Magnetic Resonance Imaging 32 (2014) 457–463
Contents lists available at ScienceDirect
Magnetic Resonance Imaging journal homepage: www.mrijournal.com
Diffusion tensor imaging of the cervical spinal cord of patients with Neuromyelitis Optica☆ René L.M. Rivero a,⁎, Enedina M.L. Oliveira b, Denis B. Bichuetti b, Alberto A. Gabbai b, Roberto G. Nogueira a, Nitamar Abdala a a b
Departamento de Diagnóstico por Imagem, Universidade Federal de São Paulo-UNIFESP, Brasil Departamento de Neurologia, Universidade Federal de São Paulo-UNIFESP, Brasil
a r t i c l e
i n f o
Article history: Received 12 June 2013 Revised 28 January 2014 Accepted 28 January 2014 Keywords: Neuromyelitis optica Diffusion tensor imaging Spinal cord Disability evaluation
a b s t r a c t Background: Previous studies have demonstrated a correlation between Expanded Disability Status Scale (EDSS) and Diffusion Tensor Imaging (DTI) metrics, but the conclusions were based on evaluations of the entire cervical spinal cord. Objectives: The purpose of this study was to quantify the FA and MD values in the spinal cord of NMO patients, separating the lesion sites from the preserved sites, which has not been previously preformed. In addition, we attempted to identify a correlation with EDSS. Methods: DTI was performed in 11 NMO patients and 11 healthy individuals using a 1.5-T MRI scanner. We measured the FA and MD at ROIs positioned along the cervical spinal cord. The mean values of FA and MD at lesion, preserved and spinal cord sites were compared with those of a control group. We tested the correlations between the mean FA and MD with EDSS. Results: FA in NMO patients was signiﬁcantly reduced in lesion sites (0.44 vs. 0.55, p = 0.0046), preserved sites (0.46 vs. 0.55, p = 0.0015), and all sites (0.45 vs 0.55, p = 0.0013) while MD increased only in lesion sites (1.03 × 10 −3 mm 2/s vs. 0.90 × 10 −3 mm2/s, p = 0.009). The FA demonstrated the best correlation with EDSS (r = − 0.7603, p = 0.0086), particularly at lesion sites. Conclusions: The results reinforce the importance of the FA index and conﬁrm the hypothesis that NMO is a diffuse disease. © 2014 Elsevier Inc. All rights reserved.
1. Introduction Neuromyelitis optica (NMO), or Devic’s disease, is a demyelinating, autoimmune disease that primarily affects the optic nerve and the spinal cord [1, 2]. This disease can be monophasic or relapsing–remitting, and whether NMO is a distinct disorder or part of a wide spectrum of multiple sclerosis (MS) has been the subject of debate for some time [3–7]. Based on the recent identiﬁcation of a speciﬁc antibody targeted to the blood– brain barrier aquaporin-4 (AQP-4) water channel in patients with NMO (NMO-IgG), this disease is now considered a central nervous system (CNS) autoimmune chanellopathy . Studies using magnetization transfer (MT) and DTI in the spinal cord of NMO patients have been performed. In the ﬁrst study, Filippi et al. demonstrated alterations in the MT histogram analysis of cervical ☆ Conﬂict of interest statement: Enedina M. L. Oliveira is a consulting neurologist for Bayer Schering Pharma of Brazil. The present study has been funded through an unrestricted grant from TEVA Pharmaceutical (Brazilian branch). ⁎ Corresponding author at: Departamento de Diagnóstico por Imagem/Universidade Federal de São Paulo (UNIFESP), Napoleão de Barros, 800, Vila Clementino, São Paulo, SP, Brasil 04024-002. Tel./fax: +55 11 5083 8963. E-mail address: [email protected]
(R.L.M. Rivero). 0730-725X/$ – see front matter © 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.mri.2014.01.023
spinal cords in NMO patients . In the second study, Benedetti et al. evaluated the spinal cord using DTI in NMO and MS patients and observed differences between the two groups with respect to the MD and FA values, suggesting a higher degree of spinal cord damage in patients with NMO than in those with MS . Recently, Qian et al. studied the spinal cord white matter in patients with NMO using a 3.0 T MRI scan and observed a signiﬁcant correlation between the FA and Expanded Disability Status Scale (EDSS) in these patients . The last two studies demonstrated a correlation between disability (EDSS) and DTI metrics, but both of these studies evaluated the entire cervical spinal cord. No studies have attempted to separate the areas of lesion (with signal change) from the preserved areas (without signal change) in the MRI T2-weighted images to observe differences in the FA or MD measurements in these two distinct regions and the importance of these areas on the degree of disability of patients with NMO. The purpose of the present study was to quantify the FA and MD values in the cervical spinal cords of NMO patients and separate the areas of the visible lesions from the apparently preserved areas of spinal cord tissue in MRI T2-weighted images, which had not been performed in previous studies. We also observed the changes in these values compared with the spinal cords
R.L.M. Rivero et al. / Magnetic Resonance Imaging 32 (2014) 457–463
of healthy individuals and identiﬁed a potential correlation with the degree of disability (EDSS) in these patients. 2. Materials and methods 2.1. Subjects A total of 11 NMO patients, diagnosed according to the Wingerchuk criteria  and followed-up at the Neuroimmunology Clinic of the Neurology Department, Universidade Federal de São Paulo, were randomly recruited in the present study. The study sample included three men and eight women between 21 and 78 years old (mean 40.2 years). All patients presented with spinal cord lesions extending over at least three vertebrate at the disease onset in the course of myelitis (observed in previous examinations of these individuals). Between 2005 and 2006, subjects in both groups in the present prospective study underwent MRI examinations using the DTI sequence. There was an interval of ﬁve years, on average, between the diagnosis and the MRI. However, there was no evidence of new episodes in this period as measured by EDSS scale in the same week of the scan. EDSS scale was also applied 3–4 years after the MRI scan. At the time of the present study, the laboratory tests to detect anti-AQP-4 antibodies were not available. The demographic and clinical data with average FA and MD values are shown in Table 1. A control group comprising two men and nine women between 24 and 58 years old (mean 40.5 years) with no underlying disease and no morphological or signal changes in the cervical spinal cord was voluntarily enrolled in the present study. All subjects provided written informed consent before the study, which was approved through the institutional committee for ethics in research at the UNIFESP-EPM. 2.2. Image acquisition The cervical spinal cords of all individuals in the patient and control groups were scanned using a 1.5-T Sonata scanner (Siemens, Erlangen, Germany), with the acquisition of the following sequences: a) T2-weighted turbo spin echo (TSE) (TR: 3400, TE: 97, turbo factor: 2, ﬁeld-of-view [FOV]: 260 × 260 mm, matrix size: 256 × 320; 12 sagittal slices; slice thickness: 3 mm) and b) Single-shot echo-planar sequence with parallel imaging (GRAPPA) to acquire diffusionweighted imaging in 12 non-collinear gradient directions with two b-values (b = 0 and 1000 s/mm2) (FOV: 230 × 230 mm; image matrix: 128 × 128; 20 sagittal slices; slice thickness: 3 mm; nominal voxel size 1.8 × 1.8 × 3.0 mm; TR: 2700; TE: 83; averages 3; EPI factor = 128). The diffusion tensor imaging acquisition time was 2 min and 3 s for each patient studied. The patients were asked to avoid moving the head, neck or limbs or swallowing during the study.
Table 1 Clinical details of the 11 patients with neuromyelitis optica. All patients mean (±SD) Age at the time of MRI (years) Disease duration (years) EDSS at ﬁrst appointment EDSS at time of MRI EDSS at last appointment Relapse Rate Progression index Average FA Average MD
40.2 (± 16.5) 10.9 (± 6.8) 3.7 (± 1.6) 4.3 (±2.6) 4.5 (± 2.7) 0.8 (± 0.5) 0.5 (± 0.4) 0.45 (±0.07) 0.99 (± 0.10)
Relapse rate: total number of relapses/disease duration; progression index: Expanded Disability Status Scale (EDSS) at the last appointment/disease duration; FA: fractioned anisotropy; MD: mean diffusivity.
2.3. Image analysis The data were processed using speciﬁc software (DTI Task Card software v1.70, Massachusetts General Hospital, Boston, MA, USA) for the quantitative analysis of diffusion tensor imaging data. This software was used for the creation of spinal cord FA and MD maps. Regions of interest (ROI) of the same size (0.3 ± 0.02 mm 2) were deﬁned using the means of the sagittal TSE T2-weighted images as a reference for determining the correct positioning and cervical level (Fig. 1). The partial-volume effect for cerebrospinal ﬂuid (CSF) was minimized on the sagittal images after identifying the image located nearest the midline when positioning the ROI on the cervical spinal cord. A single neuroradiologist (R.L.M.R), who was unaware of the patient data or the corresponding patient or control individuals, manually outlined and measured the ROIs. The MD and FA values were obtained from the spinal cord of every NMO patient and control group individual at the C2–C6 levels. The C7 level was excluded from the analysis because of the noise observed in some cases, which limited the analysis of FA and MD at that level. The intraobserver repeatability of the DTI measurements in the ROIs was also assessed. The ROIs were placed in a second session by the investigator (R.L.M.R) and again the FA and MD of each of the considered ROIs were measured in all subjects. Based on the evaluation of the conventional MRI studies, the cervical spinal cord was divided into 5 regions according to the levels of vertebral bodies (C2to C6). ROIs of the same size were positioned at each of these levels (5 ROIs), and these levels were classiﬁed as lesion sites (characterized by increased signal within the spinal cord in the T2-weighted images) or apparently preserved (normal appearing) sites of the spinal cord (with no change in the internal signal of the spinal cord on T2-weighted images). Subsequently, the mean FA and MD values of the lesion sites, the preserved sites and all sites in the NMO patients were calculated for each patient. These values were compared with the mean FA and MD values obtained in the normal spinal cord of each individual in the control group. The EDSS of all of the NMO patients at the time of their MRI examinations and the last EDSS measurement obtained in 2010 were correlated with the FA and MD values in the spinal cord. This analysis was conducted on the lesion sites, the preserved sites and along the spinal cord. 2.4. Statistical analysis 2.4.1. Reability of the DTI measurements The intra-rater reliability was tested using the intraclass correlation coefﬁcient (ICC) for the measurements performed on two different occasions. 2.4.2. Differences in the FA and MD values between the NMO and control groups Both the NMO and control groups presented a normal distribution using the D'Agostino–Pearson omnibus normality test (p = 0.509 and p = 0.434 for NMO and control group, respectively). An unpaired Student’s t test, with a signiﬁcance level of 5%, was used. 2.4.3. Relationship between the DTI measurements (FA and MD) and EDSS Spearman’s correlation coefﬁcient was used to evaluate the correlation degree (and the correlation tendency, either positive or negative) between two measurement scale variables (MD × EDSS and FA × EDSS), considering the fact that EDSS is not a continuous variable. All analyses were performed using GraphPad Prism version 5.00 (GraphPad Software Incorporated®, La Jolla, CA, USA), except for the intraclass correlation coefﬁcient, which was calculated using SPSS 14.1 (SPSS, Inc., Chicago, IL, USA), and the Bland–Altman analysis, which
R.L.M. Rivero et al. / Magnetic Resonance Imaging 32 (2014) 457–463
Fig. 1. A female 39-year-old NMO patient. At left, the T2-weighted image and the FA map at right. Hyperintense zones related to the lesion are seen on the T2-weighted image on levels C2, C3, C4 and C5, and ROIs are plotted on the FA map at the C2, C3, C4, C5, C6 and C7 levels.
was performed using Microsoft Excel 2007 (Microsoft Corporation, Redmond, WA, USA). 3. Results Hyperintense lesions were detected in the cervical spinal cords of all NMO patients (Fig. 1). Because the individuals included in the present study were chronic patients, the lesions were not always visible beyond the three vertebral bodies; however, upon the onset of myelitis, these patients presented with more extensive lesions that met the Wingerchuk criteria. An ROI was placed on each level of the cervical cord, starting on C2 until C6 level, yielding 5 ROIs for each patient. Then each level was classiﬁed as lesion site or apparently preserved area based on T2 signal. From the 55 ROIs measured on 11 patients, 33 of them were classiﬁed as lesion areas and 22 were apparently preserved. The intraclass correlation coefﬁcient (ICC) demonstrated moderate to excellent consistencies (range 0.70–0.94) between these two measurements on different occasions. The difference between the cut-off points (moderate correlation= N 0.40 and b 0.75 and excellent correlation N 0.75) of FA and MD in these two groups potentially reﬂects the low variability between these measurements. But the Bland–Altman analysis demonstrated the adequate repeatability and
Table 2 Intraclass Correlation Coefﬁcient (ICC) of two FA and MD measurements performed on different occasions in NMO and control group.
FA control group FA NMO group MD control group MD NMO group
Conﬁdence interval 95%
0.704 0.892 0.944 0.752
(0.541; (0.822; (0.905; (0.610;
FA: fractioned anisotropy; MD: mean diffusivity.
0.816) 0.936) 0.967) 0.848)
reliability of the measurements. Bland–Altman plotting charts showed good agreement between two FA measurements in control group (bias of −0.008, whereas the 95% limits of agreement were from −0.158 to 0. 0.142) and patients with NMO (bias of −0.005, whereas the 95% limits of agreement were from −0.092 to 0.081) and good agreement between 2 MD measurements in control group (bias of −0.009, whereas the 95% limits of agreement were from −0.109, to 0.090) and patients with NMO (bias of −0.006; whereas the 95% limits of agreement were from −0.089 to 0.078). The 95% limits of agreement overlap 0 in all measurements. (Table 2 and Fig. 2). The analysis of the mean FA per patient with respect to spinal cord areas with and without signal changes observed in MRI revealed statistically signiﬁcant differences in the FA values between the patients and individuals in the control group with respect to lesion sites (0.44 ± 0.09 vs. 0.55 ± 0.05, p = 0.0046), apparently preserved sites (0.46 ± 0.06 vs. 0.55 ± 0.05, p = 0.0015) and measured areas along the spinal cord (0.45 ± 0.07 vs. 0.55 ± 0.05, p = 0.0013). (Fig. 3). Regarding the MD values, the analysis of the mean of the spinal cord areas per individual revealed statistically signiﬁcant differences between the values for NMO patients and those for the individuals in the control group at the lesion sites (1.03 ± 0.09 × 10−3 mm2/s vs. 0.90 ± 0.11 × 10−3 mm2/s, p = 0.009) and along the spinal cord (0.99 ± 0.10 × mm2/s vs. 0.90 ± 0.11 × 10−3 mm2/s, p = 0.046), with no statistically signiﬁcant difference observed at the apparently preserved sites (0.94 ± 0.10 × 10−3 mm2/s vs. 0.90 ± 0.11 × 10−3 mm2/s, = 0,426) between NMO patients and the individuals in the control group (Fig. 4). There was a negative correlation between the FA of the lesion areas and EDSS at the time of the DTI acquisition in NMO patients (r = − 0.7603, p = 0.0086) but not at the time of the last evaluation (r = − 0.485; p = 0.156). In addition, a signiﬁcant correlation was observed between the FA in all spinal cord areas and the EDSS score obtained at the time of the DTI study (r = − 0.7867, p = 0.0055). No correlation was observed between the FA in all spinal cord areas
R.L.M. Rivero et al. / Magnetic Resonance Imaging 32 (2014) 457–463
Fig. 2. Bland–Altman plotting charts showing good agreement between two FA measurements in control group (A: Difference: −0.008; LOA: (− 0.158; 0.142)) and patients with NMO (B: Difference: −0.005; LOA: (−0.092, 0.081)) and MD in control group (C: Difference: −0.009; LOA:(−0.109, 0.090)) and patients with NMO (D: Difference: −0.006; LOA: (−0.089, 0.078)). LOA = Limits of Agreement.
and EDSS obtained from the last patient evaluation (r = − 0.4628, p = 0.1546) (Fig. 5). There was also no correlation in the FA of apparently preserved sites and EDSS obtained at the time of DTI study (r = − 0.2311, p =
0.4854) and from the last NMO patient evaluation (r = − 0.1078, p = 0.7545). Moreover, no correlation was observed between the MD values at the lesion sites, apparently preserved sites or along all spinal cord areas, both in the EDSS score before the DTI study (r =
Fig. 3. Box plot charts showing (A) the mean FA value in lesion sites per individual in the NMO group and the mean FA value at the spinal cord in control group, with a statistically signiﬁcant difference between the two groups(0.44 ± 0.09 vs. 0.55 ± 0.05, p = 0.0046); (B) mean FA value in apparently preserved areas per individual in the NMO group and mean FA value at the spinal cord of control group, with a statistically signiﬁcant difference between the two groups (0.46 ± 0.06 vs. 0.55 ± 0.05, p = 0.0015); (C) mean FA value in all measured areas of the spinal cord for each individual in the NMO group and mean FA value in the spinal cord of control group, with a statistically signiﬁcant difference between the two groups (0.46 ± 0.06 vs. 0.55 ± 0.05, p = 0.0015).
R.L.M. Rivero et al. / Magnetic Resonance Imaging 32 (2014) 457–463
Fig. 4. Box plot chart showing (A) mean MD values in the lesion area per NMO patient and the mean MD value in the spinal cord of control group, with a statistically signiﬁcant difference between groups(1.03 ± 0.09 × 10−3 mm2/s vs. 0.90 ± 0.11 × 10−3 mm2/s, p = 0.009); (B) the mean MD value in apparently preserved areas per NMO patient and the mean MD value in control group, with no statistically signiﬁcant difference between groups; (C) the mean MD value at all levels of the spinal cord per NMO patient and the mean MD value in the spinal cord of control group, with a statistically signiﬁcant difference between groups (0.99 ± 0.10 × mm2/s vs. 0.90 ± 0.11 × 10−3 mm2/s, p = 0.046).
0.3982, p = 0.2250; r = − 0.1689, p = 0.6147; and r = 0.4795, p = 0.1375, respectively) and in the EDSS obtained from the last NMO patient evaluation (r = 0.4497, p = 0.1635; r = −0.1683, p = 0.6147; r = 0,4090, p = 0.2141, respectively).
4. Discussion In the present study, using a different approach from that of previous studies for the evaluation of the DTI in spinal cord, we separated the
Fig. 5. Above, plotting of FA values in lesion areas of the spinal cord in correlation with EDSS score at the time of MRI examination (at left) and at the last EDSS score in NMO patients after the MRI examination (at right). The Spearman's correlation coefﬁcient indicates a strong negative correlation between FA values and EDSS scores, both at the time of the MRI examination (r = −0.7603, p = 0.0086) but not in the last evaluation (r = −0.485; p = 0,156). Below, plotting of FA values of the whole spinal cord in correlation with EDSS score at the time of MRI examination (at left) and at the last EDSS score of the patient after the MRI examination (at right). The Spearman's correlation coefﬁcient indicates a strong negative correlation between FA values and the EDSS scores at the time of the MRI examination (r = −0.7867, p = 0.0055) but no signiﬁcant correlation in the last EDSS evaluation (r = − 0.5016, p = 0.1440).
R.L.M. Rivero et al. / Magnetic Resonance Imaging 32 (2014) 457–463
sites of lesion from those apparently preserved (normal appearance) sites in the cervical spinal cord to quantify the degree of damage in these two distinct regions using DTI. The results showed that areas of signal change in the T2-weighted images in MRI were the most involved in the disease. In these areas, changes in both the FA and MD were observed, while in the apparently preserved areas, MD stability and persistent changes in the FA were detected. In addition, these results showed that the FA was well correlated with the EDSS of patients, consistent with the results of a previous study . In addition, we observed a stronger correlation of this DTI metric with the disability degree at the lesion sites. Benedetti et al. previously conducted a study using DTI on the spinal cord of NMO patients, who demonstrated alterations of MD and FA values in a different pattern from those observed in multiple sclerosis patients . However, these authors did not observe a correlation between EDSS and the FA index. In another study, Qian et al. performed DTI measurements only in the white matter of the spinal cord, but there is evidence that this approach might be limited for several reasons:  The spinal cord gray matter is anisotropic, although to a lesser degree than the spinal cord white matter [12, 13];  In the case of NMO, there is a preferential involvement of the gray matter in both acute and chronic lesions of the disease [14,16, 17] and  Golabchi et al. compared measurements of diffusion tensor with axonal clustering parameters in histological analyses of the spinal cord and observed that the correlation coefﬁcients were signiﬁcantly higher in cases in which the spinal cord white matter and gray matter were jointly analyzed than in cases in which only the white matter was analyzed. This difference potentially reﬂects the fact that the spinal cord gray matter does not predominantly comprise neurons but also contains a high number of extensive axons, which play a role in many functional connections within the spinal cord . In the present study, the image acquisition was performed in the sagittal plane focused at the most median section of the spinal cord to avoid the partial volume effect of the cerebrospinal ﬂuid. We observed a signiﬁcant decrease in the FA in NMO patients compared with the control group at both lesion sites and apparently preserved sites of the spinal cord. The MD, however, only increased in lesion areas, with no signiﬁcant differences between the NMO patients and the individuals in the control group in areas where the spinal cord signal was preserved. Benedetti et al. , Valsasina et al.  and Agosta et al.  suggested dissociation between the MD and FA values in MS patients, demonstrating a pseudonormalization of the MD and a persistent change of the FA. These results were interpreted as a consequence of the secondary glial proliferation, as glial cells do not have the same anisotropic morphology of the original tissue and, therefore, the FA does not recover. Benedetti et al. did not observe dissociation in NMO patients; however, these authors evaluated the entire cervical spinal cord. We observed that such dissociation most likely occurs in lesion sites where the tissue damage is more extensive. Dissociation is observed at the apparently preserved areas, most likely because the MD values present a pseudonormalization similar to that observed in MS patients due to the reduced degree of tissue destruction in these areas [10, 18, 19]. Furthermore, we observed a strong negative correlation between the FA and EDSS at spinal cord lesion sites; this correlation was not observed between the FA at the apparently preserved sites of the spinal cord or between the MD at any of these sites in the spinal cord. In contrast to Benedetti’s ﬁndings , the results obtained in the present study and the study of Qian et al.  showed a correlation between EDSS and FA, but not between EDSS and MD. This discrepancy could reﬂect the fact that the MD and FA values in both MS and NMO patients were compared with the EDSS in that study. Such an extrapolation could eventually induce a different result, as the pattern of their lesions and the behavior of FA and MD indices are different in these two pathologies.
The ﬁnding that FA at spinal cord lesion sites is negatively correlated with the EDSS suggests that this index might be associated with the clinical status of NMO patients. In the present study, 80% of the patients with FA ≤ 0.40 (n = 5) presented the worst EDSS scores. The small number of patients represents a limiting factor in the present study, which is a consequence of the low incidence of this disease. Therefore, larger studies will be required in the future to conﬁrm the results reported here and to explore the correlations between DTI measures and EDSS prospectively. In addition, we did not examine parallel or perpendicular diffusivities due to limitations in our post-processing software. It would be interesting to determine the pattern of change in these two DTI metrics at both lesion sites and apparently preserved sites in the spinal cord in future studies. However, the results obtained in the present study indicate a relevant ﬁnding that will promote further practical applications of DTI. In conclusion, using DTI we observed that areas showing a signal change in T2-weighted images in MRI provided evidence of increased tissue damage, with changes in both the FA and MD at these sites. However, in areas without a signal change, a lesser extent of damage was observed in the DTI analysis, with MD preservation in these regions. The FA was best correlated with the degree of patient disability. Moreover, the MRI revealed that the best correlation of FA with EDSS was observed at lesion sites in the spinal cord, and this correlation persisted over a long period of time in these regions. These ﬁndings also support the hypothesis that NMO is a diffuse disease in spinal cord. Acknowledgments This study was supported by a grant from TEVA Pharmaceutical (Brazilian branch). The authors acknowledge scientiﬁc support from the Instituto de Pesquisa e Ensino em Medicina Diagnóstica e Terapêutica (IPmed). References  Devic M. Myelite subaigue compliqueé de nevrite optique. Bull Med 1895;8:1033–4. Citado em Miyazawa I, Fujihara K, Itoyama Y, Devic Eugene. (858–930) J Neurol 1895;249(3):351–2.  Mandler RN, Davis LE, Jeffery DR, Kornfeld M. Devic's neuromyelitis optica: a clinicopathological study of 8 patients. Ann Neurol Aug 1993;34(2):162–8.  O'Riordan JI, Gallagher HL, Thompson AJ, Howard RS, Kingsley DP, Thompson EJ, et al. Clinical, CSF, and MRI ﬁndings in Devic's neuromyelitis optica. J Neurol Neurosurg Psychiatry Apr 1996;60(4):382–7.  Fazekas F, Offenbacher H, Schmidt R, Strasser-Fuchs S. MRI of neuromyelitis optica: evidence for a distinct entity. J Neurol Neurosurg Psychiatry Sep 1994;57(9):1140–2.  Weinshenker BG. Neuromyelitis optica: what it is and what it might be. Lancet Mar 15 2003;361(9361):889–90.  Wingerchuk DM, Hogancamp WF, O'Brien PC, Weinshenker BG. The clinical course of neuromyelitis optica (Devic's syndrome). Neurology Sep 22 1999;53(5):1107–14.  de Seze J, Stojkovic T, Ferriby D, Gauvrit JY, Montagne C, Mounier-Vehier F, et al. Devic's neuromyelitis optica: clinical, laboratory, MRI and outcome proﬁle. J Neurol Sci May 15 2002;197(1–2):57–61.  Wingerchuk DM, Lennon VA, Pittock SJ, Lucchinetti CF, Weinshenker BG. Revised diagnostic criteria for neuromyelitis optica. Neurology May 23 2006;66(10):1485–9.  Filippi M, Rocca MA, Moiola L, Martinelli V, Ghezzi A, Capra R, et al. MRI and magnetization transfer imaging changes in the brain and cervical cord of patients with Devic's neuromyelitis optica. Neurology Nov 10 1999;53(8):1705–10.  Benedetti B, Valsasina P, Judica E, Martinelli V, Ghezzi A, Capra R, et al. Grading cervical cord damage in neuromyelitis optica and MS by diffusion tensor MRI. Neurology Jul 11 2006;67(1):161–3.  Qian W, Chan Q, Mak H, Zhang Z, Anthony MP, Yau KK, et al. Quantitative assessment of the cervical spinal cord damage in neuromyelitis optica using diffusion tensor imaging at 3 Tesla. J Magn Reson Imaging Jun 2011;33(6):1312–20.  Fenyes DA, Narayana PA. In vivo diffusion tensor imaging of rat spinal cord with echo planar imaging. Magn Reson Med Aug 1999;42(2):300–6.  Holder CA, Muthupillai R, Mukundan Jr S, Eastwood JD, Hudgins PA. Diffusionweighted MR imaging of the normal human spinal cord in vivo. AJNR Am J Neuroradiol Nov-Dec 2000;21(10):1799–806.  Nakamura M, Miyazawa I, Fujihara K, Nakashima I, Misu T, Watanabe S, et al. Preferential spinal central gray matter involvement in neuromyelitis optica. An MRI study. J Neurol Feb 2008;255(2):163–70.  Golabchi F, Hoge W, Mamata H, Maier S, Brooks D. Techniques for automatic spinal cord histology characterization for validation of diffusion tensor imaging. Conf Proc IEEE Eng Med Biol Soc 2004;3:1640–3.
R.L.M. Rivero et al. / Magnetic Resonance Imaging 32 (2014) 457–463  Lucchinetti CF, Mandler RN, McGavern D, Bruck W, Gleich G, Ransohoff RM, et al. A role for humoral mechanisms in the pathogenesis of Devic's neuromyelitis optica. Brain Jul 2002;125(Pt 7):1450–61.  Krampla W, Aboul-Enein F, Jecel J, Lang W, Fertl E, Hruby W, et al. Spinal cord lesions in patients with neuromyelitis optica: a retrospective long-term MRI follow-up study. Eur Radiol Oct 2009;19(10):2535–43.
 Valsasina P, Rocca MA, Agosta F, Benedetti B, Horsﬁeld MA, Gallo A, et al. Mean diffusivity and fractional anisotropy histogram analysis of the cervical cord in MS patients. Neuroimage Jul 1 2005;26(3):822–8.  Agosta F, Benedetti B, Rocca MA, Valsasina P, Rovaris M, Comi G, et al. Quantiﬁcation of cervical cord pathology in primary progressive MS using diffusion tensor MRI. Neurology Feb 22 2005;64(4):631–5.