Improved Registration of 3D CT Angiography with X-ray Fluoroscopy for Image Fusion During Transcatheter Aortic Valve Implantation

Podsumowanie

The aim of this study was to improve the co-registration for image fusion (IF) of pre-interventional CT data with real-time x-ray (XR) fluoroscopy during transfemoral transcatheter aortic valve implantation (TAVI).

Streszczenie

The fusion of 3D anatomical models derived from high-fidelity pre-interventional computed tomography angiography (CTA), and x-ray (XR) fluoroscopy to facilitate anatomical guidance is of huge interest for complex cardiac interventions like TAVI procedures with cerebral protection. Co-registration of CTA and XR has been introduced either based on additional intraoperative non-/contrast-enhanced cone-beam computed tomography (CBCT) or two separate aortograms. With the related increase of radiation exposure and/or contrast agent (CA) dose, a potential additional risk for the patient is introduced. Here, we propose a modified co-registration approach making use of arteriograms of the iliofemoral arteries, routinely performed during the femoral puncture and sheath introduction. On-the-fly refinement of the co-registration during the on-going procedure enables accurate co-registration without any additional angiograms, thus reducing CA, XR dose and procedure time, while simultaneously improving operator confidence and procedure safety.

Wprowadzenie

Image fusion (IF) is the process of superimposing datasets acquired at the different time- and viewpoints on different modalities into a single-frame of reference¹. XR is the most frequently used imaging modality for intervention guidance. Even though, providing high temporal and spatial resolution, XR has low dimensionality (2D projections) and lacks anatomical details. 3D organ shape models derived from e.g. high-quality pre-interventional CTA data superimposed onto the live fluoroscopy image can augment XR by relevant anatomic soft-tissue structures. Prerequisite step for IF is the co-registration of the different imaging modalities.

Typically, co-registration of preoperative 3D image datasets with XR fluoroscopy involves one of the following techniques²: a) image-based 3D-3D registration of the preoperative 3D dataset with an intraoperative non-/contrast-enhanced CBCT dataset^3,4,5,6, or b) direct image-based 2D-3D registration, where two angiographic images with a minimum of 30° angular spacing^7,8 are used for co-registration.

With the recent introduction of fusion packages on commercial XR systems, IF can be made more readily available for a wide range of applications. Using those systems, we have previously shown that it is technically feasible and safe to overlay an aortic root model by means of direct image-based 2D-3D registration for supporting transfemoral transcatheter aortic valve implantation (TAVI)⁸. Without compromising the overall CA or XR dose, IF proved itself highly valuable during TAVI procedure by adding 3D anatomical details to the conventional XR fluoroscopy image, especially during deployment of the cerebral protection device. However, the additional acquisition of the aortograms used for the co-registration required additional CA and XR dose. Therefore, an optimized workflow providing accurate IF without the need of any additional aortograms was highly desirable.

Here, we present an approach to improved co-registration of pre-interventional CTA with real-time XR without requiring any additional CA or C-arm CT scans for IF. The femoral access TAVI is performed as described elsewhere^9,10,11. Briefly, both femoral arteries are accessed: one for the guidance of the contralateral puncture, followed by placement of a pigtail catheter through a 6F sheath to allow arteriography during placement of the valve prosthesis; the second for placement of the valve delivery system and subsequent balloon valvuloplasty and device placement. Angiographic confirmation of appropriate puncture is performed in our institution as a standard of care for localization of the puncture height (above the femoral bifurcation) and estimation of the position of the covered stent in case of access-related complications¹². For capturing embolic debris, a double-filter cerebral protection system is introduced after insertion of the TAVI delivery sheath prior to the passage of the aortic arch with any further device.

We use arteriograms performed routinely during the puncture of the femoral arteries to establish the initial co-registration. On-the-fly refinement of the co-registration is subsequently performed during the on-going procedure using the position of the pigtail catheter within the aortic root, the double-filter cerebral embolic protection system in the supraaortal vessels and the aortograms performed before the implantation of the valve prosthesis, thus ensuring accurate model overlay at any time point during the intervention.

Protokół

The study protocol is in compliance with the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institution's ethics committee. Written informed consent was obtained from all individual participants included in the study (CSI-Ulm, clinicaltrials.gov NCT02162069).

1. CT Examination

1. Perform cardiac CT angiography with retrospective ECG-gating using 80 mL of iodinated contrast media injected at a flow rate of 4.0 mL/s followed by a flushing bolus of 70 mL 0.9% saline solution.
2. Acquire data cranio-caudally extending from the femoral arteries to the subclavian arteries with the following acquisition parameters: the matrix of 512 × 512 pixels, slice thickness of 1 mm, slice spacing of 0.7 mm.
3. Reconstruct images at 30% of the R-R interval using iterative reconstruction algorithm.

2. Image Segmentation and Model Generation

1. Import CT data from an external device or an internal archiving system into image fusion software by drag-and-dropping the dataset or subsets of data for the selected patient into the Patients view within the application.
2. Start automatic segmentation of the selected CT volume by double-clicking on the image series.
   Note: Segmentation of the heart chambers (left and right ventricle, left and right atrium, myocardium) and great vessels (abdominal aorta, vena cava, and coronary sinus) is started automatically and will be displayed in the Segmentation work step.
3. Go through the image slices and verify if the edges of the aorta and left ventricle are correctly detected and if necessary edit the automatic segmentation in the Tissue window using the Edit button.
4. To perform manual segmentation of additional structures, use the Add button in the Tissue window.
5. Use existing editing tools to fill the structure (Inject Dye) and drag the edges of the structure (Drag Edge) in 2D views or remove structure parts with the Free-form Cut in the 3D view.
6. Perform manual segmentation of the main trunks of the left and right coronary arteries, main artery branches (brachiocephalic artery with right carotid and right subclavian arteries, left common carotid artery and left subclavian artery), left and right iliofemoral arteries, hip bones, and hip joints as described in steps 2.4-2.5 (Figure 1).
   Note: Depending on the quality of the CT volume this may take 20 to 30 min.

3. Image Co-registration and Fusion

1. In the Patients view merge the selected patient with the current patient in the XR system by choosing the respective action in the right-click-context-menu.
   Note: Now the Registration and Live work steps are active and can be used.
2. In the Live work step, click Add new tag point in the Tag points window to place three reference markers at aortic valve cusps for facilitating the choice of the optimal projection of the annular plane during the intervention (Figure 2).
3. Go to the Registration work step to acquire the XR runs and to perform the co-registration of the segmented models with XR.
   Note: XR runs acquired with at least 30° angular distance are required for reliable registration.
4. Copy the angiographic projection of the appropriate puncture in ~LAO 20 - 30° orientation (or ~RAO 20 - 30° depending on the initial puncture side) into the Reference view 1 by clicking the button Copy to Reference View 1.
5. Copy following x-ray projections acquired in ~AP orientation during visualization of the transition from the contralateral A. iliaca communis to the A. femoralis communis into the Reference view 2 by clicking the button Copy to Reference View 2.
6. Use the interaction tools Registration Pan, Registration Rotate (for in-plane rotation), Registration Roll (for 3D rotation) to manually align the model of the iliofemoral arteries with the acquired XR projections (Figure 3A-B).
7. Use hip bones and hip joints visible in the XR images as additional landmarks during alignment of the overlay in the iliofemoral region.
   Note: Now the model is linked to the XR system geometry, and the overlay is automatically being adapted to the current XR projection orientation, magnification, and patient table position.
8. Guide the puncture of the common femoral artery on the sheath side (Figure 3C) by using the coarse initial co-registration performed in step 3.6.
9. Record an angiographic projection of the device sheath femoral artery in ~RAO 20 - 30° (or respectively ~LAO 20 - 30°) by clicking the button Copy to Reference View and finalize the image co-registration in the iliofemoral region (Figure 3D).
   Note: Since the patient is positioned differently during CT scan and the intervention, registration based on the iliofemoral structures provides only limited accuracy in other regions. Thus, manual refinement of the co-registration in the thoracic region is required.
10. To use these data for further overlay re-alignment, copy any additional acquired projections to the “Reference View” during the transition from the iliofemoral to the thoracic region as well as any catheterization of the brachiocephalic trunk via the right radial artery.
11. After placing the pigtail catheter in the aortic arch, acquire two fluoroscopic projections without CA in LAO 30 - 40° and RAO 20 - 30° orientation and copy them in the Reference view 1 and Reference view 2.
12. Use the interaction tools Registration Pan, Registration Rotate (for in-plane rotation), Registration Roll (for 3D rotation) to manually adjust the registration in the thoracic region (Figure 4A-B).
13. Guide the placement of the protection device without administration of additional CA purely based on the anatomic overlay (Figure 4C).
14. For further refinement, manually correct the overlay position as described in step 3.12 on each acquired XR projection during the entire course of the intervention, ensuring accurate overlay at any time point.
    Note: Use aortograms acquired according to the routine procedure for verifying the correct position of the delivery system as landmarks for the overlay adjustment (Figure 4D).

Wyniki

We introduced a novel co-registration approach for image fusion during TAVI, which allows overlaying the patient-specific anatomic model onto the live XR images during the entire TAVI procedure without the need for any additional aortograms.

Several interventional steps will benefit from the IF: (1) guidance of the puncture of the common femoral artery above the femoral bifurcation on the sheath side (Figure...

Dyskusje

The main focus of this study was to investigate the feasibility of IF without modifying the clinically established TAVI workflow. Whereas the gold standard for the co-registration of the pre-interventional CTA data and XR fluoroscopy uses dedicated aortograms⁸, we propose using multiple approximate registrations with on-the-fly refinements to provide accurate 3D model overlay during the entire course of the intervention.

The continuous manual registration refinement req...

Materiały

Name	Company	Catalog Number	Comments
Philips Allura FD10	Philips Healthcare		x-ray system
EP Navigator Release 5.2.10	Philips Healthcare		image segmentation and fusion SW
Iomeron 350	Bracco Imaging Deutschland GmbH		x-ray contrast agent
Sentinel  double-filter cerebral protection system	Claret Medical, Inc.		double-filter cerebral protection system
Matlab R2013	MathWorks		statistical analysis