Arterial circulation of the lower spinal cord is supplied by a vascular network consisting of a solitary ventral and two dorsolateral trunks linked by sacral anastomoses [1]. Longer adaptation periods with chronic ischemia allow the formation of efficient collateral circulation, mainly through the dorsolateral spinal arteries [2]. In contrast, acute hypoperfusion, in particular of the ventrally located great anterior radiculomedullary artery, also known as the artery of Adamkiewicz, might lead to catastrophic ischemic complications resulting in paraparesis or paraplegia [3]. Numerous vascular conditions—such as aortic dissections [4], aortic aneurysms [5], and aneurysms of the artery of Adamkiewicz itself [6]—and therapeutic procedures—such as aortic replacements [7], aortic stent-graft implantations [8], and embolizations of vertebral metastases [9], arteriovenous dural fistulas [10], and spinal neoplasms [11]—as well as direct vascular traumas [12] endanger the integrity of the artery of Adamkiewicz and the subsequent perfusion of the distal spinal cord. The initial identification of the artery of Adamkiewicz, which allows subsequent sparing or reimplantation of the great anterior radiculomedullary artery, is usually realized by performing conventional spinal angiography, a technically difficult and risky procedure that is typically performed only in tertiary care medical facilities.

The ongoing development of MDCT technology has transformed CT into a truly 3D imaging technique with outstanding temporal and spatial resolution. CT angiography (CTA) allows precise timing of the contrast bolus for the visualization of millimeter-sized peripheral vasculature [13]. However, studies evaluating the delineation of the artery of Adamkiewicz in CTA data sets were able to at least partially identify the great anterior radiculomedullary artery in only 43-68% of all patients examined [3, 5, 14]; most recently a study identified at least a single artery of Adamkiewicz in up to 90% of patients in the study group [15]. Therefore, CT data acquisition technology alone does not seem to explain the lack of clear vascular delineation in this particular vessel, as experienced in other anatomic regions [13].

The spinal vasculature and the cord located in the spinal canal are surrounded by high-density skeletal formations. In addition to the vertebral column, this anatomic arrangement can be found only in the skull, with either the flattened and jointed bones of the cranium surrounding the brain and intracerebral vasculature at varying distances, or in the petrous bones, which closely encase the internal carotid arteries. Reconstruction algorithms for cerebral MDCT visualization were modified to suit this anatomic arrangement. In particular, filtering procedures, an inherent part of raw data reconstruction, on the one hand must eliminate beam hardening artifacts originating from high-density osseous formations and thereby obscuring the sharp delineation of brain tissue and cerebral vascular adjacent to the osseous skull. On the other hand, filtering procedures rely on autocalibrated filtering parameters to realize an equal level of inevitable image noise throughout the reconstructed axial image data set.

In this prospective study we sought to use 40-MDCT technology in combination with adapted brain reconstruction algorithms to visualize the spinal canal, in particular the spinal vasculature. This study was performed to test the hypothesis that contrast-enhanced MDCT angiography can depict the artery of Adamkiewicz if high-resolution image acquisition and adapted brain reconstruction techniques are used.

When the image data set reconstructed with the 90-mm cross-sectional field of view and the adapted brain kernel were used, we achieved successful depiction of the artery of Adamkiewicz in all 100 patients (100%) compared with continuous visualization of a contrast-enhanced vessel tract in only 46% of all examined patients when the 350-mm cross-sectional field of view image data set was reconstructed with a soft-tissue algorithm.

In 63% of patients, the great anterior radiculomedullary artery originated from the left side of the body, and in 37%, from the right side, with a segmental range from the eighth thoracic to the second lumbar intervertebral space (Fig. 3). The artery of Adamkiewicz measured 13.1-98.2 mm in length (mean, 40.1 ± 13.51 mm) and spanned one to three vertebral segments (Fig. 4). Segmental distribution, as shown in Table 1, laterality, and length did not show statistically significant sex-specific differences (p > 0.05).

Because the feeding posterior branch of the intercostal or lumbar artery measured 2.8 ± 0.71 mm (range, 1.4-5.6 mm) in diameter, intravascular CNR was determined at 49.3 ± 22.1 H (range, 16.1-130.3 H). The radiculomedullary artery was visualized as having a mean diameter of 1.8 ± 0.37 mm (range, 1.0-2.8 mm) and a foramen enhancement of 26.2 ± 11.1 H (range, 10.1-52.6 H). The base segment and the hairpin turn of the artery of Adamkiewicz measured 1.7 ± 0.35 mm (range, 1.0-2.5 mm) and 1.3 ± 0.31 mm (range, 0.7-1.9 mm), respectively. During its course, the great anterior radiculomedullary artery showed an intraluminal contrast of 19.3 ± 7.9 H (range, 3.9-44.4 H). The draining anterior spinal artery was 1.9 ± 0.54 mm in luminal diameter (range, 1.4-2.9 mm) and had an intraluminal contrast of 26.6 ± 10.9 H (range, 6.3-47.4 H). The multivariate statistical analysis of luminal diameter and contrast did not present significant differences according to sex, laterality, or segment of origin (p > 0.05).

Multivariate analysis evaluated the diameters of the posterior branches of the intercostal or lumbar arteries at the segment of origin of the artery of Adamkiewicz compared with the contralateral side and cranially and caudally located adjacent segments. No statistical differences in lumen diameter were apparent when comparing the contralateral segment with the upper and lower ipsilateral segments (p > 0.05), resulting in an average of 1.9 ± 0.32 mm (range, 1.1-3.1 mm), 2.0 ± 0.47 mm (1.0-3.3 mm), and 1.9 ± 0.39 mm (1.0-3.0 mm) in diameter, respectively. In contrast, the diameter of the posterior branch arising in the segment of origin of the great anterior radiculomedullary artery, with a diameter of 2.8 ± 0.71 mm (1.4-5.6 mm), showed a statistically significantly wider lumen compared with any other of the evaluated posterior branches (p < 0.0001) (Fig. 5).

Observed anomalies of the artery of Adamkiewicz were various forms of duplication, which were found in 5% of the evaluated study group. Ipsilateral duplication was observed originating from the same segmental radiculomedullary artery (Fig. 6) and from adjacent segments (Fig. 7). Bilateral duplication with segmental skipping of the feeding radiculomedullary artery was encountered (Fig. 8). Duplicated great radiculomedullary arteries were observed in two men and three women. The cranial artery of Adamkiewicz was uniformly supplied by the anterior radiculomedullary artery of the 10th thoracic vertebral segment.

The architecture of the intrinsic spinal vasculature is a complex network having unique vascular characteristics that are found nowhere else in the trunk or extremities of the human body. The central arterial blood is supplied by three sulcal arteries extending from the cervical spine to the conus medullaris. The solitary anterior spinal artery originates from the two vertebral arteries, and two posterior spinal arteries arise from the preatlantal section of the vertebral artery or from the posteroinferior cerebellar artery. Because of the longitudinal extension of this vascular system, a solitary cervical arterial supply is not capable of feeding the entire spinal cord and relies on reinforcement at various unpredictable segmental levels by the anterior and posterior radiculomedullary arteries.

Within this interconnected longitudinal anastomosing system, the blood flow is not unidirectional but can be observed to be directed cranially and caudally. Usually, radiculomedullary arteries supply a limited territory related to the nerve roots; however, not all intercostal and lumbar arteries have radiculomedullary feeders. In contrast, the great anterior radiculomedullary artery is characterized functionally by a significantly extended supplied territory and morphologically by a specific course of the ascending branch and the hairpin turn [17, 18].

The venous drainage of the spinal cord does not parallel the arterial supply and relies on radially symmetric intrinsic spinal cord veins, which in turn drain into the longitudinal median spinal cord vein before reaching the epidural plexus via the radicular veins [19]. Of importance is the extent of interconnections in this venous system and the fact that the median spinal cord vein, transforming into a radicular vein, shows the same hairpin shape as the artery of Adamkiewicz [17].

Three unique morphologic and functional vascular characteristics of the intrinsic spinal vasculature directly affect the method of visualization: the location of the spinal vasculature in the spinal canal surrounded by high-density skeletal formations; the bidirectional flow along its longitudinal course; and the extensive venous anastomoses, resembling spinal arteries in contrast enhancement and morphology. A similar anatomic arrangement and functionally small imaging window for solitary arterial vascular visualization without venous contamination can be found in the brain. Optimized interactions of selected image acquisition parameters, timing of contrast bolus application, and choice of image reconstruction algorithms allow successful visualization of the cerebral vasculature [20, 21].

The hypothesis that contrast-enhanced MDCT angiography can depict the artery of Adamkiewicz if high-resolution image acquisition and brain reconstruction techniques are used can be accepted because in all 100 patients examined the great anterior radiculomedullary artery was identified as a continuous vascular tract extending from an intercostal or lumbar artery via the radiculomedullary artery to the anterior spinal artery by its ascent to the midsagittal surface of the spinal cord and its characteristic hairpin turn.

Numerous vascular conditions, in combination with inherent deficiencies in collateral blood supply to the spinal cord and the iatrogenic exclusion of critical intercostal or lumbar arteries, might culminate in ischemia of the spinal cord. Therefore, spinal cord ischemia might present as a de novo symptom [22] as well as after surgical aortic repair [23]. However, the thoracolumbar region of the spinal cord is at increased risk for ischemia or infarction during periods of hypoperfusion not only because of the vascular supply: The cross-sectional area of gray matter has been histopathologically proven to be greatest in the thoracolumbar segment of the spinal cord, which also results in higher metabolic demands [23]. Spinal cord ischemia is a complex multifactorial event. Therefore, vascular mapping of the artery of Adamkiewicz is important in reducing the risk of de novo or postinterventional spinal cord ischemia, which presents clinically as paraplegia or paraparesis, disturbances of urination and defecation, and impairment of pain and temperature sensations.

Spinal cord angiography, performed with selective catheterization of the intercostal and lumbar arteries, has been the gold standard for four decades. However, a success rate of not more than 75% in locating the origin and course of the great anterior radiculomedullary artery and a frequency of up to 4.6% for neurologic complications have characterized this procedure as technically difficult and risky [24]. In patients in whom the artery of Adamkiewicz was not identified angiographically, an adequate collateral circulation and subsequent hypoplasia of the great anterior radiculomedullary artery were assumed [24].

MR angiography was able to depict the artery of Adamkiewicz in only 67% of patients in a recent study [14]. Investigating the vascular continuity of the aorta, intercostal artery, radiculomedullary artery, artery of Adamkiewicz, and anterior spinal artery, a continuous vessel tract was depicted in 85% in these patients compared with only 62.5% if CTA without an optimized reconstruction algorithm was used [14]. However, MR angiography is not only able to depict the collateral pathways of the artery of Adamkiewicz in patients with thoracoabdominal aneurysm, it can also supply vital information about the integrity and morphology of the spinal cord and surrounding dura [14].

In contrast, 40-MDCT technology, using a spatial resolution of 0.005 mm³ voxel volume in combination with an adapted brain reconstruction algorithm, allowed the successful temporal visualization of a sharply demarcated contrast bolus in the arterial spinal vasculature without venous contamination in all examined patients. Our study showed that millimeter-sized vasculature located in the spinal canal can be successfully delineated when image acquisition and reconstruction parameters, in combination with contrast bolus timing, are adapted to the special requirements necessitated by location, bidirectional flow, and the extensive anastomoses.

The asymmetric origin of the artery of Adamkiewicz, with a propensity to be supplied by left-sided radiculomedullary arteries and to have variations in length and diameter, was in agreement with the results of an autopsy study [25]. Prior studies have proven that intercostal or lumbar arteries at the vertebral segment of origin did not present with a significantly wider luminal diameter [23]. However, our study has shown that the further distal feeding posterior branch of the intercostal or lumbar arteries indeed is characterized by a significantly wider diameter than its ipsilateral and contralateral neighbors. Therefore, identification of origin and hairpin using the described methods of multiplanar reformation and vascular assessment along the diameter could accelerate the identification of the artery of Adamkiewicz.

Duplications of the artery of Adamkiewicz presented with variations of ipsilateral and bilateral origins, thereby emphasizing the importance of extensive mapping of the entire spinal cord blood supply to further reduce the frequency of neurologic complications.

Our study had limitations that must be addressed. First, our study was not validated against any imaging gold standard. The currently accepted gold standard, spinal angiography, is known to be a technically difficult and risky procedure compared with CTA. Therefore, definitive validation of arterial versus venous contrast enhancement cannot be provided by MDCT angiography. Second, the visualized longitudinal scanning volume from the eighth thoracic to the third lumbar vertebrae prohibited visualization of the duplicated great anterior radiculomedullary arteries at levels above and below those segments [24, 26]. Third, MDCT cannot show the spinal cord and the surrounding dura to the extent that MRI can for depicting its integrity and morphology. Furthermore, MRI allows precise definition of the imaged anatomic compartments and their characteristic vasculature. Finally, significant radiation exposure due to high-resolution MDCT cannot be avoided, even though adaptive dose modulation techniques were used.

In conclusion, contrast-enhanced CTA can depict the artery of Adamkiewicz and its anatomic variants using 40-MDCT technology in combination with an adapted brain reconstruction algorithm.

Address correspondence to D. T. Boll ([email protected]).