If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Division of Cardiology, Children's National Hospital, Washington, DCSheikh Zayed Institute for Pediatric Surgical Innovation, Children's National Hospital, Washington, DC
Customized Fontan designs, generated by computer-aided design (CAD) and optimized by computational fluid dynamics simulations, can lead to novel, patient-specific Fontan conduits unconstrained by off-the-shelf grafts. The relative contributions of both surgical expertise and CAD to Fontan optimization have not been addressed. In this study, we assessed hemodynamic performance of Fontans designed by both surgeon's unconstrained modeling (SUM) and by CAD.
Methods
Ten cardiac magnetic resonance imaging datasets were used to create 3-dimensional (3D) models of Fontans. Baseline computational fluid dynamics simulations assessed Fontan indexed power loss (iPL), hepatic flow distribution, and percentage of conduit surface area with abnormally low wall shear stress for venous flow (<1 dyne/cm2). Fontans not meeting thresholds were redesigned using 2 methods: SUM (ie, original venous anatomy without the Fontan was 3D printed and sent to surgeon for Fontan redesign with clay modeling) and CAD (ie, the same 3D geometry was sent to engineers for iterative Fontan redesign guided by computational fluid dynamics). Both groups were blinded to each other's results.
Results
Eight Fontans were redesigned by SUM and CAD methods. Both SUM and CAD redesigns met iPL thresholds. SUM had lower iPL, whereas CAD demonstrated balanced hepatic flow distribution and lower wall shear stress percentage. Wall shear stress percentage shared an inverse relationship with iPL, preventing oversized Fontan designs.
Conclusions
Customized Fontan conduits with low iPL can be created by either a surgeon or CAD. CAD can also improve hepatic flow distribution and prevent oversized Fontan designs. Future studies should investigate workflows that combine SUM and CAD to optimize Fontan conduits.
New manufacturing methods for Fontan grafts necessitate new design tools and collaborative approaches. In addition to computer-aided design, surgeons can also directly craft patient-specific Fontan designs with low power loss.
There is growing interest in the use of customized, patient-specific Fontan designs. Fontan designs can be created by both surgeon and engineering team. Both groups contribute to favorable hemodynamic performance and support an integrative, collaborative approach to optimize graft design and potentially improve surgical result.
An institutional experience with second- and third-stage palliative procedures for hypoplastic left heart syndrome: the impact of the bidirectional cavopulmonary shunt.
Reference values for exercise limitations among adults with congenital heart disease. Relation to activities of daily life—single centre experience and review of published data.
A multicenter, randomized trial comparing heparin/warfarin and acetylsalicylic acid as primary thromboprophylaxis for 2 years after the Fontan procedure in children.
In response to these known issues, the Fontan operation has undergone several surgical modifications, more recently a bifurcated Y-graft design to specifically improve hepatic flow distribution (HFD) and prevent pulmonary AVM formation.
However, the current era of Fontan design remains constrained by off-the-shelf grafts; for example, the Y-graft relies on polytetrafluoroethylene grafts with constraints in vessel diameter.
there is growing interest in developing customized Fontan graft designs. We have previously demonstrated an integrated approach to create customized, patient-specific Fontan conduits.
This approach included image segmentation, computer-aided designs (CAD), modifications guided by computational fluid dynamics (CFD) simulations and electrospinning to create customized Fontan conduits. CFD has been important in predicting the hemodynamic performance of the redesigns.
Several parameters have been identified in the quest to optimize Fontan conduit geometry: lower indexed power loss (iPL), which correlates with improved exercise capacity,
Of note, there has been no established CFD parameter that prevents oversizing of the Fontan conduit, which contributes to thrombosis and neointimal hyperplasia in the Fontan geometry.
Furthermore, most studies investigating Fontan design optimization utilize CAD programs that are not truly unconstrained and free-form as completely intended by the surgeon. Thus, in the next generation of bespoke surgery,
a surgeons' role in optimizing Fontan hemodynamic performances has not been clarified.
The objective of this study was to assess the design of Fontan conduits created by surgeon's unconstrained modeling (SUM) as well as the dominant hemodynamic features of Fontan conduits created by SUM and CAD. With SUM, we present a novel method where a 3D model is freely clay sculptured by the surgeon (Figure 1 and Video 1) and then 3D scanned for CFD. We also present a novel CFD criterion that prevents oversizing of Fontan designs. We hypothesize that SUM Fontan conduits can optimize iPL, whereas other hemodynamic parameters may benefit from CAD informed by CFD simulation.
Video 1Surgeon's unconstrained modeling (SUM). The novel technique of SUM incorporates clay sculpting to directly hand craft unconstrained Fontan designs as intended by the surgeon. The cardiac surgeon (NH) hand sculpted with a clay representation of the Fontan graft that connected the inferior vena cava into the superior cavopulmonary connection. Afterward, modular components were removed and the SUM Fontan was scanned with Artec Eva Lite (Artec 3D, Luxembourg) into a 3-dimensional digital format. Video available at: https://www.jtcvs.org/article/S0022-5223(20)30050-7/fulltext.
Figure 1Surgeon's unconstrained modeling (SUM) and computer-aided design (CAD) of the Fontan conduit. The novel technique of SUM incorporates clay sculpting to directly craft unconstrained Fontan designs as intended by the surgeon. In comparison, CAD relies on engineers and software to create Fontan conduit designs that are optimized based on computational fluid dynamic parameters.
This study was approved by the institutional review boards of all participating institutions. Cardiovascular magnetic resonance imaging datasets from 10 Fontan patients (Table 1) were anonymized and exported in Digital Imaging and Communications in Medicine (DICOM) format. Cardiovascular magnetic resonance imaging data included contrast-enhanced, subtracted magnetic resonance angiography and phase contrast velocity flow cines in the cavae (ie, superior vena cava [SVC] and inferior vena cava [IVC]) and the pulmonary arteries (ie, right pulmonary artery [RPA] and left pulmonary artery [LPA]).
Table 1Details of Fontan patients included in study
The patient cohort consisted of 4 extracardiac type Fontans (coded as EX; D-EX when there was also dextrocardia), 5 lateral tunnel type Fontans (coded as LT) and 1 atriopulmonary type Fontan (coded as AP). Two patients had dextrocardia, in 1 of these patients there was apicocaval juxtaposition. Two patients had bilateral superior vena cava (SVC), 1 patient had a left-sided superior cavopulmonary connection (SCPC). In 2 patients, significant stenosis was noted across the left pulmonary artery (LPA).
L-transposition of great vessels, remote ventricular septal defect
LT2
Lateral tunnel
Double-inlet left ventricle, pulmonary atresia
LT3
Lateral tunnel
Double-outlet right ventricle, remote ventricular septal defect
LT4
Lateral tunnel
Tricuspid atresia
LT5
Lateral tunnel
Tricuspid atresia
Bilateral SVC, LPA stenosis
AP1
Atriopulmonary
Double-inlet left ventricle
Fontan conversion
Cardiac magnetic resonance data were retrospectively collected and processed from this cohort in our study.
∗ The patient cohort consisted of 4 extracardiac type Fontans (coded as EX; D-EX when there was also dextrocardia), 5 lateral tunnel type Fontans (coded as LT) and 1 atriopulmonary type Fontan (coded as AP). Two patients had dextrocardia, in 1 of these patients there was apicocaval juxtaposition. Two patients had bilateral superior vena cava (SVC), 1 patient had a left-sided superior cavopulmonary connection (SCPC). In 2 patients, significant stenosis was noted across the left pulmonary artery (LPA).
Magnetic resonance angiography consisted of a late-phase, nongated, breath-held acquisition with pixel size ∼1.4 × 1.4 mm. Using commercially available image segmentation software (Mimics; Materialise, Leuven, Belgium), a 3D digital model of the original Fontan conduit was reconstructed from the blood pool, including the SVC, IVC, RPA, and LPA. The 3D model was then exported in stereolithography (STL) file format, smoothed and hollowed.
CFD Simulation
CFD simulation methodology was the same as detailed in our previous studies.
The commercial software Ansys Fluent (Ansys Inc, Canonsburg, Pa) was used for CFD simulation. Time-averaged IVC and SVC flow rates were derived from phase velocity data and prescribed as inlet boundary conditions to the CFD simulations. For the outlet boundary conditions, time-averaged RPA and LPA flow rates were prescribed by the product of outlet flow split (RPA:LPA flow) with total venous flow (summation of SVC and IVC flow rate). Other solver parameters are listed in Appendix E1.
Hemodynamic Parameters
The hemodynamic performance of the Fontan conduit and optimization goals were defined by 3 parameters: iPL across the Fontan, HFD, and percentage of Fontan surface area with below-physiologic wall shear stress for venous flow (%WSS), a novel surrogate parameter to prevent Fontan conduit oversizing.
iPL
iPL calculation is based on the methodology of Khiabani and colleagues.
The use of iPL is a form of dimensional analysis that accounts for varying systemic venous flows among Fontan patients. The absolute power loss (Ploss) is measured as the difference between the total hemodynamic energy at the inlets and the total hemodynamic energy at the outlets:
where is the static pressure, ρ is density, Q is flow, and is velocity vector. To account for variability in flow rates and patient size, absolute power loss is then indexed against cardiac output and body surface area:
where Ploss is the power loss value from equation 1, Qs is systemic venous flow, and BSA is body surface area. The denominator has the same dimensions as power (Watts or kg·m2/sec3); thus, indexed power loss is expressed as a dimensionless unit.
HFD
HFD was defined as the ratio of blood from the IVC to the LPA and RPA, respectively. The HFD was evaluated through particle tracking. A specified number of particles (Ntot) were seeded uniformly at the IVC inlet. Ntot was determined by evenly spacing the particles across the IVC, with a 0.1 mm marker size and 1 mm spacing factor. The number of particles passing through the RPA and LPA outlets (NRPA and NLPA, respectively) was calculated at the end of the simulation using the velocity field over the last 1000 time steps and averaged. The HFD was calculated as follows:
%WSS to prevent oversizing
Oversized vessels (relative to flow) will have low WSS which can lead to neointimal hyperplasia and thrombosis.
Thus, luminal surface areas with WSS <1 dyne/cm2 (ArealowWSS) were considered below physiologic. The areas of low WSS were quantified as a percentage of the total surface area of the Fontan model (AreaFontan), creating the novel parameter %WSS:
%WSS was calculated from the CFD simulation. A MATLAB script was written to exclude the superior cavopulmonary connection (SCPC) geometry from %WSS calculation, to selectively evaluate the modifiable component of the geometry.
CFD Thresholds for Conduit Optimization
Fontan models with hemodynamic parameters beyond thresholds were selected for conduit optimization. These thresholds include:
•
iPL >0.03. This benchmark was based on published literature,
In that cohort, the mean iPL was 0.037 and median iPL was 0.031.
•
HFD of RPA:LPA, outside of 40% to 60%/60% to 40% split.
•
%WSS >10%.
Methodology for SUM
The original Fontan conduit was removed from the model to create an SCPC model. For lateral tunnel (LT) and atriopulmonary Fontan conduits, a portion of atrial mass was digitally trimmed and smoothed to allow for connection via an extracardiac (EX) conduit. The SCPC, along with the rest of the heart (eg, atria, ventricles, great arteries, and pulmonary veins) were then 3D printed and mounted onto a modular scaffold that could be placed in supine or standing position. The SCPC model was sent to an experienced congenital cardiac surgeon (blinded to original CFD results). The surgeon then sculpted by hand his ideal Fontan design onto the SCPC using modeling clay (Figure 2, A and B). The rest of the heart and thoracic vascular structures remained in place during sculpting, to ensure that the SUM Fontan was anatomically feasible. The surgeon was also provided with a ruler to guide sculpting. Afterward, the rest of the heart was removed, and the SUM Fontan conduit was 3D scanned (Figure 2, C) with Artec Eva Lite (Artec 3D, Luxembourg) and exported into STL format. CFD was repeated on the SUM Fontan conduits.
Figure 2Surgeon's unconstrained modeling (SUM). This novel technique incorporates clay sculpting to directly craft unconstrained Fontan designs as intended by the surgeon. A, The designated cardiac surgeon (NH) hand sculpted with a clay representation of the Fontan graft that connected the inferior vena cava into the superior cavopulmonary connection (SCPC). The scaffold structure allowed clay sculpting to be performed in either supine or standing position. B, The SCPC and the rest of the heart remained in place during sculpting, to ensure that the sculpted Fontan conduit was anatomically feasible. The SCPC and rest of the heart were held in a modular fashion, such that components could be removed to allow for 3-dimensional (3D) scanning. C, After the modular components were removed, the SUM Fontan conduit was 3D scanned with Artec Eva Lite (Artec 3D, Luxembourg) into a 3D digital format. The SUM Fontan conduit was then evaluated by computational fluid dynamics and compared against computer-aided design Fontan conduits.
The same SCPC model and rest of the heart were independently sent to the engineering team as a digital STL format file. The engineering team was responsible for all CFD simulations and not blinded to CFD results, including the CFD results from the original Fontan. Two general strategies were employed to create new designs: flared, tube-shaped conduits or bifurcated conduits. The designs were carefully made against surrounding anatomy, avoiding collision with great arteries and pulmonary veins. All CAD models were created with Solidworks (Dassault Systems, Velizy-Villacoublay, France). The CAD models of the Fontan conduits were modified iteratively: a range of CAD models were CFD-simulated and then compared with parameter thresholds, influential designs were selected, and further modifications made. Modifications included adjusting offset and angle of the Fontan conduit into SCPC to reduce iPL and improve HFD, as well as reducing conduit size to reduce %WSS. After 3 design iterations, the final CAD model with the most optimal hemodynamic performances (lowest iPL and normal range of HFD/%WSS) was selected. The engineering team was blinded to SUM throughout design process and did not receive the SUM Fontan conduits until after final CAD selection. In 2 models (patients EX2 and LT5), significant LPA stenosis was noted; the SCPC was modified by virtually expanding the LPA before SUM and CAD were repeated.
Surgical Feedback for CAD designs
At the end of the study, the surgeon was shown the CFD results of both CAD and SUM Fontan conduits, then given a survey (Appendix E2) to provide feedback on both designs. Survey questions included likelihood of surgically implanting CAD over SUM, technical difficulty of implanting CAD, and aspects of CAD that should be corrected. The answers were provided on a Likert scale.
Statistical Comparison
All statistical analysis was performed with MedCalc version 12.2 (MedCalc Software, Ostend, Belgium). Spearman correlation analysis was used to investigate the relationship between novel parameter %WSS and iPL results across the entire group.
Results
Original Fontan CFD Results
The CFD results of the ten original Fontans are shown in Figure 3. The group consisted of 4 EX Fontans, 5 LT Fontans and 1 atriopulmonary-type Fontan. Of the 10 models, 8 Fontan conduits demonstrated at least 1 hemodynamic parameter outside of thresholds: 3 Fontan conduits with iPL >0.03, 6 Fontan conduits with suboptimal HFD (ie, beyond 40%-60%/60%-40% split), and 3 Fontan conduits with percent nonphysiologic WSS >10%. The 2 remaining models (EX1 and LT3) had parameters within thresholds and were not included for optimization. For 2 Fontan models (EX2 and LT5), significant narrowing was noted across the LPA that could not be addressed by a Fontan redesign. Because this feature influenced CFD results significantly, the LPA stenosis was virtually stented; that is, expanded to remove this effect from the results between the models.
Figure 3Three-dimensional (3D) representation and computational fluid dynamics (CFD) results of original Fontans. A cohort of Fontan patients were retrospectively collected and digitally processed into 3D models for CFD simulation. The cohort consisted of 4 extracardiac-type Fontans (coded as EX or D-EX when there was also dextrocardia), 5 lateral tunnel-type Fontans (coded as LT) and 1 atriopulmonary-type Fontan (coded as AP). The CFD results included indexed power loss (iPL), hepatic flow distribution of the left pulmonary artery (HFDLPA), and percentage of Fontan surface area with below-physiologic wall shear stress for venous flow (%WSS). CFD results in red were considered beyond thresholds. These Fontans were then selected for conduit optimization with parameter optimization goals as listed above. These Fontans underwent either surgeon's unconstrained modeling or computer-aided design.
Redesigned Fontan Conduit CFD Results (SUM and CAD Methods)
Eight sets of SUM and CAD Fontan conduits were created and compared (Figure 4). When compared with the 8 original Fontan conduits, all SUM and CAD Fontan conduits had acceptable iPL <0.03. Of note, the SUM Fontans had lower mean iPL (0.017 ± 0.007) compared with the mean CAD Fontans (0.021 ± 0.004). When compared with the mean of original Fontans (0.031 ± 0.027), both SUM and CAD Fontan conduits trended toward improvement in iPL.
Figure 4Three-dimensional (3D) representation and computational fluid dynamic (CFD) results of surgeon's unconstrained modeling (SUM) and computer-aided design (CAD). The selected cohort of 8 Fontan models underwent both CAD and SUM, with the resultant new models then re-evaluated by CFD. The CFD results, including indexed power loss (iPL), hepatic flow distribution (HFD), and Fontan areas with below-physiologic wall shear stress for venous flows (%WSS), are labeled for each SUM or CAD result. Results that are labeled in red were considered beyond acceptable threshold parameters. Both SUM and CAD had iPL within thresholds; CAD demonstrated improved HFD and lower %WSS. LPA, Left pulmonary artery.
The SUM models had larger range of HFD compared with CAD models. Only 1 of the SUM Fontan conduits had HFD within 40% to 60%/60% to 40% split; whereas all CAD models were within HFD thresholds. For EX2 and LT5, to account for changes in pulmonary flow after virtual LPA enlargement, the simulation tested across a range of RPA:LPA flow splits, a methodology proposed by de Zelicourt and colleagues.
In these 2 models, CAD models demonstrated improved balance in HFD compared with SUM models (Figure 5). As illustrated in Figure 6, A, HFD was influenced significantly by small changes in conduit offset with the SCPC. For EX2 and the patient with EX and D-EX3, bifurcated grafts were selected because they demonstrated the most stable HFD.
Figure 5Hepatic flow distribution (HFD) relative to pulmonary flow distribution, in surgeon's unconstrained modeling (SUM) and computer-aided design (CAD) in patient EX2 (extracardiac Fontan in a patient with hypoplastic left heart syndrome and left pulmonary artery [LPA] stenosis and patient LT5 (lateral tunnel Fontan with tricuspid atresia, bilateral superior vena cava, and LPA stenosis). In 2 patients within the selected Fontan cohort, significant LPA stenosis was noted that was not modifiable by a new Fontan conduit. A virtual stent was placed across the LPA to reduce power loss. Theoretically, this modification changes the outflow boundary conditions, which potentially affects the HFD results. Thus, in both models created either by SUM or CAD, computational fluid dynamic analysis was performed across a range of pulmonary flow distributions. CAD demonstrated balanced HFD across these ranges.
Figure 6Computational fluid dynamics (CFD) results of surgeon's unconstrained modeling (SUM) and all computer-aided design (CAD) iterations in patient LT4 (ie, lateral tunnel Fontan patient with tricuspid atresia). To elaborate on the significance of CFD parameters, different iterations of 1 patient are shown. A, The hepatic flow distribution (HFD) iterations across the specific patient LT 4, including SUM, CAD, and final CAD, varied significantly as a result of changes to Fontan flaring, offset, and angulation affect HFD. B, Plot distribution of indexed power loss and lower Fontan areas that had below-physiologic wall shear stress for venous flow (%WSS). There is a negative correlation between both parameters, prohibiting designs that inflated the Fontan conduit (meant to decrease indexed power loss) to disproportionate sizes (which increased percentage of wall shear stress).
SUM and CAD models demonstrated similar %WSS when compared with original Fontans. However, 3 of the SUM Fontans had %WSS beyond threshold parameters. Meanwhile, the CAD models had %WSS that were all within threshold parameters. As demonstrated in Figure 6, B, there was an inverse relationship appreciated between iPL and %WSS. There was a negative correlation between iPL and %WSS results across the entire group (r = –0.57; P = .003).
Surgical Feedback of CAD Designs
Feedback on all 8 CAD models were provided by the surgeon (Table E1). Out of the 8, 6 were considered by the surgeon to be “very likely” used without any significant changes; these models were flared, tube-shaped grafts. Two CAD models (patients EX2 and D-EX3) were considered more complex in surgical implantation, and less likely to be used by the surgeon without significant changes. In both cases, the CADs were bifurcated grafts.
Discussion
To date, this is the first CFD study that investigates the dominant hemodynamic features of both surgeon-led unconstrained designs and computer-based designs of the Fontan conduit. We developed a novel SUM methodology to generate Fontan designs literally crafted by surgical hands and assessed them by CFD. We found that the SUM Fontan conduits were hemodynamically efficient with minimal power loss, although they continued to demonstrate HFD imbalance and larger areas of %WSS. Conversely, some CAD designs were considered complex for surgical implantation and not preferred over SUM/original Fontan designs. This study confirms that a surgeon, independent of material and design constraints, can generate Fontan conduit designs with minimal power loss. However, our results suggest that a surgeon may be challenged by the influence of conduit geometry on WSS and HFD. Thus, this study also clarifies the specific benefit of CAD in terms of balancing HFD and preventing oversized Fontan conduits.
To our knowledge, this is the first known use of clay modeling whilst investigating Fontan conduit hemodynamic parameters. Historically, clay modeling is among the earliest forms of surgical simulation
Furthermore, clay modeling allows the surgeon to truly create free-form designs untethered by limitations in CAD geometrical tools (such as lofting or spline functions). For customized Fontans, we believe that SUM can be incorporated into the collaborative approach of graft design: An experienced cardiovascular surgeon can produce an initial clay model that is tested by CFD, with modifications made by CAD before electrospun into a patient-specific optimized design.
Our results are remarkable in that the surgeon's unconstrained designs had low iPL, even when compared with CAD. Hemodynamically inefficient Fontan conduit designs will lead to increased workload on the heart, increased risk for heart failure, and worse clinical outcomes,
and SUM methodology represents the application of free-form, patient-specific Fontan designs exclusively relying on surgical experience and surgical intuition
without the aid of CFD or an engineering team. In this study, a surgeon naturally considered the geometrical factors of offset, curving, and flaring, which all contribute to energy reduction.
Our results show that surgical intuition makes singular, significant contributions to optimizing Fontan hemodynamic performances, an example of the gut feeling an experienced surgeon uses in predicting postoperative outcomes.
Furthermore, the time involved for SUM was less than CAD (Figure E1). Without surgeon intuition, the engineering team instead relied on a shotgun approach to design, requiring weekly evaluation of CFD results and overall feasibility. This is consistent with other CAD processes requiring >2 weeks of processing to achieve a single optimized design.
Meanwhile, the surgeon spent <15 minutes to sculpt each model; thus, SUM could reduce the number of iterations involved. Theoretically, a surgeon could alternatively use dedicated surgical modeling software
; however, this method does not include surgeons' haptic feedback and still has geometrical constraints for complex designs.
The CFD-based differences between SUM and CAD demonstrates the additional benefit provided by CAD. As demonstrated in Figure 6, A, small adjustments to conduit-SCPC offset and angulation can cause significant changes in HFD, consistent with previous CFD-based observations of the Fontan geometry.
Incremental changes to offset and angulation (approximately 2 mm and 10°, respectively) resulted in an approximate 5% to 10% change in HFD.
CFD can also identify regions of the Fontan geometry beyond conduit implantation that contribute to hemodynamic inefficiency. As was the case in patients EX2 and LT5, the original CFD simulation determined that LPA stenosis would contribute to significant power loss. In a clinical scenario, virtually expanding the LPA would represent placement of a stent required to reduce iPL. However, the bifurcated grafts generated by CAD still required modifications based on surgical feedback, a statement toward the increased technical complexity of Y-graft Fontan designs.
In our study, %WSS is introduced as a novel CFD-based parameter to prevent oversizing of Fontan designs. Studies into the optimal extracardiac Fontan conduit sizes relative to the IVC have shown that oversized Fontan conduits can lead to flow stagnation and thrombosis.
For Fontan conduits, the role of WSS has not been validated yet. The only CFD-based study that demonstrated this potential relationship was the validation of the Fontan Y-graft with CFD by Yang and colleagues
; one of the Y-graft limbs was noted to significant thrombosis correlating to an area of low WSS.
Nevertheless for CAD, a mechanism such as %WSS is required to prevent oversized Fontan conduits. As illustrated in Figure 6, B, %WSS prevented designs that inflated the Fontan conduit to disproportionate sizes (minimizing power loss whilst introducing significant areas of low WSS). Without this parameter, conduit sizes beyond 22 mm would have been developed by CAD. We considered alternative measures, including particle residence time
or presence of low velocity fields; however, there are no physiologic standards and neither can isolate the contribution of conduit from the rest of Fontan geometry. Of note, the use of rigid body assumptions in CFD simulation potentially may lead to lower WSS values
and thus overestimation of %WSS, although areas of low WSS can still be identified.
In our study, the heart and aortic arch anatomy were included as geometrical constraints. However, the intraoperative environment is different and small variations in surgical placement may occur, potentially limiting the benefits of SUM/CAD. Trusty and colleagues
described this effect by comparing HFD results between surgical planning and postoperative anatomy in 12 Fontan patients. In their study, the HFD prediction error was 17% ± 13% and directly correlated with offset error.
Thus, prediction of hemodynamic parameters in surgical planning depends on accurate and dynamic representation of in vivo anatomy from 3D modeling.
We only simulated a resting physiologic state and did not consider exercise conditions, which increases venous flow and power loss across the Fontan conduit. At the same time, much of the work from the Cardiovascular Fluid Mechanics Laboratory in Georgia Tech justified our use of iPL, which is flow-independent,
Other parameters such as power efficiency (EnergyOut/EnergyIn) could also be considered; however, the available literature is inadequate to create benchmark parameters.
Our study was limited by small sample size, due to constraints in time (for CAD, 3-4 weeks per case) and resources (for SUM, $400 in 3D printing materials per case). Additionally, our study was limited by method bias, as CAD was directly guided by and iteratively optimized by CFD, leading to favorable CFD results. Thus, this study should not be considered a direct comparison between SUM and CAD. %WSS has not been directly validated outside of observational CFD studies
and the optimization parameter of %WSS <10% was an arbitrary cutoff. Finally, the accuracy of the hemodynamic parameters will be influenced by patient growth, even with tissue-engineered vascular-grafts–based Fontans.
Future work will entail CFD simulations of a larger Fontan cohort to clarify %WSS ranges that should be utilized in optimization, longitudinal studies to investigate changes to Fontan geometry after surgical placement, as well as 3D visualization methods that could improve surgical graft placement.
Conclusions
In the development of customized Fontan designs, we have demonstrated how cardiac surgeons can contribute to Fontan hemodynamics and create efficient Fontan designs based on surgical intuition. CAD can balance HFD and prevent oversizing. Future studies to optimize workflow in patient-specific Fontan design should investigate the combination of both surgical intuition and CAD.
Conflict of Interest Statement
Dr Krieger has a patent pending (Application and Methods of Patient-specific Tissue-engineered Vascular Graft using Electrospinning [US Patent App 62/209,990, August 2016]) and Dr Hibino has a patent issued (No. PCT/US2016/49080). All other authors have nothing to disclose with regard to commercial support.
Surgeon's unconstrained modeling (SUM). The novel technique of SUM incorporates clay sculpting to directly hand craft unconstrained Fontan designs as intended by the surgeon. The cardiac surgeon (NH) hand sculpted with a clay representation of the Fontan graft that connected the inferior vena cava into the superior cavopulmonary connection. Afterward, modular components were removed and the SUM Fontan was scanned with Artec Eva Lite (Artec 3D, Luxembourg) into a 3-dimensional digital format. Video available at: https://www.jtcvs.org/article/S0022-5223(20)30050-7/fulltext.
CFD solver assumptions included laminar flow in the Fontan, rigid body, blood as incompressible fluid, and flat uniform profile at superior vena cava and inferior vena cava. The unsteady Navier-Stokes equations were solved for 3000 time steps. The time step of each iteration was set at 0.001 seconds. A nonuniform triangular meshing with minimal size = 0.7 mm and 5 wall layers was used. Each step was solved within 40 iterations by the SIMPLE numerical solving method. The convergence values were set to 105 for x-, y-, z-velocity and mass conservation residuals. Results were averaged over the last 1000 time steps.
Appendix E2. Questions Used for Surgical Feedback
Complexity of Design
Please rate the expected complexity in surgical implantation of the computer-aided design (CAD) Fontan geometry compared with surgeon's unconstrained modeling (SUM) and native Fontan.
•
Very difficult/not feasible
•
Difficult
•
No difference
•
Easy
•
Very easy
Likelihood of Utilizing CAD Geometry
What is the likelihood of utilizing the CAD for the Fontan operation of this specific anatomy?
•
Very unlikely (unfeasible)
•
Unlikely (+ significant changes needed)
•
Possible (+ minor changes)
•
Likely (+ minor changes)
•
Very likely (no changes needed)
Modifications to CAD Geometry
What are aspects of the CAD that need to be corrected?
•
Course of the graft (to adjust for other anatomic structures)
•
Size of the graft
•
Insertion of the graft into the superior cavopulmonary anastomosis
•
Bifurcation of the graft
•
Other
Table E1Surgical feedback on Fontan geometries made by computer-aided design (CAD)
The patient cohort consisted of 4 extracardiac type Fontans (coded as EX; D-EX when there was also dextrocardia), 5 lateral tunnel type Fontans (coded as LT), and 1 atriopulmonary type Fontan (coded as AP). EX1 and LT3 were not included in the process because all their hemodynamic parameters were already within optimization thresholds.
Please rate the expected complexity in surgical implantation of CAD, compared with SUM and native Fontan
What is the likelihood of utilizing the CAD for the Fontan operation of this specific geometry?
What are aspects of the CAD that need to be corrected?
EX2
Difficult
Unlikely (+ significant changes needed)
Bifurcation of the graft
D-EX3
Difficult
Unlikely (+ significant changes needed)
Bifurcation of the graft
D-EX4
Difficult
Very likely (no changes needed)
Course of the graft (to adjust for other anatomic structures)
LT1
Very easy
Very likely (no changes needed)
LT2
Very easy
Very likely (no changes needed)
LT4
Very easy
Very likely (no changes needed)
LT5
Difficult
Very likely (no changes needed)
Insertion site of the graft into the superior cavopulmonary anastomosis
AP1
Easy
Very likely (no changes needed)
Insertion site of the graft into the superior cavopulmonary anastomosis
∗ The patient cohort consisted of 4 extracardiac type Fontans (coded as EX; D-EX when there was also dextrocardia), 5 lateral tunnel type Fontans (coded as LT), and 1 atriopulmonary type Fontan (coded as AP). EX1 and LT3 were not included in the process because all their hemodynamic parameters were already within optimization thresholds.
Figure E1Comparison of the process between surgeon's unconstrained modeling (SUM) and computer-aided design (CAD) in creation of new Fontan conduits. Patient LT2 is demonstrated in this example. The engineering team relied on a shotgun approach to design, requiring weekly meetings to evaluate computational fluid dynamics results and review overall feasibility. Meanwhile, the surgeon spent <15 minutes to sculpt each model as opposed to CAD.
An institutional experience with second- and third-stage palliative procedures for hypoplastic left heart syndrome: the impact of the bidirectional cavopulmonary shunt.
Reference values for exercise limitations among adults with congenital heart disease. Relation to activities of daily life—single centre experience and review of published data.
A multicenter, randomized trial comparing heparin/warfarin and acetylsalicylic acid as primary thromboprophylaxis for 2 years after the Fontan procedure in children.
Supported by National Institutes of Health award numbers R01HL143468 and R21HD090671 as well as University of Maryland supercomputing resources. The content is solely the responsibility of the authors and does not represent the official views of the National Institutes of Health.
The year is 2050. Freddy Jones is about to undergo a Fontan operation. Freddy is an unusual patient. Most blastocysts are genetically screened for hypoplastic left heart syndrome, and if detected, a developing patient undergoes either embryological epigenetic modification or fetal intervention, preventing or eliminating the disorder. These techniques are still not globally available, so Freddy was not screened. Thus, Freddy is managed with the older, 3-stage palliation. In preparation for the Fontan, Freddy undergoes cardiac magnetic resonance imaging with complete pressure mapping (diagnostic catheterization became obsolete by 2030).
Go is an ancient Chinese board game that is played today exactly as it was more than 3000 years ago.1 It is a game with simple rules, in which 2 players take turns putting black and white stones onto a flat piece of wood on which a grid is printed. To win, a player captures or surrounds more territories than the opponent. Despite its simplicity, Go is far more complex than chess, with possible moves and legal positions estimated to be greater than the number of atoms in the known universe. Go is a game more of intuition than anticipation or calculation.
What does clay have to do with computational flow dynamics (CFD) and the personalized Fontan (herein referred to as the TCPC)? Loke and colleagues' response1 is that a surgeon-fashioned clay TCPC actually performed reasonably well and could reduce the computational time and money by serving as starting input for the CFD iterative process. With the availability of surgical modeling software, we don't fully understand the need for clay and haptic feedback but, must admit, find the use of clay kind of appealing—can't explain it, maybe it invokes pleasant childhood memories.
30050-7&id=fx1.jpg){kind=link}
30050-7&id=fx2.jpg){kind=link}
30050-7&id=fx3.jpg){kind=link}
30050-7&id=gr1.jpg){kind=link}
30050-7&id=gr2.jpg){kind=link}
30050-7&id=gr3.jpg){kind=link}
30050-7&id=gr4.jpg){kind=link}
30050-7&id=gr5.jpg){kind=link}
30050-7&id=gr6.jpg){kind=link}
30050-7&id=fx4.jpg){kind=link}