −10.1 −12.1 −12.7 −14.7

RD

MD

AD

(26) −0.5 (−10) −0.8 (−15) −1.1 (−14) (−252)** 14.6 (156)* 7.7 (81) 1.4 (12) (−305)** 11.2 (135)* 5.2 (61) −0.5 (−5) (−188)* 12.1 (108) 6.2 (57) 0.9 (7) (−209)** 20.6 (156)* 11.6 (87) 4.0 (25) (−225)** 21.3 (160)* 12.8 (93)* 5.2 (30) (−240)** 13.7 (106) 7.0 (51) 0.7 (4)

p = 0.005), and AD (data not shown, r = −0.75, p = 0.002). Neither patient age nor disease duration correlated with any of the DTI measures in either the brainstem or the cervical cord. Spatial patterns of DTI differences in ALS Profile plot of FA along the cervical cord and brainstem reveals a lower FA (p b 0.05) among ALS patients compared to controls between C2 and C6 cervical segments (Fig. 3A). Profile plot of RD shows an

increase in RD in ALS group along a similar stretch of cord from C2–C6, except for a small segment spanning portions of the C3 and C4 cervical segments (Fig. 3B and C). Profile plot of MD and AD did not show significant differences between ALS patients and healthy controls along the entire cervical cord and brainstem, with the exception of MD at a small segment at C5 level (data not shown). A segment-by-segment comparison shows greater differences in FA between ALS patients and healthy controls at lower cervical segments, with an 11.0% reduction at C1–C2 and a 14.7% reduction at C6 (Table 1). The standardized difference in means between the two groups also generally increased from C3 to C6 segments, but was highest for C1–C2 segment. Similarly, the percent difference as well as standard difference in RD, MD and AD values between ALS and healthy control groups generally increased distally along the cervical cord, with the exception of C6 segment which may have been affected by the noisy inferior pixels. AD and MD values form ALS and healthy control groups were not statistically different along the length of the cord, except MD at the C5 segment. The FA and RD values at individual segments also correlated significantly with clinical measures of disease severity. Tapping speed correlated significantly with FA at all segments of the cervical cord (Table 2A). Furthermore, the degree of correlation steadily decreased from a Pearson's correlation coefficient of 0.60 at C1–C2 segment to 0.54 at the C6 segment. Significant correlations were also observed between tapping speed and RD at upper segments of the cervical cord

Fig. 3. Profile plot of the average (A) fractional anisotropy, and (B) radial diffusivity obtained from ALS (red line) and healthy control subjects (blue line) along the length of the cervical cord and brainstem. Y-axis crosses in the medulla oblongata inferior to the olive, as seen in (C) which is Fig. 1I, rotated and stretched to match the x-axis. The error bars denote standard deviations, and horizontal dashed line denotes regions showing significant differences in DTI values between ALS and healthy control groups (* p b 0.05; n = 14 for ALS and n = 15 for age-matched healthy controls).

G. Nair et al. / NeuroImage 53 (2010) 576–583 Table 2 (A) Fractional anisotropy (FA) and (B) radial diffusivity (RD) from individual cervical segments correlate significantly with clinical measures of disease severity such as average tapping speed (taps/10 s) and forced vital capacity (FVC, % predicted) in the ALS group. The correlation of RD with FVC is stronger at lower segments of the cord, while the correlation of both FA and RD with tapping speed is weaker in distal cord segments (* p b 0.05; ** p b 0.01; n = 14 for ALS group). A. Pearson's correlation (fractional anisotropy) C1 and C2 C3 C4 C5 C6

ALSFRS-R 0.34 0.20 0.28 0.17 0.33

FVC 0.38 0.45 0.41 0.52 0.48

Tapping speed 0.60* 0.61* 0.59* 0.54* 0.54*

Disease duration −0.05 0.07 0.14 −0.02 0.00

B. Pearson's correlation (radial diffusivity)

ALSFRS-R

FVC

Tapping speed

Disease duration

C1 and C2 C3 C4 C5 C6

−0.65* −0.37 −0.48 −0.41 −0.51

−0.55* −0.64* −0.63* −0.69** −0.71**

−0.72** −0.61* −0.54* −0.41 −0.39

0.27 0.15 0.10 0.19 0.27

(Table 2B). The degree of correlation decreased from a highly significant −0.72 at segment C1−C2 to −0.54 at segment C4 and dropped below the threshold for statistical significance at segments C5 and C6. In contrast, the correlation between FA and FVC steadily increased from 0.38 at C1–C2 segment to 0.48 at C6 cervical segment (Table 2A). A similar increase in correlation was observed between RD and FVC, with the degree of correlation increasing from a statistically significant −0.55 at C1–C2 segment to a highly significant −0.71 at C6 segment. FA and RD were not significantly correlated with any other clinical measure of ALS severity other than RD with ALSFRS-R at C1– C2 segment. Discussion DTI in ALS patients showed significant reduction in the anisotropy of water diffusion throughout the cervical cord compared to agematched healthy control subjects. The differences in FA and RD between ALS patients and healthy controls increased along the length of the cervical cord, with distal segments generally showing a larger RD and smaller FA in ALS. Furthermore, DTI parameters obtained from ALS patients were found to correlate with a variety of clinical measures of disease severity. Technical considerations for DTI of the cervical spinal cord The cross-sectional resolution of the cord could potentially have been improved, and the partial volume errors (PVE) reduced, by the use of axial slices rather than the coronal slices used herein; however, a large number of slices would be required to cover the cervical cord, increasing the scan time. Our use of coronal slices for DTI allowed coverage of almost the entire length of the cervical cord with just 19 slices, enabling the use of a relatively small TR (3200 ms) and larger number of averages within a given scan time for increased SNR. However, coronal slice placement also meant different coverage of the cervical cord for different patients depending on the size of the neck. This necessitated cropping of individual DTI maps at the inferior aspect of the C6 vertebrae using a mask drawn on the B0 image, to facilitate accurate co-registration. The coverage could potentially be improved at the expense of in-plane resolution. Moreover, the use of the skeletonize function in FSL for comparison of maps as well as in ROI generation minimized any further contamination of gray matter by CSF. Other DTI studies in the cervical cord have reported MD values that range from 0.80 × 10−3 mm2/s to 1.08 × 10−3 mm2/s (Bammer et al., 2002; Laun et al., 2009; Shanmuganathan et al., 2008; Wheeler-

581

Kingshott et al., 2002) from the cervical spinal cord of healthy control subjects, comparable to 0.89 ± 0.06 × 10−3 mm2/s reported here. The parameters used for DTI herein are similar to the optimal DTI parameters of b-value of 900 s/mm and angular sampling of 15 to 32 directions, determined using SENSE technique (Lee et al., 2006). These observations suggest that our choice of protocols were near optimal for DTI of the cervical cord. Comparison with other studies of DTI in ALS A recent study showed a reduction in cervical cord cross-sectional area and FA, as well as a significant increase in MD over time in ALS patients, but very few differences between ALS patients and healthy controls at the time of initial scanning (Agosta et al., 2009). However, the observed longitudinal changes in DTI parameters did not correlate with ALSFRS-R, perhaps because the selected ROI encompassed the entire brain and cervical cord. We have extended these previous observations to show that differences in DTI parameters encountered in ALS may be subtle with variation between cervical segments. Smaller ROIs limited to anatomical substructures, such as individual cervical segments, were helpful in establishing correlations with a variety of clinical measures in this study. Furthermore, detailed analysis of additional DTI parameters, such as AD and RD helped establish more specific correlations in ALS. For example, ALSFRS-R showed significant correlation with RD in the C1–C2 segment alone, but not with any other DTI parameter from the cervical cord ROI taken as a whole. Finally, the use of TBSS in this study, which eroded WM to remove potential PVE, may by extremely helpful in the study of cervical cord DTI. With the exception of the study by Agosta and colleagues discussed above, other studies of DTI in ALS have been limited to the brain. One of the first such studies reported increased AD and RD as well as an apparent loss of anisotropy within hyper-intense lesions of peri-ventricular WM in ALS patients (Segawa et al., 1994). A subsequent study found a significant decrease in FA and an increase in MD along the CST of ALS patients compared to age-matched healthy controls (Ellis et al., 1999). That study also showed significant correlation of FA with upper motor neuron involvement, assessed using trans-cranial magnetic stimulation, and of MD with ALS duration. Differences in FA and MD between ALS and controls were also observed by Toosy et al. (2003) in segments of the CST at and below the internal capsule. However, that study failed to find any correlation with clinical measures of disease severity. Others have also observed differences in DTI parameters extensively throughout the brain, including regions of the frontal, subgyral, and paracentral WM, precentral gyrus, corpus callosum, and the thalamus (Abe et al., 2004; Cosottini et al., 2005; Sach et al., 2004; Sage et al., 2007). Sage et al. (2009) used voxel-based analysis (VBA), TBSS, and profile plot analysis to explain these apparent inconsistencies in reports of DTI differences in ALS. Using improved VBA and TBSS techniques, these authors were able to show clear differences in DTI parameters in the CST in the brain of ALS patients compared to healthy control subjects. FA values along the length of the CST in the brain of ALS patients showed strong positive correlation with ALSFRSR using VBA and TBSS approaches in that study. The registration and analysis process used herein is very similar to those in the TBSS process described in Sage et al., with the notable exception of coregistration to a population based DTI atlas. Since no such atlas has been developed for the cervical spine, DTI images of the cord from all subjects were registered to each other before transforming to the common space to remove any bias in the registration process. Interpretation of results While the DTI metrics are thought to be highly sensitive to changes in the microstructural environment, these metrics cannot yet be used

582

G. Nair et al. / NeuroImage 53 (2010) 576–583

as a direct indicator of a specific degenerative process. Demyelinating changes in the WM are thought to be a major contributor to an increase in RD in the shiverer mouse but may also indicate local edema or the late stages of an inflammatory response (Ahrens et al., 1998; Gulani et al., 2001; Harsan et al., 2006; Lodygensky et al., 2009; Nair et al., 2005; Ono et al., 1995; Song et al., 2005). The primary pathology in ALS is that of relatively selective axonal degeneration or shrinkage in the CST with secondary demyelination and astrocytic gliosis (Hammer et al., 1979; Harvey et al., 1979; Lawyer and Netsky, 1953; Niebroj-Dobosz et al., 2007; Rafalowska and Dziewulska, 1996; Shahani et al., 1998), which could all result in the observed increase in RD. Acute axonal damage without demyelination (Song et al., 2003) or gliosis may result in a decrease in AD, which is not observed in this study from any ROI. FA, on the other hand could be affected by changes in either AD or RD or both. Larger percentage differences and better correlation with clinical measures of disease severity seen in RD than FA may reflect the more non-specific nature of FA. However, neuronal changes could be detected throughout the cervical cord using FA, while RD was sensitive to changes only in part of the cervical spinal cord (Fig. 1G and H), which may reflect the inherently complex nature of the degenerative process in ALS. While significant differences in FA and RD between ALS patients and healthy controls were seen along most of the cervical segments, both the profile plot and the ROI analysis of individual cervical segments seem to suggest larger differences in the lower segments of the cervical cord. Taken together, these results suggest that DTI of the cervical cord reliably detects neurodegenerative changes in ALS, and that these neurodegenerative changes are more prominent in the distal cervical cord, supporting the dying-back hypothesis of neurodegeneration in ALS (Fischer et al., 2004; Ince, 2000). If differences in DTI parameters are related to the neurodegenerative process in ALS, some correlation between these DTI parameters and clinical measures of disease severity is expected. Indeed, RD and MD values obtained from ALS patients showed significant negative correlation, and FA showed positive correlation with FVC and tapping speed. Moreover, while FVC was better correlated with RD at lower cervical segments, tapping speeds were better correlated with both FA and RD at the upper cervical segments, roughly to the cervical enlargement (Table 2A and B). The biological basis for these gradients is unclear. Interestingly, none of the DTI measures from the cervical cord individually correlated with ALS duration, which may be explained by the heterogeneity of the rate of disease progression in the study population.

Conclusion DTI in the cervical cord detected significantly reduced FA and increased in RD in ALS patients compared to age-matched healthy control subjects, and the DTI parameters were well correlated with many clinical measures of ALS progression. These differences were generally more pronounced in the distal cervical cord, perhaps supporting the dying-back hypothesis of neurodegeneration in ALS. This study suggests that DTI of the cervical cord may be a suitable imaging biomarker for monitoring disease progression. Longitudinal studies are underway.

Acknowledgments The authors wish to thank Dr. Longchuan Li of BITC for his technical help with susceptibility correction algorithms. This work was supported by the Amyotrophic Lateral Sclerosis Association [grant number 1712]; Muscular Dystrophy Association [grant number 132866]; and National Institute of Health [grant number PO1AG026423 and RO1EB002009].

References Abe, O., Yamada, H., Masutani, Y., Aoki, S., Kunimatsu, A., Yamasue, H., Fukuda, R., Kasai, K., Hayashi, N., Masumoto, T., Mori, H., Soma, T., Ohtomo, K., 2004. Amyotrophic lateral sclerosis: diffusion tensor tractography and voxel-based analysis. NMR Biomed. 17, 411–416. Agosta, F., Rocca, M.A., Valsasina, P., Sala, S., Caputo, D., Perini, M., Salvi, F., Prelle, A., Filippi, M., 2009. A longitudinal diffusion tensor MRI study of the cervical cord and brain in amyotrophic lateral sclerosis patients. J. Neurol. Neurosurg. Psychiatry 80, 53–55. Ahrens, E.T., Laidlaw, D.H., Readhead, C., Brosnan, C.F., Fraser, S.E., Jacobs, R.E., 1998. MR microscopy of transgenic mice that spontaneously acquire experimental allergic encephalomyelitis. Magn. Reson. Med. 40, 119–132. Andersson, J.L., Skare, S., Ashburner, J., 2003. How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging. NeuroImage 20, 870–888. Bammer, R., Augustin, M., Prokesch, R.W., Stollberger, R., Fazekas, F., 2002. Diffusionweighted imaging of the spinal cord: interleaved echo-planar imaging is superior to fast spin-echo. J. Magn. Reson. Imaging 15, 364–373. Bullmore, E.T., Suckling, J., Overmeyer, S., Rabe-Hesketh, S., Taylor, E., Brammer, M.J., 1999. Global, voxel, and cluster tests, by theory and permutation, for a difference between two groups of structural MR images of the brain. IEEE Trans. Med. Imaging 18, 32–42. Cosottini, M., Giannelli, M., Siciliano, G., Lazzarotti, G., Michelassi, M.C., Del Corona, A., Bartolozzi, C., Murri, L., 2005. Diffusion-tensor MR imaging of corticospinal tract in amyotrophic lateral sclerosis and progressive muscular atrophy. Radiology 237, 258–264. Ellis, C.M., Simmons, A., Jones, D.K., Bland, J., Dawson, J.M., Horsfield, M.A., Williams, S. C., Leigh, P.N., 1999. Diffusion tensor MRI assesses corticospinal tract damage in ALS. Neurology 53, 1051–1058. Fischer, L.R., Culver, D.G., Tennant, P., Davis, A.A., Wang, M., Castellano-Sanchez, A., Khan, J., Polak, M.A., Glass, J.D., 2004. Amyotrophic lateral sclerosis is a distal axonopathy: evidence in mice and man. Exp. Neurol. 185, 232–240. Gulani, V., Sundgren, P.C., 2006. Diffusion tensor magnetic resonance imaging. J. Neuroophthalmol. 26, 51–60. Gulani, V., Webb, A.G., Duncan, I.D., Lauterbur, P.C., 2001. Apparent diffusion tensor measurements in myelin-deficient rat spinal cords. Magn. Reson. Med. 45, 191–195. Hammer Jr., R.P., Tomiyasu, U., Scheibel, A.B., 1979. Degeneration of the human Betz cell due to amyotrophic lateral sclerosis. Exp. Neurol. 63, 336–346. Harsan, L.A., Poulet, P., Guignard, B., Steibel, J., Parizel, N., de Sousa, P.L., Boehm, N., Grucker, D., Ghandour, M.S., 2006. Brain dysmyelination and recovery assessment by noninvasive in vivo diffusion tensor magnetic resonance imaging. J. Neurosci. Res. 83, 392–402. Harvey, D.G., Torack, R.M., Rosenbaum, H.E., 1979. Amyotrophic lateral sclerosis with ophthalmoplegia. A clinicopathologic study. Arch. Neurol. 36, 615–617. Holmes, A.P., Blair, R.C., Watson, J.D., Ford, I., 1996. Nonparametric analysis of statistic images from functional mapping experiments. J. Cereb. Blood Flow Metab. 16, 7–22. Ince, P.G., 2000. Neuropathology. In: Brown, R.H., Meininger, V., Swash, M. (Eds.), Amyotrophic lateral sclerosis. Martin Dunitz, London, pp. 83–112. Jezzard, P., Balaban, R.S., 1995. Correction for geometric distortion in echo planar images from B0 field variations. Magn. Reson. Med. 34, 65–73. Laun, F., Stieltjes, B., Schluter, M., Rupp, R., Schad, L.R., 2009. Reproducible evaluation of spinal cord DTI using an optimized inner volume sequence in combination with probabilistic ROI analysis. Z. Med. Phys. 19, 11–20. Lawyer Jr., T., Netsky, M.G., 1953. Amyotrophic lateral sclerosis. AMA Arch. Neurol. Psychiatry 69, 171–192. Le Bihan, D., Mangin, J.F., Poupon, C., Clark, C.A., Pappata, S., Molko, N., Chabriat, H., 2001. Diffusion tensor imaging: concepts and applications. J. Magn. Reson. Imaging 13, 534–546. Lee, J.W., Kim, J.H., Kang, H.S., Lee, J.S., Choi, J.Y., Yeom, J.S., Kim, H.J., Chung, H.W., 2006. Optimization of acquisition parameters of diffusion-tensor magnetic resonance imaging in the spinal cord. Invest. Radiol. 41, 553–559. Lodygensky, G.A., West, T., Stump, M., Holtzman, D.M., Inder, T.E., Neil, J.J., 2009. In vivo MRI analysis of an inflammatory injury in the developing brain. Brain Behav. Immun. 24, 759–767. Mori, S., Zhang, J., 2006. Principles of diffusion tensor imaging and its applications to basic neuroscience research. Neuron 51, 527–539. Nair, G., Tanahashi, Y., Low, H.P., Billings-Gagliardi, S., Schwartz, W.J., Duong, T.Q., 2005. Myelination and long diffusion times alter diffusion-tensor-imaging contrast in myelin-deficient shiverer mice. NeuroImage 28, 165–174. Nichols, T.E., Holmes, A.P., 2002. Nonparametric permutation tests for functional neuroimaging: a primer with examples. Hum. Brain Mapp. 15, 1–25. Niebroj-Dobosz, I., Rafalowska, J., Fidzianska, A., Gadamski, R., Grieb, P., 2007. Myelin composition of spinal cord in a model of amyotrophic lateral sclerosis (ALS) in SOD1G93A transgenic rats. Folia Neuropathol. 45, 236–241. Ono, J., Harada, K., Takahashi, M., Maeda, M., Ikenaka, K., Sakurai, K., Sakai, N., Kagawa, T., FritzZieroth, B., Nagai, T., et al., 1995. Differentiation between dysmyelination and demyelination using magnetic resonance diffusional anisotropy. Brain Res. 671, 141–148. Rafalowska, J., Dziewulska, D., 1996. White matter injury in amyotrophic lateral sclerosis (ALS). Folia Neuropathol. 34, 87–91. Sach, M., Winkler, G., Glauche, V., Liepert, J., Heimbach, B., Koch, M.A., Buchel, C., Weiller, C., 2004. Diffusion tensor MRI of early upper motor neuron involvement in amyotrophic lateral sclerosis. Brain 127, 340–350. Sage, C.A., Peeters, R.R., Gorner, A., Robberecht, W., Sunaert, S., 2007. Quantitative diffusion tensor imaging in amyotrophic lateral sclerosis. NeuroImage 34, 486–499.

G. Nair et al. / NeuroImage 53 (2010) 576–583 Sage, C.A., Van Hecke, W., Peeters, R., Sijbers, J., Robberecht, W., Parizel, P., Marchal, G., Leemans, A., Sunaert, S., 2009. Quantitative diffusion tensor imaging in amyotrophic lateral sclerosis: revisited. Hum. Brain Mapp. 30, 3657–3675. Segawa, F., Kishibayashi, J., Kamada, K., Sunohara, N., Kinoshita, M., 1994. MRI of paraventricular white matter lesions in amyotrophic lateral sclerosis–analysis by diffusion-weighted images. No To Shinkei 46, 835–840. Shahani, N., Nalini, A., Gourie-Devi, M., Raju, T.R., 1998. Reactive astrogliosis in neonatal rat spinal cord after exposure to cerebrospinal fluid from patients with amyotrophic lateral sclerosis. Exp. Neurol. 149, 295–298. Shanmuganathan, K., Gullapalli, R.P., Zhuo, J., Mirvis, S.E., 2008. Diffusion tensor MR imaging in cervical spine trauma. AJNR Am. J. Neuroradiol. 29, 655–659. Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E., JohansenBerg, H., Bannister, P.R., De Luca, M., Drobnjak, I., Flitney, D.E., Niazy, R.K., Saunders, J., Vickers, J., Zhang, Y., De Stefano, N., Brady, J.M., Matthews, P.M., 2004. Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage 23 (Suppl 1), S208–S219.

583

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: voxelwise analysis of multi-subject diffusion data. NeuroImage 31, 1487–1505. Song, S.K., Sun, S.W., Ju, W.K., Lin, S.J., Cross, A.H., Neufeld, A.H., 2003. Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. NeuroImage 20, 1714–1722. Song, S.K., Yoshino, J., Le, T.Q., Lin, S.J., Sun, S.W., Cross, A.H., Armstrong, R.C., 2005. Demyelination increases radial diffusivity in corpus callosum of mouse brain. NeuroImage 26, 132–140. Toosy, A.T., Werring, D.J., Orrell, R.W., Howard, R.S., King, M.D., Barker, G.J., Miller, D.H., Thompson, A.J., 2003. Diffusion tensor imaging detects corticospinal tract involvement at multiple levels in amyotrophic lateral sclerosis. J. Neurol. Neurosurg. Psychiatry 74, 1250–1257. Wheeler-Kingshott, C.A., Hickman, S.J., Parker, G.J., Ciccarelli, O., Symms, M.R., Miller, D. H., Barker, G.J., 2002. Investigating cervical spinal cord structure using axial diffusion tensor imaging. NeuroImage 16, 93–102.