Introduction

Carotid stenosis is a major cause of stroke, which in turn is the main cause of long-term disability and the third leading cause of death1.

Usually, hemodynamically significant carotid stenosis is limited to one side, while the other side can be characterized by the absence of stenosis or the presence of mild-moderate stenosis that does not increase the risk of stroke.

Significant carotid stenosis may cause ipsilateral hemispheric stroke, and contralateral cerebral stroke is considered to be an incidental phenomenon, that however should not be neglected.

In fact, when a significant unilateral carotid stenosis is detected, hemodynamic disturbance and blood theft may occur in the contralateral hemisphere, causing poor ischemic tolerance and increasing the risk of stroke2.

Albeit the impact of stenosis treatments on cerebral perfusion has been extensively investigated from a clinical standpoint3, the biomechanical analysis of this intra-individual crosstalk between the carotids remains poorly investigated. One of the reasons is that most biomechanical studies available in the literature focus on local carotid hemodynamics of healthy subjects or isolated diseased geometries4,5.

Moreover, while it is known that both carotid endarterectomy (CEA) and carotid stenting (CAS) may induce a local remodeling on carotid morphology and hemodynamics, particularly in terms of smaller carotid bulb diameters and higher flow velocities after CAS compared to CEA-treated patients6, the assessment of the hemodynamic crosstalk between the treated and the contralateral carotid artery, after either CEA or CAS, is missing.

Based on these considerations, the present study aims to evaluate the impact of severe stenosis, before and after the vascular treatment, on the local anatomy and hemodynamics of the contralateral carotid by patient-specific data retrieved by medical image analysis and Computational Fluid Dynamics (CFD).

Results

Given the aim of the present study, five patients who did not undergo the surgical procedure were excluded. Among the 65 remaining patients, eight were excluded from preoperative computational analysis due to poor quality of CT images (n = 4), which prevented the faithful reconstruction of the vascular district of interest, or due to the absence of flow waves extracted from PC-MRI images (n = 4). Of the 57 patients analyzed in the preoperative condition, it was possible to analyze fluid dynamics through CFD simulations in only.

patients (24 CEA, 18 CAS) in the postoperative condition. Specifically, one patient died before the intervention, seven patients were lost to follow-up, and for six patients, it was impossible to analyze imaging data for geometry reconstruction or flow wave analysis. For the reasons listed, to compare the pre-and post-operative conditions between the treated stenotic side and the contralateral side, the results presented in this study considered 42 patients.

Comprehensive patient demographics and disease characteristics are reported in Table 1. Patients’ median (IQR) age was 71.00(8.20) years old. Males represented a significant majority, constituting 74% of the whole cohort of patients. They presented a median (IQR) SOI stenosis of 45.00(17.5)%, while the stenotic side showed a percentage of stenosis of 75.00(10.00)% and a median (IQR) VPS of 237.00(81.00) cm/s.

Table 1 Patient cohort information is reported: age, gender, percentage of stenosis of the SOI, percentage of stenosis of the side to treat, systolic peak velocity (PSV), and type of surgery distribution. M = male; f = female; cea = carotid artery endarterectomy; cas = carotid artery stenting.
Full size table

In Table 2, an overview of patient demographics and disease characteristics about the type of administered treatment is reported. Specifically, 24 patients underwent CEA, while 18 were treated with CAS. Notably, both groups exhibit comparability in terms of disease characteristics and age. However, an imbalance emerges within the CEA group concerning the gender variable, with 21 male and three female patients, whereas the CAS group comprises ten male and eight female patients. We also conducted statistical analysis between CAS and CEA groups on all demographic variables. Except for the gender variable (p =.02), all remaining demographic variables do not reach statistical significance.

Table 2 Considering the type of surgery, age, degree of stenosis of SOI, degree of stenosis of the side to treat, and systolic peak velocity (PSV) is shown. Gender distribution is also reported. CEA = carotid artery endarterectomy; cas = carotid artery stenting; m = male; f = female.
Full size table

The comparison between the two sides of CCA and ICA parameters considering the same treatment is reported in Figs. 1, 2, 3, 4 and 5, while the value of each parameter is reported in Table 3.

Fig. 1
Fig. 1
Full size image

Diameter of common carotid artery (CCA) is reported for each side and condition analysed for both type of treatment. Panel A shows the boxplot related to patients treated with carotid endarterectomy (CEA), while panel B illustrates the value related to subjects that performed stenting (CAS). P-value related to the comparison between SOI and STENOSIS are also reported. SOI = side of interest; PRE = pre-operative; POST = post-operative. P refers to Wilcoxon ranked-sign test.

Fig. 2
Fig. 2
Full size image

Flow rate [mL/s] of common carotid artery (CCA) is reported for each side, and the condition analysed for both types of treatment. Panel A shows the boxplot related to patients treated with carotid endarterectomy (CEA), while panel B illustrates the value related to subjects that performed stenting (CAS). P-values related to the comparison between SOI and STENOSIS are also reported. SOI = side of interest; PRE = pre-operative; POST = post-operative. P refers to Wilcoxon ranked-sign test.

Fig. 3
Fig. 3
Full size image

The diameter of internal carotid artery (ICA) is reported for each side, and condition analysed for both types of treatment. Panel A shows the boxplot related to patients treated with carotid endarterectomy (CEA), while panel B illustrates the value related to subjects that performed stenting (CAS). P-value related to the comparison between SOI and STENOSIS are also reported. SOI = side of interest; PRE = pre-operative; POST = post-operative. P refers to Wilcoxon ranked-sign test.

Fig. 4
Fig. 4
Full size image

The flow rate [mL/s] of internal carotid artery (ICA) is reported for each side and the condition analysed for both types of treatment. Panel A shows the boxplot related to patients treated with carotid endarterectomy (CEA), while panel B illustrates the value related to subjects that performed stenting (CAS). P-value related to the comparison between SOI and STENOSIS are also reported. SOI = side of interest; PRE = pre-operative; POST = post-operative. P refers to Wilcoxon ranked-sign test.

Fig. 5
Fig. 5
Full size image

The percentage of area exposed to low time average wall shear stress(taWSS) is reported for each side, and condition analysed for both types of treatment. Panel A shows the boxplot related to patients treated with carotid endarterectomy (CEA), while panel B illustrates the value related to subjects that performed stenting (CAS). P-values related to the comparison between SOI and STENOSIS are also reported. SOI = side of interest; PRE = pre-operative; POST = post-operative. P refers to Wilcoxon ranked-sign test.

Table 3 Anatomical and hemodynamic parameters are reported for each side (SOI/STENOSIS), each condition (PREOP/POSTOP), and each type of treatment (CEA/CAS). SOI = side of interest (healthy); PREOP = pre-operative; POSTOP = post-operative; cea = carotid artery endarterectomy; cas = carotid artery stenting.
Full size table

Regarding the CEA group (see Figs. 1, 2, 3 and 4 A), our preoperative assessment involved a comparative analysis of the diameter and flow of CCA and ICA on the SOI and stenotic side. The results revealed disparities, with higher values for the diameter and flow of ICA and flow of CCA in the SOI compared to the stenotic side. As reported in Fig. 4A a significant difference was identified only for ICA flow (p <.001) with values of 3.33(0.93) m/s for the SOI and 2.18(1.22) ml/s for the stenotic side. No other statistically significant differences emerged among the remaining parameters, as also reported in Figs. 1A, 2A and 3A. Concerning the percentage of area exposed to low taWSS, the stenotic side seems to reach a higher value when compared to the SOI. However, statistical significance is not reached, as illustrated in Fig. 5A.

Regarding the CAS group (see Figs. 1, 2, 3, 4 and 5B), all parameters were found to be statistically insignificant between the two sides, except for the percentage of area exposed to low taWSS, which had a lower median value (p =.029) for the percentage of the area exposed to low taWSS (SOI: 17.00(15.55)% vs. STENOSIS: 11.05(14.05)%). However, there was also a noticeable trend for the diameter and flow of ICA and CCA of SOI, which were higher than those on the stenotic side.

No differences were identified in comparing the diameter and percentage of area exposed to low taWSS between the CEA and CAS groups in preoperative conditions. However, a significant difference was observed in the flow between the two procedures; details are reported in Table 4. The CEA group exhibited higher flows of ICA (Fig. 4) and CCA (Fig. 2), and a significant difference was highlighted for CCA flow (p =.020) and ICA flow (p =.002) on the SOI side.

Table 4 P-values resulting from the comparisons performed between different surgery treatments (CEA/CAS), and conditions (PRE/POST) are reported for the anatomical and hemodynamic parameters considered: diameter (D), flow rate (Q), and percentage of area exposed to low TaWSS (% Area). TaWSS = time average wall shear stress; CCA = common carotid artery; ICA = internal carotid area; PRE = pre-operative; POST = postoperative; SOI = side of interest; CEA = carotid artery endarterectomy; CAS = carotid artery stenting.
Full size table

The same tests were conducted on the postoperative data. In both groups, the differences in diameter and percentage of area exposed to low taWSS observed between the two sides during preoperative conditions were not present anymore. It was observed that the CCA flow and ICA flow on the SOI side in the CEA group were significantly higher than in the CAS group (CCA p =.010, ICA p <.001). A significant difference was found also on the stenotic side of CCA in the CEA group (p =.042). Conversely, no statistically significant difference was noted on the stenotic side among the CAS group for ICA flow (p =.102).

Furthermore, we comprehensively analysed pre- and post-operative results, categorising them by the type of surgery (also reported in Table 4).

We found a general trend within the CEA group, which showed a reduction in flow rate, an increase in diameter, and a nearly consistent percentage of area exposed to low taWSS. However, we could only identify statistically significant differences in the SOI. The diameter of CCA increased from 6.21(1.24) mm to 6.56(1.14) mm (p =.007), and the flow rate of CCA decreased from 6.47(2.77) ml/s to 5.33(2.29) ml/s (p =.029) after the treatment. The percentage of area exposed to low taWSS in SOI increased from 15.50(10.80) % to 18.00 (20.25) % after the treatment, but this increase was not statistically significant (p =.148).

Similarly, we conducted the same analysis for the CAS group, revealing a comparable trend but no significant changes were observed. Additionally, we found that the stenotic side showed a non-significant (p =.194) increase in the percentage of area exposed to low taWSS after surgery, transitioning from a preoperative value of 11.05 (14.05) % to 18.50 (12.30) %.

Discussion

The treatment of carotid stenosis involves local anatomical and hemodynamic remodelling, which is well-known. However, it is usually only analysed on the treated side. Examining the contralateral side and comparing pre- and post-operative conditions can help us understand the changes that occur locally on the untreated side. This approach enhances our understanding of the crosstalk mechanisms between the two sides. We conducted a comprehensive analysis examining imaging data and implementing patient-specific CFD simulations to achieve this. First, we compared the results by looking at the sides (SOI vs. stenosis) in pre- and post-operative conditions, stratifying by treatment type (CEA and CAS). Then, we studied the treatment (CEA vs. CAS) in pre- and post-operative conditions, stratifying by side. Finally, we compared the pre-and post-operative results (preop vs. postop) by stratifying them by side and treatment type.

The analyses showed that flow rates were consistently higher on the SOI before surgery, regardless of the type of surgery used to treat stenosis. The stenotic side had a larger area exposed to low values of taWSS. After treatment, hemodynamic and anatomical remodelling increased diameters on the untreated side and reduced flow rates. At the same time, the area exposed to low taWSS increased. Severe stenosis can cause blood flow to favour the healthier side for adequate cerebral perfusion4. Our findings confirm this, showing uneven flow distribution between the region of interest and the stenotic side. Typically, localised hemodynamic assessments of carotid arteries focus only on the treated area, overlooking the hemodynamic aspects of the contralateral side. Our clinical and CFD analyses reveal that the treatment of the pathologic segment facilitated a hemodynamic and anatomical reconfiguration on the contralateral side, showing the existence of a cross-linking phenomenon. This mechanism, which restores normal perfusion by removing calcified plaque, is statistically significant in the CEA group. The CAS procedure restores proper perfusion but may lead to small residual stenosis that may moderate the extent of anatomical and hemodynamic remodelling on the contralateral side, explaining the differences in results between the two groups under investigation.

From the clinical point of view, a better understanding of the mechanism underlying the cross-talking between the SOI and the contralateral side may help for a better stratification of patients who should undergo carotid revascularization, especially in case of asymptomatic lesions.

Furthermore, our study aimed to also increase the knowledge about possible different impacts of CEA and CAS on the postoperative vascular remodelling.

Stenosis treatment increases low taWSS on the SOI. Vessel remodelling and flow reduction induce lower taWSS with atherogenic potential. The presence of stenosis can lead to an unfavorable hemodynamic environment and SOI overload.

To analyse WSS distribution, we used fully patient-specific CFD analysis. Our study proposes CFD simulation of a large cohort of patients using CT and MRI data before and after surgery, which is a novelty. CFD simulations have been widely used to study atherosclerosis since Milner et al.‘s pioneering study7, but our approach is unique in using both CTA and MRI data. Other studies have shown the feasibility of computing hemodynamics in stenotic carotids given in-vivo measurements8,9,10,11. However, CFD simulations have often been used to integrate flow measurements performed by (idealised) in vitro stenosed carotid replicas12,13,14,15 or with a relatively small cohort of patients16.

At the same time, from the first report about time-resolved WSS vectors estimated directly from 3D-MRI data17, several studies have proved that in-vivo carotid hemodynamics, including WSS, can be measured non-invasively by 4D flow MRI18,19,20. However, there are still several technical limitations to tackle when using this approach, mostly dealing with space and time resolution of the methods21,22.

Our work has some limitations, which will be the object of future developments for further improvements. One of these limitations is using literature-based thresholds to identify low taWSS. To overcome this, we suggest calculating the mean taWSS in the preoperative case and using it as a threshold. This approach would show a patient-specific and relative difference in the percentage of the area subjected to low taWSS. A patient-specific approach is crucial in defining the pathology, and computational simulation can be an effective supplementary tool for stenosis treatment.

In addition, in this work, only the taWSS was considered as near-wall hemodynamics descriptors. Other descriptors accounting for pulsatility and multidirectionality could be analysed. Additionally, an analysis of the topological skeleton of the wall shear stress vector field could be conducted, further extending the hemodynamic characterization. Furthermore, in this study the percent area exposed to low taWSS was calculated in the entire carotid bifurcation model, in future it could be interesting considering stratifying the analysis according to specific locations.

Although some analyzed parameters result to be significantly different, others do not reach a statistically significant difference. While the anatomical and flow-related data obtained from CTA and PC-MRI showed statistically significant differences, the CFD-derived results did not reach statistical significance in either treatment group. The lack of significance in our study could be due to the small sample size of CAS and CEA groups. For this reason, we plan to expand our work by involving more patients.

Another limitation could be the imbalance of having more female CAS, given the smaller diameter of female patients compared to males. Other studies assess the significance of gender in determining CAS procedures’ benefit and clinical outcome showing no differences in immediate, and long-term CAS outcomes between gender23 or conforming no association with higher rates of stroke after CAS24. Future investigations are needed to explore if female gender influences the hemodynamic crosstalk between carotid arteries, as there is currently no compelling evidence available to suggest this. However, our study was not planned to analyse gender differences, and the enrolled patients were initially randomized to perform either CEA or CAS.

Finally, in the presented study the cerebral perfusion and its relationship with the stenosis presence and treatment were not elucidated, thus the analysis presented lacks information regarding the vertebral arteries and the Circle of Willis.

Methods

For this study, the treated carotid artery is labeled as stenosis, while the opposite side is defined as the side of interest (SOI).

Study cohort

All included patients were referred for a subgroup of patients enrolled in a randomized study named BAROX (ClinicalTrials.gov registration number: NCT03493971). BAROX study was approved by the local ethical committee (record number 62/int/2017, 8 June 2017). Inclusion criteria of this randomized trial were patients with ≥ 70% symptomatic or ≥ 80% asymptomatic internal carotid artery (ICA) stenosis according to the European Carotid Surgery Trial (ECST) method25, deemed fit for the treatment with either CAS or CEA, according to the international guidelines26. Exclusion criteria were age > 75 years, previous disabling stroke, contralateral carotid occlusion or > 70% stenosis, and other systemic diseases judged non-compatible with the procedures or randomization.

All enrolled patients were then randomized using a dedicated software to perform either CAS or CEA with patch closure. From March 2018 to March 2021, 70 patients were recruited.

Before the intervention, all patients underwent a Carotid Duplex ultrasound, computed tomography angiography (CTA), and phase-contrast magnetic resonance imaging (PC-MRI). After the operation, patients were followed up with Duplex ultrasound at 6, 12, and 24 months, and CTA and MRI at 12 and 24 months. According to the study protocol, all patients also performed a mini-mental test evaluation and a clinical evaluation before the procedure and at 6, 12, and 24 months after.

From these data of the BAROX study, the subanalysis of the present study included data from 42 patients (24 CEA and 18 CAS) that completed the follow-up with CTA and MRI and had high-quality images.

Imaging modalities

CTA of the supra-aortic trunks was performed (Siemens SOMATOM Definition AS, Erlangen, Germany, 64 slices). Acquisition parameters were 120 kV, 100 mAs, pixel size 0.45 mmx0.45 mm, and acquisition matrix 512 × 512. The images were reconstructed using a slice-thickness of 1 mm with an incremental factor of 0.5 mm. MRI was performed using a 1.5-T unit with 45-mT/m gradient power (Magnetom Aera Maestro Class, Siemens, Erlangen, Germany) and a head coil. Electrocardiographically triggered free-breathing through-plane phase-contrast (PC) sequences were performed for phase-velocity mapping with the following parameters: TR/TE = 4/3.2 ms, thickness 5 mm, and temporal resolution 41 ms. The flow was measured with PC sequences in the three vessels of interest CCA, ECA, and ICA. To this aim, two axial planes were acquired, one 20 mm below the bifurcation on the CCA, and one 30 mm above the bifurcation, including both the ICA and ECA lumens27. Velocity encoding (VENC) has been varied in different cases from 80 cm/s to 120 cm/s to exploit the maximum dynamic range of the velocity field and minimize image artifacts such as aliasing. All data were processed according to the workflow shown in Fig. 6 and described in the following paragraph.

Fig. 6
Fig. 6
Full size image

Workflow is presented. All data were processed to obtain the anatomical and hemodynamic parameters of interest. CTA scans were segmented to reconstruct the 3D models of stenotic and controlateral carotid arteries. The models were used to extract patient-specific diameters. The phase contrast magnetic resonance images were processed to extract patient-specific flow waves of the common carotid artery (CCA), internal carotid artery (ICA), and external carotid artery (ECA); flow waves were used to analyze patient-specific flow rates and define the BCs. A surface mesh was defined to import the fluid domain into Simvascular to perform CFD simulations. The simulations were run, and the results were post-processed to evaluate the percentage of the luminal area exposed to low-time average wall shear stress (taWSS).

Imaging and data processing

CTA scans were segmented to obtain right and left carotids using the Vascular Modeling Toolkit 8 (VMTK toolbox v1.4). The segmentation starts from the CCAs origin up to the base of the skull for ICAs and up to their first branch for ECAs, which usually takes off between 20 and 30 mm downstream of the bifurcation. CCAs were clipped proximally at a minimum distance of 7 diameters below the bifurcation, according to Hoi et al.27.

For both sides, luminal diameters were measured 20 mm below the bifurcation in the CCA, 30 mm above the bifurcation in the ICA, and 10 mm above the bifurcation in the ECA, as referred in Markl M18., to ensure a consistent comparison among patients and to avoid confounding factors related to the presence of the plaque.

PC-MRI data were analyzed with Medis software (Medis medical imaging systems BV, Leiden, Netherlands) to measure velocity and time-averaged flow waves.

Computational fluid dynamics simulations

CFD simulations were carried out in every patient’s stenotic and SOI carotids to find the luminal area exposed to atherogenic biomechanical stimuli, considering the near-wall hemodynamics through wall shear stress-based descriptors. After the lumen segmentation, non-structured uniform tetrahedral meshes were created with TetGen v1.513, with a target element size of 0.01–0.02 cm27, which yielded meshes between 0.7 and 2 million elements28, resolving satisfactorily the WSS field29. Pulsatile flow simulations were run with the software SimVascular v2019.08.0915 to solve the incompressible Navier-Stokes equations in a rigid-wall domain. Simulations were run for three cardiac cycles to damp transitory states, and the last cycle was used to inform the results26,30. Variations in the quantities of interest were < 5% along all the carotid bifurcation between the second and third cardiac cycles. Blood density was set at 1.06 g/cm3, viscosity 0.04 Poise, and a no-slip condition was imposed at the vessel wall31,32. For each patient, the cardiac cycle duration was based on the measurements at the PC-MRI scan; the time step was imposed as one-thousandth of the cycle period; thus 3000 time steps were run to simulate three cardiac cycles, as mentioned before.

We used patient-specific inlet and outlet flow boundary conditions (BC) following the strategy used by Cibis and colleagues19, where PC-MRI velocity is directly imposed as a flat profile at the origin of the CCA as inflow. Then, to fulfill the conservation of mass law on the outflow BC, both PC-MRI derived flows of ICA, and ECA were corrected by a constant. In particular, in the branch of the carotid bifurcation with the highest flow, we imposed a Dirichlet BC with a parabolic velocity profile. In contrast, a stress-free BC was used in the branch with a lower flow.

CFD temporally discretized results were analyzed. The time-averaged wall shear stress (taWSS) in the cardiac cycle was evaluated and processed according to cut-off values reported in the literature28,33 to identify three regions along the lumen: 1) low taWSS, defined as 0 < taWSS < 4 dyn/cm² and associated with atherosclerosis-prone regions; (2) physiological taWSS (10 < taWSS < 70 dyn/cm²); and (3) high-shear thrombogenic regions, where taWSS > 70 dyn/cm². The results are reported as a percentage of the total lumen area exposed to low taWSS, considering 20 mm below the bifurcation of the CCA, 30 mm above the bifurcation of the ICA, and 10 mm above the bifurcation of the ECA.

Statistical analysis

According to the type of treatment (CAS or CEA), the population was divided into two groups. Geometric and hemodynamic data between the two carotids were compared in preoperative and postoperative conditions. The Shapiro-Wilk test tested the normality of the distribution of values, and the variables sample was defined as non-normally distributed. The Wilcoxon signed-rank test compared CCA and ICA parameters, considering the same treatment. Concurrently, the Mann-Whitney test and Chi-Square test were used to analyze the differences between the two groups (CEA vs. CAS), used respectively for continuous and categorical variables. Accordingly, the percentage of areas exposed to low taWSS was also included in the analyzed parameters. All results are presented as median and interquartile range. Statistical significance was assumed at p <.050.