Imaging of spinal cord vascular malformations

Imaging of spinal cord vascular malformations

Imaging of Spinal Cord Vascular Malformations John Jackson, MD, and Shahram Partovi, MD Spinal vascular malformations are an important cause of myelo...

969KB Sizes 0 Downloads 26 Views

Imaging of Spinal Cord Vascular Malformations John Jackson, MD, and Shahram Partovi, MD

Spinal vascular malformations are an important cause of myelopathy in patients. They are often overlooked in the differential diagnosis because of their rarity. It is critical, however, to make a correct diagnosis in the case of spinal vascular malformations, which typically require specific treatment that can reverse or stop the progression of symptoms. Left untreated or treated incorrectly, they can cause further neurological deterioration that may be irreversible. The clinical findings in patients with spinal vascular malformations are often nonspecific and can be confused with the presenting symptoms of more common disease entities such as acquired spinal stenosis or even peripheral vascular disease. The radiologic evaluation of these lesions often shows characteristic findings that will lead to the proper diagnosis or that will at least limit the differential diagnosis and lead to definitive diagnostic testing. This article describes the various radiologic tools available for the evaluation of spinal vascular malformations. It also describes the characteristic radiologic findings of each type of spinal vascular malformation and which modality is best suited for its evaluation. Through the appropriate use and interpretation of imaging techniques, patients with myelopathy caused by a spinal vascular malformation can be accurately diagnosed and treated. Copyright 2003, Elsevier Science (USA). All rights reserved.

went two lumbar decompressive surgeries and an additional operation for release of a “tethered spinal cord.” Another patient underwent a multilevel lumbar laminectomy. Of the patients who underwent inappropriate surgical procedures, most had imaging changes on preoperative studies that were either diagnostic or strongly suggestive of dural AVF.1 These cases indicate that it is always necessary to consider a vascular lesion as a cause of myelopathy. Because imaging plays a central role in reaching a correct diagnosis, clinicians must be familiar with the various imaging modalities available, when to order each modality, and the typical appearance of the various vascular lesions of the spine on the different modalities. This knowledge is essential for diagnosing vascular lesions of the spine correctly when the rare occasion presents itself. This article first describes the different modalities available for imaging vascular abnormalities of the spine. It also reviews the advantages and disadvantages of each modality and when each modality may be useful in the evaluation of a suspected spinal vascular lesion. Finally, each type of vascular lesion and its characteristic imaging findings are delineated.

Classification of Vascular Lesions of the Spine ascular abnormalities of the spine encompass a broad range of pathologic conditions ranging from dural arteriovenous fistulas (AVFs), arteriovenous malformations (AVMs), cavernous malformations, hemangioblastomas, and aneurysms. Most of these abnormalities are rare and therefore, tend to be a diagnostic challenge. The presenting symptoms are often nonspecific and could be attributed to a number of other causes. Because the vascular causes of myelopathy are rare, they may be overlooked in the differential diagnosis. Consequently, radiologic evaluation plays a central role in obtaining a correct diagnosis and in instituting the appropriate therapy. In a series of 94 patients with myelopathy referred to the Mayo Clinic, the mean length of time between symptom onset and the eventual diagnosis of dural AVF was 23 months (range, 2-120 months). In this same series, two patients underwent open spinal cord biopsies for presumed tumor. Four patients underwent two or more lumbar laminectomy operations, including repeated extended lumbar laminectomy. One of these four patients under-


From the Division of Neuroradiology, Barrow Neurological Institute, St. Joseph’s Hospital and Medical Center, Phoenix, AZ. Address reprint requests to Shahram Partovi, MD, Neuroscience Publications, Barrow Neurological Institute, 350 W. Thomas Road, Phoenix, AZ 85013. E-mail: [email protected] Copyright 2003, Elsevier Science (USA). All rights reserved. 1092-440X/03/0603-0005$35.00/0 doi:10.1053/S1092-440X(03)00044-6

Historically, spinal vascular malformations were grouped into three major categories: dural AVFs, intramedullary AVMs, and intradural perimedullary fistulas. Intramedullary AVMs have been divided further into glomus and juvenile AVMs. In the widely accepted classification of AVMs established by Anson and Spetzler,2 type I lesions were dural AVFs. They were further divided into types IA and IB, depending on whether the dural AVF was supplied by one or multiple feeding vessels. Type II AVMs were glomus-type, true intramedullary AVMs. Type III AVMs were large juvenile AVMs, which often have both intraand extramedullary components and can even have extradural and extraspinal extensions. Type IV lesions were intradural extramedullary AVFs. This article uses the revised classification scheme recently proposed by Spetzler and co-workers3 This scheme includes three broad categories of spinal vascular malformations—neoplasms, aneurysms, and arteriovenous lesions—with subtypes within each category. Neoplasms include hemangioblastomas and cavernous malformations. The second category, isolated aneurysms of the spinal axis, are rare in the absence of an associated arteriovenous lesion. They are usually related to blood flow or dissection. The third category, arteriovenous lesions, is split into AVFs and AVMs. AVFs are then subdivided into intra- and extradural lesions. Intradural lesions are further split into dorsal and ventral lesions. AVMs are categorized as extradural-intradural or as intradural. Intradural lesions are

Operative Techniques in Neurosurgery, Vol 6, No 3 (September), 2003: pp 125-140


further subdivided into intramedullary, intramedullary-extramedullary, and conus medullaris lesions.3

Imaging Modalities Plain Radiography Conventional radiographs have such a low sensitivity for detecting spinal cord vascular malformations that they have no role in the diagnostic examination of a patient suspected of harboring a spinal cord vascular malformation. Rarely, intramedullary AVMs widen the intrapedicular distance, and extensive extradural-intradural AVMs cause scalloping of the posterior vertebral body. These changes may be seen on radiographs.1

Myelography and Computed Tomographic Myelography As noted by Bemporad and Sze,4 in 1927 Perthes diagnosed the first patient with a spinal angioma before surgery using myelography. Before magnetic resonance imaging (MRI), prone-supine myelography was the standard screening test for patients suspected of having a dural AVF. Myelography can be used in patients who are too large to be imaged with either computed tomography (CT) or MRI or who cannot undergo MRI for other reasons. Classically, dilated tortuous vessels and beading of the cauda equina have been considered suggestive of a dural AVF. Gilbertson and co-workers5 reviewed patients who had or were suspected of having a dural AVF who all eventually underwent spinal angiography: 25 patients had angiographically proven dural AVFs (true-positives) and 12 patients who did not have dural AVFs at angiography (false-positives) underwent pronesupine myelography. In all of the 25 patients who had dural AVFs, myelography showed prominent vessels extending over multiple levels (range, 3-20 levels; mean, 8 levels). Beading of the cauda equina was present in 61% of this true-positive group. Vessels extending over a mean of six vertebral bodies were identified on the myelograms of 11 of the 12 (92%) patients with false-negative angiograms who did not have an AVF. Beading of the cauda equina was present in two patients who did not have a dural AVF (17%). Although the prominence of the vessels, the length over which they extended, and their tortuosity on supine myelography differed between the two groups, there was considerable overlap. In fact, the 12 patients with negative spinal angiograms were referred to spinal angiography because of the vessels depicted on their supine myelogram. This study demonstrates both a disadvantage and an advantage of supine myelography. The advantage is that it is very sensitive. No patients who did not have dilated vessels on supine CT myelography were discovered to have dural AVFs. The disadvantage of CT myelography is the considerable overlap between the imaging findings of normal patients and those with underlying pathology. Findings such as dilated or tortuous vessels on the surface of the spinal cord or beading of the cauda equina on CT myelography can be seen in normal patients without an underlying AVF. Based on these results, the overall accuracy and specificity of CT myelography are considered inferior to high-quality MRI for the evaluation of a possible AVF.5 CT myelography, however, may have a complementary role when findings on MRI are inconclusive. Myelography


alone is also unable to define the vascular supply of an AVF. Furthermore, the procedure is invasive and associated with risks such as spinal headache, nerve root damage, allergic reaction to contrast, arachnoiditis, infection, or epidural hematoma (albeit the incidence of these adverse outcomes is very low).

MRI MRI is the study of choice for the evaluation of spinal vascular malformations. Many typical and suggestive findings on MRI allow a spinal cord vascular malformation to be diagnosed. Even when MRI alone does not support a conclusive diagnosis, the findings often suggest a level of abnormality that can direct spinal angiography. In turn, the angiography becomes safer for the patient because the dye load can be lowered and the length of the procedure can be shortened. MRI has many advantages. First, it can rule out other more common causes of myelopathy or radiculopathy. It is a wellestablished method for evaluating disk disease and acquired spinal stenosis. Unlike myelography and spinal angiography, MRI is noninvasive. Therefore, there are no associated procedural-related risks of complications. And unlike other radiographic modalities, no ionizing radiation is involved. Finally, newer MRI techniques may allow studies to be tailored to further define the vascular supply of lesions. Several different techniques in MR angiography (MRA) of the spine, including gadolinium-enhanced or nonenhanced threedimensional (3D) time of flight (TOF), two-dimensional (2D) and 3D phase-contrast (PC), and dynamic gadolinium-enhanced MRA, have all been used with various levels of success. When clinicians are attempting to determine the level of dural AVFs, these techniques may add benefit by helping to perform a more directed spinal angiogram. Recently, Saraf-Lavi and co-workers6 found that neither the sensitivity nor specificity of MRA for detecting the presence or absence of a spinal dural AVF was increased compared with those of conventional MRI techniques. However, the ability of three reviewers to determine the level of a dural AVF with gadolinium-enhanced spinal 3D TOF MRA compared with conventional MRI alone was significantly better. In patients who were correctly diagnosed as having dural AVFs with conventional MRI, that modality predicted the level of the dural AVF in only 15% of the patients. In contrast, the addition of gadolinium-enhanced 3D TOF spinal MRA predicted the correct level in 50% of the patients. The level of the dural AVF plus or minus one vertebral segment was predicted in 43% of the conventional MRI group compared with 73% with both conventional MRI and spinal MRA.1,6 In 11 consecutive patients with vascular malformations of the spine, Mascalchi and co-workers7 used 2D PC MRA and identified the arterial feeder in the three patients who had an intramedullary AVM and in two patients who had a perimedullary AVF. However, they were able to identify the fistulous site of the dural AVF in only two of six patients.7 Their findings demonstrate the difficulty of using MRA techniques to detect small vessels involved in a dural AVF and of detecting low-flow lesions with PC and TOF techniques compared with the more easily detected high-flow lesions such as AVMs. The disadvantages of MRI reflect individual patient differences rather than disadvantages of the modality itself. Patients may be unable to undergo an MRI for a variety of reasons. Some patients may have implanted devices such as pacemakers, defiJACKSON AND PARTOVI

brillators, or older aneurysm clips that are incompatible with MRI. Some people are unable to tolerate examination with a high-field strength magnet because they are severely claustrophobic. Others have physical limitations such as excessive weight that precludes them from fitting into the bore of the magnet. Limitations such as claustrophobia or an inability to remain still can be overcome by using conscious sedation or general anesthesia. The benefit of obtaining the study, however, must outweigh the additional risk to the patient posed by the sedation.

Spinal Angiography Spinal angiography is the gold standard for the radiologic diagnosis of spinal AVMs and AVFs. A major advantage of spinal angiography is its superior temporal and spatial resolution and definition of arteries too small to visualize on CT myelography or MRI. Neither MRI nor CT myelography can give the same direct information about the arterial supply to an AVM as readily as spinal angiography can. In addition to spinal MRA, MRI can suggest the level of a spinal dural AVF, plus or minus one vertebral body level. However, the definitive diagnosis rests with spinal angiography. Spinal angiography defines the actual feeding arteries and draining veins and their location for preoperative planning. In some cases, embolization therapy can be performed at the same time, making the study both diagnostic and therapeutic. Additional important information for preoperative or pre-embolic planning, such as defining the collateral blood supply, the location and level(s) of major supply to the anterior and posterior spinal arteries, and anatomic vascular variants, are also best defined with spinal angiography. The disadvantages of spinal angiography are related to its invasiveness. Because lesions that require spinal angiography are rare, many centers perform too few spinal angiograms to develop specific expertise in this area of angiography. Many competent cerebral angiographers may be less adept at performing spinal angiography because they do not perform a high volume of cases. A complete spinal angiogram of the neural axis requires a high load of contrast dye, and a two-stage procedure is often necessary. It also involves significant radiation to the patient and angiographer. A high degree of organization and documentation is required to assure that a vertebral level is not missed during the examination or later misinterpreted as a different level. Because of the length of the procedure and the need for patients to remain still throughout the study to define very small vessels, many clinicians advocate general anesthesia for spinal angiography, introducing yet another risk. Some authors advocate an aortogram with a low injection rate such as 8 to 10 mL/sec for a total of 30 mL at the start of the procedure. This study is usually performed with a pigtail catheter near or immediately upstream from the region of interest based on review of the patient’s MRI or CT myelogram. The low injection rate allows the contrast to stream down the dorsal wall of the aorta, filling the intercostal and lumbar radicular arteries. The operator can get the “lay of the land” first, allowing more focused selective vertebral level injections.8 Other authors initially perform a bilateral iliac high-injection rate “flush” aortogram (40 mL/sec for a total of 75 mL) before selective level injections.5 This technique requires accessing both common femoral arteries. Many reports have been published about the complications associated with spinal angiography. However, the exact statisSPINAL VASCULAR MALFORMATIONS

tics are difficult to assess because the level of expertise, techniques used, and thoroughness vary at each center. Consequently, complication rates at different centers are far from uniform. The statistics on complications also have changed over time. Historically, certain ionic contrast agents were implicated in direct neurotoxicity to the spinal cord related to their hyperosmolar effects. These effects and risks may still be quoted erroneously today. Newer nonionic contrast agents do not exert a direct hyperosmolar effect on the spinal cord. Recent reports suggest that the rate of major complications associated with spinal angiography is very low. Champlin and co-workers9 reported 61 patients who underwent spinal angiography. There were no major complications and only two minor complications (small groin hematomas).8 Forbes and co-workers10 reported the outcomes of 134 angiograms performed in 96 patients. Only three patients (2.2%) experienced neurologic symptoms related to the angiography, two of which recovered fully within 24 hours. The third recovered fully in less than 1 week. Eleven patients (8.2%) had local complications related to the puncture site. Ten were hematomas and one was a vascular occlusion. Five (3.7%) patients had systemic, nonneurologic complications. Four patients had nausea, vomiting, and transient hypotension, and one patient developed a rash. These results demonstrate that spinal angiography tends to be well tolerated even though it is not entirely risk free.8,10

Spinal Vascular Abnormalities Intradural Dorsal AVFs Intradural dorsal AVFs (most frequently referred to as type I AVMs in prior classifications) are the most common vascular abnormalities of the spinal cord. They represent 68 to 80% of all vascular malformations involving the spinal axis.4,5 Although their etiology is controversial, they are usually considered to be acquired lesions. They most commonly afflict men ranging from 40 to 70 years old. The male-to-female predominance is approximately 4:1. Their typical clinical presentation consists of slowly progressive, sometimes fluctuating symptoms that initially include weakness, numbness, pain, and nonspecific symptoms. Eventually, bowel, bladder, and sexual dysfunction; spastic paresis; and loss of pain and temperature sensation follow. The latency between the onset of symptoms and the time of definitive diagnosis can be considerable. Because these lesions are a reversible cause of myelopathy and surgical treatment is associated with very low morbidity rates, it is desirable to diagnose the lesion early in the course of the disease. The long-term prognosis after surgical ablation of the fistula tends to correlate with the severity of the patient’s preoperative symptoms. Complete paraplegia is a predictor of poor outcome.1 Diagnosis may be delayed for many reasons. Because of their age group, patients typically afflicted with an intradural AVF often have other common medical conditions whose early symptoms could be confused with those of an intradural AVF (eg, lower extremity vascular disease and acquired spinal stenosis). Furthermore, many findings on MRI can suggest the presence of an intradural AVF. Few of the findings, however, are specific. They could be confused with transverse myelitis, a neoplasm, multiple sclerosis, changes related to trauma, or possibly spinal cord infarction. Because dural AVFs are rela-


tively rare, clinicians or radiologists may not entertain the diagnosis early in the patient’s course. Intradural dorsal AVFs are abnormal fistulous connections. They have been further subclassified into two subtypes: type A, with a single feeding artery, and type B, with multiple feeding arteries. The usual scenario is a single feeding radicular artery in the thoracic or lumbar spine with a fistulous connection that empties intradurally at the dural nerve root sleeve into a medullary vein.4 A single, dilated medullary vein then fills in a retrograde fashion and communicates with the perimedullary venous plexus. The coronal venous plexus often becomes congested over several levels. Because radicular arteries are small vessels, dural AVFs are typically low-flow lesions. However, the retrograde arterial blood seeking an exit from the perimedullary spinal venous system creates high pressure. The myelopathy is thus caused by chronic venous hypertension. Myelography and CT Myelography. The findings on myelography and CT myelography reflect the enlarged coronal venous plexus, with enlargement and increased tortuosity of the anterior or posterior median spinal veins, or both. Prominent serpiginous vessels are seen most frequently on the dorsal surface of the spinal cord but are occasionally visible on the ventral surface in addition to dorsal venous enlargement. The vessels may extend over several levels. Beading of the cauda equina, another classic myelographic finding, represents dilated veins coursing along the nerve roots of the cauda equina. The myelographic images and the CT should be obtained with the patient placed supine to increase sensitivity for detecting the enlarged draining veins most often found on the dorsal surface of the spinal cord. Many clinicians advocate obtaining both prone and supine myelographic views, which may add to the conspicuity of the prominent veins on the supine films compared with the less prominent findings on prone images. Atkinson and co-workers1 reported 94 patients with eventual angiographically proven intradural AVFs treated at the Mayo Clinic from June 1985 to December 1999. They imaged 27 patients with prone and supine myelography followed by CT and also performed anteroposterior tomography during the myelographic part of the study. All patients had enlarged vessels on the dorsal surface of the spinal cord, and 10% also had dilated veins on the ventral surface. None had abnormal vessels only ventrally. They also reviewed 10 myelographic studies performed elsewhere before the patient’s referral. Either because of absent findings or an inadequate study, only 5 of the 10 demonstrated abnormal vasculature suspicious for a dural AVF.1 Gilbertson and co-workers5 reviewed a subset of patients with AVFs treated at the Mayo Clinic from January 1985 to December 1993 who underwent both myelography and eventually angiography. All patients eventually found to have an intradural AVF on angiography (angiogram-positive group) also had vessels visualized on the surface of the spinal cord with CT myelography. However, 92% of the patients who had CT myelography followed by complete angiography did not reveal an AVF (angiogram-negative group). Myelography also visualized vessels on the surface of the spinal cord, with wide overlap in the degree of tortuosity and number of vertebral body levels involved. Beading of the cauda equina on myelography was a less sensitive finding present in 61% of the eventually proven angiogram-positive patients. On myelography, however, beading of the cauda equina was more specific. It was present in only 17% of the spinal angiogram-negative group.5 In summary,


Fig 1. Intradural dorsal AVF. Sagittal T2-weighed MRIs of the thoracic spine in a patient with an intradural AVF demonstrates prominent flow voids dorsal to the spinal cord (arrowheads). Abnormal T2-weighted signal is also present within the spinal cord itself and extends down to the conus medullaris (arrow).

myelographic findings of prominent vessels on the surface of the spinal cord is a very sensitive but nonspecific sign of an underlying intradural AVF. In contrast, the classic finding of beading of the cauda equina is not very sensitive but is a fairly specific sign of an underlying AVF. MRI. On conventional MRI, the typical findings associated with intradural dorsal AVFs include increased T2-weighted signal in the spinal cord, intramedullary enhancement with gadolinium, serpiginous flow voids on the pial surface of the spinal cord, mass effect involving the dorsal aspect of the spinal cord with a scalloped appearance caused by enlarged perimedullary veins, a T2-weighted hypointense ring around the central T2-weighted hyperintensity, spinal cord edema or swelling, decreased T1-weighted signal in the spinal cord, and enhancement of the perimedullary venous plexus (Fig 1).1,4,5,11 Of these JACKSON AND PARTOVI

signs, increased T2-weighted signal in the spinal cord is the most sensitive finding. It has been present in all cases in a number of series.4,5,11,12 Increased T2-weighted signal alone, however, is a nonspecific finding associated with abnormalities including neoplasms, degenerative processes, multiple sclerosis, trauma, infarct, transverse myelitis, and myelomalacia from prior insult to name a few. Without ancillary findings, increased T2-weighted signal is not particularly suggestive of a dural AVF. The location of the abnormal signal is also helpful diagnostic information. Most intradural AVFs involve the lower thoracic spinal cord and are associated with abnormal signal extending to the conus medullaris (involving the tip of the conus in 87% of patients in the series by Gilbertson and co-workers).5 Typically, the abnormal T2-weighted signal is seen over several levels. In the series by Gilbertson and co-workers,5 abnormal signal extended over a mean of seven levels. The signal is usually homogeneous and centrally located in the spinal cord, sparing the periphery. In some cases there is patchy enhancement in the area of abnormal signal after gadolinium administration. Hurst and Grossman11 considered peripheral T2-weighted hypointensity around the area of increased T2-weighted signal to be a good sign. This finding can help determine that the abnormal signal is likely from venous hypertensive myelopathy rather than from a tumor, demyelinating process, or other cause. They evaluated 11 patients who had either a typical dural AVF of the spine or an intracranial dural AVF associated with spinal drainage that created venous hypertensive myelopathy. All 11 patients had peripheral low-intensity T2-weighted signal in addition to the central increased signal. In the absence of hemorrhage, a T2-weighted hypointense ring is not associated with neoplasms, transverse myelitis, or multiple sclerosis. It is a good ancillary finding to suggest a dural AVF. Hurst and Grossman11 pointed out that the T2-weighted hypointense area was seldom conspicuous on fast-spin echo (FSE) T2-weighted imaging. Conventional, spin-echo T2-weighted, or gradient T2weighted sequences emphasized the abnormality. On most routine spinal imaging, sagittal FSE T2-weighted MRI is usually used rather than conventional T2-weighted spin-echo or gradient techniques. However, axial gradient T2-weighted sequences are often obtained routinely. Therefore, when a bright signal is present in the central spinal cord on T2-weighted images, axial gradient sequences should be inspected carefully for a T2-weighted hypointense rim. Prominent flow voids in the intradural space are highly suspicious, but they can be mimicked by flow artifact from cerebrospinal fluid (CSF). Therefore, a flow void should not be interpreted as a dural AVF without ancillary findings to support the diagnosis. A scalloped appearance involving the dorsal aspect of the spinal cord, enhancement of prominent perimedullary veins, serpiginous flow voids, and T1-weighted hypointensity are additional, less common findings also highly suggestive of an intradural AVF.1,4,5,11,12 If conventional imaging is suggestive of an intradural AVF, spinal MRA can help predict the level of the fistula and direct the spinal angiogram. In the absence of other findings suggestive of an intradural AVF on conventional imaging alone, the additional cost, time, and contrast needed to perform MRA make it unlikely that this modality will play a role in the actual diagnosis of intradural AVFs.6 Subarachnoid hemorrhage (SAH) is not a finding associated SPINAL VASCULAR MALFORMATIONS

with intradural AVFs in the thoracic spinal cord and conus medullaris, and an intradural AVF should not be considered if SAH is present. However, some cervical intradural AVFs do manifest with SAH because they can drain cephalad into the pontomesencephalic venous drainage. Because outflow for the fistula is available, the fistula changes from its typical low-flow, high-pressure state to a high-flow, low-pressure state. Subsequently, venous varices form and can rupture, resulting in SAH.1 Spinal Angiography. Spinal angiography is the gold standard for the diagnosis of an intradural AVF. The angiographic study shows a small focal nidus at the site of fistula. The fistulous arterial-to-venous shunting is usually demonstrated in the nerve root sleeve (Fig 2). Drainage from a radicular artery into a radicular or medullary vein then drains intradurally back toward the spinal cord into the perimedullary venous system. Most intradural AVFs drain cephalad through the posterior median vein and coronal venous plexus. The rare intradural AVF has more than one radicular artery supplying the fistula (type B lesions). The fistulous radicular artery seldom arises from the same vertebral level as the artery of Adamkiewicz. In the Mayo series reported by Atkinson and co-workers,1 this configuration occurred in only 1 of 94 patients. Niimi and co-workers13 reported 80 intradural fistulas, three cases of which had venous drainage directly from a radicular vein into a dilated anterior spinal vein. Although a rare presentation angiographically of an intradural AVF, all three of these cases presented diagnostic challenges because the appearance of the dilated anterior spinal vein could be mistaken for the anterior spinal artery (of Adamkiewicz). Subtle but constant angiographic features such as distortion of the hairpin turn, branching of the anterior spinal vein, drainage of the anterior spinal vein into an epidural vein, opacification of other medullary veins connected with the anterior spinal vein, and identification of the anterior spinal artery at the segment where the anterior spinal vein was opacified allowed the anterior spinal vein to be distinguished from the anterior spinal artery. To appreciate the vascular supply and drainage of these AVFs, careful angiographic technique including oblique and lateral views as well as delayed imaging with a large field of view was necessary. Although rare, such cases illustrate the need for a neuroradiologist experienced in spinal angiography to evaluate spinal vascular malformations. Posttreatment Imaging. Willinsky and co-workers14 examined 17 patients with intradural AVFs, 10 of who underwent pretreatment and posttreatment MRI and clinical follow up. In three patients, prominent vessels seen in the subarachnoid space were less evident on the posttreatment MRI. T2-weighted spinal cord hyperintensity was no longer evident in 6 patients and was less evident in three. At 9 months one patient had persistent T2-weighted hyperintensity that was less evident at 34 months. Two of the 10 patients did not improve clinically after treatment of their intradural AVF. However, there were no clear-cut differences in the findings on the posttreatment MRI studies of the two patients who did not improve clinically. Bowen and co-workers12 used MR angiography to evaluate three patients before and after surgical treatment of an intradural AVF. In two patients the enlarged medullary vein draining the fistula seen preoperatively resolved. In the third patient the enlarged dorsal coronal venous plexus resolved. However, visualization of the anterior median vein persisted, although its size and tortuosity had decreased on the posttreatment imaging


Atkinson and co-workers1 obtained postoperative MRIs in 32 patients 3 to 48 months after surgery. Clinical improvement correlated with total resolution of T2-weighted signal changes in the spinal cord seen preoperatively. When the T2-weighted hyperintensity in the spinal cord did not resolve or only resolved partially, there was no specific correlation with clinical improvement. In patients whose fistula was obliterated successfully, the abnormal dorsal spinal cord vascularity seen on preoperative imaging had resolved completely on postoperative MRI. In summary, the resolution of T2-weighted spinal cord hyperintensity after treatment of an intradural AVF seems to correlate not only with obliteration of the fistula but with clinical improvement. Failure of the T2-weighted hyperintensity to resolve is not predictive of whether the fistula was obliterated. If clinical deterioration persists in these cases, repeat spinal angiography is warranted because the fistula may still be patent. If enlarged dorsal vessels are present on the pretreatment MRI, they should resolve or become inconspicuous on posttreatment MRI when the fistula is obliterated successfully. MRA may be a noninvasive method for assessing the patency of the fistula as Bowen and co-workers12 demonstrated in a small number of patients. Further evaluation of this method will determine whether it should be a mainstay in posttreatment imaging of intradural AVFs.

Intradural Ventral AVFs

Fig 2. Intradural dorsal AVF (type A). (A) Anteroposterior and (B) lateral projections of a spinal angiogram demonstrating an intradural dorsal AVF. Only a single feeding artery was found during the angiogram (arrow); therefore, this lesion would be considered a type A intradural dorsal AVF. Both angiography and dynamic imaging showed that the venous drainage of this lesion was in a cranial direction (arrowhead).

studies. Another patient was studied before and after endovascular treatment. MR angiography performed 1 and 5 months after treatment showed persistence of the dominant, enlarged medullary vein. The fistula was obliterated surgically and subsequently the enlarged medullary vein was not visualized.


Intradural ventral AVFs are perimedullary fistulas occurring directly between the anterior spinal artery and a draining vein. They result in an enlarged perimedullary venous network and have no intervening nidus. In the original Anson and Spetzler2 classification, intradural ventral AVFs were referred to as type IV AVMs and were subclassified into types IV-A, IV-B, and IV-C. The A, B, and C subclassifications have been retained in the revised classification system of Spetzler and co-workers3 Type A lesions are slow-flow AVFs that do not enlarge the posterior or anterior spinal artery. Type B lesions represent a gradation between type A and type C lesions. Type C lesions are large fistulas with dilated feeding arteries coursing into dilated, tortuous draining veins. They are high-flow lesions and may have multiple feeders. The high-flow state of type B and C lesions can cause myelopathy related to vascular steal and ischemia or spinal cord compression from venous varices. Intradural ventral AVFs can be found anywhere along the spinal axis but typically involve the thoracolumbar area. Age at presentation ranges from 1 to 70 years old, but they most often manifest in the third decade.1 The imaging features of intradural ventral AVFs depend on the type of flow. Because of their slow flow and high pressure, the imaging characteristics of type A lesions are similar to those of intradural dorsal AVFs and they reflect abnormalities mostly related to venous hypertension. These findings include dilated vessels on the surface of the spinal cord, high T2-weighted signal, and patchy enhancement. However, the dilated draining veins involve the ventral rather than the dorsal surface of the spinal cord. The higher flow type B and C lesions are more likely to behave like AVMs: 10 to 20% manifest with SAH.1 When intradural ventral AVFs are large (ie, type C lesions), the tortuous flow voids on the pial surface can sometimes be confused as being intramedullary, especially when enlarged vessels exert JACKSON AND PARTOVI

Fig 3. Intradural ventral AVF. (A) Anteroposterior and (B) lateral spinal angiographic projections show an intradural AVF supplied by the anterior spinal artery, thereby confirming the ventral location of this lesion.

mass effect and compress the adjacent spinal cord. Because these lesions are located on the ventral surface of the spinal cord, dilated veins can be difficult to perceive as vascular flow voids rather than as CSF flow voids.1 Consequently, if clinical suspicion is high and MRI is negative, prone-supine myelography should be performed to determine if angiography is warranted. Again, angiography is key once an abnormality is detected on MRI or CT-myelography. Angiography can assess the flow characteristics of the lesion and accurately depicts the fistulous site and feeding vessel(s) (Fig 3). Arterial aneurysms and venous varices are also evaluated well with angiography.

Extradural AVFs Extradural or epidural AVFs are direct connections between an extradural artery and vein. This configuration leads to a highflow fistula associated with engorgement of the epidural venous plexus. The enlarging epidural venous plexus can cause proSPINAL VASCULAR MALFORMATIONS

gressive myelopathy from spinal cord compression. High venous pressure can be transmitted back into the intradural venous system, preventing adequate venous outflow and causing venous hypertensive changes in the spinal cord. The high-flow state of these lesions can shunt large quantities of blood through the AVF creating a component of vascular steal from the spinal cord.3 Extradural AVFs have been reported after trauma and surgery (iatrogenic) and as congenital lesions. They can occur anywhere along the spinal axis but are most common in the cervical spine with an arterial supply from the vertebral artery, ascending muscular branches, thyrocervical branches, or costocervical branches. Risk factors besides trauma and instrumentation include conditions associated with dysplastic vessels, which are likely predisposed to rupture/injury with subsequent formation of a fistula. These conditions include neurofibromatosis type I, fibromuscular dysplasia, Marfan’s


syndrome, or Ehler’s Danlos type IV15 Unlike intradural AVFs, extradural AVFs seem to occur in a slightly younger age group and over a greater range of ages. On MRI or CT these lesions typically appear as extradural, enhancing masses. On MRI15 they have been confused with an epidural hematoma or as an intradural/extradural AVM.16 They have even been angiographically occult (either because of a technically inadequate angiogram or extremely slow flow through the shunt with the high intravenous pressure preventing the tortuous veins from filling during an arterial injection).17 In the one report of an “angiographically occult” extradural AVF,17 highly suggestive MRI findings allowed a targeted surgical exploration at the site of the fistula. On MRI prominent venous flow voids may be visualized in the epidural space (Fig 4). Some patients with an extradural AVF have high T2-weighted signal changes in the spinal cord and no significant mass effect from engorged epidural veins, supporting venous hypertension as the cause of myelopathy.17 Rather than an abnormal signal from the spinal cord, other extradural AVFs have been associated with mechanical compression of the spinal cord because of an enlarged epidural venous plexus. These findings demonstrate the variable presentations associated with extradural AVFs. An interplay between blood flow through the lesion and variable intactness of the antivenous reflux mechanism of the radicular vein at the dural nerve root sleeve likely determines how these lesions manifest. These lesions also likely exhibit a natural progression. Some patients have presented with initial symptoms of mechanical compression, including localized radicular symptoms from mass effect, and progressed to symptoms and imaging features suggestive of venous hypertensive changes.15,16,18,19 Purely epidural fistulas with subsequent reflux into the intradural space distant from the fistulous site have exhibited imaging features typical of intradural AVFs.20,21 Extradural AVFs have also manifested with dilated perimedullary, intradural vessels but no clinical or imaging features of spinal cord compromise related to elevated venous pressure.22 In summary, extradural AVFs have a spectrum of imaging and clinical features. Their imaging features are related to their pathophysiology. The most typical finding is low T2-weighted signal expanding the epidural space (representing flow voids from dilated epidural venous plexus). The dilated “mass” of vessels in the epidural space can cause spinal cord compression. High T2-weighted signal within the spinal cord can also reflect compressive myelopathy or edema related to elevated venous outflow pressure in the intradural venous system because of the lack of outflow into the epidural space. Angiography can show single or multiple fistulous connections between radicular, vertebral, or paraspinal arteries and epidural veins. On angiography, these fistulas are typically high flow. There are also rare reports of intradural venous reflux and drainage, frequently distant from the extradural fistulous site.20-22

Intramedullary AVMs These lesions, once referred to as type II AVMs, are intramedullary glomus-type malformations with a discrete nidus. Unlike extra- and intradural AVFs, patients with intramedullary AVMs usually manifest with acute symptoms caused by sudden intramedullary hemorrhage or SAH. When patients exhibit progressive myelopathy associated with sudden bouts of deterioration, the clinical presentation can rarely be confused with that


of a spinal AVF. This scenario can be caused by mass effect on the spinal cord from a larger lesion or venous varix, by spinal cord ischemia related to vascular steal, or by medullary venous congestion.8 Intramedullary AVMs are high-flow, high-pressure lesions, usually with low resistance. Consequently, they often cause aneurysms to develop on the feeding arteries and varices of the draining veins. Aneurysms have been reported associated with intramedullary AVMs in 40 to 50% of the cases.1 Intramedullary AVMs have also been associated with cutaneous angiomas, Klippel-Trenaunay-Weber syndrome, and Rendu-Osler-Weber syndrome. Although these lesions can occur anywhere in the spine, they may have a slight preponderance for the cervical spine, especially the cervicothoracic junction. These lesions typically occur in younger patients.1,4 Cervical AVMs can drain cephalad into the pontomesencephalic veins and thus can potentially manifest with intracranial SAH caused by rupture of a venous varix.8 CT and Myelography. CT and myelography can be used to image intramedullary AVMs, but CT and myelography are less sensitive and specific than MRI (Fig 5). Because intramedullary AVMs drain through the coronal venous plexus, enlarged pial vessels may be seen on myelography. Another finding on CT myelography may be focal enlargement of the spinal cord at the site of the AVM. Occasionally, atrophy of the distal spinal cord is detected. However, this intermittent finding is nonspecific.1,8 With newer multislice detector technology and CT angiography, the vascular supply to AVMs could potentially be defined with CT in patients unable to undergo MRI. To the authors’ knowledge, no data have been published using this technique specifically for spinal AVMs even though experience with imaging cerebral AVMs has been reported. If the study could be directed to a specific area of interest based on noncontrast CT or myelography, CT angiography of spinal AVMs would likely be feasible. CT imaging has been performed with selective catheterization of a spinal AVM to define the vasculature in relationship to the surrounding soft tissue.23 MRI. MRI can show round and serpiginous flow voids, particularly at the nidus. The spinal cord tissue surrounding the nidus usually appears normal but not within the nidus. Dilated vessels are seen in the subarachnoid space around the nidus. MRI is sensitive for the evaluation of glomus-type AVMs. The vascular supply to intramedullary AVMs can be defined with PC MRA much easier than the fistulous site in an intradural AVF can be identified.24 If gliosis from ischemia or compression is present, MRI can also reveal high-intensity signal in the spinal cord surrounding the AVM. Hemorrhagic products may surround the nidus, particularly on gradient sequences (Fig 6A and B).1 Although MRI is very good at defining the size, location, and extent of the abnormality in an intramedullary AVM and the surrounding soft tissues, angiography remains a critical part of the initial evaluation. Spinal angiography can define both the arterial feeding vessel(s) and the collateral blood supply. MRI does not visualize the collateral supply well, even though it may visualize the direct feeding vessel(s). Angiography is better than MRI for determining the presence of aneurysms involving the feeding vessels or nidus. Angiography also is used to define the vascular supply to the normal spinal cord, vertebrae, and paraspinal tissues (Fig 6C). This information aids in pre-embolic and preoperative planning.8 JACKSON AND PARTOVI

Fig 4. Extradural AVF. (A) Sagittal and (B) axial T1-weighted MRIs of the cervical spine and (c) single view of the corresponding spinal angiogram. Large extradural flow voids are seen on the MRIs (arrowheads). On the sagittal image the flow voids mimic an epidural mass. Spinal angiography confirms the lesion to be entirely extradural with enlarged draining veins. (With permission from Barrow Neurological Institute.) SPINAL VASCULAR MALFORMATIONS


rior and posterior spinal arteries leading to a glomus-like nidus. The nidus is typically extramedullary or pial, but it can have intramedullary components. The AVM may extend from the conus medullaris to involve part or all of the filum terminale. On MRI these lesions appear as areas of abnormal T1-weighted signal within the conus–flow voids with possible hemorrhage. Spinal angiography is essential in characterizing the complex vascular supply and drainage and in visualizing each nidus.

Cavernous Malformations

Fig 5. Intramedullary AVM. Postmyelography CT of the spine in a patient with an intramedullary AVM shows abnormal vascular structures within the thecal sac, distorting the normal configuration of the spinal cord.

Extradural-Intradural Arteriovenous Malformations On MRI, angiography, and CT, the imaging features of extradural-intradural AVMs are similar to those of intramedullary lesions except the former are much larger than the latter. These lesions are the rarest of the spinal AVMs, accounting for about 7% of all spinal AVMs. They are fed by multiple arterial feeders. They differ from intramedullary lesions in that normal spinal cord tissue is interspersed within the tangle of abnormal vessels. They often extend along the long axis of the spinal cord over multiple levels. An extramedullary component of the AVM and even an extradural component are common, with extension into the surrounding vertebral bodies and paraspinal soft tissues. When these lesions involve all structures of a single metamere, including the skin, they have been called metameric AVMs.1 MRI is extremely useful in imaging these lesions as it can define the full extent of the lesion and its relationship to surrounding tissues, in particular the extradural component (Fig 7). This knowledge aids in operative planning. CT with myelography plays an ancillary role to MRI but can be employed if the patient is unable to undergo MRI. In this case, iodinated contrast may help to better define the extradural components of the lesion. Because of the large size and extensive, infiltrative nature of these lesions, they are frequently treated with a multidisciplinary approach, using both intraarterial embolization and surgery. Because of the presence of the intervening normal spinal cord tissue, complete surgical resection is often difficult and associated with a high risk of neurologic morbidity.3 As with intramedullary lesions, angiography is used to define the feeding vessels, venous outflow, associated aneurysms or venous varices, collateral blood supply, and normal spinal vascular anatomy.

Conus Medullaris Arteriovenous Malformations AVMs of the conus medullaris are a newly proposed category characterized by multiple feeding arteries, multiple niduses, and a complex pattern of venous drainage (Fig 8).3 These lesions demonstrate a mixed arterial supply from both the ante-


Cavernous malformations are vascular malformations that consist of multiple sinusoidal, blood-filled channels or “caverns.” They have weak walls and a propensity to hemorrhage. Spinal cavernous malformations were once thought to be exceedingly rare. Before 1985 only 19 cases had been reported. With the advent of MRI, however, they have been recognized increasingly.25,26 In the general population the incidence of cavernous malformations is estimated to be 0.02% to 4%. Of all cavernous malformations, 3% to 5% occur in the spine.1 The male-to-female ratio is about 5:7. About half have been reported in the thoracic spine. Based on a review of 117 spinal cavernous malformations by Zevgaridis and co-workers,26 the peak age of presentation is the fourth decade. Patients may become symptomatic with stepwise episodes of clinical deterioration, slow progressive neurologic decline, or an acute onset with subsequent rapid or gradual decline over weeks to months. Typically, spinal cord cavernous malformations are intramedullary lesions. Occasionally, they can be exophytic and become symptomatic with SAH. In addition to intramedullary lesions, both intradural-extramedullary and extradural lesions have been reported.27,28 Spinal cord cavernous malformations can occur anywhere along the spinal axis but have been shown to occur more frequently in the thoracic spine, followed by the cervical and then lumbar regions.29 CT and myelography are neither sensitive nor specific for spinal cord cavernous malformations. On CT or myelography, the most frequent abnormal finding is spinal cord swelling or local widening at the site of the lesion. Occasionally the lesion appears dense on CT. There is sometimes faint enhancement on contrast– enhanced CT imaging and very rarely an associated syrinx. CT and myelography are insensitive, however, with reported cases of normal CT and myelography when subsequent MRI scans showed the intramedullary lesion25,29-31 There are very rare case reports of densely calcified spinal cord cavernous malformations.32,33 Spinal cord cavernous malformations are angiographically occult with no abnormality seen on angiograms obtained in several different patients from different reported series.25,30,31,34 MRI is the best modality for imaging spinal cord cavernous malformations, which often look similar to intracranial cavernous malformations. They tend to be oblong with heterogeneous T1-weighted and T2-weighted signal characteristics, reflecting blood products of various ages (Fig 9). A rim of T1- and T2weighted signal representing hemosiderin often surrounds spinal cavernous malformations. Visualization of the lesion can be enhanced with gradient-recalled sequences sensitive to blood products. On these sequences, a “blooming” effect makes the lesion appear slightly larger than on other sequences. Turjman and co-workers25 reviewed 11 patients, and the most common finding on MRI was a mixed, heterogeneous JACKSON AND PARTOVI

Fig 6. Intramedullary AVM. (A) Sagittal T2-weighted MRI of the cervical spine, (B) axial gradient-echo (GRE) MRI through the lesion, and (C) corresponding angiogram. Abnormal T2-weighted signal in the cervical spinal cord is mixed with areas of low signal intensity, some of which represent flow voids and others hemorrhagic products. The spinal cord is slightly expanded. The axial GRE image (B) shows the intramedullary hemorrhagic products to best advantage as areas of pronounced signal loss. The spinal angiogram (C) shows two arterial feeders to the lesion, both of which arise from the thyrocervical trunk.

T1-weighted and T2-weighted signal intensity mass with a reticulated appearance (7 cases). Surprisingly, a low-signal hemosiderin border was clearly visible in only 2 of the 11 reported cases. Fusiform low-signal intensity extending away from the lesion on both ends of an oblong lesion oriented along the spinal axis was found in 2 cases. Intramedullary spinal cord malformations do not appear to have been reported to enhance on MRI, and enhancement was found in none of the patients in the study by Turjman and co-workers25 However, MRI contrast agent was administered in only 3 of these 11 patients. The use of SPINAL VASCULAR MALFORMATIONS

contrast was not mentioned specifically in other series.30,31,34 On CT cavernous hemangiomas enhance after contrast administration.29,34 Given the histologic composition of the lesions, it is likely that contrast enhancement would be variably present on MRI if enough cases were studied. However, it may be difficult to perceive enhancement in the presence of blood products that cause high T1-weighted signal on noncontrast images. On MRI the major differential diagnostic considerations include neoplasms and intramedullary AVMs. The lack of flow voids would exclude AVMs, although a partially throm-


Fig 7. Extradural and intradural AVM. (A) Spinal angiogram and (B) axial T1-weighted MRI of the thoracic spine. The spinal angiogram shows a complex vascular lesion with both intradural (arrow) and extradural (arrowheads) components. The MRI (B) shows large flow voids both within and outside the spinal canal in the paraspinal tissues.

bosed AVM could not be excluded entirely. The mixed, heterogenous signal characteristics representing blood products of various ages as well as a rim of low signal intensity representing hemosiderin (if present) help distinguish cavernous malformations from spinal cord neoplasms. Follow-up imaging can show a change in signal intensity, reflecting the evolution of the blood products, which is another helpful differentiating feature. Several epidural cavernous angiomas have been reported. The MRI appearance of a purely epidural cavernous malformation or hemangioma can mimic that of a meningioma or nerve sheath tumor. Shin and co-workers27 described five cases, all of which had a paravertebral extension and lobulated contour. In four of the cases, the lesion partially encircled the spinal canal. Three of those four cases had a larger dorsal component. All five cases demonstrated high T2-weighted signal and marked, homogeneous enhancement.27

Hemangioblastomas Intramedullary neoplasms constitute 10 to 20% of all tumors of the spinal axis. Hemangioblastomas are rare, representing 2 to 14.8% of all intramedullary neoplasms.8,35 Although typically intramedullary, hemangioblastomas have been reported to occur in the intradural extramedullary and extradural compartments.35,36 Approximately 30% of patients with spinal cord hemangioblastomas have von Hippel-Lindau disease, and these patients may harbor more than one of these highly vascular


neoplasms. Patients with spinal cord hemangioblastomas are typically young to middle-aged males (age range, 16-40 years; mean age, 32 years; 5.5:1 male-to-female ratio). Hemangioblastomas are most often located in the cervical spinal cord followed by the thoracic spinal cord.35 On MRI, hemangioblastomas often have the classic appearance of a cyst associated with a markedly enhancing mural nodule (Fig 10). About 20 to 25% of the time, they appear as solid-enhancing nodules without a cystic component.37 Even in the absence of a tumoral cyst, most hemangiomas are associated with syringomyelia by the time of presentation. Flow voids are associated with the hemangioblastomas itself or are present on the surface of the spinal cord adjacent to the neoplasm. They represent dilated venous and arterial vessels associated with the neoplasm. On imaging the major differential diagnoses to rule out are intramedullary spinal ependymomas, spinal cord AVMs or AVFs, and hydromyelia. On MRI, ependymomas are not as well defined as encapsulated hemangioblastomas. Flow voids on noncontrast MRI and marked tumor blush and delineation of the feeding arteries and main draining vein on spinal arteriography would not be expected with ependymomas but are usually associated with hemangioblastomas.35 Unlike hemangioblastomas, spinal cord AVMs or AVFs typically have no associated cyst or syringomyelia. Hydromyelia involves the central spinal canal. The syringomyelia associated with hemangioblastomas tends to be lateral to the central aspect of the JACKSON AND PARTOVI

Fig 8. AVM of the conus medullaris. Spinal angiograms with injections from (A) right L3 and (B) left L1 vertebral levels shows an intramedullary AVM at the level of the conus medullaris. Multiple arterial feeders are seen (arrows).

Fig 9. Cavernous malformation. (A) Sagittal T2-weighted MRI of the cervical spine and (B) axial GRE through the lesion. On T2-weighted MRI, cavernous malformations exhibit mass effect and heterogenous signal and contain areas of both high and low signal intensity. The hemorrhagic products are best appreciated on GRE imaging where signal loss is greatest.

Fig 10. Thoracic hemangioblastoma. (A) Contrast-enhanced sagittal T1-weighted MRI at the midthoracic level and (B) its corresponding spinal angiogram. The MRI (A) shows a solid enhancing intramedullary lesion with associated flow voids. Angiography (B) confirms the finding. The flow voids seen on MRI represent the arterial and venous structures associated with this lesion. (Fig 10B with permission from Thieme.)

spinal cord. In addition, simple hydromyelia would not be expected to have any associated enhancement or flow voids.

Aneurysms There are case reports of false or pseudoaneurysms following trauma or surgical manipulation with damage to spinal vessels.38,39 On angiography, these lesions appear as irregular outpouchings of the parent vessel wall (Fig 11). In the absence of an associated AVM, isolated saccular aneurysms involving the spinal arteries are rare. In 1999 Kawamura and co-workers40 found that only 10 true cases of “isolated” saccular aneurysms involving the spinal arteries had been reported. They excluded cases associated with vascular malformations or those in which angiography was not performed. At presentation, the mean age of these 10 patients was 38 years. Eight patients presented with SAH from aneurysm rupture. In five cases the aneurysms were


at C1 or C2. In seven cases the anterior spinal artery was the parent vessel that gave rise to the aneurysm. Saccular aneurysms involving the posterior spinal artery have also rarely been reported. One aneurysm arising from a radicular artery coursing along the cauda equina was imaged with both MRI and conventional angiography.3,41 Aneurysms involving spinal arteries are often caused by increased blood flow or pressure related to an AVM or to aortic or vertebral artery occlusive disease. The anterior spinal artery serves as a collateral pathway that becomes dilated with increased blood flow, and the aneurysm subsequently forms.40,42 A congenital variation with an unusual origin of the anterior spinal artery off the vertebral arteries, presumably altering flow dynamics, has also been associated with an aneurysm of the anterior spinal artery.43 Spinal aneurysms can be imaged with MRI or contrast-enhanced CT but are best defined with spinal JACKSON AND PARTOVI

Fig 11. Spinal artery aneurysm. (A) Magnified anteroposterior (AP) view of a spinal angiogram shows a fusiform aneurysm (arrow) of the anterior spinal artery. (B) AP view of a spinal angiogram shows a feeding artery aneurysm (arrowhead) supplying a spinal AVM. (With permission from Lippincott Williams & Wilkins.)

angiography. Their appearance depends on their vascular origin and size.

Conclusions There are three main modalities and methods available to image spinal vascular lesions. Because of the rarity of many of these lesions, they may be overlooked in the differential diagnosis of patients with myelopathic or radicular symptoms. It is important to be familiar with the different modalities available for imaging, when to employ each modality, as well as the typical or suggestive imaging features of each type of lesion. MRI is the principle modality that should be used to evaluate a suspected spinal vascular lesion. MRI can exclude other more common causes of myelopathy and radiculopathy such as disk disease and acquired stenosis. Because the sensitivity of MRI for detecting spinal vascular lesions is high, it is a good screening test. Myelography followed by CT is a good ancillary and noninvasive test that can be used in patients unable to undergo MRI because of physical constraints (such as implanted devices inSPINAL VASCULAR MALFORMATIONS

compatible with MRI). In the case of intradural dorsal AVFs, CT myelography, although nonspecific, is very sensitive to dilated venous structures and should be performed in a patient with a high clinical suspicion but inconclusive or negative MRI findings. In the absence of any suggestive features of a spinal vascular lesion on MRI or CT myelography, spinal angiography is unlikely to uncover a lesion that is occult on the other imaging modalities. Spinal angiography serves to define the feeding vessels, collateral supply, normal spinal vascular anatomy, and venous drainage of lesions when discovered on MRI or CT myelography. Newer angiographic MRI techniques can help to focus angiography by predicting the site of the lesion and feeding vessels, thus limiting the patient’s exposure to radiation, contrast load, and procedural time. There seems to be little benefit to MRA of the spine if no abnormality is found on conventional MRI (with and without gadolinium). Many entities, such as neoplasms, demyelinating diseases, hydromyelia, infarcts, or trauma, must be distinguished from a vascular lesion when an abnormality is detected on MRI or CT myelography. By understanding the pathophysiology and the


characteristic imaging features of the spinal vascular lesions described in this article, clinicians will better distinguish among the diagnostic possibilities. The differential diagnosis can then be limited, which can further focus an angiographic study (except in the case of spinal cavernous malformations, which are angiographically occult). The most important thing to remember is that even though rare, many spinal vascular lesions are treatable causes of radiculopathy and myelopathy. They should always be considered in a patient with otherwise unexplained symptoms. If a vascular lesion is entertained in the differential diagnosis, available imaging modalities will increase the probability of obtaining an accurate diagnosis.

References 1. Atkinson JLD, Miller GM, Krauss WE, et al: Clinical and radiographic features of dural arteriovenous fistula, a treatable cause of myelopathy. Mayo Clin Proc 76:1120-1130, 2001. 2. Anson JA, Spetzler RF: Classification of spinal arteriovenous malformations and implications for treatment. BNI Quarterly 8:2-8, 1992. 3. Spetzler RF, Detwiler PW, Riina HA, et al: Modified classification of spinal cord vascular lesions. J Neurosurg 96:145-156, 2002. 4. Bemporad JA, Sze G: Magnetic resonance imaging of spinal cord vascular malformations with an emphasis on the cervical spine. Neuroimaging Clin N Am 11:111-129, 2001. 5. Gilbertson JR, Miller GM, Goldman MS, et al: Spinal dural arteriovenous fistulas: MR and myelographic findings. AJNR Am J Neuroradiol 16:2049-2057, 1995. 6. Saraf-Lavi E, Bowen BC, Quencer RM, et al: Detection of spinal dural arteriovenous fistulae with MR imaging and contrast-enhanced MR angiography: Sensitivity, specificity, and prediction of vertebral level. AJNR Am J Neuroradiol 23:858-867, 2002. 7. Mascalchi M, Bianchi MC, Quilici N, et al: MR angiography of spinal vascular malformations. AJNR Am J Neuroradiol 16:289-297, 1995. 8. Nelson PK, Setton A, Berenstein A: Vertebrospinal angiography in the evaluation of vertebral and spinal cord disease. Neuroimaging Clin N Am 6:589-605, 1996. 9. Champlin AM, Rael J, Benzel EC, et al: Preoperative spinal angiography for lateral extracavitary approach to thoracic and lumbar spine. AJNR Am J Neuroradiol 15:73-77, 1994. 10. Forbes G, Nichols DA, Jack CR Jr, et al: Complications of spinal cord arteriography: Prospective assessment of risk for diagnostic procedures. Radiology 169:479-484, 1988. 11. Hurst RW, Grossman RI: Peripheral spinal cord hypointensity on T2-weighted MR images: A reliable imaging sign of venous hypertensive myelopathy. AJNR Am J Neuroradiol 21:781-786, 2000. 12. Bowen BC, Fraser K, Kochan JP, et al: Spinal dural arteriovenous fistulas: Evaluation with MR angiography. AJNR Am J Neuroradiol 16:2029-2043, 1995. 13. Niimi Y, Setton A, Berenstein A: Spinal dural arteriovenous fistulae draining to the anterior spinal vein: Angiographic diagnosis. Neurosurgery 44:999-1003, 1999. 14. Willinsky RA, terBrugge K, Montanera W, et al: Posttreatment MR findings in spinal dural arteriovenous malformations. AJNR Am J Neuroradiol 16:2063-2071, 1995. 15. Kahara V, Lehto U, Ryymin P, et al: Vertebral epidural arteriovenous fistula and radicular pain in neurofibromatosis type I. Acta Neurochir (Wien) 144:493-496, 2002. 16. Asai J, Hayashi T, Fujimoto T, et al: Exclusively epidural arteriovenous fistula in the cervical spine with spinal cord symptoms: Case report. Neurosurgery 48:1372-1376, 2001. 17. Inci S, Bertan V, Cila A: Angiographically occult epidural arteriovenous fistula of the craniocervical junction. Surg Neurol 57:167-173, 2002.


18. Willinsky R, terBrugge K, Montanera W, et al: Spinal epidural arteriovenous fistulas: Arterial and venous approaches to embolization. AJNR Am J Neuroradiol 14:812-817, 1993. 19. Coric D, Branch CL Jr., Wilson JA, et al: Arteriovenous fistula as a complication of C1-2 transarticular screw fixation: Case report and review of the literature. J Neurosurg 85:340-343, 1996. 20. Arnaud O, Bille F, Pouget J, et al: Epidural arteriovenous fistula with perimedullary venous drainage: Case report. Neuroradiology 36:490491, 1994. 21. Pirouzmand F, Wallace MC, Willinsky R: Spinal epidural arteriovenous fistula with intramedullary reflux: Case report. J Neurosurg 87:633-635, 1997. 22. Cognard C, Semaan H, Bakchine S, et al: Paraspinal arteriovenous fistula with perimedullary venous drainage. AJNR Am J Neuroradiol 16:2044-2048, 1995. 23. Hasegawa M, Fujisawa H, Kawamura T, et al: The efficacy of CT arteriography for spinal arteriovenous fistula surgery: Technical note. Neuroradiology 41:915-919, 1999. 24. Bowen BC, Pattany PM: Spine MR angiography. Clin Neurosci 4:165-173, 1997. 25. Turjman F, Joly D, Monnet O, et al: MRI of intramedullary cavernous haemangiomas. Neuroradiology 37:297-302, 1995. 26. Zevgaridis D, Medele RJ, Hamburger C, et al: Cavernous haemangiomas of the spinal cord: A review of 117 cases. Acta Neurochir (Wien) 141:237-245, 1999. 27. Shin JH, Lee HK, Rhim SC, et al: Spinal epidural cavernous hemangioma: MR findings. J Comput Assist Tomogr 25:257-261, 2001. 28. Rao GP, Bhaskar G, Hemaratnan A, et al: Spinal intradural extramedullary cavernous angiomas: Report of four cases and review of the literature. Br J Neurosurg 11:228-232, 1997. 29. Ogilvy CS, Louis DN, Ojemann RG: Intramedullary cavernous angiomas of the spinal cord: Clinical presentation, pathological features, and surgical management. Neurosurgery 31:219-230, 1992. 30. Fontaine S, Melanson D, Cosgrove GR, et al: Cavernous hemangiomas of the spinal cord: MR imaging. Radiology 166:839-841, 1988. 31. McCormick PC, Michelsen WJ, Post KD, et al: Cavernous malformations of the spinal cord. Neurosurgery 23:459-463, 1988. 32. Naim-Ur Rahman, Jamjoom A, al-Rayess M: Intramedullary ossified cavernous angioma of the spinal cord: Case report. Br J Neurosurg 12:267-270, 1998. 33. Tyndel FJ, Bilbao JM, Hudson AR, et al: Hemangioma calcificans of the spinal cord. Can J Neurol Sci 12:321-322, 1985. 34. Cosgrove GR, Bertrand G, Fontaine S, et al: Cavernous angiomas of the spinal cord. J Neurosurg 68:31-36, 1988. 35. Xu QW, Bao WM, Mao RL, et al: Magnetic resonance imaging and microsurgical treatment of intramedullary hemangioblastoma of the spinal cord. Neurosurgery 35:671-675, 1994. 36. Neumann HP, Eggert HR, Weigel K, et al: Hemangioblastomas of the central nervous system: A 10-year study with special reference to von Hippel-Lindau syndrome. J Neurosurg 70:24-30, 1989. 37. Rebner M, Gebarski SS: Magnetic resonance imaging of spinal-cord hemangioblastoma. AJNR Am J Neuroradiol 6:287-289, 1985. 38. Chan KT, Korivi N: Lumbar artery pseudoaneurysm in traumatic spinal cord injury: A case report. Arch Phys Med Rehab 84:455-457, 2003. 39. Prabhu VC, France JC, Voelker JL, et al: Vertebral artery pseudoaneurysm complicating posterior C1-2 transarticular screw fixation: Case report. Surg Neurol 55:29-34, 2001. 40. Kawamura S, Yoshida T, Nonoyama Y, et al: Ruptured anterior spinal artery aneurysm: A case report. Surg Neurol 51:608-612, 1999. 41. Goto Y, Kamijyo Y, Yonekawa Y, et al: Ruptured aneurysm of the posterior spinal artery of the upper cervical spinal cord: Case report. Neurosurgery 22:558-560, 1988. 42. Kito K, Kobayashi N, Mori N, et al: Ruptured aneurysm of the anterior spinal artery associated with pseudoxanthoma elasticum: Case report. J Neurosurg 58:126-128, 1983. 43. Yonas H, Patre S, White RJ: Anterior spinal artery aneurysm: Case report. J Neurosurg 53:570-573, 1980.