3D-Printing to Plan Complex Transcatheter Paravalvular Leaks Closure

Background: Percutaneous closure of paravalvular leak (PVL) has emerged as an alternative to surgical management in selected cases. Achieving complete PVL occlusion, while respecting prosthesis function remains challenging. A multimodal imaging analysis of PVL morphology before and during the procedure is mandatory to select an appropriate device. We aim to explore the additional value of 3D printing in predicting device related adverse events including mechanical valve leaflet blockade, risk of device embolization and residual shunting. Methods: From the FFPP registries (NCT05089136 and NCT05117359), we included 11 transcatheter PVL closure procedures from three centers for which 3D printed models were produced. Cardiac CT was used for segmentation for 3D printed models (3D-heartmodeling, Caissargues, France). Technology used a laser to fuse very fine powders (TPU Thermoplastic polyurethane) into a final part-laser sintering technology (SLS) with an adapted elasticity. A simulation on 3D printed model was performed using a set of occluders. Results: PVLs were located around aortic prostheses in six cases, mitral prostheses in four cases and tricuspid ring in one case. The device chosen during the simulation on the 3D printed model matched the one implanted in eight cases. In the three other cases, a similar device type was chosen during the procedures but with a different size. A risk of prosthesis leaflet blockade was identified on 3D printed models in four cases. During the procedure, the occluder was removed before release in one case. In another case the device was successfully repositioned and released. In two patients, leaflet impingement was observed post-operatively and surgical device removal had to be performed. Conclusion: In a case-series of complex transcatheter PVL closure procedures, hands-on simulation testing on 3D printed models proved its usefulness to plan and facilitate these challenging procedures.


Introduction
Transcatheter closure of paravalvular leaks (PVL) has emerged as an alternative to surgical management in selected cases [1][2][3][4][5]. A self-expanding occluder with a memory of shape property is deployed within the leak. The main challenges are achieving complete leak occlusion, achieving stable occluder implantation while maintaining valve prosthesis function [3]. A wide spectrum of devices is used, of which two types are specifically labelled for this procedure, namely the paravalvular leak occluder (PLD) (Occlutech GmbH, Jena, Germany) and Amplatzer paravalvular leak plug (AVP) 3 (Abbott Medical, Plymouth, MN, USA) with multiple device sizes available [6,7]. A multimodal imaging analysis before the procedure and transesophageal echocardiography during the procedure allows to choose one of these devices based on PVL shape and measurements [8,9]. Device selection is a crucial step for procedural success. Indeed, complications have been reported following these procedures including mechanical valve leaflet blockade, device embolization, mechanical hemolysis when residual shunt occurs and pericardial effusion [4,5,10]. Similarly, hands on simulation on 3D printed models has emerged as a useful tool to plan complex cardiac interventions [11][12][13][14][15][16].
These 3D printed models incorporate all the anatomical variables, the geometry of the defect but also the surrounding features of the valve and offers devices testing. Already, pre-procedural tests on 3D printed models have shown to be useful to plan transcatheter left atrial appendage closure [17], percutaneous pulmonary valve implantation [12] and transcatheter correction of sinus venosus atrial septal defect [18]. We aim in this study to assess the benefit of hands-on simulation on 3D printed models to plan complex transcatheter PVL closure.

Methods
We retrospectively selected PVLc procedures in which 3D printed models were used prior to the procedure from the prospective FFPP registry that included 238 procedures in 213 patients between 2017 and 2019 (NCT05089136 and NCT05117359). A single operator (Vlad Ciobotaru, 3D-heart modeling, Caissargues, France) performed all 3D modeling and 3D printed models. We collected clinical data, anatomical data, procedural data and outcomes. A descriptive analysis was performed. All patients had a multimodality imaging analysis including 3D transesophageal echocardiography, cardiac CT and a 3D printed model based on the CT scan.

The 3D Printing Process
A multiphase cardiac CT was performed before the procedure which was used for segmentation (Philips IntelliSpace Portal V11, Best, The Netherlands) based on an intelligent recognition of the tissues from point to point, including muscle and wall parts, enhanced cavities and the cardiac prosthesis and surrounding calcifications. A volume rendering was obtained from segmented structures and was then transformed into a stereolithography (STL) file. Post processing of the STL file was performed offline: for smoothing, hollowing, trimming or thickening ( Figure 1). We employed powder-based 3D printing technology that uses a laser to fuse very fine powders (TPU, thermoplastic polyurethane) into a final part-laser sintering technology. The elasticity of the 3D model was adapted according to the selecting-laser-sintering process parameters and the thickness of each component.  . Each cardiac structure of interest has been segmented: aortic wall, mechanical aortic valve prothesis, left myocardium, annular calcification (*). A diastasis is observed between the mechanical aortic prosthesis (P) and the aortic wall (Ao) corresponding to a paravalvular leak (PVL) (red arrow). (B) CT full volume rendering of the segmented structures. The mechanical aortic valve is displayed in green. The PVL is seen from a left ventricular view. (C) A standard triangle language (STL) file was created from segmented structures. The PVL is seen from the aorta (black arrow). (D) Threedimensionally printed model derived from the STL. Visualization of the aortic root, aortic valve (P) and the PVL (black arrow). PVL: paravalvular leak; Ao: aortic; LV: left ventricle; P: prosthesis.

Simulation and Testing on 3D Printing Models
A set of devices including PLD Rectangular Waist and AVP 3, Amplatzer Septal Occluder (Abbott Medical, Plymouth, MN, USA), Amplatzer Vascular Plug 2 (Abbott Medical, Plymouth, MN, USA), Amplatzer Duct Occluder (Abbott Medical, Plymouth, MN, USA), Muscular VSD Occluder (Abbott Medical, Plymouth, MN, USA) was used to test PVL closure on the 3D printed models. A Sapien valve 29 mm in diameter (Edwards Lifesciences, Irvine, CA, USA) was used to simulate valve-in-valve and valve-in-ring as an alternative strategy. The risk of leaflet impingements, device embolization and residual leak were investigated during the manipulations.

Results
A total of 11 patients were included from three centers (Table 1). Patients had mitral prostheses in four cases, of which three were mechanical; aortic prosthesis in six cases of which two were bioprosthesis, two were Perceval sutures-less valves, two were mechanical valves. A transcatheter PVL closure was performed post tricuspid valve-inring implantation. . Each cardiac structure of interest has been segmented: aortic wall, mechanical aortic valve prothesis, left myocardium, annular calcification (*). A diastasis is observed between the mechanical aortic prosthesis (P) and the aortic wall (Ao) corresponding to a paravalvular leak (PVL) (red arrow). (B) CT full volume rendering of the segmented structures. The mechanical aortic valve is displayed in green. The PVL is seen from a left ventricular view. (C) A standard triangle language (STL) file was created from segmented structures. The PVL is seen from the aorta (black arrow). (D) Three-dimensionally printed model derived from the STL. Visualization of the aortic root, aortic valve (P) and the PVL (black arrow). PVL: paravalvular leak; Ao: aortic; LV: left ventricle; P: prosthesis.

Results
A total of 11 patients were included from three centers ( Table 1). Patients had mitral prostheses in four cases, of which three were mechanical; aortic prosthesis in six cases of which two were bioprosthesis, two were Perceval sutures-less valves, two were mechanical valves. A transcatheter PVL closure was performed post tricuspid valve-in-ring implantation.

Choice of Size and Type of PVLc Plug
One device was used in six patients, two devices in two patients and three devices in three cases. A device implanted was removed before release in one patient. An AVP 3 plug was implanted in 6/11 patients and an AVPII was used in two cases in combination with another plug in one case. A VSD plug was used in two patients. The choice of the optimal device used during the simulation on the 3D model matched with the type and size of plug used during the procedure in eight cases (see Table 1 case 1). In one patient, the inserted device was smaller than the one chosen during simulation and this patient required two additional plugs to cover a residual leak (see Table 1, case 2). In three patients, the inserted device was larger than the one chosen during the simulation on the 3D model; two of these patients required surgery for leaflet blockage. (see Table 1 case 4,5).
An alternative strategy to PVL closure with occluder was considered in two cases and successfully applied. A valve-in-valve in sutures-less Perceval prothesis was performed (see Table 1 case 3). In one patient with valve-in-tricuspid ring, two VP2 were implanted (see Table 1 case 6).

Detection of Prosthetic Leaflet Blockage
The prospective 3D printing simulation detected a risk of prosthetic blockage in four cases. Interestingly, in a patient with mechanical mitral PVL, the appropriate choice of the device was an AVP3, but a risk of leaflet blockade was identified that depended on the orientation of this oval device. This issue occurred during the procedure and was solved by proper orientation of the AVP3 in a similar position as during the simulation (Figures 2-4). In one case with a large mitral PVL ( Figure 5), an ASO that was initially deployed, was removed due to leaflet blockade. In two other patients, the plug-induced blockage was anticipated during the simulation on the 3D model and occurred a few hours after the end of the procedure (Figures 7 and 8) and required emergent surgery for leaflet blockade.

Detection of Residual Shunt
A risk of residual shunt was anticipated in six cases and observed after the procedure in three cases.
The targeted PVL was successfully sealed in 9 of 11 cases with minor residual leak in 4 cases. In one case, the residual PVL was very small and was not expected on the simulation (Figures 3 and 4).

Discussion
In this multi-center case-series, we illustrate how hands-on simulation testing on 3D printed models can be integrated into the planning process of transcatheter PVL closure in addition to multimodal imaging improving communication within the heart team: Imager/Interventional/Surgeon for feasibility of percutaneous addressing. Although complex to achieve, a 3D printed model of PVL was demonstrated to be feasible and useful to predict device-related adverse events such as residual leak and valve leaflet blockade.
A CT scan is not routinely performed to assess PVL given its limits inherent to valve artifacts. However, integrating these limits, we successfully obtained cardiac segmentation of the valve, the PVL and the surrounding structures with morphological assessment complementary to the one provided by echocardiography. Accurate 3D printing models require adapted cardiac CT protocol to optimize the temporal and spatial resolution. Technical specifications included low-pitch spiral acquisition and thin sections with low incremental steps, paying particular attention to the suppression of dynamic, hardening, or metallic artifacts (beam-hardening artifact) by adapting windowing, adjustments of specific algorithms and beam energy, but also by performing additional hard-filter reconstructions and iterative post-processing reconstructions [19]. The process of segmenting the heart is time consuming and may be a limitation of the technique if not performed in conjunction and agreement with other imaging techniques especially 3D TEE. An accurate, fully automated, CT anatomical segmentation would facilitate the process and reproducibility [20].
The appropriate choice of materials and printing techniques are another important point. Multi-material 3D printed models have already been used to explore the role of annular calcification in PVL formation post-TAVR in a retrospective study by Zen Qian et al. by using 3D printed tissue-mimicking phantoms of aortic root, in vitro [21]. SLS technique has been chosen because of its very high spatial resolution and efficient cost-quality ratio [22]. The SLS technique has the advantage of being executed without separate feeder for support material. This allows the printing of complex geometries that were not feasible with other techniques. Moreover, the thickness can be as thin as 0.5 mm, allowing a high flexibility for the aortic wall, whereas a larger thickness makes the material stiffer to mimic a prosthetic ring or leaflet. The TPU material is tear-resistant, permitting leaflets movement around the hinge points.
Three-dimensional printing has the advantage of displaying all the elements required to adequately select an occluder that will match with the complex PVL shape. Location, size, number and morphology of PVL, together with the relationships with the valve and the surrounding structures, are depicted in a condensed manner in the 3D printed models [15,16,22,23].
The PVL closure procedures correspond to very diverse, sometimes complex situations, with their own specificities depending on the type of prosthesis and the PVL location. The strategy and pre-operative planning are of paramount importance with a tailored approach facilitated by 3D model testing.
In our series, the prediction of the risk of leaflet blockade was anticipated in all cases during testing, whereas it was uncertain using imaging alone. Indeed, interference with the mitral prosthesis can occur on the ventricular side, which is not accessible to direct analysis by TEE. Furthermore, the expansion of the plug and its alignment with the axis of the prosthetic wing is hard to predict solely by imaging., depending strongly on the shape of the PVL.
Interestingly, in two patients, the obstruction occurred a few hours after the procedure despite a result appearing optimal with a normal symmetrical movement of the leaflet initially at the end of the procedure, requiring a secondary surgical valve replacement. This delayed leaflet blockage was due to a change in preload conditions and not because of a displacement or migration of the prosthesis which was confirmed during the per-surgical examination showing an adequate localization of the PVLc device. It should be noted that all PVLc procedures were validated by a heart team and planned on TEE and preoperative CT, however this obstruction risk had been anticipated solely by the 3D printing simulation and dynamic testing using the same device but not by imaging alone.
Very complex, unusual hybrid strategies have been planned on 3D models. A severe tricuspid regurgitation was treated by SAPIEN valve-in-ring implantation. The ring was semi-circular and a residual severe PVL was expected on the simulation, adequately resolved by implantation of two VP 2 in a strategy similar to what was planned on preprocedural testing. In this case, simulation on 3D printing gives us confidence of the feasibility of this complex strategy and makes the procedure fast and straightforward following the plan. Two other cases of PVL on Perceval valves due to stent infoldings, illustrate the importance of preoperative testing for procedure planning. In these cases, the simulation on the 3D printed models allowed us to identify that the valve-in-valve strategy would adequately address the issue versus a plug implantation, which was shown as unstable, inefficient and with risk of obstruction towards the coronary ostium. The simulation allowed a fast and efficient procedure without losing time, energy and devices in a wrong approach. Additional costs per model are required, however, these costs could be balanced by the reduction in the number of devices used and the reduction in adverse events in complex cases [23]. When considering the cost of a 3D printed model, a global cost-analysis of the procedure should be performed. On-screen simulation is being developed for many cardiac interventions. However, simulation on 3D printed models, in addition to providing a unique comprehensive anatomical analysis, allows to understand the interaction between devices, the valvular prosthesis and adjacent structures taking into account the deformations of the devices and the models that cannot be assessed by the usual imaging modalities [24].

Limits
We present a limited series of cases. However, to our knowledge, this is the largest study to date to investigate the additive value of 3D printing to plan PVL closure. Furthermore, the diversity of cases illustrates the usefulness of 3D printing planning in a wide spectrum of PVL closure.
The simulation on the 3D model was not standardized, given continuous improvement in materials and techniques for 3D printing as well as evolution in PVL closure techniques.
The segmentation of the CT and 3D printing models were performed by a single operator cumulating experience in cardiac imaging, 3D printing and interventional cardiac echocardiography. Reproducibility of the technique with other operators remains to be demonstrated.
Larger, prospective, multicenter studies should confirm the contribution of 3D printing to multimodality imaging in complex PVL closure procedures.

Conclusions
In a case-series of complex transcatheter PVL closure procedures, hands-on simulation testing on 3D printed models proved its usefulness to plan and facilitate these challenging procedures.

Key Question
Prosthetic paravalvular leaks leading to heart failure and/or hemolysis can be treated by interventional cardiology or open-heart surgery. Transcatheter closure remains challenging, and 3D printed models may be helpful in selected cases to guide these complex procedures.

Key Findings
Hands-on simulation testing on 3D printed models predicted the risk of device related adverse events such as leaflet blockage or residual paravalvular leak.

Take-Home Message
Three-dimensional printing can be integrated into the multimodal planning of transcatheter paravalvular leaks closure procedures to improve outcomes.

One-Sentence Summary
A series of 11 transcatheter paravalvular leaks closure cases were planned on 3D laser sintering printed soft TPU model, permitting device implantation testing and prediction of leaflet blockage or residual leak.