Intra-Aneurysmal High-Resolution 4D MR Flow Imaging for Hemodynamic Imaging Markers in Intracranial Aneurysm Instability

Abstract

BACKGROUND AND PURPOSE:

Prediction of aneurysm instability is crucial to guide treatment decisions and to select appropriate patients with unruptured intracranial aneurysms (IAs) for preventive treatment. High-resolution 4D MR flow imaging and 3D quantification of aneurysm morphology could offer insights and new imaging markers for aneurysm instability. In this cross-sectional study, we aim to identify 4D MR flow imaging markers for aneurysm instability by relating hemodynamics in the aneurysm sac to 3D morphologic proxy parameters for aneurysm instability.

MATERIALS AND METHODS:

In 35 patients with 37 unruptured IAs, a 3T MRA and a 7T 4D MRI flow scan were performed. Five hemodynamic parameters—peak-systolic wall shear stress (WSS_MAX) and time-averaged wall shear stress (WSS_MEAN), oscillatory shear index (OSI), mean velocity, and velocity pulsatility index—were correlated to 6 3D morphology proxy parameters of aneurysm instability—major axis length, volume, surface area (all 3 size parameters), flatness, shape index, and curvedness—by Pearson correlation with 95% CI. Scatterplots of hemodynamic parameters that correlated with IA size (major axis length) were created.

RESULTS:

WSS_MAX and WSS_MEAN correlated negatively with all 3 size parameters (strongest for WSS_MEAN with volume (r = −0.70, 95% CI −0.83 to −0.49) and OSI positively (strongest with major axis length [r = 0.87, 95% CI 0.76–0.93]). WSS_MAX and WSS_MEAN correlated positively with shape index (r = 0.61, 95% CI 0.36–0.78 and r = 0.49, 95% CI 0.20–0.70, respectively) and OSI negatively (r = −0.82, 95% CI −0.9 to −0.68). WSS_MEAN and mean velocity correlated negatively with flatness (r = −0.35, 95% CI −0.61 to −0.029 and r = −0.33, 95% CI −0.59 to 0.007, respectively) and OSI positively (r = 0.54, 95% CI 0.26–0.74). Velocity pulsatility index did not show any statistically relevant correlation.

CONCLUSIONS:

Out of the 5 included hemodynamic parameters, WSS_MAX, WSS_MEAN, and OSI showed the strongest correlation with morphologic 3D proxy parameters of aneurysm instability. Future studies should assess these promising new imaging marker parameters for predicting aneurysm instability in longitudinal cohorts of patients with IA.

SUMMARY

PREVIOUS LITERATURE:

Studies have shown a correlation of 3D morphologic parameters, including size-related parameters, as well as flatness, curvedness, and shape index with aneurysm instability. Identifying additional imaging markers is crucial for assessing aneurysm instability and optimizing treatment decisions. The current progress in high-resolution 4D MR flow imaging enables in vivo detection of hemodynamics, including peak systolic wall shear stress, time-averaged wall shear stress, oscillatory shear index, mean velocity, and velocity pulsatility index in the aneurysm sac itself. To our knowledge, no study identified 7T 4D MR flow imaging markers for aneurysm instability by relating hemodynamics in the aneurysm sac to the aneurysm 3D morphology proxy parameters for aneurysm instability.

KEY FINDINGS:

In our study, wall shear stress calculated at peak systole showed the strongest correlation with area (r = −0.69, 95% −0.83 to −0.47), wall shear stress showed the strongest correlation with volume (r = −0.70, 95% −0.83 to −0.49), and oscillatory shear index showed the strongest correlation with major axis length (r = 0.87, 95% CI 0.76–0.93).

KNOWLEDGE ADVANCEMENT:

This study is an important first step to a better understanding of hemodynamics in the pathophysiologic process of aneurysm instability and demonstrates possible new hemodynamic imaging markers for aneurysm instability.

Unruptured intracranial aneurysms (IAs) have a prevalence of approximately 3% in the general adult population.¹ While most aneurysms stay stable during a lifetime, rupture leads to subarachnoid hemorrhage with severe consequences, including a 35% case fatality rate and 35% dependency in survivors.² Preventive treatment by endovascular or neurosurgical approaches can prevent subarachnoid hemorrhage but carries a significant risk of complications.³ Therefore, in clinical decision-making, the aneurysm rupture risk has to be weighed against the treatment complication risk. Currently, it is difficult to predict IA rupture. While IA size is one of the main predictors of aneurysm instability, a significant portion of ruptured aneurysms is small, because small aneurysms by far outnumber larger ones.^3–5 Identifying additional imaging markers beyond size is crucial for assessing aneurysm instability and optimizing treatment decisions.

Because hemodynamics play an important role in the development of aneurysms and aneurysm instability,⁶ high-resolution 4D MR flow imaging performed on 7T MRI might serve as a possible method to identify aneurysm instability.⁷ The current progress in 4D MR flow imaging enables in vivo detection of hemodynamics in the aneurysm sac itself. Recent advancements have also enabled 3D quantification of aneurysm morphology. Studies have shown a correlation of 3D morphologic parameters, including major axis length, volume, and surface area (size parameters) as well as flatness, curvedness, and shape index with aneurysm instability. In this way, these morphology parameters can play a role as proxy parameters for aneurysm instability.^4,5,8,9

The aim of this study is to identify 4D MR flow imaging markers for aneurysm instability by relating hemodynamics in the aneurysm sac to the aneurysm 3D morphology proxy parameters for aneurysm instability.

MATERIALS AND METHODS

Study Population

This cross-sectional study consisted of a consecutive series of patients undergoing clinical routine follow-up imaging for an unruptured and untreated IA in the period between June 2021 and July 2022. All patients met the inclusion and exclusion criteria of this study. Inclusion criteria were patients over 18 years old with unruptured IAs larger than 4 mm. Patients who were pregnant or who had previous treatment for an additional IA were excluded. Clinical care consisted of a 3T TOF-MRA follow-up scan. As part of this study, an additional 7T 4D MR flow imaging scan was performed at the University Medical Center Utrecht, the Netherlands. This was on the same day as the 3T TOF-MRA scan in patients who gave informed consent for this study and did not have a contraindication for 7T MRI. The local ethics review committee approved the study.

MRI Protocol

The 3T (Philips Healthcare) MRA scans were performed by using a 32-channel head coil. The protocol included a 3D T1-weighted gradient-echo TOF-MRA acquisition with the following scan parameters: FOV 200 (Feet-Head) × 200 (Left-Right) × 80 (Anterior-Posterior) mm³, acquired spatial resolution 0.5 × 0.5 × 0.5 mm³, TR/TE = 22/3.4 ms, flip angle =18°, and acquisition duration around 5 minutes. The 7T MRI (Philips Healthcare) examinations were performed by using an 8-channel volume transmit coil in quadrature mode and 32-channel receive coil (Nova Medical) for the acquisition of 4D flow data. The 4D flow scan was performed using the Amsterdam University Medical Center PROspective Undersampling in multiple Dimensions (PROUD) software patch, which enables a pseudospiral k_y/k_z-plane acquisition scheme designed for incoherent undersampling with a variable sampling attenuation.¹⁰ A 4D phase-contrast MRI acquisition was performed with the field-of-view encompassing the IA and its parent vessel. Retrospective gating was performed by using a peripheral pulse unit for heartbeat detection. Parameters for the 4D phase-contrast MRI scan were as follows: angulated coronal FOV 190 (Feet-Head) × 190 (Right-Left) × 20 (Anterior-Posterior) mm³, acquired and reconstructed resolution of 0.5 × 0.5 × 0.5 mm³, TR/TE = 7.0/2.2, velocity encoding sensitivity 50 cm/s, flip angle 10°, nominal acceleration factor 7, and 12 reconstructed cardiac phases (ie, reconstructed temporal resolution 83 ms for a heart rate of 60 bpm). A previous study¹¹ investigated that wall shear stress (WSS) increases with more spatial resolution in experimental phantom studies. For this reason, we used the high acquired spatial resolution of 0.5 × 0.5 × 0.5 mm, which is made possible by the undersampling technique used and more importantly by the surplus of signal from 7T. To improve field homogeneity, vendor-supplied image-based shimming was performed by using second-order terms, before acquiring the 4D flow scan. Acquisition duration was 10 minutes. Data sets were reconstructed with the pipeline developed for the PROUD data by Amsterdam University Medical Center, by using the Berkeley Advanced Reconstruction Toolbox.¹² The reconstruction pipeline included phase subtraction and background phase correction in MRecon (Gyrotools).

Morphology Proxy Parameters

In total, 6 3D-quantified morphology parameters were derived: major axis length, volume, surface area (size parameters), flatness (axis length ratio), curvedness (local morphology), and shape index. 3D morphologic measurements were performed as described in detail in a previous study and morphology parameters are visualized in this same study.⁵ Shape index is a rotation and translation invariant measure which describes the topology of the unruptured intracranial aneurysm surface. The values of the shape index range from −1 to 1, which correlates with a concave (or “cup” shape) or convex (or “dome” shape) respectively. For example, a shape index of 0.3 (on a scale between −1 to 1) indicates a more subtle convex shape of the aneurysm in comparison with a shape index of 0.9 that describes a more definite convex shape.⁵ Segmentations of the IAs were made on the 3T TOF-MRAs by manually drawing contours per imaging slice around the IA, by using an in-house–developed software implemented in MeVisLab (MeVis Medical Solutions), by an experienced neuroradiologist (I.v.d.S. with 15 years of experience). Following the segmentation, the software used a marching cubes algorithm to fit a mesh around the IA. This mesh was then used to calculate the aneurysm’s volume and surface area. Next, the major, minor, and least axis of the IA were calculated by using principal component analysis. From these principal components, various morphology parameters were calculated according to the Image Biomarker Standardization Initiative (IBSI) guidelines.¹³

Hemodynamic Parameters

We quantified 5 hemodynamic parameters (peak systolic wall shear stress [WSS_MAX], time-averaged wall shear stress [WSS_MEAN], the oscillatory shear index [OSI], mean velocity, and velocity pulsatility index [vPI]) in the aneurysmal sac. The reconstructed 4D flow data sets were analyzed by R.J.v.T. (with more than 4 years of 4D flow analysis experience) by using CAAS MR Solutions v5.1.2 software (Pie Medical Imaging) (Fig 1). CAAS automatically generates the centerline and perpendicular slices for the region of interest along the complete circle of Willis from the internal carotid artery toward the smaller cerebral arteries. These perpendicular slices were visually checked and automatically propagated over all time points of the cardiac cycle to create volumetric flow rate traces and velocity traces over the cardiac cycle. The blood flow vPI (vPI = (Velocity_max − Velocity_min)/Velocity_mean) was calculated from each velocity trace. The inter- and intraobserver reliability of the analysis with the CAAS software in the circle of Willis has been tested previously with good to very good reproducibility and repeatability (intraclass correlation coefficient = 0.65–0.96).¹⁴ WSS was calculated by using the segmentations performed for the 3D aneurysm morphology quantification, and previously described software in Matlab R2018a, by multiplying the wall shear rate by the dynamic viscosity of blood (3.2 × 10⁻³Pa·s).¹¹ We performed 2 types of WSS calculations: WSS calculated at peak systole (WSS_MAX) and time-averaged WSS (WSS_MEAN), in which the WSS is expressed as the average over the cardiac cycle. The WSS_MAX and WSS_MEAN were expressed as the spatial mean value (Fig 1). The OSI was defined as the fluctuation of WSS over 1 cardiac cycle. OSI is a nondimensional parameter and ranges from 0 (no change) to 0.5 (oscillating flow).¹⁵

FIG 1.

An included subject with an anterior communicating artery aneurysm with a major axis length of 8.5 mm.

Sample Size Calculation

To detect a minimum correlation coefficient value of 0.5, with a desired type II error of 10% and α of 0.05, a sample size of 37 aneurysms is sufficient.¹⁶

Statistical Analysis

The normality of the data were tested by using the Kolmogorov-Smirnov test of normality. We calculated Pearson correlation coefficient r with 95% CI to study relationships between the 5 hemodynamic parameters (WSS_MAX, WSS_MEAN, OSI, mean velocity, and vPI) and the 6 morphologic proxy parameters (major axis length, volume, surface area, flatness, shape index, and curvedness). Additionally, we visualized the relationship between the hemodynamic parameters that correlated with IA size (major axis length) by creating scatterplots. Statistical analyses were performed in Statistics Package for Social Sciences (SPSS, version 25, IBM).

RESULTS

Baseline Characteristics

In total, 35 patients (10 men, mean age 66 ± 9) with 37 unruptured IAs were scanned. The aneurysms had a mean major axis length ± standard deviation of 7.0 ± 1.7 mm (range 4.4–10.3 mm). The most common aneurysm location was the middle cerebral artery (46%), followed by the anterior communicating artery (27%), the internal carotid artery (22%), and the basilar artery (5%). Other baseline patient and IA characteristics (location, morphology, and hemodynamics) are summarized in Table 1. The outcomes of IA morphologic and hemodynamic parameters are given in mean values and standard deviations.

Table 1:

Patient and IA characteristics

Patient Characteristics
No. of patients	35
No. of aneurysms	37 (2 patients with 2 IAs)
Men	10 (29%)
Age (years, mean ± SD)	66 ± 9
Mean aneurysm size (mm, mean ± SD)	7.0 ± 1.7
Hypertension	19 (54%)
Systolic/diastolic blood pressure before MRI (mm Hg, mean ± SD)	142 ± 19/85 ± 10
IA location (number, proportion)
Anterior communicating artery	10 (27%)
Basilar artery	2 (5%)
Internal carotid artery	8 (22%)
Middle cerebral artery	17 (46%)
IA morphology (mean ± SD)
Major axis length (mm)	7.0 ± 1.7
Area (mm2)	153 ± 70
Volume (mm3)	132 ± 91
Flatness	0.81 ± 0.05
Shape index	0.30 ± 0.11
Curvedness	2.39 ± 0.18
IA hemodynamics (mean ± SD)
WSSMAX (Pa)	3.9 ± 1.5
WSSMEAN (Pa)	0.94 ± 0.27
OSI	0.27 ± 0.06
Mean velocity (cm/s)	23 ± 7
vPI	0.73 ± 0.11

Note:—SD indicates standard deviation.

Correlations between Hemodynamics and Morphology

All hemodynamic and morphologic parameters were normally distributed. WSS_MAX and WSS_MEAN correlated negatively with the 3 size-related parameters: major axis length (r = −0.67, 95% CI −0.82 to −0.44 and r = −0.64, 95% CI −0.80 –to −0.40, respectively), volume (r = −0.67, 95% CI −0.82 to −0.44 and r = −0.70, 95% CI −0.83 to −0.49, respectively), and area (r = −0.69, 95% CI −0.83 to −0.47 and r = −0.67, 95% CI −0.82 to −0.44, respectively). OSI correlated positively with the size-related parameters, including major axis length (r = 0.87, 95% CI 0.76–0.93), volume (r = 0.77, 95% CI 0.59–0.88), and area (r = 0.76, 95% CI 0.58–0.87).

WSS_MAX and WSS_MEAN correlated positively with shape index (r = 0.61, 95% CI 0.36–0.78 and r = 0.49, 95% CI 0.20–0.70, respectively) and OSI correlated negatively with shape index (r = −0.82, 95% CI −0.90 to −0.68). OSI correlated positively with flatness (r = 0.54, 95% CI 0.26–0.74). WSS_MEAN correlated negatively with flatness (r = −0.35, 95% CI −0.61 to −0.029). No correlations were found between mean velocity or vPI and the morphologic parameters, except for mean velocity, which correlated negatively with flatness (r = −0.33, 95% CI −0.59 to −0.007).

Correlation between hemodynamic parameters and morphologic 3D proxy parameters of aneurysm instability is shown in Table 2 and is visualized in Fig 2. Scatterplots for the relationships of WSS_MEAN, WSS_MAX, and OSI with aneurysm size are shown in Fig 3.

Table 2:

Correlation between hemodynamic parameters and morphologic 3D proxy parameters of aneurysm instability

		Major Axis Length	Volume	Area	Flatness	Shape Index	Curvedness
WSSMAX	Pearson ρ	−0.67	−0.67	−0.69	0.03	0.61	0.04
	95% CI	−0.82 to −0.44	−0.82 to −0.44	−0.83 to −0.47	−0.3 to 0.35	0.36 to 0.78	−0.29 to 0.36
WSSMEAN	Pearson ρ	−0.64	−0.70	−0.67	−0.35	0.49	0.08
	95% CI	−0.80 to −0.40	−0.83 to −0.49	−0.82 to −0.44	−0.61 to −0.029	0.2 to 0.7	−0.25 to 0.39
OSI	Pearson ρ	0.87	0.77	0.76	0.54	−0.82	−0.12
	95% CI	0.76 to 0.93	0.59 to 0.88	0.58 to 0.87	0.26 to 0.74	−0.9 to −0.68	−0.43 to 0.21
Mean velocity	Pearson ρ	−0.05	0.04	0.01	−0.33	−0.08	−0.13
	95% CI	−0.37 to 0.28	−0.29 to 0.36	−0.32 to 0.33	−0.59 to −0.007	−0.39 to 0.25	−0.44 to 0.2
vPI	Pearson ρ	−0.09	−0.06	−0.03	0.19	0.03	0.12
	95% CI	−0.4 to 0.24	−0.38 to 0.27	−0.35 to 0.3	−0.14 to 0.48	−0.3 to 0.35	−0.21 to 0.43

Note:—The bold fond in table indicates the correlations between the hemodynamic and morphological parameters.

FIG 2.

Visualization of correlations for hemodynamic versus morphologic parameters.

FIG 3.

Scatterplots for peak-systolic WSS_MAX, time-averaged WSS_MEAN (pascal), and OSI on the y-axis, related to major axis length on the x-axis.

DISCUSSION

In this study, we demonstrated that the 4D flow hemodynamic parameters WSS_MAX, WSS_MEAN, and OSI showed the strongest correlation with 3D morphologic parameters that are proxy-parameters of aneurysm instability, as described in previous literature. No correlations were found between mean velocity or vPI and the morphologic parameters, except for a modest negative correlation of mean velocity correlating with flatness.

A study by using 3T 4D flow imaging to correlate aneurysm hemodynamics to IA morphologic parameters size and size ratio in 70 unruptured IAs, showed a negative correlation between the WSS_MEAN and IA size-related parameters.^14,17 This supports our findings with 7T MRI. In previous literature, both low WSS and high WSS have been associated with aneurysm instability, a paradox that has yet to be clarified.^6,18 Our findings support the low WSS theory, which states that low WSS along the vessel wall causes damage to endothelial cells and that vascular remodeling occurs due to pro-inflammatory changes, enhancing the development of IAs.¹⁸

One study¹⁹ on morphologic and hemodynamic risk factors for aneurysm rupture investigated the morphologic and hemodynamic parameters of 23 aneurysms ≥7 mm, consisting of 11 unruptured and 12 ruptured aneurysms. This study used 3D rotational angiography and computational fluid dynamics to analyze all IAs. The WSS_MEAN was found to be lower and the OSI to be higher in ruptured (instable) aneurysms in comparison to the unruptured aneurysms, and therefore, they were regarded as indicators of rupture risk.¹⁹ The findings in this study are in line with our findings that WSS_MEAN and OSI are correlated significantly with morphology proxy parameters of IA instability.

Whereas the literature is consistent regarding the size parameters (IA major axis length, volume, and area) and shape index in relation to IA instability, the interpretation of flatness remains challenging due to contradictory findings in the literature. One study found that an increase in flatness, measured over time, is associated with aneurysm instability,⁵ while another study found low flatness, measured at baseline, to be associated with unstable aneurysms.⁴ In our study WSS_MEAN, OSI, and mean velocity correlated significantly with flatness although the correlations between the hemodynamic parameters with flatness were less strong compared with other morphologic parameters. This might be due to the contradictory association of flatness as a proxy parameter for IA instability. In addition, according to the sample size calculation, the minimum correlation coefficient of significance was 0.5. Both WSS_MEAN and mean velocity had a correlation coefficient falling below 0.5, hence indicating unclear clinical relevance.

Mean velocity was not correlated to other morphologic parameters, which is in line with another study in which they could not demonstrate any significant association with velocity and IA size.¹⁷ Therefore mean velocity might be less valuable as a new imaging marker for IA instability.

A correlation between vPI and IA rupture was described in a retrospective cohort of 4 ruptured IAs, whereby pulsatility and nonsphericity were the only consistent predictive risk factors before and after rupture for all 4 IAs.²⁰ We found no correlations between vPI and any of the morphology parameters. The small number of IAs and the difference in study design of the aforementioned study regarding the precise location of measurements and the pulsatility calculation steps could explain the differences.

No studies have been performed relating hemodynamic parameters to shape index of the aneurysms. A reason for this might be that shape index is not one of the IBSI parameters, but rather describes the local shape of the aneurysm. However, because shape index is related to IA instability,⁵ our study provides new insights and information by demonstrating that WSS_MAX and OSI both correlated to the parameter shape index.

Gadolinium enhancement of the IA wall or aneurysm wall enhancement (AWE) is another potential imaging marker that is associated with aneurysm instability in cross-sectional studies. A recent multicenter longitudinal cohort study on AWE and risk of aneurysm growth and rupture showed that AWE predicts aneurysm growth or rupture during short-term follow-up, but not when adjusting for aneurysm size.²¹ To our knowledge, 3 studies relating IA wall enhancement to both 3D quantified morphology and hemodynamics have been performed on this topic.^22-24 These 3 studies all had a different study design and the 3D morphology characteristics were not defined according to the IBSI guidelines. Furthermore, the hemodynamic assessment was performed with computational fluid dynamics modeling and not with 4D flow imaging, impeding comparison with our findings. Currently no consensus exists on the relationship between morphology, hemodynamics, and wall enhancement. The correlation between the latter is of added value for developing an optimized imaging strategy for IAs and should be assessed in future studies.

A recent systematic review²⁵ investigated the relationship between computational fluid dynamics in the circle of Willis and IA formation, growth, or rupture in 3 case series of 13 aneurysms in total. Due to limited number of cases and the heterogeneity between the 3 different case series, no conclusions were drawn on the relation between hemodynamics, the development, and instability or rupture of unruptured IAs. The reviewed studies in this systematic review all had a cross-sectional design, which does not allow one to determine if the differences in hemodynamics are the cause or the consequence of IA rupture.

The strengths of this study were the ultra-high-field 7T MRI 4D flow with 0.5 mm isotropic 4D flow data sets within 10 minutes of scan time with increased SNR and resolution compared with 3T MRA. The advantage of 7T 4D flow imaging in measuring hemodynamics is that it provides, in vivo quantification of hemodynamics with inclusion of the time dimension that allows the calculation of time-resolved parameters such as oscillatory shear index and WSS_MAX. This would not have been possible in comparison with 3D methods such as TOF. By providing supporting evidence that 4D hemodynamic measures correlate with morphologic parameters, we aimed to elucidate the pathologic process of aneurysm formation.¹⁴ Moreover, we correlated our in vivo intra-aneurysmal hemodynamics with 3D-quantified IA morphology parameters that are correlated to aneurysm instability. These 3D-quantified morphology parameters were calculated based on definitions in accordance with the IBSI guidelines.¹³

This study has some limitations. The cross-sectional study design does not allow us to draw conclusions on the relationship between hemodynamic parameters and IA instability over time. Future studies should be of longitudinal design and investigate the dynamic interaction between hemodynamics and IA instability over time and study the prognostic value of all hemodynamic and IBSI standardized 3D morphology parameters in prediction of IA growth and IA rupture. Next, future translational studies should explore the applicability of 7T 4D flow results on the clinically used 3T scanner, enabling the acquisition of hemodynamic studies in larger populations. This study included IAs with a size >4 mm, and therefore, it does not allow us to draw conclusions for aneurysms with a size smaller than 4 mm. Finally, according to our sample size calculation, we had statistical power to demonstrate a correlation of 0.5 or higher. Therefore, the clinical relevance of correlation coefficients being lower than 0.5 is unclear.

CONCLUSIONS

In this study we demonstrate that 4D flow hemodynamics WSS_MAX, WSS_MEAN, and OSI correlated strongest with morphologic parameters that are proxy-parameters of aneurysm instability. This study is an important first step to a better understanding of hemodynamics in the pathophysiological process of aneurysm instability and demonstrates possible new hemodynamic imaging markers for aneurysm instability.

ABBREVIATIONS:

AWE

aneurysm wall enhancement

IA

intracranial aneurysm

IBSI

Image Biomarker Standardization Initiative

OSI

oscillatory shear index

PROUD

PROspective Undersampling in multiple Dimensions

WSS

wall shear stress

WSS_MAX

peak systolic wall shear stress

WSS_MEAN

time-averaged wall shear stress

vPI

velocity pulsatility index