3D Virtual Reality Imaging of Major Aortopulmonary Collateral Arteries: A Novel Diagnostic Modality

Background Major aortopulmonary collateral arteries (MAPCAs), as seen in patients with pulmonary atresia, are arteries that supply blood from the aorta to the lungs and often require surgical intervention. To achieve complete repair in the least number of interventions, optimal imaging of the pulmonary arterial anatomy and MAPCAs is critical. 3D virtual reality (3D-VR) is a promising and upcoming new technology that could potentially ameliorate current imaging shortcomings. Methods A retrospective, proof-of-concept study was performed of all operated patients with pulmonary atresia and MAPCAs at our center between 2010 and 2020 with a preoperative computed tomography (CT) scan. CT images were reviewed by two congenital cardiac surgeons in 3D-VR to determine additional value of VR for MAPCA imaging compared to conventional CT and for preoperative planning of MAPCA repair. Results 3D-VR visualizations were reconstructed from CT scans of seven newborns where the enhanced topographic anatomy resulted in improved visualization of MAPCA. In addition, surgical planning was improved since new observations or different preoperative plans were apparent in 4 out of 7 cases. After the initial setup, VR software and hardware was reported to be easy and intuitive to use. Conclusions This study showed technical feasibility of 3D-VR reconstruction of children with immersive visualization of topographic anatomy in an easy-to-use format leading to an improved surgical planning of MAPCA surgery. Future prospective studies are required to investigate the clinical benefits in larger populations.


Introduction
In pulmonary atresia with ventricular septal defect (PA-VSD), major aortopulmonary collateral arteries (MAPCAs) are often present. These MAPCAs can be essential in patients with diminutive pulmonary vascular abnormalities. 1 As a consequence, lung areas can be vascularized via single, noncommunicating blood supply (MAPCAs only) or via dual, communicating blood supply (both MAPCA and main pulmonary artery [MPA]). 2 If adequate antegrade pulmonary arterial circulation is present, dual-supply MAPCAs can be occluded (surgically or therapeutically), while MAPCAs with single lung segment perfusion to the pulmonary circulation should be relocated from the aorta to the native pulmonary arterial system (unifocalization procedure). 3 The goal of surgical correction is to eventually create a biventricular circulation with an adequate pulmonary arterial system, to which all MAPCAs with singular blood supply are unifocalized and in which the dual-supply MAPCAs are ligated, preferably in the least number of operations. Due to patient-specific (under)-development of the pulmonary arterial system and variance in the number of MAPCAs, systemic-to-pulmonary artery (PA) shunt placement or right ventricular outflow tract (RVOT) procedure to support pulmonary arterial outgrowth and/or multistage MAPCA unifocalization may be necessary. 4,5 Unifocalization may be a complex surgical procedure and can be associated with increased morbidity and mortality. 6 Nevertheless, recent studies report improved survival by a higher percentage of complete repair and fewer operations per patient. 7 Complete repair with a limited number of interventions depends on optimal visualization of the pulmonary arterial anatomy and MAPCAs and precise preoperative planning. Currently, in our center, preoperative planning is routinely performed with catheterization angiography (CA) and computed tomography angiography (CTA) where the emphasis on three-dimensional (3D) topographical anatomical relationships is sometimes lacking. Furthermore, two-dimensional (2D) CA and CTA imaging require an advanced level of 3D imaging review expertise on the part of the cardiac surgeon and cardiovascular radiologist. Previous studies have shown that identification of anatomical structures and their course is superior in 3D visualization compared with 2D. 8 3D volume rendering of CT scans is a novel technology, and can be time-consuming and requires extra resources such as dedicated imaging software, a radiologist, or radiology technician. 9 In addition to this, 3D-CT images are still shown on a flat, 2D computer screen, which lacks depth perception and interaction. 10 3D-Virtual Reality (3D-VR) visualization is a promising and innovative new technology that could potentially complement the shortcomings of 2D imaging. 11 Recent studies, including publications by our group, have shown that VR can provide true in-depth perception, intuitive model manipulation and interaction, structure segmentation, and can enable turning on/off surrounding structures during the assessment [12][13][14] . In other surgical departments, 3D-VR surgical planning has been shown to reduce operation duration and decrease the length of stay in hospital. 15 In this retrospective single-center pilot study, we aimed to study the technical feasibility, clinical usefulness, and experience of multiple users on the application of 3D-VR for the preoperative planning of patients with PA-VSD and MAPCAs.

Patient Selection
Seven patients diagnosed with PA-VSD with MAPCAs who underwent corrective repair between 2010 and 2020 at our center were eligible for inclusion since both preoperative CTA and CA images were available. An overview of the baseline characteristics is provided in Table 1. Median age at time of CTA was 2 days (range 2-388 days). CA was planned in the week before surgery, median age 213 days (range 29-506 days). The study was approved by the local medical ethical committee (MEC-2020-0891) of Erasmus University Medical Center. Informed consent was obtained from the legal representatives of all living patients before inclusion.
Computed Tomography in 3D, Image Segmentation, and Virtual Reality Rendering To create volumetric 3D-VR rendering of the CT scans, preoperative CTA scans were used (Supplemental Material A). There were no specific technical requirements for CT scans and all scans were performed at our center according to the local protocol. Semiautomatic 3D segmentation was performed to highlight anatomical structures (eg bronchi, PA, and MAPCAs) by using a previously published protocol. 16 Digital imaging and communications in medicine (DICOM) files and corresponding 3D segmentation files were loaded into our CardioVR software (MedicalVR, Amsterdam, the Netherlands), which creates an automatic 3D-VR visualization. Afterwards, the segmentation was checked by an experienced physician. 3D-VR was directly available for the user by using a VR head-mounted display and associated controllers.

Study Design, Objectives, and Data Analysis
Within this pilot study, our objective was to study the technical feasibility, clinical usefulness, and user experience of 3D-VR for preoperative planning of newborns with PA-VSD with MAPCAs. Technical feasibility was defined as the possibility to create and immersively review 3D reconstructions of (low-resolution) CTA scans of newborns in VR. Clinical usefulness, the additional value of 3D-VR over 2D-CT, was determined via (1) if specific characteristics were found with 3D-VR and (2) if surgical preoperative planning would be altered after 3D-VR reconstruction. Thirdly, the user experience of 3D-VR environment was evaluated via a nonstandardized, written questionnaire (Supplemental Material B).
Two experienced congenital cardiothoracic surgeons were included as study participants. A brief (10 min) audiovisual hardware and software instruction was provided. During the assessments, an experienced VR user was on site for technical support. Both participants reviewed all imaging modalities (CA, 2D-CT, and 3D-VR) individually per case and described various MAPCA-related parameters, after which a surgical plan was drafted. The participants were blinded to each other's descriptions and surgical plans. Original surgery reports were obtained from the electronic health record to compare the surgeons' devised plan.

Data Visualization and Analysis
2D computer screen recordings of the 3D-VR environment were captured and stills of this video were used for figures. Additional 3D model reconstructions, only used for visualization, were made in 3D Slicer. 17 All data was analyzed using Microsoft Office Excel (Microsoft).

Technical Feasibility
It was technically feasible to create 3D reconstructions of segmented CTA scans of all 7 newborns in VR (Figures 1 and 2, animated GIF figure in Supplemental Material C). A total of  6 scans had adequate spatial resolution and contrast enhancement for VR imaging, but CT scan and consequently the VR experience of patient 2 was suboptimal. Segmentation of the esophagus, which provides useful information about the course of MAPCAs around this structure, was possible in f4 patients.

Clinical Usefulness
The usefulness and additional value of 3D-VR over 2D-CT is shown in Table 1 Figure 5A). During surgery, no intrapulmonary connection of the left lung was seen. 3D-VR visualization of patient 7 showed that the modified Blalock-Taussig (mBT) shunt had been placed onto the MAPCA (Figure 6). This anastomosis had not been noticed after CA and 2D-CT review at time of reporting by cardiac surgeon or by an experienced radiologist ( Figure 6C).

VR User Experience
Questionnaire scores are shown in Table 2. Both surgeons found the VR environment easy to use, and the VR model easy to interact with. Although VR assessment takes more  time (∼15-20 min) than conventional imaging techniques, mainly due to setting up the VR hardware and software, the 3D-VR model was of additional value to assess the number of MAPCAs, the offspring of MAPCAs and the course of MAPCAs. Classification of (non-)communicating arteries scored low. Lastly, the threshold to use VR is still too high

Comment
This study showed that 3D-VR for imaging review and preoperative planning of newborns with PA-VSD-MAPCAs was possible and addressed the fact that, albeit still time-consuming, surgeons can retrieve important patient-specific details to improve surgical procedure, with an excellent user experience. 3D-VR reconstruction of CTA scans could be created of all newborns, which is more difficult than in adults due to smaller patient dimensions and suboptimal contrast enhancement. 3D-VR can be used as a complementary modality for preoperative planning of MAPCA repair to provide better overview of the structures of interest for the surgeon. The VR environment was easy to use by both surgeons and had additional value in the assessment of number, origin, and course of MAPCAs. Presence of a nasogastric tube is a recommendation, due to possible segmentation of the esophagus and visualization of the course of the MAPCAs in relation to the esophagus in 3D-VR.
The less optimal CT scan and VR experience of patient 2 could be explained by a thicker slice width and contrast timing. The approach of scanning newborns with congenital heart disease in our center is manually triggered CT scanning due to aberrant and patient-specific anatomy, especially with MAPCAs, plus combining pulmonary and systemic arterial phase to lower total radiation dose (3 cc/kg contrast), which is currently suboptimal for VR purposes. Because of this suboptimal contrast injection, suggestion of intrapulmonary connection between MAPCAs and native pulmonary system based on the CT scan (either 2D-CT or 3D-VR) should be viewed with suspicion, as a false-positive connection was observed in patient 6. CA remains a golden standard to determine single or dual supply.
All patients included in our study were treated with a staged approach to reach corrective surgery. Midline one-stage complete unifocalization claims to lower potential adverse consequences compared to a staged approach (ie increased perioperative risks due to multiple hospitalizations and operations). 18,19 This is the case despite single-stage unifocalization  being a more complex operation, with risks including incomplete repair. 18,19 Due to the latter, midline one-stage unifocalization has not (yet) been performed in our center. 3D-VR could possibly, via more optimal MAPCA visualization, contribute to a higher percentage of complete repair in both approaches and could potentially facilitate one-stage midline repair in some patients. Despite this, MAPCA repair (choice of approach and unifocalization/ligation) will remain difficult, due to the variation in pulmonary arterial and MAPCA anatomy as well as method of repair due to differences in experience and approaches worldwide. 3D-VR is, after an initial cost for essential software and hardware components, readily available and offers a low-cost method to visualize specific anatomic structures. Costs for "off the shelf" hardware, consisting of a high-performance computer and VR devices, starts at $1500 to 2500. Disadvantages of other visualization techniques for preoperative planning, such as 3D printing, are that these models are static and lack interaction and multiangle sliced views, without information about surrounding structures, and require printing time and material. 20,21 This proof-of-concept study showed the additional value of implementation of 3D-VR imaging in the preoperative planning of MAPCA repair, which assists surgeons with the difficult unifocalization procedure. The retrospective character of this study is a limitation, since we could not obtain real-life information about preoperative planning nor real-life feedback about the MAPCA segmentation in 3D-VR, although surgical reports were available. A prospective case series would be valuable to overcome this limitation.

Future Perspectives
In the future, we are planning to study the implementation of 3D-VR for various (congenital) heart diseases, and to collaborate with multiple medical centers. Furthermore, in the near future, we hope to be able to provide detailed information about blood supply to specific lung segments and classify intrapulmonary communications, based on the 3D-VR visualization. This feature is already available in adults, but spatial resolution and contrast enhancement in pediatric cardiac patients is still inadequate to visualize bronchial segmental branches and specific pulmonary segmental arteries. Implementing additional image modalities in our VR environment, such as CA or even rotational angiography (RA) or magnetic resonance angiography (MRA), is another future goal, which may comprise the information of both CTA and CA and result in fewer scans necessary and lower radiation exposure. 22,23 However, RA and MRA are still under development and not (yet) performed in our center. 22,23 Lastly, we are exploring the possibilities of augmented reality for visualization and segmentation as perioperative navigation, which is already used in hepatobiliary surgery, since perioperative guidance can help surgeons to find the origin and course of multiple MAPCAs without the need to rely on their memory of preoperative images. 24