Three-Dimensional Holographic Guidance, Navigation, and Control (3D-GNC) for Endograft Positioning in Porcine Aorta: Feasibility Comparison With 2-Dimensional X-Ray Fluoroscopy

Abstract

Objectives

Intraprocedural deployment of endovascular devices during complex aortic repair with 2-dimensional (2D) x-ray fluoroscopic guidance poses challenges in terms of accurate delivery system positioning and increased risk of x-ray radiation exposure with prolonged fluoroscopy times, particularly in unfavorable anatomy. The objective of this study was to assess feasibility of using an augmented reality (AR) system to position and orient a modified aortic endograft delivery system in comparison with standard fluoroscopy.

Materials and Methods

The 3-dimensional guidance, navigation, and control (3D-GNC) prototype system was developed for eventual integration with the Intra-Operative Positioning System (IOPS, Centerline Biomedical, Cleveland, OH) to project spatially registered 3D holographic representations of the subject-specific aorta for intraoperative guidance and coupled with an electromagnetically (EM) tracked delivery system for intravascular navigation. Numerical feedback for controlling the endograft landing zone distance and ostial alignment was holographically projected on the operative field. Visualization of the holograms was provided via a commercially available AR headset. A Zenith Spiral-Z AAA limb stent-graft was modified with a scallop, 6 degree-of-freedom EM sensor for tracking, and radiopaque markers for fluoroscopic visualization. In vivo, 10 interventionalists independently positioned and oriented the delivery system to the ostia of renal or visceral branch vessels in anesthetized swine via open femoral artery access using 3D-GNC and standard fluoroscopic guidance. Procedure time, fluoroscopy time, cumulative air kerma, and contrast material volume were recorded for each technique. Positioning and orientation accuracy was determined by measuring the target landing-zone distance error (δ_LZE) and the scallop-ostium angular alignment error (θ_SOE) using contrast-enhanced cone beam computed tomography imaging after each positioning for each technique. Mean, standard deviation, and standard error are reported for the performance variables, and Student’s t tests were used to evaluate statistically significant differences in performance mean values of 3D-GNC and fluoroscopy.

Results

Technical success for the use of 3D-GNC to orient and position the endovascular device at each renal-visceral branch ostium was 100%. 3D-GNC resulted in 56% decrease in procedure time in comparison with standard fluoroscopic guidance (p<0.001). The 3D-GNC system was used without fluoroscopy or contrast-dye administration. Positioning accuracy was comparable for both techniques (p=0.86), while overall orientation accuracy was improved with the 3D-GNC system by 41.5% (p=0.008).

Conclusions

The holographic 3D-GNC system demonstrated improved accuracy of aortic stent-graft positioning with significant reductions in fluoroscopy time, contrast-dye administration, and procedure time.

Introduction

In the present standard of care, complex endovascular aortic surgery is performed by visualization of 3-dimensional (3D) vascular structures with 2-dimensional (2D) imaging using fluoroscopy. This is associated with significant limitations and exposes the operator to significant health risks from long term x-ray radiation exposure. Fusion overlay of 3D imaging on fluoroscopy has allowed for intraoperative guidance of endovascular devices thereby reducing radiation exposure and contrast administration.^1,2 However, despite this advantage, the fused images are displayed and visualized on a 2D monitor.

Augmented or mixed reality (AR/MR) imaging allows for simultaneous visualization of the surrounding physical environment overlaid with digital information. The operator wears a digital head-mounted display (HMD) with optical see-through lenses that project stereoscopic images, referred to as holograms. Anatomically correct 3D holograms of vascular structures can be generated using preoperative tomographic imaging. ³ These holograms can then be projected onto the operative field in spatial registration with the patient’s location and landmarks. Thus, for example, when the surgeon looks directly at the operative field, the hologram of the subject’s abdominal aorta, branch vessels, and relevant bony structures are visualized in the anatomically correct locations.

Navigation and visualization of endovascular devices inside vascular structures is predominately carried out under 2D fluoroscopy with the inherent risk of radiation exposure to the operator and the patient. Intravascular ultrasound is an alternative to this modality but is mainly used during postdeployment and is used extensively in aortic dissections. Electromagnetic (EM) tracking provides real-time (ie, live) locational tracking of endoluminal devices without the harmful effects of ionizing radiation. ⁴ The EM navigation system is attached to the surgical table and generates a weak magnetic field around the patient. The delivery system is equipped with an EM sensor that detects relative changes in a set of magnetic field generator coils to track its position and orientation (P&O) of the delivery system.

We have developed a 3D holographic guidance, navigation, and control (3D-GNC) system for endograft positioning inside porcine aorta. The 3D-GNC prototype system was developed to project spatially registered 3D holographic representations of the subject-specific aorta for intraoperative guidance and an EM-tracked delivery system for intravascular navigation. The objective of this study was to assess safety, efficiency and feasibility of using this AR system to position and orient a modified, tracked aortic endograft delivery system in comparison with standard 2D fluoroscopy.

Materials and Methods

3D-GNC System Overview

Anatomically correct 3D holograms of vascular structures were generated using preoperative subject imaging. This hologram was projected onto the operative field using an optical see-through HMD, and co-registered the actual anatomy, allowing the 3D digital image to be viewed and overlaid on the physical subject in augmented reality for intraoperative guidance. A holographic representation of a modified stent graft equipped with EM tracking sensors was created. The EM-tracked device was navigated intravascularly as the fused holographic representation was visualized simultaneously by the operator wearing the AR headset. The operator also received real-time numerical feedback for controlled deployment of the endograft landing zone distance and ostial alignment.

Anatomic Hologram Production

Virtual anatomic models of the subject porcine aorta, aortic branches, spine, pelvis, and the endograft were developed using Unity3D (Unity Technologies, San Francisco, CA) and Visual Studio 2015 (Microsoft Corporation, Redmond, WA) software for the HoloLens mixed reality headset (Microsoft Corporation, Redmond, WA). Subject-specific hologram of the porcine aorta with its branches and other anatomical structures (spine and pelvis) were generated by segmentation of preoperative multidetector computed tomography (MDCT) angiogram of porcine abdomen and pelvis being studied (Figure 1). To limit preoperative changes in body morphology leading to potential spatial misregistration between image-based and tracked-device holograms during 3D-GNC, the time period between the MDCT data acquisition and the endovascular procedure was less than 3 days.

Figure 1.

Three-dimensional (3D) reconstructed preoperative multidetector computed tomography (MDCT) angiogram of porcine abdomen and pelvis depicting the aorta and spine.

Endograft Modification for Electromagnetic and Fluoroscopic Tracking and Visualization

A Zenith Spiral-Z AAA Iliac Leg Graft (Cook Medical, Bloomington, IN) was modified with a scallop. The modified endograft was equipped with three gold radiopaque markers for fluoroscopic visualization and with an electromagnetic tracking sensor to allow for EM navigation of the device in the vasculature (Figure 2). The 14-French (5.4-mm outer diameter) profile of the endograft introducer sheath allowed for introduction of the device via the porcine femoral artery.

Figure 2.

(A) Partially unsheathed 14-French Zenith Spiral-Z AAA Iliac Leg Graft (Cook Medical, Bloomington, IN). (B) Endograft modified with a scallop and 3 radiopaque gold markers. (C) Resheathed modified device equipped with electromagnetic tracking sensor.

Endograft Hologram Production

A virtual representation (hologram) of the modified endograft was also developed using Unity3D (Unity Technologies, San Francisco, CA) software. A holographic vector was added to originate at the center of the scallop to assist with intraoperative holographic orientation of the scallop with vessel ostia (Figure 3).

Figure 3.

(A) Hologram of porcine aortic model and the endograft with holographic vector from the center of the scallop not aligned with the left renal artery. (B) Alignment of the scallop with vessel ostium. Additional digital numeric feedback is provided to the operator to assist with accurate positioning and orientation.

System Integration and Function

The holographic 3D-GNC system was developed as a research prototype for eventual integration with the Intra-Operative Positioning System (IOPS, Centerline Biomedical, Cleveland, OH). The 3D P&O of the modified endograft delivery system was tracked with an Aurora 6 degree-of-freedom (6-DOF) electromagnetic sensor (Northern Digital, Waterloo, Ontario, Canada). The sensor and conductive wires were coated with polyethylene terephthalate (PET) and sealed at the proximal and distal ends of the delivery device. We developed spatial registration methods for holographic projection of digital aortic and delivery device models as well as holographic control indicators to the operative field. The platform was used to transform the preoperative CT-to-HMD and EM-to-HMD coordinates for anatomical and tracked-device models, respectively. This registered both types of models in a common coordinate frame. Preoperative CT coordinates of the aorta (and other anatomical models) were transformed to HMD coordinates using semiautomated 3D transformations based on the locations of fiducial skin markers (CT-SPOT, Beekley Medical, Bristol CT) placed on the ventral abdominal skin surface. These fiducials were located in HMD space, in EM space, and in the preoperative CT scan to compute the registration matrices. After the anatomical and device digital models were registered in the HMD coordinate frame, the headset tracks head movements using an internal inertial measurement unit (IMU) and computes stereoscopic co-projection of GNC holograms to the surgical field (Video 1). The EM-to-HMD transformation was based on a trackable Vuforia (PTC, Boston, MA) image targets selectively placed on the ventral skin surface (Figure 4).

Figure 4.

Coordinate frames used for 3-dimensional guidance, navigation, and control (3D-GNC) registration and projection.

Experimental Methods

In vivo, endovascular specialists consisting of vascular surgeons and interventional radiologists independently positioned and oriented the scallop of the modified endograft to the ostia of renal, superior mesenteric, and celiac arteries in anesthetized swine. The device was advanced via open femoral artery exposure inside porcine aorta. The delivery system was partially unsheathed and the scallop was oriented at each renal and visceral vessel ostium using only fluoroscopy in one arm of the experiment. In the other arm of the experiment, the scallop was oriented at each renal and visceral vessel ostium using the 3D-GNC system with the operator wearing the HoloLens headset. The operator was provided the option of confirmation of scallop P&O with fluoroscopy as a supplement. Final positioning and accuracy of the scallop relative to the vessel ostium was assessed by cone beam computed tomography (CBCT; Figures 5 and 6).

Figure 5.

Workflow for each experimental arm comparing fluoroscopy (top row) with 3D-GNC system (bottom row). C+, injection of contrast media; CBCT, cone-beam CT scan.

Figure 6.

Intraoperative image of the operator’s view from the HoloLens headset with the 3D-GNC system after advancement of endograft inside porcine aorta prior to alignment of the scallop with vessel ostium.

For each arm of the experiment the following variables were measured: fluoroscopy time, cumulative air karma, contrast volume administered, and procedure times. CBCT was used to measure the distance error associated with proximal landing zone deployment (δ_LZE) and the angular misalignment of the scallop with vessel ostium (θ_SOE). Landing zone error was measured as the Euclidian distance between inferior edges of the scallop and the visceral vessel ostium on multiplanar reformatted contrast-enhanced CBCT images.

Continuous variables are expressed as averages ± standard deviation. Student’s t test was used to compare the outcomes of the 2 arms of the experiment. P values <0.05 were considered statistically significant. The study was approved by the Cleveland Clinic Institutional Review Board. The endovascular specialists participated in the study after completion of Institutional Animal Care and Use Committee’s (IACUC) training modules, health and safety requirement for individuals working with live animals for research.

Results

The scallop of the modified endograft was aligned a total of 80 times at the ostium of renal and visceral vessels by 10 endovascular specialists in 3 anesthetized swine. Each endovascular specialist aligned the scallop of the modified endograft at each renal, superior mesenteric, and celiac arteries (4 vessels in total) by advancing the device positioned in the infrarenal aorta at time zero using fluoroscopy and the 3D-GNC system. Technical success for the use of fluoroscopy and 3D-GNC to orient and position the endograft at each renal-visceral branch ostium was 100%. The operators did not use any supplemental fluoroscopy or contrast administration to confirm the final position of the scallop with the vessel ostia (Table 1).

Table 1.

Radiation and Contrast Exposure With Imaging Modalities. ^a

	Fluoroscopy	Holographic 3D-GNC	p
Fluoroscopy time (min)	1±0.6	0	<0.001
Cumulative air kerma (mGy)	161.7±14.4	0	<0.001
Contrast volume (mL)	14.9±10.2	0	<0.001

Abbreviation: 3D-GNC, 3-dimensional guidance, navigation, and control.

^aContinuous variables reported as mean ± standard deviation.

In comparison with standard fluoroscopic guidance, the 3D-GNC system resulted in a 56% decrease in time required to complete the tasks of inserting, advancing, positioning, and orienting the endograft delivery system to align the scallop with 1 of 4 targeted visceral vessel ostia (p<0.001). Positioning accuracy (target vessel landing zone distance error) was comparable for both techniques (p=0.86), while the overall orientation accuracy was improved with the 3D-GNC system by 41.5% (p=0.008; Table 2).

Table 2.

Technical Accuracy With Imaging Modalities. ^a

	Fluoroscopy	Holographic 3D-GNC	p
Target landing-zone distance error (mm)	4.1 ± 3.4	4.2 ± 2.8	0.86
Scallop-ostium angular alignment error (deg)	26.4 ± 20.1	15.5 ± 14.2	<0.008
Procedure time (min)	4.4 ± 2.9	1.9 ± 0.9	<0.0001

Abbreviation: 3D-GNC, 3-dimensional guidance, navigation, and control.

^aContinuous variables reported as mean ± standard deviation.

Per-Vessel Analysis

Procedure times for orientation of the scallop at each vessel was consistently decreased. At the right renal artery, procedure time was reduced from 4.8±3.2 minutes using fluoroscopy only to 2.3±1.2 minutes using 3D-GNC (p=0.041). At the left renal artery, procedure time was reduced from 3.6±1.7 minutes using fluoroscopy only compared with 1.7±0.6 minutes using 3D-GNC (p=0.007). At the superior mesenteric artery, procedure time was reduced from 4.3±2.9 minutes using fluoroscopy only to 1.9±0.8 minutes using 3D-GNC (p=0.035). At the celiac artery, procedure time was reduced from 5±3.5 minutes using fluoroscopy only to 1.8±0.8 minutes using 3D-GNC (p=0.017).

The target landing-zone distance error (δ_LZE) for each vessel remained comparable between the 2 methods. At the right renal artery, the δ_LZE measured 3.1±1.5 mm vs 3.9±1.7 mm for fluoroscopy alone vs 3D-GNC, respectively (p=0.3). At the left renal artery, the δ_LZE measured 3.4±2.4 mm vs 4.3±2.0 mm for fluoroscopy alone vs 3D-GNC, respectively (p=0.39). At the superior mesenteric artery, the δ_LZE measured 6.4±5.3 mm vs 5.2±4.5 mm for fluoroscopy alone vs 3D-GNC, respectively (p=0.59). At the celiac artery, the δ_LZE measured 3.1±1.8 mm vs 3.2±1.7 mm for fluoroscopy alone vs 3D-GNC, respectively (p=0.87).

The mean scallop-to-vessel ostium angular alignment error (θ_SOE) was lower with 3D GNC relative to fluoroscopy for all 4 branch arteries but was not statistically significantly different at renal and superior mesenteric arteries. At the right renal artery, θ_SOE measured 28.4°±24.6° vs 22.6°±12.7° for fluoroscopy alone vs 3D-GNC, respectively (p=0.54). At the left renal artery, θ_SOE measured 18.8°±16.3° vs 9.1°±10.1° for fluoroscopy alone vs 3D-GNC, respectively (p=0.15). At the superior mesenteric artery, θ_SOE measured 27.7°±15.5° vs 20.3°±18.7° for fluoroscopy alone vs 3D-GNC, respectively (p=0.35). Significant improvement in θ_SOE was noted at the level of celiac artery. At the celiac artery, θ_SOE measured 30.8°±24.0° vs 9.3°±8.8° for fluoroscopy alone vs 3D-GNC, respectively (p=0.03).

Discussion

Advancements in imaging technology have had a significant impact on complex endovascular therapy. However, reliance on 2D displays and ionizing radiation for visualization of 3D vasculature and navigation of endovascular devices poses significant limitations and health risks to the operator and the patient from long-term radiation exposure.^5,6 This is most evident in complex endovascular procedures such as fenestrated and branched endografts used for repair of complex aortic aneurysms. From a technical perspective, fenestrated endografts require precise alignment of the fenestration(s) with the renal or visceral vessel ostia to allow for cannulation and subsequent stenting using 2D fluoroscopy imaging. A significant learning curve is associated with 3D alignment of these endografts inside the aorta using 2D imaging. Adjunctive use of EM navigation systems can reduce radiation exposure and be beneficial in vessel cannulation and endovascular device positioning. ⁷

Intraoperative technical components of complex endovascular aortic surgery consist of 3 main components, which we refer to as guidance, navigation, and control. These broadly involve pre- and intraoperative imaging of vascular anatomy to guide endovascular repair, navigation of wires, catheters, sheaths and endovascular devices inside the vasculature to the target vessels, and accurately controlled deployment of endovascular devices to allow for complete repair. In order to accomplish these tasks, MDCT angiogram of the porcine aorta being studied was used to generate a 3D hologram. The hologram of porcine aorta was projected onto and fused to the operative field in its anatomically correct location. The EM-tracked endograft was then visualized in its holographic representation via the AR headset. As the physical device was advanced intravascularly, the real-time synced hologram was also visualized as advancing inside the porcine aorta by the operator wearing AR headset. The operator further receives real-time numerical feedback for controlling the advancement and alignment of the scallop with the vessel ostia while wearing a mixed reality headset.

We have developed a unique AR system that allows for real-time visualization of vascular anatomy and tracked endovascular devices coupled with an EM navigation system. This holds potential for complete endovascular aortic intervention without reliance on ionizing radiation. Other investigators have demonstrated feasibility of endovascular aortic interventions in phantom models.^8–10 In a small randomized controlled trial, Manstad-Hulaas et al ¹¹ demonstrated the feasibility of endovascular aortic aneurysm repair using 3D EM navigation system in conjunction with fluoroscopy to reduce contrast use and radiation exposure during the procedure. Application of this technology has remained limited in humans. In a swine model, complex endovascular procedures such as in situ fenestration of endografts has been performed using EM navigation systems. ¹²

The endovascular specialists in this trial had no prior experience or training with the AR headset or EM tracking systems. The intuitive nature of this technology is associated with a minimal learning curve. The real-time quantitative feedback provides further information to the operator via a digital readout in the AR headset. This allows the operator to determine and control the distance and angular alignment of the scallop from the vessel ostium of interest. These functions have significant potential to improve safety and effectiveness in complex endovascular aortic procedures, which often require accurate alignment of fenestrations with the vessel ostium. This effectively eliminates the need for many procedure steps that include fluoroscopy such as aligning the fenestration and vessel ostium perpendicular to the C-arm gantry.

In this study, a single scallop device was used in order to establish the proof of concept. In future work, the same guidance, navigation, and control techniques with AR and EM tracking may be evaluated for its suitability for fenestrations, as well as devices with multiple scallops and/or fenestrations.

As the EM navigation system relies on a magnetic field around the subject, it is contraindicated with devices such as a pacemaker. The system relies on merger of preoperative MDCT to generate the hologram of the subject aorta. At the present, similar to fusion fluoroscopy imaging, vessel deformation and motion would not be visualized with the 3D hologram in this study. However, real-time EM tracked device position would reflect these changes regardless of the projected anatomical holograms.

Future work will address modeling deformations due to forces associated with interventional devices and motion of the cardiac and respiratory cycles. Future work may also extend 3D-GNC to track the deployment process of the stent-graft, or test the suitability of pre-operative MRI angiography as an alternative to multidetector row CT imaging.

To advance 3D-GNC from this proof-of-concept study to clinical use, there are several technical limitations to overcome. Attaching electromagnetic sensors increased the endograft delivery system profile by approximately 4 French; tighter integration will be necessary to maintain the low profile of present delivery systems. Preoperative processing of tomographic imaging datasets to generate anatomical holograms will need to be further streamlined and automated for efficient clinical use. Components used in the spatial registration process will be revised to maintain the sterile field and operating room environment. Image data set processing for advanced vessel mapping and sterile fiducial registration techniques used in the Intra-Operative Positioning System (IOPS, Centerline Biomedical, Cleveland, OH), that has been released to the United States market since this study was performed, will be leveraged for translation of 3D-GNC to clinical use.

We demonstrated the feasibility of completely radiation-free endovascular aortic procedure in vivo with improved results compared with current standard of 2D fluoroscopy.

Conclusions

In a porcine model, holographic 3D-GNC demonstrated improved overall accuracy of aortic stent graft positioning without the use of fluoroscopy or contrast-dye administration. A significant reduction in procedure time was associated with the 3D-GNC system compared to fluoroscopy. The 3D-GNC system demonstrated feasibility as an imaging modality for endovascular surgical procedures.