Patient selection
Patients who underwent venous sinus stenting procedures with a 3D MRV-fusion overlay between April and December of 2017 were included in this study, which was approved by our Institutional Review Board with a waiver of informed written consent. Cases with poor quality pre-procedural MRVs (with motion degradation or inadequate visualization of the sinuses) were excluded. 10 consecutive cases fit these criteria during this time period and were included in the study.
Procedural stages to create and integrate fusion images
Building the 3D MRV model
As part of the standard pre-procedural assessment, all patients from the study group had a thin-slice, contrast-enhanced MR Venogram. All institutional MRV exams were performed on 3 Tesla units (Siemens MAGNETOM Skyra and GE Healthcare Signa Architect) with coronal 2D time of flight, sagittal 3D phase contrast (INHANCE/NATIVE), and 3D thin section T1 pre and postcontrast sequences (SPGR or MPRAGE). Contrast enhanced sequences were acquired following weight-based intravenous administration of gadobutrol (Gadavist). The fusion images used for MRV-guidance relied on two 3D models: a vessel model and a skull model. The vessel model was created using vessel segmentation software (Vessel ASSIST, GE Healthcare, Chicago, IL) for 3D extraction of the venous sinuses on a preprocedural MR venography series. The segmentation required the selection of two end points, one in the superior sagittal sinus and one in the sigmoid sinus, on the multi-planar reformatted MRV images (Fig. 1, a-b). The skull model was used to facilitate accurate registration between the pre-procedural MRV models and live fluoroscopy (Figs. 1, d and 2).
Registering the 3D MRV model with live fluoroscopy
Registration between the 3D model and fluoroscopy was performed using a process involving two fluoroscopic acquisitions performed at different obliquities in the frontal and lateral planes. The 3D skull model from the preoperative MRV was aligned with the cranium in each fluoroscopic view (Fig. 3), and the registration was confirmed by imaging the vessels with either an inserted guidewire or a contrast injection. The registered model of the venous sinus was fused with live fluoroscopy so that the operator could visualize the 3D overlay on both the frontal and lateral planes by using the related foot pedal. Landmarks such as Points of Interest that were marked on the 3D model were also displayed on the overlay and were used to denote the location of the cortical veins, stenosis, and other landmarks as necessary.
Registration of the 3D model could be easily refined by the operator during the procedure, if necessary, from the tableside or from the workstation. The opacity and threshold of the 3D model were also adjustable at tableside to allow for better visualization of catheters and wires during the procedure.
The registration required 1–2 min of procedural time and was performed with the patient on the table. The time spent in the registration process mostly consisted of image capture, and the additional time to register the 3D model with the lateral plane as opposed to only the frontal plane was negligible. Building the models and optimizing their appearance for an overlay required approximately 3–5 min by the technologist in the control room without additional impact on procedural time, as the model could be built prior to the case itself.
The 3D model and cortical vein ostia markers were used primarily during stent deployment to optimize the placement of the proximal and distal landing positions of the stent and make sure there was no overlap of the stent with cortical vein ostia as seen in Fig. 3. The overlaid 3D model provided the benefit that it could be seen at all bi-plane angulations without re-registration modification, unlike a 2D roadmap.
Image review
Images from the 3D MRV overlay cases included in the study were retrospectively assessed by three neuro-interventionalists with 12, 2, and 2 years of experience, who were not involved with the treatments. First, the neuro-interventionalists reviewed the 2D pre-stent deployment images without MRV-fusion from each case to evaluate their confidence levels in their understanding of the cortical venous anatomy by consensus on a 3-point Likert scale, from 1 - low confidence to 3 - high confidence. They then reviewed images with the 3D overlay to evaluate the same and ranked their confidence levels on the same scale. Cases were reviewed in a random order and blinded fashion.
Statistical analysis
Categorical variables were presented as numbers and percentages and were compared using a Paired T Test at a 99% confidence interval to compare the mean operator confidence levels for the 2 groups. Analyses were conducted using Microsoft® Excel® (Build 11,929.20708) and Minitab® 17 statistical software (2010) (version 17.3.1; Minitab Inc.).