Virtual test occlusion for assessing ischemic tolerance using computational fluid dynamics.

BACKGROUND: Ischemic tolerance has been evaluated by the balloon test occlusion (BTO) for cerebral aneurysms and tumors that might require parent artery occlusion during surgery. However, because of its invasiveness, a non-invasive evaluation method is needed. In this study, we assessed the possibility of virtual test occlusion using computational fluid dynamics (CFD) as a non-invasive alternative to BTO for evaluating ischemic tolerance.
METHODS: Twenty-one patients who underwent BTO were included in the study. Virtual test occlusion was performed using CFD analysis, and the flow rate (FR) and wall shear stress (WSS) of the middle cerebral artery on the occlusion side were calculated. The correlations between these parameters and examination data including the parameters of computed tomography perfusion during BTO were assessed and the cutoff value of CFD parameters for detecting the good collateral group was calculated.
RESULTS: The FR was strongly correlated with mean transit time (MTT) during BTO and moderately correlated with collateral flow grade based on angiographic appearance. The WSS was moderately correlated with collateral flow grade, mean stump pressure (MSP), and MTT. Furthermore, the FR and WSS were strongly correlated with the total FR and the diameters of the inlet vessels. The cutoff value of FR for detecting the good collateral group was 126.2 mL/min, while that of the WSS was 4.54 Pa.
CONCLUSION: The parameters obtained through CFD analysis were correlated with collateral flow grade and MSP in addition to MTT. CFD analysis may be useful to evaluate ischemic tolerance as a non-invasive alternative to BTO. Copyright:

INTRODUCTION

Ischemic tolerance has been evaluated by balloon test occlusion (BTO) for large and giant cerebral aneurysms[6,16] and head and neck tumors[2,23] that might require parent artery occlusion during surgery. Using BTO, the risk of acute ischemic stroke can be evaluated to observe symptom progression during temporary occlusion of the parent artery as the internal carotid artery (ICA) or vertebral artery (VA) since the first report of its use in 1974.[25] A previous review showed that the ischemic stroke rate for abrupt ICA occlusion was 26% without BTO versus 13% with BTO.[18] The risk of delayed ischemic stroke has also been assessed by various objective parameters including angiographic appearance, stump pressure, near-infrared spectroscopy cerebral oximeter, xenon-enhanced computed tomography (CT), single-photon emission CT, positron emission tomography, and digital subtraction angiography (DSA) equipment during BTO.[1,3,5,11,18,22,31,32,34-36] However, the complication rate of this procedure was reportedly 3.7% in 300 cases,[33] and its invasiveness cannot be ignored. Therefore, a non-invasive alternative to BTO is needed.

Image-based computational fluid dynamics (CFD) are able to extract patient-specific hemodynamic information based on the imaging data obtained through CT angiography (CTA), 3-dimensional (3D) DSA, and magnetic resonance angiography.[10,27,28] Their utilities have been reported in cerebral aneurysm, atherosclerotic lesion, moyamoya disease, trigeminal neuralgia, and hemifacial spasm.[9,12,13,24,27,29,30,37,38] Only one study to date have evaluated ischemic tolerance using image-based CFD as an alternative to BTO,[4] and the correlation between flow rate (FR) obtained through CFD and cerebral blood flow (CBF) on 123I-IMP single-photon emission CT during BTO has been revealed. In this study, BTO has been performed with CT perfusion, stump pressure, and near-infrared spectroscopy cerebral oximeter in vivo optical spectroscopy (INVOS) (Nihon Kohden Corp., Tokyo, Japan) examinations in addition to the conventional neurological examination and collateral angiography. Then, we evaluated possibility of virtual test occlusion using CFD analysis versus BTO, comparing the parameters obtained through virtual test occlusion with several modalities during BTO.

MATERIALS AND METHODS

Patients

The study protocol was approved by the ethics committee of Sapporo Medical University Hospital. As this study had a retrospective design, patient consent was obtained with an opt-out policy using a website. From March 2015 to February 2020, consecutive patients who underwent BTO of the ICA at our hospital were enrolled in the study. The patients were diagnosed with cerebral aneurysms, head and neck tumors, and other conditions (intracranial infection) that might require parent artery occlusion during surgery. Patients under 10-years-old and those who required BTO of other arteries were excluded from the study. A total of 21 patients (seven males and 14 females) were examined. The median age (interquartile range) of the patients was 61.0 (50.0–69.5) years (range, 11–74 years). The diagnoses of the 21 patients included cerebral aneurysm in four; head and neck tumor in 15; and other in two.

BTO

All procedures were performed using single-plane DSA equipment (Infinix Celeve-i INFX-8000C, Canon Inc., Tokyo, Japan) as previously described.[14] The clinical, stump pressure, and INVOS examinations were performed in addition to collateral angiography during the temporal occlusion. Following collateral angiography, keeping the balloons inflated, CT perfusion was performed with interventional radiology CT equipment without patient room-to-room transfer. If the patients presented with ischemic symptoms during the occlusion, the balloons were deflated immediately regardless of the collateral angiography or CT perfusion. The appearance on collateral angiography was divided into five grades (0–4) based on the collateral flow grading system defined by the American Society of Intervention and Therapeutic Neuroradiology/Society of Interventional Radiology.[8] This scale was designed for stroke recanalization. We modified this scale and evaluated the collateral angiographic appearance during BTO as previously described.[14] The Grade 4 group was defined as the good collateral group with rapid collateral flow (within 2 s of the contralateral side), while the groups with Grade 3 or less were defined as the poor collateral group with slow collateral flow (more than 2 s slower than the contralateral side). Patients who could not undergo collateral angiography because of ischemic symptoms during temporal occlusion were classified into the poor collateral group.

Stump pressure was measured using a 3.3-Fr MASAMUNE balloon catheter (Fuji Systems Corp., Tokyo, Japan) placed in the ICA on the lesion side. The ratio of the mean stump pressure (MSP) before and after the temporal occlusion adjusted by mean blood pressure taken from a cuff on the left upper arm was calculated as MSPBR ([MSP 10 min after occlusion/mean blood pressure 10 min after occlusion]/[MSP before occlusion/mean blood pressure before occlusion]) (BR; ratio before and after temporal occlusion).

The INVOS examination was performed during the temporary occlusion. The ratio of INVOS before versus after the temporal occlusion adjusted by INVOS on the contralateral side was calculated as INVOSBR = ([INVOS on the lesion side 10 min after the occlusion/INVOS on the contralateral side 10 min after the occlusion]/[INVOS on the lesion side before the occlusion/INVOS on the contralateral side before the occlusion]).

If the temporary occlusion was <10 min due to ischemic symptoms, MSP and INVOS values just before balloon deflation were used.

CT perfusion during BTO

CT perfusion was performed using interventional radiology CT equipment with 16 rows (Aquilion 16 LB; Canon Inc.). The CT perfusion scanning parameters were described previously.[14] The CT perfusion information was transferred to a workstation (ZIOSTATION 2; Ziosoft Inc., Tokyo, Japan), and 3D multicolored images were produced. Standard single-value decomposition was implemented for the deconvolution algorithm. A region of interest (ROI) showing arterial input function was manually applied at a single branch of an insular segment of the middle cerebral artery (MCA) in the unimpaired hemisphere. A ROI showing the venous output function was also established on the superior sagittal sinus. The CBF (mL/100 g/min) in the MCA territory on the lesion and contralateral sides was measured by the manual creation of an ROI over this area along the level of the basal ganglia. The asymmetry ratio (AR) of the CBF was then calculated as CBFAR = ([CBF on the lesion]/[CBF on the contralateral side]) × 100 (%). Similarly, the cerebral blood volume (CBV) (mL/100 g), mean transit time (MTT) (seconds), and time to peak (TTP) (seconds) were measured, and their AR was calculated as CBVAR = ([CBV on the lesion side]/[CBV on the contralateral side]) × 100 (%); MTTAR = ([MTT on the lesion side]/[MTT on the contralateral side] – 1) × 100 (%); TTPAR = ([TTP on the lesion side]/[TTP on the contralateral side] – 1) × 100 (%).

Virtual test occlusion based on CFD analysis

The CTA examination was performed before BTO using a multidetector-row CT scanner with 320 rows (Aquilion ONE; Canon Inc.). The 3D-CTA scanning parameters were as previously described.[7] The CTA information was transferred to a workstation (ZIOSTATION 2). The 3D images that included bilateral ICA (from the petrous portion), MCA (the first segment: M1), anterior cerebral artery (the first and second segments: A1 and A2, respectively), VA (from the intradural portion), basilar artery, posterior cerebral artery (the first and second segments: P1 and P2, respectively), posterior communicating artery (PcomA), and anterior communicating artery (AcomA) were produced at a threshold of 30% based on the CT value of M1 on the lesion side by a volume-rendering algorithm [Figure 1a]. The ophthalmic, superior cerebellar, anterior inferior cerebellar, and posterior inferior cerebellar arteries and other small branches were excluded in the image processing. In cases of vessel adhesion, manual processing was performed, especially in bilateral A2.

Figure 1:

A case of left middle fossa tumor. Three-dimensional imaging data were obtained through CTA using a volume-rendering algorithm. (a) The white arrow shows the ICA on the lesion side. Extraction of the center lines and segmentation and labeling of the associated vessels were performed (b). ICA on the lesion side was set as a wall (white arrowhead), while that of the contralateral ICA and bilateral VA were set as the inlet, and the on-site analysis was run (c). After the analysis, the steam line contributing to M1 on the lesion side is shown (d). CTA: Computed tomography angiography, ICA: Internal carotid artery, M1: First segment of the middle cerebral artery, VA: vertebral artery.

In ZIOSTATION 2, the 3D image data were transformed to stereolithography (STL) format and Hemoscope Ver.1.5 (EBM Corp., Tokyo, Japan) was used for further image processing and analysis. In the vessel module, extraction of the center lines, segmentation, and labeling of the associating vessels (bilateral ICA, M1, A1, A2, VA, P1, P2, PcomA) were performed. These vascular geometries were filled with unstructured cells mainly consisting of hexahedrons approximately 0.25 mm in the far-wall regions and 0.125 mm (width) and 0.05 mm (height) in the near-wall regions. In the near-wall regions, the meshes were aligned to fit the boundary with three layers. The inlet (contralateral ICA, bilateral VA) and outlet vessels (bilateral M1, A2, P2) were extended as long as possible based on imaging data and trimmed their proximal and distal ends [Figure 1b]. Then, an automatic measurement was performed to confirm the anatomical information of the associated vessels. In the validated CFD module, the ICA on the lesion side was set as the wall, while that of the contralateral ICA and bilateral VA were set as the inlet and the on-site analysis was run [Figure 1c]. The boundary conditions were determined according to a constant WSS theory (1.5 Pa). Specifically, FR of the inlet and outlet vessels was calculated using the Hagen-Poiseuille equation:

where Q, τ, µ, and D denote FR (mL/min), WSS (Pa), fluid viscosity, and vascular diameter, respectively. After total FR at the inflow vessels was calculated, the amount of FR was distributed at each outlet vessel according to equation;[1] the specified boundary conditions of the inlet and outlet vessels were the pressure and velocity boundary conditions, respectively. The present computation adopted a steady flow, and the inlet pressure was set at 100 mmHg.

CFD analysis was performed using a finite volume method to solve the 3D unsteady Navier-Stokes equations and equation of continuity. Blood was assumed to be an incompressible and Newtonian fluid with a density of 1050 kg/m3 and a viscosity of 0.004 Pa s. The Euler and second-order upwind schemes were utilized to discern unsteady and convective acceleration terms. The convergent criteria were set to 10−4. Mean FR (mL/min) and mean WSS of the target vessels was obtained thorough this analysis, and the values (FR and WSS of M1 on the lesion side and FR of all inlet vessels) were evaluated. We named this simulation method virtual test occlusion. This method included vessels that were associated with perfusion on the lesion side. The virtual test occlusion with a steam line contributing to M1 on the lesion side is shown in [Figure 1d]. In case neither the AcomA nor the PcomA was sufficient for the CFD analysis, we defined their FR and WSS of M1 on the lesion side and FR of all inlet vessels as 0.

Statistical analysis

The data are expressed as median (interquartile range). Spearman’s rank correlation coefficient was used to confirm the correlation between the parameters of virtual test occlusion and those of other modalities. A correlation coefficient (ρ) larger than 0.7, between 0.4 and 0.7, and between 0.2 and 0.4 indicated strong, moderate, and weak correlations, respectively. Receiver operating characteristic (ROC) curve analysis was used to determine the most suitable cutoff value of FR and WSS to detect good collateral groups based on the shortest distance from the curve to the upper left corner. Statistical analyses were performed using SPSS software (version 25; IBM Corporation, Armonk, NY, USA). P < 0.05 was considered statistically significant.

RESULTS

Correlations between parameters of virtual test occlusion and those of other modalities

Patient characteristics are shown in Table 1. Collateral angiography, CT perfusion, stump pressure, and INVOS examinations were successfully performed in 19, 11, 17, and 18 patients, respectively. The main reasons for not performing these examinations were presentation of ischemic symptoms during the test occlusion in five patients and insufficient room and device setting in the remaining patients. The median diameter of the ICA and M1 on the lesion side, contralateral ICA, and bilateral VA and the total diameter of the inlet vessels included in the virtual test occlusion were 5.4 mm (5.0–5.9 mm), 3.1 mm (2.9–3.4 mm), 5.3 mm (4.6–6.0 mm), 3.2 mm (2.7–3.6 mm), and 11.2 mm (8.9–12.6 mm), respectively. The median FR, WSS of M1 on the lesion side, and FR of all inlet vessels were 80.4 mL/ min (0.0–151.6 mL/min), 3.4 Pa (0.0–5.9 Pa), and 317.0 mL/ min (0–574.6 mL/min), respectively. The distribution of collateral flow grade was as follows: Grade 0, 1 = 0; Grade 2 = 4; Grade 3 = 9; and Grade 4 = 6. The median CBFAR, CBVAR, MTTAR, TTPAR, MSPBR, and INVOSBR were 95.0% (79.5– 96.5%), 106.9% (101.1–110.7%), 10.5% (8.0–37.0%), 3.6% (0.2–19.1%), 50.9% (31.2–65.2%), and 98.4% (92.0–101.1%), respectively.

Table 1:

Patient characteristics.

A summary of the correlations is shown in Table 2. Compared with the parameters obtained during BTO, FR of M1 on the lesion side showed a strong correlation with MTTAR (ρ = −0.784, P = 0.004); moderate correlations with collateral flow grade (ρ = 0.684, P = 0.001), CBFAR (ρ = 0.482, P = 0.133), and MSPBR (ρ = 0.447, P = 0.072); and a weak correlation with TTPAR (ρ = −0.333, P = 0.318) with the exception of CBVAR (ρ = 0.145, P = 0.670) and INVOSBR (ρ = 0.079, P = 0.756). WSS of M1 on the lesion side showed moderate correlations with collateral flow grade (ρ = 0.647, P = 0.003), MTTAR (ρ = −0.697, P = 0.017), TTPAR (ρ = −0.528, P = 0.095), and MSPBR (ρ = 0.571, P = 0.017), and a weak correlation with CBVAR (ρ = −0.245, P = 0.467) with the exception of CBFAR (ρ = 0.173, P = 0.612), and INVOSBR (ρ = −0.043, P = 0.865). In the parameters obtained during BTO, MTTAR was most strongly correlated with FR and WSS of M1 on the lesion side. Scatter diagrams showing the notable correlations are shown in [Figure 2]. FR was significantly correlated with collateral flow grade and MTTAR, and WSS was significantly correlated with collateral flow grade, MTTAR, and MSPBR in the parameters obtained through BTO.

Table 2:

Correlations between parameters of CFD analysis and those other modalities during BTO.

Figure 2:

Scatter diagrams showing the correlations between FR and collateral flow grade (a), MTTAR obtained through CT perfusion (b) and MSPBR (c), WSS and collateral flow grade (d), and MTTAR (e) and MSPBR (f). CT: Computed tomography, FR: Flow rate; AR: Asymmetry ratio, BR: Ratio before and after temporal occlusion, MTT: Mean transit time; MSP: Mean stump pressure, WSS: Wall share stress.

We subsequently evaluated the anatomical factors that affect the parameters of M1 on the lesion side obtained through virtual test occlusion. The FR of M1 on the lesion side showed strong correlations with the diameter of the contralateral ICA (ρ = 0.709, P = 0.007), total diameter of the inlet vessels (ρ = 0.886, P < 0.001), and total FR of the inlet vessels (ρ = 0.962, P < 0.001) and moderate correlations with ICA diameter on the lesion side (ρ = 0.648, P = 0.012), diameter of M1 on the lesion side (ρ = 0.556, P = 0.039), and mean diameter of the bilateral VA (ρ = 0.567, P = 0.054). WSS of M1 on the lesion side showed strong correlations with total diameter of the inlet vessels (ρ = 0.833, P < 0.001) and total FR of the inlet vessels (ρ = 0.957, P < 0.001) and a moderate correlation with contralateral ICA diameter (ρ = 0.626, P = 0.022) and weak correlations with ICA diameter on the lesion side (ρ = 0.358, P = 0.208) and mean diameter of the bilateral VA (ρ = 0.354, P = 0.259) with the exception of M1 diameter on the lesion side (ρ = 0.059, P = 0.840). Regarding anatomical factors, total FR of the inlet vessels was most strongly correlated with FR and WSS of M1 on the lesion side. It was also obtained through virtual test occlusion depending on anatomical information.

ROC curve analysis of good collateral group

The good collateral group consisted of six patients, while the poor collateral group consisted of 15 patients. The ROC curve analysis of FR and WSS of M1 on the lesion side to detect the good collateral group is shown in [Figure 3]. The most suitable cutoff value of FR obtained through ROC curve analysis was 126.2 mL/min (sensitivity: 83.3%; and specificity: 86.7%), and that of WSS was 4.54 Pa (sensitivity: 83.3%, and specificity: 86.7%).

Figure 3:

The gray and black lines showing a ROC curve analysis of the good collateral group detected by FR and WSS, respectively. FR: Flow rate, ROC: Receiver operating characteristic, WSS: Wall share stress.

DISCUSSION

Utility of virtual test occlusion for assessing ischemia

In this study, the CFD parameters were correlated with collateral flow grade, MSP, and MTT. Collateral angiographic appearance, stump pressure, and MTT obtained through CT perfusion during BTO are reportedly predictors of ischemic tolerance,[1,14,34,35] and these results might indicate that the CFD parameters represent the ischemic tolerance for noninvasively detecting the risk of delayed ischemic stroke. In the previous report, Charbel et al. indicated a decrease in FR in both the ipsilateral M1 and A1 segments, calculated by image based CFD model, that was greater than 20% was 100% sensitive and 100% specific in identifying patients who could not tolerated the BTO.[4] We confirmed the results also using the other parameter, WSS, in addition to FR, comparing with several modalities during BTO. To date, the utility of WSS has been reported mainly in cerebral aneurysm in that low WSS might facilitate the growing phase and trigger cerebral aneurysm rupture by causing degenerative changes in the aneurysm wall.[27] On the other hand, FR and WSS obtained through CFD analysis reportedly reflect the actual cerebral hemodynamic condition and are useful for detecting the risk of ischemic stroke. Mori et al. reported low WSS and WSS gradient values of the ipsilesional lateral striate artery (LSA) compared with that of the contralateral side in patient with LSA-territory infarcts, and CFD factors of LSA could be a risk factor for LSA-territory infarctions.[19] In patients with symptomatic intracranial atherosclerotic disease, Leng et al. reported that translesional WSS ratio was independently associated with recurrent ischemic stroke in the same territory within 1 year.[17] In this manner, FR and WSS obtained through CFD may be used to non-invasively detect the risk of delayed ischemic stroke.

Regarding the items influencing CFD parameters, FR and WSS might be reflected, particularly inlet vessel diameter. FR obtained through CFD analysis reportedly reflects actual cerebral hemodynamic condition depending on anatomical structures. Kataoka et al. reported that vessel diameter and FR of intracranial arteries might be controlled so WSS remains constant after bypass surgery.[12] Zhu et al. reported that FR, vessel diameter, and pressure decrease in ICA on the surgical side declined significantly following surgery for moyamoya disease.[38] FR is calculated using the aforementioned equation [1] in image-based CFD analysis, primarily depending on vascular diameter. In the present study, FR of M1 on the lesion side was strongly correlated with the diameter of the contralateral ICA and the total diameters and FR of the inlet vessels. Consequently, the diameters of the associating vessels might be important for ischemic tolerance if the diameters of the AcomA or PcomA were sufficient for the CFD analysis.

Indications for recanalization with parent artery occlusion

In the present study, the most suitable cutoff values of FR and WSS for detecting the good collateral group obtained through ROC curve analysis were 126.2 mL/min and 4.54 Pa, respectively. These values represent the requirement for good collateral screening, and not indicating the need for revascularization surgery. The quantification of ischemic tolerance with FR and WSS obtained through virtual test occlusion using CFD analysis is expected to indicate ICA occlusion without high-flow bypass. However, the indications for revascularization surgery should be considered carefully with an appropriate safety margin to avoid delayed ischemic stroke. Regarding the requirement for revascularization surgery, various parameters using BTO have been reported previously. Abud et al. reported that carotid sacrifice without bypass was possible when the delay between the venous drainage of the injected and occluded hemisphere was <3 s.[1] Shimizu et al. reported no ischemic complications for endovascular parent artery occlusion in patients with a venous phase delay <1 s during BTO.[26] Labeyrie et al. reported delayed ischemic events due to hemodynamic context for endovascular parent artery occlusion in patients who had a well-tolerated BTO with a relative delay of venous filling of <4 s were most often transient and had a good prognosis.[15] Based on these findings, patients whose FR and WSS obtained through virtual test occlusion using CFD analysis more than 126.2 mL/min and 4.54 Pa might be treated by ICA occlusion without high-flow bypass with or without low-flow bypass, although prospective validation is needed to confirm this.

Limitations and future work

This study has several limitations. First, visualization of the small vessels such as the AcomA and PcomA was difficult because of partial volume effect and transformation to an STL format. The mean diameters of the AcomA and PcomA were reportedly 1.5 mm and 1.4–1.6 mm, respectively,[20,21] and vessels whose diameters were <1.0 mm were difficult to detect using our method. However, these cases all showed acute ischemic symptoms or poor collateral appearances, and there was no need to detect the risk of delayed ischemic symptoms using CFD analysis. Second, the technical quality of CTA imaging could affect the result. Finally, a difference must exist between CFD and the actual hemodynamic condition because no biological information was available. Blood flow must be distributed on demand according to peripheral vascular resistance in the actual mechanisms. In addition, blood cells could not be ignored in terms of the blood flow in small vessels such as the AcomA and PcomA. Despite these limitations, our findings indicate the utility of FR and WSS obtained through virtual test occlusion using the CFD analysis method is a non-invasive alternative to BTO. It will be necessary to confirm our results with future prospective and large sample studies.

CONCLUSION

Among the parameters obtained through CFD analysis, FR and WSS were correlated with collateral flow grade and MSP in addition to MTT. The good collateral group could be distinguished using virtual test occlusion and might be treated by ICA occlusion without high-flow bypass, although prospective validation is needed to confirm this. Most researchers focus on angiographic appearance and clinical data during BTO for detecting delayed ischemic stroke after parent artery occlusion, with little research having been done on non-invasively analyzing hemodynamic mechanisms. Virtual test occlusion might be used to evaluate ischemic tolerance as a non-invasive alternative to BTO.