Vertebrobasilar Artery Geometry and Basilar Artery Plaques: A High-Resolution Magnetic Resonance Imaging Study

Atherosclerotic plaques are often present in regions with complicated ow patterns. Vascular morphology plays a role in hemodynamics. In this study, we investigate the relationship between the geometry of the vertebrobasilar artery system and the basilar artery (BA) plaque prevalence. We enrolled 290 patients with posterior circulation ischemic stroke. We distinguished four congurations of the vertebrobasilar artery: Walking, Tuning Fork, Lambda, and No Conuence. The diameter of the vertebral artery (VA) and the number of bends in the intracranial VA segment was assessed using three-dimensional time-of-ight magnetic resonance angiography. We differentiated between multi-bending ( ≥ 3 bends) and oligo-bending (< 3 bends) VAs. High-resolution magnetic resonance imaging was used to evaluate BA plaques. Logistic regression models examined the relationship between the geometry type and BA plaque prevalence.


Background
Intracranial atherosclerosis is an important cause of stroke worldwide [1] and accounts for almost 33-50% of ischemic strokes in the Chinese population [2]. Posterior circulation strokes account for about 20-30% of all ischemic strokes [3,4], and vertebrobasilar atherosclerosis is a common cause of posterior circulation ischemic strokes.
Many conventional imaging modalities exist for identifying luminal stenosis, including digital subtraction angiography, computed tomography angiography, and three-dimensional time-of-ight magnetic resonance angiography (3D-TOF-MRA). However, identifying the morphology and composition of plaques has been shown to provide an incremental bene t over luminal stenosis alone when the aim is to de ne vulnerable lesions [5,6] and predict subsequent cardiovascular ischemic events [7,8], both in the coronary and carotid circulation. High-resolution magnetic resonance imaging (HRMRI) is a novel noninvasive technique to examine blood vessels and plaques, speci cally cerebral vessel wall plaques [9].
Several recent studies have con rmed the feasibility of using HRMRI to evaluate intracranial arterial walls [9,10]. However, most studies mainly focused on the relationship between plaque distribution, plaque enhancement, artery remodeling patterns, intraplaque hemorrhage, and clinical ischemic events [11][12][13][14]. There is limited research that investigates the association between the vascular geometry and the plaque prevalence in posterior circulation based on magnetic resonance imaging.
Atherosclerosis is not only associated with systemic risk factors such as hypertension, smoking, hyperlipidemia, and diabetes mellitus, but also focal vessel geometry [15]. The vertebrobasilar artery (VBA) system is unique in human anatomy in that two arteries (the left and right vertebral artery (VA)) merge into one artery (basilar artery (BA)). The diameters of the left and right VA differ in up to 50% of people [16], and their anatomical course is also different. Vertebrobasilar geometry can be classi ed into distinct con gurations according to the anatomical course of the VA on either side and the difference in their diameters. Wake-Buck et al. [17] described the relationship between vertebrobasilar geometry and differences in hemodynamic distribution. Atherosclerotic plaques often develop in regions with low wall shear stress, such as the inner wall of a curved artery or the apex of a junction [18].
We hypothesized that different vertebrobasilar geometries with their speci c hemodynamics will in uence the presence of atherosclerotic plaques. This study aimed to explore a potential association between the presence of BA plaques and different vertebrobasilar geometries using HRMRI in vivo.

Patients
We selected 303 consecutive patients who presented with a posterior circulation ischemic stroke to the Department of Neurology at our hospital between July 2017 and June 2018. Patients were included if they met the following criteria: (1) ischemic stroke or transient ischemic attack in the posterior circulation supplied by basilar artery presented with dizziness, unilateral limb weakness/ataxia, Gait ataxia and so on; (2) quality of the high-resolution magnetic resonance images su cient for analysis. Our exclusion criteria were as follows: (1) nonatherosclerotic vasculopathy, such as dissection, arteritis or Moya-Moya disease; (2) evidence of a cardioembolic stroke (atrial brillation); (3) contraindications to MR imaging; (4) poor image quality due to motion artifact; (5) vascular geometry not classi ed. Based on these criteria, seven patients with motion artifacts, and six patients in whom vascular geometry could not be classi ed were excluded.
Eventually, 290 patients were enrolled in our study.

Image analysis
A plaque was de ned as an eccentric wall thickening, whereas the thinnest part was estimated to be < 50% of the thickest part by visual inspection on axial CUBE images [19] (Fig. 1). The presence or absence of BA plaque was reviewed by two experienced readers (Zheng and Xue) blinded to MRA ndings. The differences between the two observers were solved by consensus. To assess intra-observer reproducibility, CBUE images were reevaluated by one reviewer (Zheng) one month later.
The diameter of the VA was measured on 3D-TOF-MRA. The diameter of each vessel was calculated as the average of the measurements made at three consecutive points, 3 mm apart, starting from the vertebrobasilar junction (for both VAs and the BA) [16,20]. The dominant VA was de ned as: the diameter of the VA on the dominant side was wider than that of the contralateral VA (difference in diameter ≥

mm) [16, 20].
Based on the 3D-TOF-MRA images, VBA geometry was qualitatively classi ed into four basic geometric con gurations: Walking, Tuning Fork, Lambda, and No Con uence (Fig. 2). The Walking geometry is distinguished by two VAs with a diameter difference of less than 0.3 mm that bend in the same direction before merging into the BA [16,17]. The Tuning Fork shows two VAs of equal diameter (difference in diameter < 0.3 mm) that bend in opposite directions to form a symmetrical con uence from which the BA emerges [16,17]. The Lambda geometry is de ned as two VAs with a diameter difference of at least 0.3 mm that merge into the BA [16,17]. In the No Con uence con guration, the VAs are not merged, but one VA continues as the BA, and the other VA feeds into other arteries, mostly the posterior inferior cerebellar artery.
Vertebral artery limited by posterior fossa bends mainly to left or right, the number of bends in the intracranial segments of the VAs was assessed on 3D-TOF-MRA images. Two 5 mm long lines were drawn starting from the vertex of a vascular curve to both sides, which intersect with the VA to form an angle. If this angle was ≤ 150 °, it was de ned as a vascular bend (Fig. 3). According to the total number of bends in the VAs' intracranial segments, patients were divided into a multi-bending group (total number of bends ≥ 3) and an oligo-bending group (total number of bends < 3).Vascular curvature was measured by the same two observers (Zheng and Xue) one month later blinded to HRMRI ndings. We took the consensus results from the measurements of two radiologists. To assess intra-observer reproducibility, vascular curvature was remeasured by one reviewer (Zheng) one month later.

Statistical analysis
Quantitative data were expressed as the mean ± standard deviation (SD), and qualitative data were expressed as percentages. The Shapiro-Wilk test for normality was used to investigate the distribution of data. Intraobserver or interobserver variability for the identi cation of BA plaque and vascular curvature measurements was performed using the intraclass correlation coe cient (ICC) analysis. The comparison of the BA plaque prevalence in the four different vascular geometries was performed using the chi-square test. Logistic regression models further examined the relationship between the geometry type and BA plaque prevalence. The comparison between multi-bending and oligo-bending VAs was performed by the chi-square and Fisher exact test. In the case of the Lambda con guration, differences in the diameter between the right and left VA for patients with and without BA plaques were compared with the Mann-Whitney U test, and a receiver operating characteristic curve analysis was used to identify the best cutoff value correlated with the presence of BA plaque. A P-value < 0.05 was considered to be statistically signi cant.
The statistical analysis was performed using IBM SPSS Statistics for Windows, version 19.0 (IBM Corp., Armonk, NY, USA).

Results
The clinical characteristics of the study population are summarized in Table 1.
The mean age of the 290 patients was 68.5 ± 10.3 years, and 60% were males. A total of 86 patients (mean age, 70.2 ± 10.6 years) had BA plaques, and 204 patients (mean age, 67.7 ± 10.1 years) had no BA plaques. A body mass index ≧ 28, hypertension, and diabetes mellitus were more frequent in patients with BA plaques as compared to patients without BA plaques. The comparison of clinical risk factors between patients with and without BA plaques is shown in Table 2. Based on their VBA geometry, 49 patients were assigned to the Walking group, 73 patients were in the Tuning group, 144 patients were classi ed into the Lambda group, and the last 24 patients belong to No Con uence group.
The inter-observer and intra-observer reproducibility for identi cation of BA plaque were ICC = 0.905 (95% CI 0.779-0.961) and ICC = 0.903 (95%CI 0.774-0.960), respectively. The inter-observer and intra-observer reproducibility for vascular curvature measurement were ICC = 0.905(95% CI 0.778-0.961) and ICC = 0.883(95%CI 0.728-0.952), respectively. Values are presented as mean ± standard deviation or number (%); BMI, body mass index. Correlation between VBA geometry and the presence of BA plaque The BA plaque prevalence was highest in patients who had a Walking con guration (53.06%) and lowest in patients with a Tuning Fork con guration (15.07%). There was a signi cant difference in the BA plaque prevalence among the four basic geometric con gurations (χ 2 = 21.265, P < 0.001, Table 3). Table 4 shows the association between VBA geometry and BA plaques. In the binary logistic regression analysis, Walking (odds ratio, 6.372; 95% con dence interval, 2.718-14.937; P < 0.001), Lambda (odds ratio, 2.168; 95% con dence interval, 1.037-4.533; P = 0.04), and No Con uence con guration (odds ratio, 3.382; 95% con dence interval, 1.188-9.625; P = 0.022) were found to be signi cantly associated with presence of BA plaque. After adjusting for age, sex, body mass index > 28, hypertension, and diabetes mellitus, these associations remained statistically signi cant (all P < 0.05; Table 4).   In the Lambda group, the mean diameter difference between the VAs was 1.4 mm (0.9 mm -1.6 mm) in patients with BA plaques and 0.9 mm (0.6 mm-1.3 mm) in patients without BA plaques (P < 0.001). The receiver operating characteristic curve analysis identi ed a cutoff value for the difference in VA diameter related to BA plaque formation of 1.35 mm.

Discussion
This study investigated the relationship between VBA geometry and BA plaque prevalence. We found that BA plaque prevalence was highest in the Walking geometry (53.06%) and lowest in the Tuning Fork geometry (15.07%). Moreover, the number of vascular bends in the intracranial segments of the VAs and the difference in diameter between the right and left VA also affected the presence of BA plaque.
Yu et al. [16] have investigated the relationship between the geometry patterns of vertebrobasilar artery and atherosclerosis. In their study, the vertebrobasilar artery geometry was qualitatively classi ed into four basic con gurations: Walking, Tuning, Dominant-Lambda, and Hypoplasia-Lambda. They didn't focus on the No Con uence Geometry, which plays an important role in the presence of BA plaque.
However, in their study, they only explored the correlation between the geometric con gurations and BA plaque distribution. In the present study, we investigated the relationship between the geometry of the vertebrobasilar artery system and the presence of BA plaque. In the future study, we also want to determine the relationship between geometric con gurations (Walking, Tuning Fork, Lambda, and No Con uence) and the distribution of BA plaque. Besides, in the Yu et al's study, 84 patients were included in the nal analysis. In contrast, our study recruited 290 patients in the nal analysis which will increase the power of statistical analysis. Ravensbergen et al. [18,21,22] in their studies employed autopsy and a series of junction models, and demonstrated that vertebrobasilar geometry affects hemodynamics and that atherosclerotic plaques are often found in regions with complex ow patterns and/or low wall shear stress. In the previous study [17] high-eld MRI was used in conjunction with computational uid dynamics (CFD) modeling to investigate the hemodynamics of subject-speci c con uence models (n = 5, two with Walking, two with Tuning Fork, and one with Dominant-Lambda geometry), and showed that vertebrobasilar geometry strongly in uences both the skewing of velocity pro les and wall shear stress distribution in the VBA system. In Walking geometry, the BA ow resulting from the merging of two VAs that bend in the same direction (right) makes the BA ow curve to the opposite direction (left). These chronic processes may induce a BA curvature. The shear stress is low at the inner wall of the BA curvature, and atherosclerotic plaques are prone to form in regions with low shear stress. Also, in the Walking geometry, the BA ow resulting from the VA ows swirling upward makes the ow distribution more complex, which can also induce plaque formation [17]. In the Tuning Fork geometry, the ows in the BA are roughly parallel, and the velocity pro le peak in the BA is rather central [17], resulting in the low BA plaque prevalence in this geometric con guration.
In our study, BA plaque prevalence in the Lambda con guration (27.78%) was higher than that in the Tuning Fork con guration (15.07%). Compared to Lambda patients without BA plaques, those with BA plaques had a larger difference in the diameters of their VAs. In a study by Hong et al. [20] BA curvature was found to be associated with a diameter difference between the VAs. The BA ow resulting from VAs with a diameter difference ≥ 0.3 mm makes the BA ow curve to the side of the weaker VA, and the chronic processes caused by the asymmetric VA ow induce greater curving of the BA wall, which consequently may cause atherogenesis.
We found that BA plaque prevalence in the No Con uence geometry (37.50%) was also higher than that in the Tuning Fork geometry (15.07%). We hypothesize that BA ow coming from one VA also causes the BA ow curve to the wall opposite to the VA, and as a chronic process may consequently cause a curving of the BA wall. Subsequently, the deformation of the BA wall makes it prone to atherogenesis, which may lead to ischemic stroke in the posterior circulation. To date, there are no hemodynamic studies in patients with No Con uence geometry.
In our study, BA plaque prevalence was higher in patients with multi-bending VAs as compared to patients with oligo-bending VAs. We think that the ow patterns in the rst group are more complex than that in the latter.
Our results have important clinical implications. First, we further classi ed VBA geometry based on the difference in diameter between the VAs and the VA course, which may help to improve the understanding of the vertebrobasilar system. Second, we demonstrated that the Walking, Lambda, and No Con uence geometry, multi-bending of the intracranial VA segment, and a large difference in the diameters of the VAs are high-risk factors for BA atherosclerosis initiation. Therefore, people with these geometric factors should be careful to prevent BA plaque formation.
Our study has several limitations. First, we did not measure the hemodynamics and ow distribution in the four geometric con gurations. The underlying mechanism of the basilar artery geometry in uencing the development of atherosclerotic plaques needs to be investigated from the hemodynamic aspect in future studies. Second, vascular curvature measurements were done manually from 2D images, since 2D had limited orientation, the vascular curvature at other dimensions cannot be assessed and the processing of vascular curvature measurements may be affected as the effect of ow artifact on 3D-TOF MRA. Third, coronal scanning was performed using the three-dimensional-CUBE sequence, and axial CUBE images were reconstructed. Wall thickness measurements from 3D MPR to determine plaque could be affected by volume averaging artifacts as the image resolution is comparable to the artery wall size, especially, when the basilar artery wall-thickening was not signi cant. Four, this is a pilot study with a limited sample size. Future studies recruiting larger populations are warranted. Five, even though a cut-off of 0.3 mm was used to de ne a dominant vertebral artery in previous studies (16,17), the spatial resolution 0.625 × 0.625 × 1.2 mm maybe was not high enough in this study. Finally, this study is an observational cross-sectional design, thus may be in uenced by uncontrolled confounding.

Conclusion
In conclusion, this study explored the correlation between vertebrobasilar geometry and the presence of BA plaque. The BA plaque prevalence was highest in the Walking and lowest in the Tuning Fork con guration. Further, the extent of vascular bending in the intracranial VA segments and the diameter difference between the two VAs also were associated with the presence of BA plaque. The present study was approved by the Ethics Committee of Fujian Medical University Union Hospital,and all patients provided their written informed consent.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Figure 1
Example gures of plaques on 3D CUBE T1WI. An atherosclerotic plaque was present on coronal image (A) and corresponding reconstructed axial image(B) from a 79 years old male patient. Images C and D are from another male patient with 64 years old. An atherosclerotic plaque can be found on coronal image (C) and corresponding reconstructed axial image(D).