Spinal Arterial Aneurysm

Yes, they happen.  Very rarely, but probably more than we know, since the only practical way to make a diagnosis of one is (still) to have a catheter spinal angiogram, and not every patient with spinal hemorrhage gets one.  I believe the vast majority of isolated spinal aneurysms to be hemorrhagic dissections (ruptured dissecting aneurysms).  They have nothing in common with saccular aneurysms of the brain, except that both bleed.

Spinal aneurysm is an aneurysm arising from any artery inside the dural sac.  Our experience, as well review of literature, suggests that most of these aneurysms arise from the proximal intradural radiculomedullary or radiculopial arteries, and only rarely from the anterior or posterior spinal arteries. Their morphology is often fusiform and, in the minority of cases where they appear to be “saccular”, there is a feeling that these are pseudoaneurysms — meaning holes in the artery with a surrounding hematoma holding back the flood.  All of these considerations suggest, to me, a dissecting etiology — the vessel is torn at the point where it traverses the dura, and the dissection propagates into the intradural (subarachnoid) segment of the vessel.  If the dissection ruptures, the result is spinal subarachnoid hemorrhage, which is how these aneurysms present.  If, however, the dissection thromboses, the outcome will depend on the availability of collateral support.  Closure of a dominant radiculomedullary artery (Adamkiewicz, for example) with no effective collaterals will result in an anterior spinal infarct syndrome (acute paraplegia, bowel, bladder, sexual dysfunction, pain and temperature sensory loss, sparing proprioseption and vibration).  If effective radiculomedullary collaterals exist, the patient will likely do well.

The aneurysms herein discussed are considered as “isolated” — meaning that no other vascular abnormalities are found.  There are, of course, plenty of spinal arterial and venous aneurysms encountered as part of spinal vascular shunts, such as spinal arteriovenous fistulas and spinal arteriovenous malformations.  It is possible that some spinal aneurysms, particularly involving the anterior or posterior spinal arteries, may in fact be associated with other lesions such as spinal AVMs.  The proposed scenario is that the AVM or fistula ruptures, with resulting hemorrhage compressing or obliterating the bleeding lesion, leaving only the associated aneurysm visible on angiography.  This, in my opinion, is a rather unlikely scenario, for several reasons. First, the lesion has to have bled rather than the aneurysm itself.  Second, most spinal shunts remain visible even after hemorrhage. Third, the vessels feeding spinal shunts are usually quite large and abnormal-looking in other ways, which should trigger suspicion even if the original shunt is no longer visible and “only” the aneurysm remains.

In summary, we angiogram most patients with spinal hemorrhage.  An abnormal-appearing spinal artery (fusiform dilatation, most commonly), especially if matched to the epicenter of the hematoma, is likely an aneurysm (probably dissecting or pseudoaneurysm).


Embolization is the primary method of attack.  For the radiculomedullary artery, embolization is only feasible if collateral reconstitution of the anterior spinal axis can be demonstrated.  This means performing a complete spinal angiogram to identify additional radiculomedullary contributors, and injecting those levels while temporarily occluding the radiculomedullary artery giving rise to the aneurysm.  These arteries are too small to accommodate balloons, so a “balloon test occlusion” is not likely to help.  It is pointless and misleading to inflate the balloon in the proximal segmental artery, because of the rich collateral potential in the spine.  The radiculomedullary artery will be reconstituted  by via contralateral epidural arterial network or via the paraspinal anastomoses (see Spinal Arterial Anatomy section for illustrations of such collaterals).  The test occlusion has to involve the radiculomedullary artery itself.  This is best accomplished by “wedging” a microcatheter into it (gently advancing a microcatheter into the radiculomedullary artery until it is wedged there and cannot be advanced any more).  Even without wedging (if one is worried about dissection), a microcatheter of size comparable to the diameter of the radiculomedullary artery should result in significant decrease in flow through the vessel. With the catheter in this position, the neighboring radiculomedullary artery is injected to see if there is sufficient anterior spinal artery reconstitution.

If the patient passes test occlusion, the aneurysm can be embolized with either coils or liquid embolics.  This is one instance where we prefer Onyx to nBCA in the spine.  We feel that the risk of n-BCA spilling into the anterior spinal artery, however good our catheter position might be, is simply too great and consequences too severe.  Even if Onyx does not reach the aneurysm, or does not seal it entirely, substantial diminution of forward flow though the radiculomedullary artery is very helpful.  The remaining radiculomedullary artery is likely to thrombose, as no retrograde runoff vessel remains to keep it open.  The same is true for coil embolization proximal to the aneurysm, provided that coils are placed into the radiculomedullary artery itself and not so proximally that collateral radiculomedullary reconstitution is possible.

Aneurysms of the anterior spinal artery itself do not appear to have a feasible endovascular solution.  Spetzler et. al. describes wrapping one such aneurysm, with good results.

Recovery depends entirely on extent of damage caused by index hemorrhage.

Case Illustration

Case courtesy of my mentor Dr. Tibor Becske of NYU Langone Medical Center; unearthed from archives by Dr. Eytan Raz

This middle-aged man presented with acute back pain, followed quickly by lower extremity paresis, sensory level, and incontinence — typical spinal hemorrhage presentation.  Notably, nonhemorrhagic spinal artery infactions present in the same manner — onset is usually painful, even when there is no bleeding — unlike most cerebral ischemic infarcts.

MRI of the thoracic spine shows an intradural, extramedullary hematoma (orange arrows), which is best seen when outlined by a normal ventral spinal vein (light blue) on post-contrast T1-weighted images.

Injection of the right T10 segmental artery identifies a typical fusiform aneurysm (red) of the radiculomedullary artery (Adamkiewicz, pink), giving rise to the anterior spinal artery (purple).  The catheter arrow is black, and dorsal division of the segmental artery is yellow

Fortunately, another sizable contribution to the anterior spinal axis was located from the right T7 segmental artery.  With the microcatheter wedged in the right T10 level proximal radiculomedullary artery (white arrow), thus obstructing its flow (test occlusion), the right T7 segmental artery is injected through a second diagnostic catheter.  You can see robust reconstituion of the entire spinal axis, including the portion below T10.  Therefore, the patient “passed” test occlusion, enabling sacrifice of the right T10 radiculomedullary artery with coils (green arrows).  Notice that coils do not reach the level of the pseudoaneurysm.  We felt this to be acceptable, as no other branches arise from the T10 radiculomedullary artery to enable its continued patency.  With no outflow, the artery and its dissecting aneurysm will thrombose.  In fact, the post-coiling image on the right no longer opacifies the dissecting pseudoaneurysm.  The patient recovered well and has not re-hemorrhaged.

CASE 2 — Adamkiewicz aneurysm

Once again, the aneurysms appear to have a predilection for the intradural portion of the radiculomedullary artery.  The reason for this, as far as i am aware of, is unknown.  The location is quite distal to the vessel’s entry tru the dura, and therefore mechanical / frictional / dissecting mechanisms by which the aneurysm would develop are hard to support.

This post-partum patient presented with back pain, followed by complete sensory, motor, and autonomic deficit below the waist.  MRI demonstrates extensive blood products in the intradural space, what seems like a dorsal epidural collection at the upper thoracic level (black arrow) and a discrete oval lesion at T11 level (white arrows) which is extramedullary.  Is it indradural or extradural?


Angiogram demonstrates a partially thrombosed fusiform pseudo=aneurysm of the Artery of Adamciewicz, beautifully (indeed!) seen in this stereo pair


Native image

The next closest radiculomedullary artery, at left T7 level, is hopelessly small.

Test occlusion was deemed impractical.  The patient was taken to the OR for decompression of intrathecal hematoma.  Recovery has been quite partial.

CASE 3 — this patient’s aneurysm required surgery

This young patient presented with multifocal extradural, extramedullary (epidural) and intradural extramedullary (subarachnoid) hemorrhages involving the cervicothoracic cord, with extensive cord signal changes that were ultimately given no better diagnosis than transverse myelitis.  Spinal angiography identified a radiculopial spinal aneurysm arising from the right T9 segmental artery, like so:

An obligatory stereo pair, so useful in the spine:

Microcatheter run, better showing the aneurysm and its off-midline relationship to the spinous processes.  Notice unusually large size of the posterior spinal artery segment.  In fact, prominent surface cord vessels, hyperemia, and arteriovenous shunting were seen throughout the cervical and thoracic spine, consistent with an acutely inflammatory process.

We felt that neither nBCA nor Onyx embolization were worth the risk; to durably chose this aneurysm the embolic must fill the aneurysm itself, and close the radiculopial artery, which is usually surprizingly well tolerated; in this case we felt that the prominent size of the PSA, and presence of major myelopathic symptoms, were too much to risk another hit.  Coils were therefore placed in the intercostal artery (to mark the level of laminectomy) and into the proximal ventral division to slow down intra-aneurysmal flow.  Our feeling was that proximal coils alone may achieve aneurysm thrombosis this case, which however did not turn out to be the case in surgery:


As suspected, the patient remains severely neurologically impaired from whatever it is we call “transverse myelitis.”

Further reading:

Lasjaunias, Berenstein, and Ter Brugge; Surgical Neuroangiography, 2nd edition, Volume II — the best reference. Note that academic institutions and other organizations may have access to full book PDF version through SpringerLink.

Ansgar BerlisKai-Michael ScheuflerChristian SchmahlSebastian RauerFriedrich Götz and Martin Schumacher.  Solitary Spinal Artery Aneurysms as a Rare Source of Spinal Subarachnoid Hemorrhage: Potential Etiology and Treatment Strategy. AJNR 2005; 26, 405-10.  Open access.

Carlos Garcia, Stella Dulcey, Jose Dulcey. Ruptured aneurysm of the spinal artery of Adamkiewicz during pregnancy. Neurology 29: 394-8; 1979  Open access