CT-Fluoroscopy Image-Fusion Guidance for Embolization of Aortopulmonary Collaterals

Background. In adults with congenital heart disease, anatomically complex culprit collateral vessels may cause life-threatening hemoptysis and require catheter-based embolization. Techniques using conventional 2-dimensional (2D) fluoroscopy can be challenging. Technique. We describe a technique using 2D/3-dimensional (3D) image fusion for intraprocedural guidance to embolize aortopulmonary collaterals. Two fluoroscopic images of the thorax at least 30° apart with pigtail catheter in the ascending aorta were used for image fusion with preprocedural computed tomography (CT) angiography using the spine and pigtail catheter as landmarks. 3D planning information was overlaid on 2D fluoroscopy for cannulation and embolization. Results. Between November 2018 and June 2019, a total of 6 sessions of aortopulmonary collateral embolization using CT-fluoroscopy image-fusion guidance were conducted in 3 patients with adult congenital heart disease. In 3/6 sessions, the indication for embolization was hemoptysis. Common target vessels were left and right bronchial arteries (4 and 3 sessions, respectively). The spine and a pigtail catheter in the aorta were frequently used as landmarks for image fusion (67%). Particle embolization was used in 100% of cases. Mean procedure and fluoroscopy times were 3 hours, 23 minutes and 1 hour, 3 minutes, respectively. On average, 169 mL (350 mg iodine/mL) of contrast material was used in each session and total skin dose of radiation exposure was 1538 mGy. Successful collateral embolization was confirmed by postprocedure angiography that showed negligible or no flow through culprit collaterals. Conclusion. Use of CT-fluoroscopy image-fusion guidance can aid in embolization of aortopulmonary collaterals with complex anatomy in 3D space.

J INVASIVE CARDIOL 2021;33(6):E451-E456. doi:10.25270/jic/20.00498

Key words: coil, congenital heart disease in adults, device, electron beam CT, embolization, imaging, multidetector CT, transcatheter

Aortopulmonary collateralization often develops in patients with congenital heart disease (CHD). These thin-walled, fragile collateral vessels easily rupture; hence, CHD is the leading cause for non-inflammatory hemoptysis in the United States.^1,2 Hemoptysis can be a life-threatening emergency and aortopulmonary embolization is considered the first-line treatment for cases with massive and recurrent hemoptysis.¹ Embolization can be used to treat hemoptysis or as prevention in patients with known collaterals undergoing high-risk cardiac surgery to prevent perioperative bleeding complications. Aortopulmonary collateral networks can arise from the descending aorta, as well as the bronchial, intercostal, and subclavian arteries. With the evolving technological advancements in cross-sectional imaging techniques, such as computed tomography (CT) or magnetic resonance (MR) angiography, 3-dimensional (3D) morphology of the target collateral vessels, including their aortic origin and lung parenchyma, can be very well imaged and visualized. Anatomy of the collateral vessels including bronchial arteries in the normal population is highly variable to begin with, and the collaterals that require embolization may lie in different anatomic planes, making it difficult to rely solely on 2-dimensional (2D) fluoroscopic images for guiding intervention in the cardiac catheterization laboratory.^3,4

This series included 6 sessions of aortopulmonary collateral embolization using CT-fluoroscopy image-fusion guidance in 3 patients with adult CHD at an academic quaternary center over an 8-month period from November 2018 to June 2019. The hospital institution review board approved the study and all patients provided informed consent. Data collected included patient demographics, cardiac history, anatomic details of collaterals, indication for collateral embolization, image-fusion technique, and procedural details (embolization material, procedure time, fluoroscopy time, radiation dose, contrast usage). 

2D/3D CT-fluoroscopy image-fusion technique. Preoperatively acquired 3D datasets including CT angiography or MR imaging can accurately map the patient-specific heart model and provide guidance to the interventionalist. ⁵ Furthermore, 2D/3D CT-fluoroscopy image fusion is being increasingly used in the cardiac catheterization laboratory for 3D planning, marking vascular regions of interest and overlaying them on 2D fluoroscopy for intraprocedural guidance. In the presented cases, the following 3 steps were involved in 2D/3D image fusion: 

(1) Preprocedural planning. 3D multiplanar reformatting of CT angiography images were performed to identify and electronically mark collateral-vessel centerlines and their takeoffs; C-arm working projection angle was determined for each target vessel by optimal visualization of collaterals in 2D (Figures 1 and 2). It is a common prerequisite to have 1 mm thin-cut CT slices for this purpose to enable better visualization and rendering of these smaller-caliber aortopulmonary collateral vessels. 

(2) 2D/3D image fusion. X-ray images from the optimal angle and a second view >30° apart were acquired. The image-fusion system rotated the 3D image to the same angle as the x-ray images, creating digitally reconstructed radiographs for both views. Volume-rendered CT images were then fused semi‐automatically by aligning landmarks such as spine and aorta from CT images with spine and pigtail catheter from 2 cine images (Figures 3-7). It is important to acquire the images in the same respiratory phase and electrocardiographic gating between two x-ray images by breath-hold and electrocardiographic gating during acquisition. 

(3) Overlay of vascular landmarks on 2D fluoroscopy and interventional guidance. This allows overlay of electronic annotations of vascular centerlines and ostia on 2D fluoroscopy that is automatically tracked and synchronized with changes in C-arm angulation, table position, and image zoom, to enable intraprocedural guidance for vessel cannulation. In our sessions, this meant facilitation of collateral embolization (Figures 8-11). Currently, such image-fusion techniques often require an experienced imaging technologist to be working on these cross-sectional images simultaneously during the catheterization procedure in the control room to provide image guidance.

We report 6 aortopulmonary collateral artery embolization procedures in different sessions for 3 CHD patients using 2D/3D CT-fluoroscopy image-fusion guidance.

Table 1 summarizes our procedure details, including 2D/3D registration landmarks, procedural time, contrast volume, and radiation exposure. The 3 patients included 2 women and 1 man (age range, 34-66 years). Adult CHD diagnoses and cardiac history included patent ductus arteriosus with Eisenmenger syndrome; single-chamber atrium, heterotaxy (right atrial isomerism) with interrupted inferior vena cava, and pulmonary hypertension; congenital peripheral pulmonary arterial stenosis status post prior stenting in the left lower pulmonary artery (bifurcation A9/A10) and right upper pulmonary artery (truncus anterior), as well as in-stent complete total occlusion of the right coronary artery. Indication for embolization was hemoptysis for 3 sessions (50%) and preoperative optimization for heart-double lung transplant in the other 3 sessions. Target collateral origins, in order of frequency, were left bronchial artery (4 sessions), right bronchial artery (3 sessions), right internal mammary artery (2 sessions), left internal mammary artery (2 sessions), and right pectoral artery (1 session). The spine and a pigtail catheter in the aorta were the most common landmarks used for image fusion (4 sessions). Other landmarks used were a wire in the right atrium, area of pulmonary hemorrhage, bronchial tree, a catheter in the innominate artery, and pulmonary artery stents (Figures 3-7). Particle embolization was done in all sessions (700 and/or 900 micron Embozene microparticles; Boston Scientific), followed by coil embolization in 4 sessions (2 mm x 2 cm, 2 mm x 4 cm, 3 mm x 15 cm, 4 mm x 10 cm, 4 mm x 15 cm, 5 mm x 20 cm, and 5 mm x 30 cm Ruby coils and 30 cm and 60 cm packing coils; Penumbra) and vascular plug deployment in 1 session (8 mm Amplatzer Vascular Plug II; St. Jude Medical). Mean procedure and fluoroscopy times were 3 hours, 23 minutes and 1 hour, 3 minutes, respectively. Median number of aortograms performed was 2 and an average of 169 mL (350 mg iodine/mL) of contrast material was used in each session. Mean radiation exposure per session was 1538 mGy of total skin dose. Successful embolization was confirmed in each session by postembolization angiography showing minimal or absent flow through culprit collaterals. 

Aortopulmonary collaterals can follow any course in 3D space. The most common collaterals we encountered were from the bronchial artery, and bronchial artery anatomy is variable, with around 36% of the population having at least 1 bronchial artery of ectopic origin.⁴ Attempting to understand 3D orientation and embolization of vessels in this complex anatomy solely with the aid of 2D fluoroscopic imaging is challenging. Due to this factor, at least 20% of cases have recurrent hemoptysis after bronchial artery embolization, with a common cause being unrecognized collaterals.^1,6 The use of CT-fluoroscopy image-fusion guidance helps identify optimal C-arm angle, allowing thorough visualization of all potential culprit vessels. This facilitates cannulation, increasing the chance of successful embolization and decreasing the risk of recurrent hemoptysis.

There are limited reports in the literature discussing the use of image fusion in aortopulmonary collateral embolization. In 2016, a case report from Poland reported the use of 2D/3D image fusion for aortopulmonary collateral identification in a 3.5-year-old CHD patient. In their case, a segmentation was performed manually to expose all vessels that provided pulmonary blood support. Color ring markers were used to mark take-off of vessels. They then aligned the 3D segmented datasets with 2D fluoroscopy using spinal bodies and vertebrae as fiducials.⁷ In our cases, we were able to mark the takeoffs without prior segmentation. Multiplanar reformatting was used to mark both the origins and landmarks as well as to identify the central line of the arteries and collaterals. 

Apart from preprocedural and intraprocedural guidance, other advantages of this technique are reduced procedural time and contrast volume, based on previous studies that used this technique for vascular surgery and gynecologic procedures. Use of CT overlay in patients undergoing fenestrated and/or branched endovascular repair showed significantly lower contrast material volume (159 mL vs 199 mL) and procedural time (5.2 hours vs 6.3 hours) compared with a control group.⁸ The same fusion technique was used with MR angiography for embolization of uterine arteries and showed accurate overlay in 100% of cases and absence of contrast use for uterine artery catheterization in 85% of cases.⁹ 

Since we were being aggressive in identifying and embolizing all possible collaterals, our fluoroscopy time was not considerably lower than standard procedures. 

The limitations for 2D/3D registration include the following: (1) patient movement during the procedure can affect accuracy of image fusion because 3D images are not acquired in real time; (2) this process is time and resource intensive; and (3) imported 3D CT datasets cannot be synchronized to respiratory and cardiac motion.

Accurate overlay of CT image on fluoroscopy requires time and effort of a dedicated imaging scientist. Our group is currently developing an artificial intelligence (AI) multimodality image-fusion platform that will not only decrease the time and manpower required for this modality, but also make it more easily accessible to interventionalists.

Aortopulmonary collateral anatomy is highly variable and complex. Use of CT-fluoroscopy 2D/3D fusion is a powerful tool to aid in preprocedural planning and intraprocedural guidance for successful embolization of these collaterals. Use of multimodality imaging tools in the cardiac catheterization laboratory has the potential to improve the safety and efficiency of complex transcatheter interventions and conceivably facilitate novel interventions in the future.