Ultrahigh-Field MR (7 T) Imaging of Brain Lesions in Neuromyelitis Optica

Background. Brain lesions are common in neuromyelitis optica spectrum disorder (NMOsd) and may resemble lesions of multiple sclerosis (MS). Objectives. To describe the imaging characteristics of supratentorial lesions in NMOsd on ultrahigh-field (7 T) MRI with special attention to vessel-lesion relationship. Methods. Ten NMOsd patients, all women and all seropositive for NMO IgG, with mean age of 51.3 ± 15.4 years and disease duration of 9.2 ± 6.4 years, were scanned at a 7 T whole-body human MR system with high-resolution 2D gradient echo sequence optimized to best visualize lesions and venous structures, T2- and T1-weighted imaging. Results. In 10 patients with NMOsd, a total of 92 lesions were observed (mean: 9.2 ± 8.8; range: 2–30), but only 8 lesions (9%) were traversed by a central venule. All lesions were <5 mm in diameter, and 83% were located in subcortical white matter. There were no lesions in the cortex or basal ganglia. Two patients exhibited diffuse periependymal abnormalities on FLAIR. Conclusions. Small, subcortical lesions without a central venule are the most consistent finding of NMOsd on 7 T MRI of the brain. Ultrahigh-field imaging may be useful for differentiating between NMOsd and MS.


Introduction
Neuromyelitis optica (NMO) has been traditionally considered a predominantly opticospinal disorder, but recent studies have shown that clinical spectrum of NMO includes cerebral and brainstem syndromes as well [1][2][3]. Brain lesions on magnetic resonance imaging (MRI) are found in the majority of NMO patients with long-standing disease [1,2,[4][5][6], which makes it sometimes difficult to differentiate this disorder from multiple sclerosis (MS).
Ultrahigh-�eld MRI (4 Tesla or above) affords an unprecedented view of brain structures and pathology in vivo on a submillimeter scale due to high signal-to-noise ratio [7,8]. e largely enhanced susceptibility of ultrahigh-�eld MR effect provides unique contrast between lesions and veins. is feature makes ultrahigh-�eld MR a valuable tool for differentiating plaques of MS, which frequently contain a venule, from lesions of other etiologies that rarely do [9].

Methods
Inclusion criteria in the study were the diagnosis of NMOspectrum disorders (NMOsd) de�ned either as NMO by 2006 Wingerchuk et al. criteria [20], or recurrent longitudinally extensive transverse myelitis (LETM) with NMO IgG seropositivity, or recurrent optic neuritis (ON) with NMO IgG seropositivity. NMO IgG was measured at the Mayo Clinic Laboratory (Rochester, MN) by either immunohistochemical or enzyme-linked immunosorbent assay [21]. All nonbedbound NMOsd patients followed by the neurologists of the NYU-MS Center (I. K. and J. H.) were invited to participate in the study. For each patient, we collected the following information: gender, ethnicity, country of birth, age at symptom onset, disease duration at MRI, index event, number of optic neuritis and of transverse myelitis relapses, visual acuity and ambulatory status at last followup, NMOsd therapy, and presence and number of vascular risk factors (de�ned as history of smoking, hypertension, hypercholesterolemia, or cardiovascular disease).
Infratentorial volume could not be reliably evaluated because our 7 T receive coil yielded poor signal-to-noise ratio in this region, and the examination was therefore limited to supratentorial region. Two experienced observers (Y. G. and I. K.) examined T2-weighted and 2D gradient echo (T2 * ) imaging sequences for presence of lesions. To reduce the possibility of misidenti�cation, only the lesions that were present on both sequences were counted. For each lesion, we recorded lesion size, con�guration, location, and presence or absence of central venule and of hypointense peripheral ring on T2 * sequence. Lesion localization was de�ned as periventricular (<2 mm of ventricular border), juxtacortical (<2 mm from cortical ribbon), subcortical/deep white matter (>2 mm from cortex or ventricle, excluding corpus callosum), corpus callosum, cortex, and deep grey matter nuclei. Lesions were described as "perivenous" if they were seen to be traversed by a vein on T2 * sequence [9]. We also recorded the pattern of periependymal abnormalities.
Groups were compared using Mann-Witney test for continuous variables.
Data collection was carried out according to protocol approved by the Institutional Review Boards of NYU-Langone Medical Center, and all patients gave informed consent in writing before participation.

Results
Ten patients agreed to participate in the study, six of whom were diagnosed with NMO [20] and four with limited forms of NMO. All patients were women and NMO IgGseropositive. Demographic, disease-related, and radiologic characteristics of our NMOsd cohort are shown in Table 1.
We observed lesions in every patient. e total number of lesions was 92 (mean number was 9.2 ± 8.8; median: 6; range: 2-30). Seventy-six lesions (83%) were subcortical, 8 (9%) were juxtacortical, 3 lesions (3%) were found in corpus callosum, and 5 (5%) were periventricular (4 of which were adjacent to frontal horns of the lateral ventricles). We did not detect any lesions in the cortex or basal ganglia. All but three of the 92 lesions were <3 mm in cross-sectional diameter, and all lesions were <5 mm. e lesions generally appeared round on axial cross-sections (Figures 1(a)-1(f)). None evidenced hypointense ring on T2 * sequence. We also observed several "thread-like" lesions, and these were invariably hypointense on T1-weighted sequences (Figures 2(a)-2(d)). Several lesions, which were <3 mm in crosssectional diameter, could be seen on three consecutive slices, indicating that they were at least 6 mm long. In all, eight lesions (9%) were classi�ed as "thread-like. " Two patients had diffuse periependymal abnormalities on T2-weighted sequences (Figures 3(a) and 3(b)). We did not observe any tumefactive or longitudinally extensive lesions in our patients.
Only eight lesions (9%) were traversed by a venule. Figure  1(d) shows an example of such lesion. ree perivenous lesions were seen in one patient with NMO who had a total of four lesions, and �ve patients had one perivenous lesion each.
e group with limited form of NMOsd, comprised of 4 patients, tended to have more lesions (mean: 13.4), compared to the six-patient NMO group (mean: 4.8), but the difference was not statistically signi�cant (Mann-Whitney test, = 16.5 -two tailed = 0.35). We did not observe a trend for increasing lesion number with age, disease duration, or the number of vascular risk factors (Table 1).

Discussion
T2-weighted hyperintense lesions on 7 T MRI were observed in all ten NMOsd patients in our cohort. All but three of the 92 lesions were <3 mm in largest diameter. Nearly all lesions were round (Figure 1), but several lesions had a distinct, "thread-like" appearance on axial cross-section ( Figure 2). We consider it unlikely that "thread-like" lesions represent dilated Virchow-Robin spaces (VRS) because of their location, orientation, and signal characteristics. VRS are typically found in basal ganglia with penetrating arteries mostly from circle of Willis. ey appear as round dots on axial plane with similar signal as CSF (i.e., brightly hyperintense spots on T2-weighted sequence) [22]. In contrast, "thread-like" lesions found in our NMO patients were observed mostly in juxtacortical white matter and corpus callosum, do not follow arterial course, and were not as bright as CSF on T2weighted sequence. read-like lesion appearance has not, to   Native American/Irish/African-American.
All lesions were <3 mm in diameter, except for 3 lesions in patient number 9 (see Table 1), which were ∼5  our knowledge, been reported in other conditions, including MS, which is characterized by perivenous ovoid lesions [23]. With respect to lesion distribution, the overwhelming majority of lesions in NMOsd (∼85%) were subcortical. e few discrete periventricular lesions were mostly located anterior to the frontal horns (Figure 1(d)) and were more extensive than the typical "periventricular caps" seen in healthy subjects [24]. Two of our patients exhibited diffuse periventricular abnormalities (Figure 3), which is notable in view of known rich expression of Aquaporin-4, the target of NMO IgG, in the ependymal layer [25]. In one patient ( Figure  3(a)), splenium of the corpus callosum was affected in the "diffuse spreading" manner described as being characteristic of NMO in a Japanese cohort [26]. e discrepancy in the prevalences of diffuse spreading lesions in corpus callosum in our series (10%) and the Japanese series (71%) could perhaps be due to lesser degree of disability of our cohort, in which only 30% of patients required any kind of ambulatory assistance compared to the Japanese cohort, in which half of the patients needed ambulatory assistance. It is also possible that genetic background impacts phenotypic and radiographic expression of NMO. A patient from the Far East appears to manifest brain symptoms and large "NMO-speci�c" brain lesions more commonly than in the �est [1,2,5]. Cortical lesions were not found in any of our patients in agreement with the recent radiographic studies that utilized double inversion recovery sequence [27] and ultrahigh-�eld MR [28] to image the cortex in NMO as well as a histopathologic study [29]. Absence of focal cortical lesions in NMOsd need not imply absence of other cortical pathology [30], but is helpful for differentiating NMO from MS, in which cortical lesions are increasingly recognized, especially with ultrahigh-�eld MR [11][12][13][14].
Importantly, only 9% of NMOsd brain lesions in our series contained a central vessel. is is in stark contrast with MS, in which 60-80% of cerebral lesions visualized with ultrahigh-�eld MRI are traversed by a venule [9,10,15,16]. Paucity of intralesional venules in our series is not contradictory to results of a study in which venules were detected in a high proportion of white matter lesions of any etiology on susceptibility weighted imaging [31]. e latter study focused exclusively on lesions >3 mm in diameter, while nearly all lesions in our series were ≤3 mm in diameter.
Absent histopathological correlation, it is impossible to be certain that brain lesions seen in our cohort are due to an NMO-speci�c pathogenic process. Some of the lesions could theoretically represent stigmata of "small vessel disease" (SVD), in itself a pathologically heterogeneous entity [32]. For a number of reasons, SVD origin of observed lesions seems to us to be less likely explanation. First, the �nding of juxtacortical and corpus callosum lesions and the absence of basal ganglia lesions militate against SVD etiology. Secondly, "thread-like lesions" (Figures 2(a)-2(d)) have not been described in SVD. ird, we could �nd no correlation between vascular risk factors and the number of brain lesions (Table 1) and observed small lesions in patients in their 20s and 30s with no vascular risk factors. Lastly, in at least three of our patients, a lesion on 7 T MRI was seen to enhance with gadolinium previously on a conventional MRI study (unpublished data). is observation supports in�ammatory etiology for at least some of the lesions herein described. Whatever the pathogenesis of small subcortical lesions NMO may be, these "non-speci�c" lesions represent the most consistent feature of NMO on brain MRI [5].
Our data is broadly consistent with recently published �ndings of 7 T MRI of the brain in 10 German NMOsd patients [28]. e German NMOsd cohort was somewhat older than ours (46 years versus 42 in our group) but had a slightly shorter disease duration (7 years versus 9 years). Both groups were exclusively female, and all but one patient were NMO IgG seropositive. In neither ours nor the German NMOsd cohort were any cortical lesions observed, and the lesions that were present were predominantly small and subcortical. e proportion of perivenous lesions among the German NMOsd patients was higher than in our group (35% versus 9%), which may be related to the higher average number of lesions in their cohort (14 lesions versus 9 lesions in our group). Higher lesion number increases the probability of chance �nding of a traversing vein. Importantly, in both cohorts, the number of perivenous lesions was signi�cantly smaller than what is typically reported in MS (>80%) [8]. Finally, although hypointense ring, which is indicative of iron or hemosiderin deposition, is found in a signi�cant minority of MS lesions [17,28], it was only observed in one out of ten NMOsd patients in the German NMO cohort and none of our ten patients. is further supports differential brain lesion pathogenesis in MS and NMO.
Our work and that of Sinnecker et al. [28] demonstrate that small round subcortical lesion without a central venule or hypointense ring is the most characteristic feature on NMOsd on ultrahigh-�eld MRI. In contrast, the signature lesion of MS is oblong, periventricular, and traversed centrally by a venule [9,10,15,16], corresponding to "Dawson's �ngers", described pathologically almost a hundred years ago [33]. We suggest that the absence of "Dawson's �ngers" and cortical lesions on 7 T MRI in a patient with multiple, >9, T2 hyperintense lesions should alert the radiologist that diagnosis of MS is unlikely.
Ultrahigh-�eld imaging of brain in NMOsd provides strong supportive evidence that MS and NMO constitute two distinct nosological entities despite their predilection for the same organs. Ultrahigh-�eld brain MRI may prove useful for differentiating NMOsd from MS.