Regional Aneurysm Wall Enhancement is Affected by Local Hemodynamics: A 7T MRI Study

Image Data

Twenty patients with 23 unruptured intracranial aneurysms were prospectively enrolled in our study of aneurysm characterization with 7T MR imaging. The MR imaging studies included a TOF-MRA and 3D T1-weighted MR imaging sequences acquired before and after intravenous injection of gadolinium (Gd) (Online Supplemental Data). Additionally, 13 patients were imaged with 3D rotational conventional angiography (3DRA) and another 4 with CTA as the standard of care and as part of their presurgical assessment if they were scheduled for an intervention. Otherwise, 3DRA or CTA or both were performed during their clinic visit if the aneurysm was not going to be treated. 3DRA, CTA, and 7T MR imaging were performed within 2 weeks as patients were recruited into the study during presurgical planning. For patients who were not treated, imaging was also completed within 2 weeks for consistency. Patient and aneurysm characteristics are summarized in the Online Supplemental Data.

Hemodynamics Modeling

Using previously described methods,²³ patient-specific aneurysm models were reconstructed from the available 3D images (in order of preference, 13 from 3DRA, 4 from CTA, and 6 from 7T TOF-MRA), and computational fluid dynamics (CFD) simulations were performed by numerically solving the Navier-Stokes equations. Pulsatile inflow conditions were prescribed using waveforms measured in the internal carotid and vertebral arteries of older patients scaled with an empirical power law of the vessel area.²⁴ Outflow boundary conditions were specified consistently with the Murray’s law. Vessel walls were approximated as rigid. Simulations were performed for 2 cardiac cycles with a time-step of 0.01 seconds, and results from the second cycle were used to characterize the local hemodynamics of the aneurysm and AWE regions.

Mapping AWE Regions

The methodology to map AWE regions to CFD models is illustrated in Fig 1. Step 1 is the construction of a vascular model from the available 3D images, as described above. Bilateral 3DRA images (Fig 1A, -B) are used to construct a model of an anterior communicating artery aneurysm (Fig 1C).²⁵ Step 2 consists of performing a rough segmentation of the vasculature from the TOF-MRA (Fig 1D) and aligning it with the 3D model (Fig 1E, segmentation = red, model = transparent) using the image-processing software Amira Software (Thermo Fisher Scientific). In step 3, pre- and post-Gd T1-weighted images (Fig 1F, -G) are first subtracted (Fig 1H); then, AWE regions are manually segmented Fig 1I). Areas of AWE were subjectively determined by 2 teams of experienced investigators (E.A.S. and J.A.R., and S.H. and J.R.C.) and fused together by overlapping the regions. The high spatial resolution of 7T imaging allows visual determination of areas of enhancement (Fig 1F, -I) within the aneurysmal wall. In step four, 3D surfaces are recovered by extracting isosurfaces in Amira. These AWE region surfaces were thus already aligned with the 3D vascular model (Fig 1J) and were then projected onto the vascular model and used to label its surface elements as “enhanced” or “nonenhanced”. Finally, AWE regions are split into connected components and assigned different labels (colors) (Fig 1K). Note that in cases in which the vascular model was constructed from the TOF-MRA images, the AWE regions are automatically aligned and step 2 is skipped.

• Download figure
• Open in new tab
• Download powerpoint

FIG 1.

Illustration of methodology with the anterior communicating artery aneurysm. Step 1: model construction from 3DRAs (A–C). Step 2: rough segmentation of TOF-MRA and alignment with the vascular model (D and E). Step 3: pre- and post-Gd MR image subtraction and AWE region segmentation (F–I). Step 4: 3D-AWE region isosurface extraction, alignment, projection to the vascular model, and splitting into connected components (J and K). Models constructed from CTAs follow a similar approach, while models from TOF-MRA are automatically aligned to the Gd MR imaging series. VWI indicates vessel wall imaging.

AWE Region Characterization

The local hemodynamic environment in the aneurysm regions with and without enhancement were then quantitatively characterized by computing the region average of a number of hemodynamic variables,²⁶ including the following: peak systole and mean (time-averaged) wall shear stress magnitude (pWSS, mWSS, respectively), oscillatory shear index, relative residence time, mean gradient oscillatory number, mean WSS divergence, and mean WSS gradient.

The location of the AWE regions on the aneurysm sac as well as with respect to the inflow stream were characterized as described below and illustrated in the Online Supplemental Data.

As in previous studies,²⁶ the aneurysm neck was delineated by interactively selecting points on the neck and connecting them along geodesic lines (lines of minimum distance on the surface) (Online Supplemental Data). The geodesic distance to the neck was then computed for all points in the aneurysm. The aneurysm was then subdivided into 3 subregions (Online Supplemental Data) corresponding to the neck (pink), body (red), and dome (yellow).²⁶ Then, the location of each single-connected AWE region was defined as in the neck, body, or dome, depending on the maximum percentage of overlap of the area of the AWE region and each aneurysm region. The example in the Online Supplemental Data shows 3 regions of wall enhancement (white contours) superimposed with the aneurysm regions. In this case, 1 region is located at the neck, and the other 2, at the aneurysm body.

The location of AWE regions was also characterized according to their distance to the inflow. On the basis the labels defining each region (Online Supplemental Data), close tetrahedral elements of the CFD grid were identified (Online Supplemental Data). These elements were then used to seed streamlines that pass near the AWE region. Starting at one of these seed elements, we constructed and concatenated 2 randomly selected streamlines, one in the negative velocity direction (ie, toward the inflow) and another in the positive velocity direction (ie, toward the outflow) (Online Supplemental Data). The time that it takes a particle to flow from the inflow to the seed and from the seed to the outflow was calculated. This time is given as a percentage (ie, the time from inflow to outflow is 100%). The process was repeated for 10 different seeds, and the average “time-to-inflow” was assigned to the region. AWE regions were then classified as “close” or “far” from the inflow, depending on whether their time-to-inflow was smaller or larger than the median of this variable over all AWE regions in our sample.

Data Analysis

Paired 2-sample nonparametric Wilcoxon (Mann-Whitney) tests were used to compare the means of hemodynamic variables between aneurysm regions with and without enhancement. In this case, variables were normalized with the value corresponding to the entire aneurysm, and the AWE region of each aneurysm was considered a single region (union of all connected components).

Next, single-connected AWE regions located at the neck, body, or dome were compared with the nonenhancement regions of the same aneurysm. Thus, the ratio of hemodynamic variables in each AWE region over the nonenhanced region of the same aneurysm was computed, and single-sample Wilcoxon tests were performed to test whether the ratios were significantly different from 1. Similar comparisons were also made between AWE regions located near and far from the inflow and the nonenhanced regions of the same aneurysm.

Finally, hemodynamic characteristics of AWE regions at different locations on the aneurysm (neck, body, or dome) and relative to the inflow (close, far) were compared. Thus, the ratio of variables computed in AWE regions divided by the nonenhanced region of the same aneurysm were compared using 2-sample Wilcoxon tests.

All statistical tests were performed in R statistical and computing software (http://www.r-project.org/). Differences were considered significant with P < .05 for 2-sided tests, while trends were considered marginally significant at P < .05 in corresponding 1-sided tests.