J Neurol (2004) 251 : 818–824 DOI 10.1007/s00415-004-0434-z Wilhelm Küker Michael Weller Uwe Klose Hilmar Krapf Johannes Dichgans Thomas Nägele ■ Abstract Infarction is a rare cause of spinal cord dysfunction. Whereas diffusion-weighted (DW) MRI has been established as a highly sensitive technique for assessing acute cerebral ischemia, its role in spinal cord infarction re- Received: 12 September 2003 Received in revised form: 21 January 2004 Accepted: 9 February 2004 Dr. W. Küker (쾷) · U. Klose · H. Krapf · T. Nägele Department of Neuroradiology University Hospital Tübingen Hoppe-Seyler-Str. 3 72076 Tübingen, Germany Fax: +49-7071/295392 E-Mail: wmkueker@med.uni-tuebingen.de M. Weller · J. Dichgans Department of Neurology University of Tübingen Medical School Tübingen, Germany ORIGINAL COMMUNICATION Diffusion-weighted MRI of spinal cord infarction High resolution imaging and time course of diffusion abnormality mains to be determined. The purpose of this study is to present the signal characteristics of acute spinal cord ischemia using DWMRI within the first two days and after one week. MRI including DW imaging (DWI) was performed in three patients with acute spinal cord dysfunction 8, 12 and 30 hours after the onset of symptoms and repeated after one week in two patients. Two initial scans included EPI DW sequences in transverse and sagittal orientation. The remaining examinations were performed with an optimised highspatial resolution DWI sequence in the transverse plane. The diagnosis of spinal cord ischemia was established by imaging, clinical history and CSF analysis. T2 signal abnormality and restricted diffusion was demonstrated in all initial examinations. Transverse DW sequences JON 1434 Introduction Spinal cord infarctions are rare compared with brain ischemia, but may cause severe neurological symptoms. Diagnosis mainly rests on history and clinical symptoms. Acute, often painful, rapidly progressive sensorimotor deficits without prior history or clinical signs of an inflammatory disorder are encountered in the initial phase of spinal cord infarction. The symptoms depend on the affected level. Most infarctions occur in the distal part of the cord and the ensuing deficits are flaccid paraparesis of the legs and loss of sphincter control. Because had the highest sensitivity. The spinal infarctions were mainly located in the centre of the spinal cord and the grey matter. Contrast enhancement was absent. After one week, the restricted diffusion had normalised (pseudo normalisation) whereas the T2 signal changes had become more prominent. Restricted diffusion in the course of spinal cord ischemic infarction can be demonstrated using DW-MRI. Whereas a diffusion abnormality can be found after few hours, it does not last for longer than one week. At this time, the establishment of the diagnosis has to rely mainly on T2-weighted images with additional post contrast T1weighted images being useful. ■ Key words spinal infarction · DWI · MRI · spinal cord the dorsal part of the spinal cord is rarely involved, sensory symptoms most frequently affect the protopathic qualities with a relative sparing of epicritic sensation. The symptoms of cervical spinal cord infarction are dominated by an irreversible flaccid paresis in the affected spinal segments, due to destruction of the motor neurons, and an initially flaccid, later spastic, paraparesis due to the damage to the corticospinal tract. Protopathic sensory symptoms are also more pronounced than the epicritic. The clinical presentation is, however, variable and often unspecific in the initial phase. Relevant differential diagnoses include inflammatory and demyelinating le- 819 sions, spinal cord compression due to tumour, spondylosis, disk herniation or haematoma as well as venous congestion due to pathological arterio-venous shunts [13]. Diagnosis is achieved by CSF analysis and by spinal MRI, which is the essential imaging tool with which to visualise the medullary cord. Before the advent of diffusion weighted imaging, the diagnosis of acute spinal cord infarction rested solely on the exclusion of other forms of pathology. Although no effective treatment for spinal cord infarction has been established, the exclusion of other spinal cord lesions is mandatory and best achieved by unequivocal identification of spinal cord ischaemia. Most frequently, the territory of the anterior spinal artery is affected. There have been few reports of lesions of the postero-lateral spinal cord arteries [7].Spinal cord infarction has been attributed to vertebral artery dissection, fibrocartilaginous embolization, vasculitis, aortic aneurysms and surgery or coagulatory disorders [13]. However, the aetiology remains obscure in most patients. The MRI differential diagnosis of spinal cord lesions is beset with serious difficulties. Because of the small diameter of the spinal cord, lesions are usually small and may lack a discernible pattern. Topographic disease patterns, which are fairly characteristic in the brain, e. g., of multiple sclerosis (MS) or herpes encephalitis, have been identified in the spinal cord for some forms of pathology such as MS [6, 9], vitamin B12 deficiency [14] or some autoimmunological disorders. However, in daily practice they are not as reliable as in the brain. The most powerful tool for the detection of cytotoxic oedema in neural tissue, heralding infarction, is DW MRI [11], which visualises even small areas of ischaemia [4]. However, the application of DW MRI in spinal cord lesions has been very limited for technical reasons [1, 2, 8].A new generation of MR scanners featuring improved coil and gradient designs for spinal cord imaging has provided a new opportunity to implement this imaging method for spinal cord ischaemia. We here report clinical and imaging findings, including temporal changes of the diffusion abnormality, in three patients with ischaemic infarction of the spinal cord. Case reports ■ Clinical findings Patient #1 was a 41-year-old woman who noted pain,sensory loss and weakness of both legs progressing to inability to stand and walk within one day. The neurological history was otherwise unremarkable. Neurological examination on admission showed a flaccid, distal para- paresis with decreased tendon reflexes, urinary retention, and no pyramidal signs. CSF was normal, and MRI was suggestive of spinal ischaemia affecting the sacral cord.There was no evidence of aneurysm or dissection of the aorta, and overall no cause for spinal cord ischaemia was identified. The patient gradually improved with no further episodes of neurological deterioration. Patient #2 was a 67-year-old woman who developed painful sensorimotor deficits, left more than right, with a level at L4, and urinary retention within 3 hours. The clinical history did not indicate a relevant prior disorder. Examination on admission revealed flaccid paresis, distal more than proximal, left more than right, with decreased patellar tendon reflexes and lost ankle reflexes. There was sensory loss in both legs and in the perianal area. CSF contained a mildly elevated albumin level and MRI suggested ischaemia of the sacral cord. As in patient #1, no cause for spinal cord ischaemia was identified,and the patient gradually improved over weeks with no further episodes of neurological dysfunction. Patient #3 was a 62-year-old woman who had plasmacytoma diagnosed in 1999 and underwent allogenic bone marrow transplantation in 2002. The patient developed sudden paraplegia with a level at D9, and MRI suggested ischaemia of the thoracic cord. CSF analysis was not done because of thrombocytopenia. Within few days, the patient showed rapid neurological deterioration with impairment of consciousness attributable to multiple cerebral ischaemic lesions. In the presence of multiple necrotic skin lesions,the infarctions most likely resulted from vasculitis in the context of graft versus host disease. Imaging ■ Imaging technique MRI was performed at 1.5T (Siemens Magnetom Sonata, Siemens Medical Systems, Erlangen, Germany) using a standardised protocol on a phased-array spinal coil. All examinations encompassed T2-weighted turbospinecho images in sagittal and transverse orientation as well as T1-weighted images in the sagittal plane. Paramagnetic contrast agent was injected during all initial scans (Gd-DTPA, Magnevist®) at 0.1 mmol/kg body weight. DW-MRI was performed with an isotropic single shot EPI sequence. Images were acquired with b values of 0, 500 and 1000 s/mm2 and diffusion sensitivity in three orthogonal directions, from which ADC maps were calculated. The initial examinations of patients #1 and #2 included DW sequences in transverse and sagittal orientation (repetition time (TR) 160 ms, echo time (TE) 87 ms, slice thickness (SL) 5 mm, number of acquisitions (AC) 820 3, field of view (FoV) 230 mm, matrix 128, b-value = 0, 500, 1000 s/mm2). The follow-up scans and the sole examination of the third patient were carried out using a transverse DW sequence which had been optimised for spinal diffusion imaging on a healthy volunteer (TR 220, TE 130, SL 5,AC 30, FoV 152, matrix 128, b 1000 s/mm2). Improved spatial resolution was achieved by a reduced field of view. The signal-to-noise ratio was amplified by an increased number of acquisitions (30 instead of 3). In all patients, diffusion-weighted images were acquired in the area of suspected ischaemia as well as in normal appearing parts of the spinal cord as a reference. ■ Imaging findings T2-weighted images disclosed abnormal hyperintense signal abnormality predominantly of the grey matter in all patients.In patient #1,who was examined 8 h after the onset of symptoms, the signal change was very subtle in sagittal orientation (Fig. 1a), but clearly visible in transverse slices (Fig. 1b). White matter involvement was absent. The infarction was located in the conus medullaris. Swelling was not noted. DW-MRI disclosed restricted diffusion in the areas of T2 signal abnormality (Fig. 1c, d) with a decreased calculated ADC value of 0.8x10–3 mm2/s. These changes were present in transverse and sagittal images, but better visible in transverse slices (Fig. 1c, d). Reference images of normal spinal cord tissue did not indicate altered diffusion (ADC 1.1x10–3 mm2/s). Contrast enhancement was absent. Patient #2 was imaged after 12 h (Fig. 2). The images were very similar to those of patient #1. The infarction was located in the conus medullaris with preferential involvement of the grey matter of the spinal cord which showed subtle swelling (Fig. 2a). The DW images in transverse and sagittal (Fig. 2c, d) orientation clearly delineated areas of restricted diffusion with a decreased ADC value of 0.9x10–3 mm2/s.The remainder of the spinal Fig. 1 Patient #1 8h after the onset of symptoms. a T2-weighted TSE image in the sagittal plane (TR 4020, TE 114, SL 3, FoV 300). There is a faint hyperintensity visible in the central part of the conus medullaris. b This T2-weigted TSE image (TR 4250, TE 99, SL 4, FoV 160) in transverse orientation at the level of Th 12 shows hyperintensity of the grey matter of the medullary cord. A peripheral rim of white matter is preserved. c This diffusion-weighted single shot EPI image (TR 160, TE 87, SL 5, AC 3, FoV 230, matrix 128, b 1000 s/mm2) at the level of Th 12 shows restricted diffusion in the central part of the spinal cord. d The ADC map calculated from b values of 0, 500 and 1000 in the same position as Fig. 1c shows the infarcted tissue hypointense, thereby excluding T2shine through. The calculated ADC value was 0.8x10–3 mm2/s a b c d 821 a b c Fig. 2 Patient #2 12h after the onset of symptoms. a T2-weighted TSE image in the sagittal plane (TR 4020, TE 114, SL 3, FoV 300). There is a hyperintense lesion in the central part of the conus medullaris. b DW-EPI sequence (TR 160, TE 87, SL 5, AC 3, FoV 230, matrix 128, b 1000 s/mm2) in the sagittal plane. The infarcted spinal cord tissue appears bright while the not involved parts of the cord are dark. c In this corresponding ADC map to Fig 2 g, the infarcted area appears dark cord did not display any diffusion abnormality (ADC 1.1x10–3 mm2/s). There was no contrast enhancement. The spinal cord lesion in patient #3 was located in the mid-thoracic area. At the time of the MRI 30 h after the onset of symptoms, there was a moderate hyperintensity of the spinal cord in sagittal and transverse T2-weighted images. In contrast to unaffected parts of the spinal cord, the grey matter was hyperintense in transverse T2weighted images. The spinal cord appeared slightly swollen. DW images were acquired according to the initial (TR 160, TE 87, SL 5, AC 3, FoV 230, matrix 128, b 1000 s/mm2) and to the optimised (TR 220, TE 130, SL 5, AC 30, FoV 152, matrix 128, b 1000 s/mm2) protocols in transverse orientation. Restricted diffusion of protons was demonstrated by both sequences. The high-resolution sequence delineated the morphology of the infarction in more detail. The calculated ADC values did not differ significantly (0.9 ⫻10–3 mm2/s). Contrast uptake was not noted. Follow-up MRI was performed in patients #1 and #2 (Fig. 3) one week after the onset of symptoms. The spinal cord lesions were clearly visible on sagittal and transverse T2-weighted images (Fig. 3a, b). The affected areas mainly encompassed the grey matter and the adjacent parts of the white matter.A peripheral rim of white mat- ter in the vascular territory of the circular vessels of the medullary cord surface was preserved. Now an involvement of the dorsal columns was clearly noted. DW-MRI was performed using the optimised DW sequence. In patient #1, the diffusion abnormality had disappeared and the ADC value had consequently returned to normal. High resolution DW-MRI in patient #2 showed a small area of reduced diffusion in the right dorsal column of the medullary cord (Fig. 3c, d). Otherwise the DW image had also normalised. Discussion The incidence of spinal cord ischaemic infarction has not yet been established. However, it is rare compared with cerebral ischaemia. The low frequency of spinal infarction has been attributed to the vascular supply of the spinal cord. The main spinal cord arteries are small compared with cerebral vessels and with low rates of flow [10]. The probability of cardiogenic emboli entering these arteries seems to be low. Furthermore, there is an extensive collateral network between the main medullary arteries at the level of the spinal cord surface [10].This nearly always compensates for atherosclerosis, 822 Fig. 3 Patient #2 7d after onset of symptoms. a The T2-weighted image (TR 4020, TE 109, SL 4, FoV 280) of the conus medullaris now demonstrates the extent of the infarction more clearly. b This T2-weighted image (TR 4250, TE 115, SL 4, FoV 160) at the level of Th 11 displays the persistant hyperintensity of the grey matter in the medullary cord. c This DW image (TR 220, TE 130, SL 5, AC 30, FoV 152, matrix 128, b 1000 s/mm2) at the Th 11 level does not indicate restricted diffusion in most parts of the medullary cord. However, there is a small area of residual hyperintensity in the right dorsal part of the conus. The ADC map in the same position has also nearly returned to normal (ADC-value 1.1 x10–3 mm2/s) a b c d i. e., slow occlusion of the large radiculary arteries. Spinal cord infarction is probably mostly caused by small emboli gaining access to medullary vessels distal to the ring of collaterals or by sudden occlusion of the radiculary arteries. This may happen due to vessel wall dissection or after vascular surgery. However, in about half of the patients, no cause of medullary cord infarction is identified [13, 15]. Spinal cord ischaemia nearly always occurs in the anterior spinal artery territory or of the sulco-commissural arteries, the terminating branches of the anterior spinal artery. This may be due to the large vascular territory of the anterior spinal artery and a larger calibre of these vessels compared with the posterior spinal artery. Frequent locations of clinically apparent spinal infarctions are the cervical and lumbar swellings of the cord [13]. This may in part be due to the vascular anatomy and in part to the high demand on oxygen and glucose in these areas with a particularly high concentration of motoneurons. The clinical presentation in spontaneous cases is suggestive, but not pathognomonic. Clinical signs are dominated by a rapidly progressive, mostly bilateral paresis of the legs, arms or all four extremities. The initial symptoms also often include severe pain. The degree of sensory deficits varies according to the size of the affected vascular territories. Imaging is performed to rule out treatable causes of acute paraparesis. Whereas demyelinating disorders of the medullary cord are rarely painful, few neoplastic conditions are as rapidly progressive as medullary cord infarction. However, spinal epidural or subdural haematomas, which may occur spontaneously, have to be excluded by MRI [3]. As in ischaemic cerebral infarction, T2-weighted images are sensitive to the total amount of tissue water in the spinal cord.A signal abnormality is not apparent before a considerable net inflow into the infarcted tissue has occurred. T2-weighted images do not therefore display infarcted nervous tissue within the first hours after the onset of ischaemia [13]. Intracellular oedema, the 823 initial sign of cytotoxic, i. e. ischaemic cell damage, can be demonstrated with DW-MR sequences [11]. However, the application of DW-MRI for spinal cord lesions is much more difficult than in the brain for the following reasons: The signal-to-noise ratio of MRI depends on the size of the object studied and the coil used. In cerebral ischaemia, the examinations are performed using a dedicated head coil with a good ratio of head size to coil volume (coil load). For imaging the spinal cord, spinal surface coils are used. For anatomical reasons, the field homogeneity is unfavourable in these coils compared with head examinations. CSF and blood pulsation have a detrimental influence on image quality. Spinal cord infarctions are small in size compared with most instances of cerebral ischaemia. Given these obstacles, and the low incidence of spinal ischaemia, it is not surprising that few reports on DWimaging in spinal infarction have been published [1, 2, 8, 12]. Whereas it seems to be established that spinal cord ischaemia can be identified using DW-MRI, the best imaging sequences and the dynamic evolution of diffusion abnormalities remain to be identified. The application of DW sequences has been performed in the sagittal [2, 12] as well as in the transverse [2, 8] plane. However, the acquisition of isotropic DW images and the calculation of ADC maps was later optimised using transverse sequences because image distortion in sagittal orientation is more marked. Diffusion abnormality has been observed beginning 4 h after the onset of clinical symptoms of spinal cord ischaemia [12]. Most examinations were performed much later, between 12 h [8] and 30 h [2], all with a strong diffusion abnormality. Our patients were first examined between 8 and 30 h after the onset of symptoms. A strong diffusion abnormality was demonstrated in the presence of T2-signal abnormality as well as a decrease of the calculated ADC values. As a reference, a normal appearing part of the spinal cord was examined in all patients. No diffusion-abnormality or alteration of the ADC value was detected there [8]. Early follow-up MRI after 5 days was reported once [8]. These authors observed a significant increase in the ADC values compared with the initial examination. After 20 days, the ADC values had normalised in the infarcted tissue. We performed follow-up MRI in two patients 7 days after the initial scans. Although the T2-weighted images showed a typical infarction in the anterior spinal artery territory, the diffusion abnormality and the ADC values had mostly normalised. These findings are in line with a prior report of Gass et al. [2]. As in cerebral infarctions in newborns [5], this pseudonormalisation of the diffusion abnormality occurs much earlier than in the cerebral infarctions of adults where it is observed as early as after two weeks. Up to now, the spinal infarctions visualised using DWI were located in the lower part of the medullary cord, in the territory of the A. radicularis magna. In one of our patients, the mid-thoracic part of the medullary cord was affected.Whereas the lower medullary cord infarction has a characteristic appearance in MRI, at least some hours after the onset of symptoms, mid-thoracic infarctions may be more difficult to identify. Furthermore, MRI with EPI sequences is impeded by susceptibility artefacts created by the lungs and by CSF pulsation. However, even in this location, DW-MRI using EPI sequences in the transverse plane can demonstrate medullary cord infarction, too. In conclusion, acute medullary cord infarction can be detected using DW-MRI with EPI sequences in the transverse plane. However, the diffusion restriction as detected by MRI may show a rapid decline giving rise to an early signal normalisation (pseudonormalisation). References 1. Bammer R, Fazekas F, Augustin M, et al. (2000) Diffusion-weighted MR imaging of the spinal cord. AJNR Am J Neuroradiol 21(3):587–591 2. Gass A, Back T, Behrens S, Maras A (2000) MRI in spinal cord infarction. Neurology 54:2195 3. Kuker W, Thiex R, Friese S, et al. (2000) Spinal subdural and epidural haematomas: diagnostic and therapeutic aspects in acute and subacute cases. Acta Neurochir (Wien) 142(7):777–785 4. Kuker W, Weise J, Krapf H, Schmidt F, Friese S, Bahr M (2002) MRI characteristics of acute and subacute brainstem and thalamic infarctions: value of T2and diffusion-weighted sequences. J Neurol 249(1):33–42 5. Mader I, Schoning M, Klose U, Kuker W (2002) Neonatal cerebral infarction diagnosed by diffusion-weighted MRI: pseudonormalization occurs early. Stroke 33(4):1142–1145 6. Oppenheimer DR (1978) The cervical cord in multiple sclerosis. Neuropathol Appl Neurobiol 4(2):151–162 7. Reich P, Muller-Schunk S, Liebetrau M, Scheuerer W, Bruckmann H, Hamann GF (2003) Combined cerebellar and bilateral cervical posterior spinal artery stroke demonstrated on MRI. Cerebrovasc Dis 15(1–2):143–147 8. Stepper F, Lovblad KO (2001) Anterior spinal artery stroke demonstrated by echo-planar DWI. Eur Radiol 11(12): 2607–2610 9. Tartaglino LM, Friedman DP, Flanders AE, Lublin FD, Knobler RL, Liem M (1995) Multiple sclerosis in the spinal cord: MR appearance and correlation with clinical parameters. Radiology 195(3):725–732 10. Thron A (1988) Vascular anatomy of the spinal cord. Berlin, Heidelberg, New York: Springer 11. Warach S, Gaa J, Siewert B, Wielopolski P, Edelman RR (1995) Acute human stroke studied by whole brain echo planar diffusion-weighted magnetic resonance imaging. Ann Neurol 37: 231–241 824 12. Weidauer S, Dettmann E, Krakow K, Lanfermann H (2002) Diffusionsgewichtete MRT bei spinalen Infarkten: Darstellung von zwei Fällen und Literaturübersicht. Nervenarzt 73(10): 999–1003 13. Weidauer S, Nichtweiss M, Lanfermann H, Zanella FE (2002) Spinal cord infarction: MR imaging and clinical features in 16 cases. Neuroradiology 44(10):851–857 14. Wiendl H, Strayle-Batra M, Schulz JB (2002) Very bright dorsal columns: spinal magnetic resonance imaging in funicular myelosis. Arch Neurol 59(1): 147–148 15. Yuh WT, Marsh EE 3rd, Wang AK, et al. (1992) MR imaging of spinal cord and vertebral body infarction. AJNR Am J Neuroradiol 13(1):145–154