This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Cardio, is properly cited. The complete bibliographic information, a link to the original publication on https://cardio.jmir.org, as well as this copyright and license information must be included.
Patients with single ventricle heart defects receive 3 stages of operations culminating in the Fontan procedure. During the Fontan procedure, a vascular graft is sutured between the inferior vena cava and pulmonary artery to divert deoxygenated blood flow to the lungs via passive flow. Customizing the graft configuration can maximize the long-term benefits. However, planning patient-specific procedures has several challenges, including the ability for physicians to customize grafts and evaluate their hemodynamic performance.
The aim of this study was to develop a virtual reality (VR) Fontan graft modeling and evaluation software for physicians. A user study was performed to achieve 2 additional goals: (1) to evaluate the software when used by medical doctors and engineers, and (2) to explore the impact of viewing hemodynamic simulation results in numerical and graphical formats.
A total of 5 medical professionals including 4 physicians (1 fourth-year resident, 1 third-year cardiac fellow, 1 pediatric intensivist, and 1 pediatric cardiac surgeon) and 1 biomedical engineer voluntarily participated in the study. The study was pre-scripted to minimize the variability of the interactions between the experimenter and the participants. All participants were trained to use the VR gear and our software, CorFix. Each participant designed 1 bifurcated and 1 tube-shaped Fontan graft for a single patient. A hemodynamic performance evaluation was then completed, allowing the participants to further modify their tube-shaped design. The design time and hemodynamic performance for each graft design were recorded. At the end of the study, all participants were provided surveys to evaluate the usability and learnability of the software and rate the intensity of VR sickness.
The average times for creating 1 bifurcated and 1 tube-shaped graft after a single 10-minute training session were 13.40 and 5.49 minutes, respectively, with 3 out 5 bifurcated and 1 out of 5 tube-shaped graft designs being in the benchmark range of hepatic flow distribution. Reviewing hemodynamic performance results and modifying the tube-shaped design took an average time of 2.92 minutes. Participants who modified their tube-shaped graft designs were able to improve the nonphysiologic wall shear stress (WSS) percentage by 7.02%. All tube-shaped graft designs improved the WSS percentage compared to the native surgical case of the patient. None of the designs met the benchmark indexed power loss.
VR graft design software can quickly be taught to physicians with no engineering background or VR experience. Improving the CorFix system could improve performance of the users in customizing and optimizing grafts for patients. With graphical visualization, physicians were able to improve WSS percentage of a tube-shaped graft, lowering the chance of thrombosis. Bifurcated graft designs showed potential strength in better flow split to the lungs, reducing the risk for pulmonary arteriovenous malformations.
Congenital heart disease is the most common birth defect found in nearly 1% of births worldwide [
Advances in medical imaging scanning and 3D-printing techniques have been showing great potential for customizing Fontan grafts. One of the customization approaches is known as tissue-engineered vascular grafts (TEVGs), which uses biocompatible material to facilitate the growth of neotissue, including collagen, vascular muscle, and endothelial cells [
3D-printed scaffolds can be modeled using various approaches including computer-aided design (CAD) software [
The ability to produce graft designs alone is not sufficient to optimize Fontan procedures. Without accounting for the flow inside each graft design, a patient may experience increased risk of medical complications. Multiple studies have emphasized the importance of a low indexed power loss (iPL) [
This study was approved by the investigational review board at Children’s National Hospital (reference number: Pro00009721).
One anonymized post-Fontan procedure imaging data set was acquired via magnetic resonance imaging. The data set was exported as a DICOM file and then manually segmented into two 3D models using Mimics software (Materialise): a (1) TCPC model and (2) a heart model without the TCPC anatomy (
Patient’s Fontan anatomy. (a) The 3D models of the anonymized patient anatomy: heart (dark gray) and total cavopulmonary connection (light gray). (b) Patient anatomy with Fontan IVC to Glenn conduit removed and 10-mm inferior extrusion on the IVC. LPA: left pulmonary artery; IVC: inferior vena cava. RPA: right pulmonary artery; SVC: superior vena cava.
The VR surgical planning software, CorFix, was developed based on the Unity 3D engine. The software-running platform was an Alienware Aurora R8 (Dell) with an Intel Core i7-9700 processor, a NVIDIA GeForce RTX-2080Ti, and 16 GB of RAM. An Oculus Rift S was used for displaying CorFix in full-immersive VR. Touch controllers (Oculus Rift) were integrated into the system for interacting with the interface. CorFix was previously designed to perform simple diagnosis (ie, zoom, rotation, label, ruler, and clipping) and modeling (ie, cutting vessels, parametric modeling, and free-form modeling) tasks. This version of CorFix had a modified user interface to accommodate clinicians untrained in VR, modeling software (eg, CAD), or CFD. The interface was adapted to allow users to intuitively design patient-specific vascular grafts in a short amount of time and integrate image analysis in the workflow.
The Corfix interface was designed to support simple memory recall, allowing for a short, 1-time, 10-minute tutorial. A virtual clipboard was used as an access point for menu and Oculus controller information. The top row contained icons that support the designing of a tube-shaped or bifurcated graft. Icons were designed to match the color and shape of the corresponding geometry and anatomy. In the center of the clipboard, a diagram of the Oculus controller and its functionality were visualized. The bottom row contained menus that are necessary when the design process is completed (
Screenshot of a user creating a bifurcated graft using CorFix. IVC: inferior vena cava
The import process saves all information needed to reconstruct the graft designs using the same algorithm used to construct the conduits. It first saves the transform information of the heart, Glenn, and the graft. The location and radii of the Bezier curve are then stored. These data are then exported into 1 CSV (comma-separated value) file in the aforementioned order. The design import feature works by parsing the saved file from top to bottom and then reconstructing the scene in that order.
The minimum design parameters for a bifurcated graft were 2 anastomosis regions and 1 split region (
Schematic of Fontan graft designs. (a) Minimum design parameters for a bifurcated graft and (b) the cubic Bezier curve and radii interpolation diagram. IVC: inferior vena cava.
These calculated points were connected to the center of the ellipse to make triangular meshes, forming a surface. Two cubic Bezier curves were used to define the pathways and girths of the bifurcated graft. The first Bezier curve used the center of the native IVC surface and a user-defined anastomosis region. The second Bezier curve used the center of another anastomosis region and a user-specified split region of the graft. The formula for the pathways was as follows:
where
where ∆ is the difference between the radii from one center point to another. For example,
Hemodynamic simulation results were outputted in .h5 format. A data import and transform script was developed using MATLAB (MathWorks) since the .h5 format is not supported in Unity. The script consisted of 3 parts: data size, hemodynamic performance summary, and raw WSS values. The data size rows summarized the number of graft designs that were simulated and the total length of the raw WSS values. The hemodynamic performance summary contained information on iPL, %WSS, and HFD on each graft. The raw WSS values are composed of actual WSS values on each x, y, and z coordinate of a graft. These parts are concatenated into 1 CSV file, which is then imported into Unity. As a default, minimum and maximum WSS are set to 0 and 1
Screenshot of the hemodynamic simulation results of clipboard and point cloud rendering using the percentage of nonphysiologic wall shear stress output values.
The hemodynamic performance parameters included iPL, %WSS, and HFD. iPL is a dimensionless value of a pressure difference between the Fontan graft and the PA. It is normalized using a patient’s body surface area. High iPL values have an increased chance of deteriorated cardiac performance and exercise capacity [
where BSA is the body surface area of the patient,
where
The number of total particles varies and depends on the surface area of the inlets. The particles are equally spaced from each other. This study set the healthy ranges of each benchmark parameter as below 0.03 for iPL, below 10% for %WSS, and within the range of 40% to 60% for the HFD ratio.
Ansys Fluent 19 (ANSYS Inc) was used to make extensions at inlet and outlet boundaries. The inlet, IVC, and SVC, were extruded by 10 times their largest diameter. The outlets, that is the left and right PA, were extruded by 50 mm. These extensions acted as a mechanism for developing a stable blood velocity profile. The CFD simulation was performed by solving steady 3D Navier-Stokes equations with Newtonian fluid and rigid wall assumptions. A calculation for the Reynolds number was implemented to assess the laminar flow of a patient’s anatomy.
The institutional review board at the Children’s National Hospital in Washington, DC, approved this study. The study was advertised by sending emails to the groups of residents, fellows, cardiac specialists, and medical engineers. A total of 5 voluntary participants were recruited including 1 fourth-year resident, 1 third-year cardiac fellow, 1 pediatric intensivist, 1 pediatric cardiac surgeon, and 1 biomedical engineer. All participants gave informed consent prior to their participation.
Before the experiment, all participants were queried about their knowledge on the Fontan procedure and vascular grafts. Those who did not have a strong understanding about the topics were given a short tutorial. The tutorial covered anatomy of patients with SVHD, surgical repair for SVHD, and the shapes of Fontan grafts in 3 PowerPoint (Microsoft) slides. All participants then received a tutorial on the 3 benchmark parameters that would be calculated to identify the performance of their Fontan graft designs. This tutorial did not include any information about the relationships between each benchmark parameter and the graft design parameters. The participants were informed that %WSS is negatively correlated with iPL. Healthy ranges of each benchmark parameter were visually provided inside the VR environment as a reference. The last tutorial was for the CorFix interface and took about 10 minutes. None of the participants had prior experience with VR, requiring the CorFix tutorial to include information about the hardware (Oculus). During the CorFix tutorial, participants wore the gear and went through the following topics with verbal feedback: importing anatomies, interacting with the anatomies, designing basic tube-shaped and bifurcated Fontan grafts, making anatomies transparent, visualizing CFD results, and modifying the existing tube-shaped design.
After it was confirmed there were no further questions about the VR gear or CorFix software, participants created and exported 3D models of 1 bifurcated and 1 tube-shaped Fontan graft (
A participant creating a (a) tube-shaped and a (b) bifurcated Fontan graft on CorFix during the experiment.
All participants filled out a digital demographic survey prior to the experiment, including questions about their position, level of VR experience, knowledge and experience on the Fontan procedure, and the level of training on fluid dynamics. Three hard-printed surveys were provided at the end of the study. The System Usability Scale (SUS) and the Usefulness, Satisfaction, and Ease of Use Questionnaire (USE) were used to measure the usability of the system. To identify the level of sickness when using VR gear, the Simulator Sickness Questionnaire (SSQ) was provided.
Participants spent an average of 5.49 minutes creating 1 tube-shaped graft and 13.40 minutes creating 1 bifurcated graft. An average of 2.92 minutes was spent modifying the tube shape after it was created. This time includes reviewing the native patient model and the 6 design variations. The summary of design times and actual designs are provided in
Summary of graft design and modification times.
|
Tube-shaped | Modified tube-shaped | Bifurcation |
Time (min), mean (SD) | 5.49 (2.35) | 2.92 (1.67) | 13.40 (3.48) |
Minimum time (min) | 2.50 | 2.49 | 9.45 |
Maximum time (min) | 8.10 | 5.07 | 16.57 |
Summary figure of the Fontan graft designs.
The patient Fontan data set without modifications showed suboptimal hemodynamic performance. with 55.36% of the Fontan anatomy under nonphysiologically optimal WSS, unbalanced HFD with 72.72% of hepatic flow going to the left PA, and an iPL of 0.0086, indicating minimal flow change within the anatomy (
Hemodynamic performance of the provided Fontan data set without any modifications. %WSS: percentage of nonphysiologic wall shear stress; HFD: hepatic flow distribution; iPL: indexed power loss; LPA: left pulmonary artery.
Each participant produced 1 tube-shaped and 1 bifurcated Fontan graft. CFD simulations were performed on each of the graft designs. The detailed hemodynamic results are provided in
Summary of computational fluid dynamics simulations on the participants’ Fontan graft designs. %WSS: percentage of nonphysiologic wall shear stress; HFD: hepatic flow distribution; iPL: indexed power loss; LPA: left pulmonary artery.
All participants were asked to review 7 Fontan graft design variations based on the native Fontan surgical case. None of 7 design variations were considered optimal for the patient. The variations were created to assist the participants in identifying important design parameters that contribute to each hemodynamic benchmark parameter. After evaluating the design variations, the participants were given the freedom to modify their tube-shaped graft design to attempt to optimize the hemodynamic parameters. All those who modified their tube-shaped Fontan graft were able to reduce %WSS with an average improvement of 7.02%, ranging from 2.32% (cardiac interventionalist) to 13.28% (biomedical engineer;
Summary table of computational fluid dynamic values after participants were presented with a set of prompt design variations; the %WSS values improved. %WSS: percentage of nonphysiologic wall shear stress; HFD: hepatic flow distribution; iPL: indexed power loss; LPA: left pulmonary artery.
CorFix scored an average of 57 on the SUS questionnaire with a minimum score of 42.5 and a maximum of 67.5. The average SUS value suggests that the usability of our prototype was marginal. The average total score of USE (
The SSQ (
Summary table for the Usefulness, Satisfaction, and Ease of Use Questionnaire.
|
Usefulness | Ease of use | Ease of learning | Satisfaction | Overall |
Score, mean (SD) | 3.75 (1.03) | 4.47 (1.38) | 5.10 (1.13) | 4.60 (1.64) | 4.38 (1.10) |
Maximum score | 2.63 | 2.36 | 3.25 | 1.71 | 2.57 |
Minimum score | 5.38 | 5.64 | 6.00 | 5.71 | 5.53 |
Summary table for the Simulator Sickness Questionnaire survey.
|
Nausea | Oculomotor | Disorientation |
Score, mean (SD) | 11.45 (10.45) | 24.26 (26.48) | 16.70 (30.18) |
Maximum score | 0 | 0 | 0 |
Minimum score | 28.62 | 68.22 | 69.60 |
All participants were able to successfully design patient-specific conduits using the VR software with limited training. Although none of the participants had VR experience and CorFix was rated with marginal acceptable usability, designing tube-shaped and bifurcated grafts took less than 6 and 14 minutes, respectively. We used the time spent on a task as a surrogate for task difficulty and assessment of user adoption since there is sound literature indicating that among adult learners, time spent on a task is commensurate with task difficulty [
All participants expressed that if a real-time hemodynamic analysis of their designs were available, they would be able to better pinpoint the flaws of their designs. We, therefore, plan on further developing the CorFix software to add real-time simulation and visualization features. Our system has implemented button and pointer color changes and tactile feedback (ie, vibration) to bolster the interactivity inside the virtual scene. However, many participants struggled with depth perception and interactivity. Grabbing design control points or even clicking buttons on the virtual menu were frequently observed. Developing a feature or a device that could better support tactile feedback may enhance the usability and the innate learnability of the software.
The bifurcated graft designs were more successful in improving the hepatic flow distribution to a healthy range compared to the tube-shaped graft designs. During the experiment, all participants were asked to review 7 different tube-shaped Fontan grafts, which were derived from the actual surgical case although none of these design variations were surgically optimal. We hypothesized that the participants would be able to find patterns between design parameters and hemodynamic performance. Our study showed that when participants decided to modify their designs after reviewing other cases, they were able to design a more optimal graft by lowering %WSS. On average, %WSS was reduced by 7.02%. A biomedical engineer with a strong fluid dynamics education background showed the maximum %WSS reduction of 13.28%. Considering how lower %WSS is related to a lower risk of thrombosis for Fontan grafts, this design could provide a significant long-term improvement for the patient. We therefore infer that showing problematic regions in color, like a contour map, may help doctors without an engineering background to sufficiently identify low %WSS. iPL and HFD improvements were not consistent throughout the participants. Unlike %WSS, these hemodynamic parameters were presented only in numerical format. We hypothesize that with supplementary graphical visualization, users may be able to improve iPL and HFD more easily.
With the development of graft modeling and evaluation software like CorFix, physicians may be able to easily customize Fontan grafts and find an optimal graft configuration for long-term benefits. We plan to further develop CorFix by adding real-time CFD simulation and automatic graft optimization features for bolstering the graft design and evaluation process. Our next study will incorporate many of these changes and focus on recruiting more cardiac surgeons and testing against a larger number of patient surgical cases.
This study had a small sample for recruitment due to the limited number of doctors and their time availability despite 3 months of advertising and 2 additional months during the data collection period. We were able to include individuals with various levels of medical experience, which provides a broad spectrum of users and supports important preliminary insights. Our future study will involve greater participation and a larger number of patient cases to supplement the current results.
This paper reports the design of a VR software for patient-specific designs of vascular grafts that demonstrated feasibility and initial usability in a pilot usability study. All participants were able to create patient-specific graft designs with minimal training, needing on average only 5.49 minutes to design 1 tube-shaped graft and 13.40 minutes to design 1 bifurcated graft. Participants rated the design software with a good degree of usefulness, satisfaction, and ease of use. Further design improvements are needed to visualize hemodynamics during the design process, and a larger study is required to fully compare the VR design to current state-of-the-art surgical procedures.
percentage of nonphysiologic wall shear stress
computational fluid dynamics
comma-separated value
hepatic flow distribution
indexed power loss
inferior vena cava
Simulator Sickness Questionnaire
System Usability Scale
superior vena cava
total cavopulmonary connection
Usefulness, Satisfaction, and Ease of Use Questionnaire
virtual reality
wall shear stress
This paper was supported by the National Institutes of Health (award #R01HL143468 and #R21HD090671).
BK, XL, and AK are founders of and hold shares of stock options in CorFix Medical, Inc. The results of the study discussed in this publication could affect the value of CorFix Medical Inc. This arrangement has been reviewed and approved by the Johns Hopkins University in accordance with its conflict of interest policies.