Introduction

Intracranial vertebral artery dissecting aneurysms (VADAs) are associated with acute ischemic stroke, subarachnoid hemorrhage, and other neurological symptoms [1]. A ruptured VADA is often followed by rebleeding, resulting in severe clinical outcomes in the acute stage [2, 3]. Therefore, early intervention is required to prevent rerupture of the VADA. Endovascular trapping of a ruptured VADA is a widely accepted treatment in obtaining reliable hemostasis by protecting the ruptured site from both collateral and antegrade blood flow [4, 5]. However, endovascular trapping of the VADA carries a certain risk of medullary infarction [6,7,8,9] (Fig. 1). The ischemia in such cases has been considered to be caused by coiling and/or thrombosis of the orifice of unrecognized perforating arteries. However, these fine branches are not readily depicted on conventional vertebral angiography. Recent advances in angiographic technology have allowed us to evaluate the anatomy of the perforating arteries [10, 11].

Fig. 1
figure 1

Illustrative case with vertebral artery dissecting aneurysm (VADA) who presented medullary infarction after internal trapping. a Volume rendering image shows VADA at distal segment of the right vertebral artery (VA) (arrow). The right posterior inferior cerebellar artery (PICA) was originated from the distal side of VA (arrowheads). b Internal trapping using detachable microcoils was performed. Right vertebral arteriogram shows disappearance of VADA and stagnation of contrast material in proximal segment of right VA. c Contralateral vertebral arteriogram showed no retrograde filling of VADA and good patency of right PICA and basilar trunk. d Diffusion-weighted image 2 days after treatment revealed acute infarction in right half of medulla oblongata (arrow)

Full size image

The microanatomical features of perforating arteries arising from the cerebral artery have been previously evaluated by gross anatomical methods [12,13,14]. According to these reports, the vertebral artery (VA) gives off several perforating branches at the intracranial segment. However, the anatomical features of these perforating arteries have not been evaluated using radiological methods. In this study, the detectability and anatomical variations of perforating arteries arising from the VA were evaluated using reconstructed images of three-dimensional (3D) rotational vertebral angiography.

Materials and methods

Patients

This retrospective study was approved by the Institutional Review Board of Kurume University Hospital. Written informed consent was obtained from all patients before angiography; however, the need for informed consent for this study was waived because of the retrospective study design. From January 2016 to December 2017, 167 consecutive patients underwent 3D rotational vertebral angiography at two different major stroke centers equipped with the same angiographic machine and workstation. Among them, 47 patients were excluded from this study because of the following factors, which prevented adequate evaluation of the normal angiographic anatomy: insufficient field of view on rotational angiography (26 patients), inappropriate image quality (3 patients), VA truncal lesions such as VA dissections or VA aneurysms (7 patients), arteriovenous fistulas involving VA perforators (9 patients), and brain tumors supplied by VA perforators (2 patients). The inappropriate image quality was due to the patient motion artifact (2 patients) and massive subarachnoid hematoma surrounding the medulla oblongata (1 patient). The remaining 120 patients (87 females, 33 males; age range, 11–83 years; mean, 57.9 years) were retrospectively evaluated. All 120 patients had undergone one-side vertebral arteriography; thus 120 vertebral arteries in total were analyzed. The underlying disease was an unruptured intracranial non-VA aneurysm in 57 patients, subarachnoid hemorrhage from a non-VA aneurysm or unknown etiology in 28, arteriovenous malformation in 8, intracranial dural arteriovenous fistula in 3, and other condition in 2.

Angiography

All angiography examinations were performed in an angiographic suite equipped with a biplane flat-panel detector system (Innova IGS 630; GE Healthcare, Milwaukee, WI, USA) in the two hospitals. The 4- or 5F-sized catheter was positioned at the proximal segment of the Vas, and non-ionic iodinated contrast material was injected through an automated injector. The imaging parameters for rotational angiography were set as follows: rotation angle, 200°; rotation speed, 40°/s; matrix size, 512 × 512 by 20-cm flat-panel detector; and non-ionic iodinated contrast material (Iopamidol 300) with a flow rate of 1.5 to 3.0 ml/s (total volume of 9–24 ml). The low flow rate injection was attempted for the non-dominant type VA having a small diameter. The raw data obtained from the biplane and rotational vertebral angiography were stored in the picture archiving and communication system (PACS) server in each hospital. The data were re-transferred to the workstation (Advantage Workstation VolumeShare; GE Healthcare) to compose the volume rendering reconstruction and multiplanar reconstruction with settings of 0.5- to 5.0-mm thickness and 1-mm interval.

Image interpretation

All angiographic and multiplanar reconstruction images were simultaneously evaluated by experienced neuroradiologist and endovascular neurosurgeon with 22 and 12 years of clinical experience to reach a consensus regarding the presence and anatomical configuration of the VA and perforating arteries. The baseline anatomical characteristics of the VA were examined with a particular focus on the dominancy of the VA in terms of its diameter and then classified into the dominant-side VA, the non-dominant-side, and equally sized VA. The existence of the origin of the anterior spinal artery (ASA), existence of the origin of the posterior inferior cerebellar artery (PICA), and distance from the VA union to the origin of the PICA were also evaluated. The dominancy of VA and the origin of ASA were evaluated based on the information from the retrograde filling of the contrast agent into the distal segment of the contralateral VA during the rotational vertebral arteriography.

The anatomical configuration of the perforating arteries was evaluated with a focus on the numbers and types of detectable perforators. The numbers of perforators were recorded in each patient. In addition, the relationships between the number of perforators and the baseline anatomy of the VA and its branches, including the origins of the anterior spinal artery (ASA) and the posterior inferior cerebellar artery (PICA), were evaluated. The types of perforating arteries were classified into terminal or longitudinal types depending on their course and into ventral, dorsal, or dorsolateral types depending on the location of their termination. The terminal type was defined as a perforator terminating at the same level of the medullary surface showing a short trunk, and the longitudinal type was defined as a perforator running longitudinally along the medullary surface (Fig. 2a). When the longitudinal type had multiple origins at the VA or other branches and these arterial roots join to be one longitudinal type perforator at the medullar surface, the numbers and locations of the origins were also recorded. The ventral type was defined as a perforator terminating at the ventral aspect of the medulla, the lateral type as one terminating at the lateral aspect, and the dorsolateral type as one terminating at the dorsolateral aspect (Fig. 2b). The distance between the VA union and the origin of the PICA or perforating arteries was also measured using the measurement tool on the workstation.

Fig. 2
figure 2

Schematic drawings of anatomical configurations of perforating arteries (left anterolateral view of the medulla oblongata). a The terminal type was defined as a perforator terminating at the same level of the medullary surface, showing short trunk (arrow), and the longitudinal type was defined as a perforator running longitudinally along the medullary surface (double arrows). b The ventral type was defined as a perforator terminating at the ventral aspect of the medulla (arrow), the lateral type as one terminating at the lateral aspect (double arrows), and the dorsolateral type as one terminating at the dorsolateral aspect (triple arrows). AICA, anterior inferior cerebellar artery; BA, basilar artery; VA, vertebral artery; PICA, posterior inferior cerebellar artery; ASA, anterior spinal artery

Full size image

Statistical analysis

Statistical analysis was performed using the SPSS statistics version 21 (IBM, Armonk, NY, USA). The relationships between each type of VA and the numbers of detectable perforating arteries were statistically evaluated (Fisher’s exact test). In addition, the differences among the types of perforating arteries were evaluated by comparing the distance between the VA and perforating arteries in each type (Mann–Whitney U test). A p-value of 0.05 or lower was considered to be statistically significant.

Results

The baseline anatomical characteristics of the VA were as follows: the dominant-side VA was present in 66 patients, the non-dominant-side VA was present in 10, and equally sized VAs were present in 44 patients. The ASA was given off from the ipsilateral VA in 69 patients, the contralateral VA in 31, and the bilateral VAs (double origin) in 11. The ASA was not detectable in the remaining nine patients. Therefore, 80 VAs served as the origin of the ASA and 40 VAs did not. The PICA originated from the intracranial VA in 87 patients, the extracranial VA in 26, and the anterior inferior cerebellar artery (AICA) in 7 (i.e., non–PICA-type VA) (Table 1).

Table 1 Baseline anatomical characteristics of the VA
Full size table

Zero, 1, 2, 3, 4, and 6 perforators were detected in 2, 51, 56, 9, 1, and 1 patient, respectively. Therefore, a total of 200 perforating arteries were detected (median, 2; interquartile range [IQR], 1–3). Among them, 146 perforating arteries were of the terminal type (Fig. 3a) and 54 were of the longitudinal type (Fig. 3b). In addition, 32 perforating arteries were of the ventral type, 151 were of the lateral type, and 17 were of the dorsolateral type (Fig. 4). All ventral type perforating arteries were of the terminal type. However, the lateral type perforating arteries comprised both the terminal type (71.5%) and longitudinal type (28.5%). Dorsolateral type perforating arteries were classified into the terminal type (35%) and longitudinal type (65%) (Fig. 5).

Fig. 3
figure 3

Illustrative cases of detected perforating arteries. a A 60-year-old woman with a basilar top aneurysm. Axial reconstruction of a partial maximum intensity projection (MIP) image of rotational right vertebral angiography shows a perforating artery (arrow). This artery was classified as lateral and terminal type. b A 50-year-old man with a brain tumor. Coronal reconstruction of a partial MIP image of rotational right vertebral angiography shows a perforating artery (arrow). This artery was classified as lateral and longitudinal type. This artery rostrally anastomosed with a branch of the posterior inferior cerebellar artery (arrowhead). c Axial reconstruction image of identical rotational angiography shows same perforating artery running along the lateral surface of upper cervical spinal cord (arrow). Note the anterior spinal artery indicating anteromedian border of spinal cord (arrowhead). The ventral, lateral, and dorsolateral borders of the spinal cord and medulla could be assumed by positional relationships of these surrounding vessels

Full size image
Fig. 4
figure 4

Cumulative numbers of each type of all 200 perforating arteries. a Termination. b Course

Full size image
Fig. 5
figure 5

Ratios of termination type in each course type

Full size image

The left VA gave off zero to six perforators (median, 2 [IQR, 1–2]), and the right VA gave off one to four perforators (median, 2 [IQR, 1–2]) (Table 2). There was no significant difference in the number of detected perforating arteries according to the laterality of the VA (p = 1.0000). The dominant-side VA gave off zero to three perforating arteries (median, 2 [IQR, 1–2]), the non-dominant-side VA gave off zero to six perforating arteries (median, 2 [IQR, 1–2]), and the equal type VAs gave off one to three perforating arteries (median, 2 [IQR, 1–2]) (Table 2). There was no significant difference in the dominancy of the VA (p = 0.6046). The VAs containing the origin of the ASA gave off zero to three perforating arteries (median, 2 [IQR, 1–2]), whereas the VAs without the origin of the ASA gave off zero to six perforating arteries (median, 2 [IQR, 1–2]) (Table 2). There was no significant difference between the VAs with and without the origin of the ASA (p = 1.0000). Finally, the VAs with the PICA originating from the intracranial segment gave off zero to three perforating arteries (median, 1 [IQR, 1–2]), the VAs with the PICA originating from the extracranial segment gave off one or two perforating arteries (median, 1 [IQR, 1–2]), and the VAs without the origin of the PICA (i.e., the PICA originated from the AICA) gave off one to six perforating arteries (median, 2 [IQR, 2–3]). The VAs without the origin of the PICA showed a significantly higher number of branching perforating arteries than those with the origin of the PICA (p < 0.001).

Table 2 Relationships between VA branch types and numbers of detectable perforators
Full size table

The longitudinal type perforating arteries had a single origin in 13 patients, double origins in 37, and triple origins in 4. The cumulative numbers of origins according to location were 56 from the VA, 39 from the PICA, and 4 from the C1 radicular artery.

The distance from the VA union to the origins of all perforating arteries ranged from 1.3 to 57.5 mm. The distance between the VA union and the origins of ventral type perforating arteries ranged from 1.3 to 57.5 mm (median, 5.1 mm [IQR, 5.6–10.375]), that between the VA union to the origins of lateral type perforating arteries ranged from 7.4 to 48.3 mm (median, 22.6 mm [IQR, 16.6–34.5]), and that between the VA union to the origins of dorsolateral type perforating arteries ranged from 33.2 to 48.0 mm (median, 41.1 mm [IQR, 33.2–46.8]). All distances between the VA union and each type of perforating artery showed a significant difference (Fig. 6).

Fig. 6
figure 6

Distance from the vertebral artery union to the origins of all perforating arteries

Full size image

Discussion

Endovascular treatment for a ruptured VADA is traditionally performed to trap the dissected segment of the VA while avoiding occlusion of the origins of the PICA and ASA. The risk of medullary infarction due to occlusion of the perforating arteries has been previously discussed [6, 9, 15, 16]. According to these previous reports, the incidence of medullary infarction ranges from 19 to 47%. However, Endo et al. indicated that it is difficult to recognize ischemic complications in the acute stage of subarachnoid hemorrhage [6]. The ischemic tolerance of the perforating arteries is also difficult to assess by the balloon occlusion test. Therefore, clinicians must have adequate knowledge of the normal anatomy of the perforating arteries arising from the VA and the ability to detect them using conventional and 3D angiography. Recently, the flow-diverting stent placement has been reported as an alternative treatment technique to preserve the vertebral arterial flow as well as its branches [17,18,19]. The detectability of perforators might be an important information for the consideration of the flow-diverting stent placement.

The anatomy of the perforating arteries has been evaluated using gross anatomical methods [13, 14, 20,21,22]. However, the perforators from the VA have not been fully discussed. Marinkovic et al. evaluated the perforating arteries from the VA and its side twigs [13]. In their study, the perforators that were given off from the VAs ranged in number from 1 to 3 (mean, 1.2) and in diameter from 210 to 520 μm. In the present series, the median number of detectable perforating arteries in all VAs was 2, which is consistent with the abovementioned anatomical study. In addition, only three (1.8%) patients were excluded from the original 167 cohort due to the inadequate image quality; the two of them was due to the patient motion artifact, and the remaining one was due to the massive hematoma. Our results indicated that the reconstructed images from the 3D angiography demonstrated acceptable imaging quality for the detection of the VA perforating arteries. As the advantages of the radiological evaluation, the overall morphology of the arteries can be evaluated. The morphological characteristics could be well delineated and originally classified into terminal or longitudinal types in relation to their course and into ventral, dorsal, or dorsolateral types in relation to the location of their termination.

Regarding the relationship between the anatomical type of VA and the number of detectable perforating arteries, non–PICA-type VAs (i.e., the PICA arose from the AICA) gave off significantly higher numbers of perforating arteries than other types of VAs. The previous anatomical evaluations reported ASA itself and the proximal segment of PICA constantly gives off perforating arteries [13, 20]; however the anatomy of perforators from VA in relation to the origins of ASA and PICA was not fully discussed previously. Our result indicates that the endovascular trapping of non–PICA-type VAs might be associated with a risk of perforating artery occlusion and supports the higher incidence of ischemic complications in previous reports of trapping of this type of VA [6, 8]. Special attention should be paid to the existence of perforating arteries in the treatment of VADAs in non–PICA-type VAs. In such cases, the reconstructive techniques including the stent-assisted coil embolization and the flow-diverter stent placement might be alternative techniques to preserve the blood flow of the branches as well as the VAs [17,18,19, 23,24,25,26,27].

The detected perforating arteries were classified into three types (ventral, lateral, and dorsolateral) depending on their terminations and into two types (terminal and longitudinal) depending on their course (Figs. 1 and 2). A longitudinal type perforating artery originated from the PICA and/or the proximal segment of the intradural VA might correspond to the LSA in terms of its course. Lasjaunias et al. defined the LSA as originating from either the PICA or the intradural VA and running along the lateral aspect of the medulla posterior to the dentate ligaments and anterior to the posterior spinal nerve roots [28]. These authors also reported that the LSA rostrally anastomosed with the PICA branches and showed variations of the connections with not only the intracranial VA but also various origins of the PICA, including the extracranial VA and radicular arteries [28]. In the present study, most of the longitudinal type perforating arteries showed multiple origins at the PICA, VA, and C1 radicular artery. This result supports the idea that this type of perforating artery corresponds to the LSA. Regarding the terminations, all ventral types were terminal type perforators; in contrast, most of the dorsolateral types were longitudinal type perforators. This result clearly indicates the posterolateral territory of the blood supply from the LSA.

The distances between the VA union and the origins of the perforating arteries were significantly different for the ventral type, lateral type, and dorsolateral type. The ventral type perforators originated from the VA close to the union; in contrast, the dorsolateral type perforators originated from the VA distant from the union. The lateral type perforators originated from any segment of the VA. According to these results, impairment of blood flow in the perforators close to the VA union causes ischemia on the ventral (medial) side of the medulla, whereas impairment of blood flow in the perforators arising from the proximal side of the VA causes an infarction on the dorsolateral side.

The anatomical characteristics of the perforating arteries have been evaluated by radiological methods in a few studies. Lescher et al. evaluated the detailed imaging anatomy of basilar artery perforators and found that 3D DSA was able to display the perforating branches from the basilar artery [11]. Shimada et al. reported higher visualization of the perforating arteries in patients with VADAs using high-resolution cone-beam computed tomography (VasoCT; Philips Healthcare, Best, Netherlands) than conventional 3D DSA [29]. These recently developed angiographic imaging modalities allow us to reliably visualize tiny perforating arteries and evaluate their anatomy.

Our study had three main limitations. First, our study design was retrospective and involved only two centers. However, the number of patients was sufficient to discuss the anatomy in comparison with gross anatomical methods. Second, the reconstructed images did not allow for visualization of the brain parenchyma itself; therefore, it was difficult to evaluate the absolute anatomical relationships between the perforating arteries and medulla. Further evaluation using the reconstructed images combined with the multimodality fusion technique should be considered. Third, we did not have a gold standard, the gross anatomical information, to compare with the imaging anatomical descriptions.

Conclusion

Non–PICA-type VAs, in which the PICA arose from the AICA, gave off significantly higher numbers of perforating arteries. The anatomical configuration of the perforating arteries could be well delineated and classified according to their courses and locations of the terminations. The currently available angiographic systems and reconstructed images can facilitate visualization of small perforating arteries from the VAs.