SYSTEM AND METHOD FOR RECONSTRUCTING IMPLANTABLE DEVICE GEOMETRY FROM PATIENT-SPECIFIC IMAGING SCAN - Patent application

Abstract

A model of the in vivo heart valve device implanted in a patient is generated by providing a known model of the pre-implanted heart valve device, performing an imaging scan on the patient having the implanted heart valve device to obtain at least one patient-specific landmark, deforming the known model to fit the at least one patient-specific landmark to obtain a constructed patient-specific valve model, and simulating the leaflets and skirt for the constructed patient-specific valve model by finite element analysis, wherein the constructed patient-specific valve model accurately represents the geometry of the implanted heart valve device in its current in vivo configuration with an error of less than 0.5 mm.

Description

[Technical field]

[0001] This application claims priority to U.S. Provisional Patent Application No. 63 / 187,057, filed May 11, 2021, which is incorporated by reference in its entirety. [Background technology]

[0002] Due to the inevitable structural deterioration of the biological valve leaflets, patients who have undergone surgical aortic valve replacement (SAVR) or transcatheter aortic valve replacement (TAVR) may require a second intervention. For example, successful TAV implantation in deteriorated SAV or TAV has been reported as a minimally invasive solution to this problem. As the application of TAVR expands to intermediate and even low-risk populations, more patients will require a second or possibly third TAVR. To avoid complications related to the ViV procedure, thorough pre-procedure planning must be performed to address valve anchor fixation, paravalvular leak, and coronary artery occlusion associated with the existing valve. Summary of the Invention [Problem to be solved by the invention]

[0003] Advanced pre-procedural planning typically involves modeling valve deployment using finite element analyses (FEA) and / or computational fluid dynamics (CFD) simulations to assess hemodynamics. It is beneficial to accurately determine the existing valve geometry for high-fidelity pre-operative simulations. However, due to blooming artifacts caused by metallic stents, valve stents segmented from patient CT scans appear several times thicker. Furthermore, due to low tissue signal intensity, segmenting the biological leaflets from CT scans is challenging, making prediction of coronary artery obstruction by the leaflets difficult. One method to address this issue is to attempt to capture the final valve geometry using PEA-based simulations that deploy the valve in the patient's pre-TAVR anatomy. However, the error associated with this method typically falls within the range of 2–4 mm. Attempts have been performed to reconstruct clean stent frames, but prior art methods have been complex and valve-specific or did not include the leaflets or skirt in the model. [Means for solving the problem]

[0004] Exemplary methods and systems are disclosed that use an image-based matching framework to accurately recover implanted device geometry (e.g., implanted TAV or SAV geometry) from patient-specific CT scans (e.g., with blooming artifacts eliminated). The exemplary methods and systems are configured to accurately reconstruct patient-specific valve geometry from CT scans by utilizing known valve models and by employing an image-based matching framework.

[0005] In some aspects, the exemplary systems and methods start by obtaining a clean model of the implanted device (e.g., SAV / TAV) from the device manufacturer or by reverse engineering from a micro-CT scan. Then, a few corresponding landmarks between this reference model (e.g., valve model) and the patient-specific implanted model (patient-specific stent) are used as an initial pass of registration. Subsequently, an intensity-based non-rigid B-spline matching may be performed and the clean stent model may be modified to accurately match those in the patient's CT scan. Finally, a transformation may be applied to the valve leaflets and skirt to generate a patient-specific valve model.

[0006] By manually selecting only a few landmarks, a reconstructed patient-specific implant model (eg, a patient-specific valve model) can be generated that is highly accurate, for example, with an error of less than 0.5 mm.

[0007] The model may be used for post-procedure evaluation or in pre-procedure planning (e.g., ViV pre-procedure planning). The model may also be used for post-TAVR assessment and high-fidelity pre-procedure ViV evaluation. It may also be used as ground truth to validate patient-specific valve deployment simulations.

[0008] In one aspect, a model of an in vivo heart valve device implanted in a patient is generated by providing a known model of the pre-implanted heart valve device, performing an imaging scan on the patient having the implanted heart valve device to obtain at least one patient-specific landmark, deforming the known model to fit the at least one patient-specific landmark to obtain a constructed patient-specific valve model, and simulating the leaflets and skirt for the constructed patient-specific valve model by finite element analysis, wherein the constructed patient-specific valve model accurately represents the geometry of the implanted heart valve device in its current in vivo configuration with an error of less than 0.5 mm.

[0009] In one aspect, a model of an in vivo device implanted in a patient is generated by providing a known model of the device pre-implant, performing an imaging scan on the patient having the implanted device to obtain at least one patient-specific landmark, and deforming the known model to fit the at least one patient-specific landmark to obtain a constructed patient-specific model, wherein the constructed patient-specific model accurately represents the geometry of the implanted device in its current in vivo configuration with less than 0.5 mm error.

[0010] In one aspect, a method for generating a patient-specific heart valve device model includes providing a known model of a pre-implanted heart valve device, performing an imaging scan on a patient having an implanted heart valve device to obtain at least one patient-specific landmark, deforming the known model to conform to the at least one patient-specific landmark to obtain a constructed patient-specific valve model, and simulating valve leaflets and a skirt for the constructed patient-specific valve model by finite element analysis, wherein the constructed patient-specific valve model accurately represents a geometry of the patient-specific heart valve device in a current in vivo configuration of the patient-specific heart valve device with an error of less than 0.5 mm. [Brief description of the drawings]

[0011] [Figure 1] 1 is a flowchart of an image matching based method for reconstructing TAV / SAV geometry from patient-specific CT scans. [Figure 2A] Figure 1 shows a clean CAD model of the TAVR obtained by reverse engineering the micro-CT scan. [Figure 2B] FIG. 1 is a cross-sectional view of a voxelized model and selected landmarks. [Figure 2C]FIG. 1 is a cross-sectional view of a voxelized model and selected landmarks. [Figure 2D] FIG. 1 shows a patient-specific valve segmented by thresholding. [Figure 2E] FIG. 2C is a cross-sectional view of a CT scan of a patient with locations and landmarks corresponding to FIG. 2B. [Figure 2F] FIG. 2E is a cross-sectional view of a CT scan of a patient with locations and landmarks corresponding to FIG. 2D. [Figure 3A] FIG. 11 shows a comparison between the stent model and the patient's CT scan during steps of the matching process, showing after an initial rigid registration using the provided landmarks. [Figure 3B] FIG. 11 shows a comparison between the stent model and the patient CT scan during steps of the matching process, showing after landmark-based non-rigid registration. [Figure 3C] FIG. 11 shows a comparison between the stent model and the patient CT scan during steps of the matching process, showing the final intensity-based B-spline deformable matching. [Figure 4A] Cross-sectional view showing a comparison between the collated stent model and a patient CT scan, showing annotated slices of the sample. [Figure 4B] Various cross-sectional views showing a comparison between the verified stent model and the patient's CT scan, showing slices at various locations. [Figure 5A] FIG. 1 shows a side view of a sample patient-specific (self-expanding) valve model obtained using the method of the present invention. [Figure 5B] FIG. 1 shows a top view of a sample patient-specific (self-expanding) valve model obtained using the method of the present invention. [Figure 5C] FIG. 1 shows a bottom view of a sample patient-specific (self-expanding) valve model obtained using the method of the present invention. [Figure 6A]FIG. 1 shows a sample patient-specific (balloon-expandable) valve model obtained using the method of the present invention, showing the original model obtained from a micro-CT scan. [Figure 6B] FIG. 1 shows a sample patient-specific (balloon-expandable) valve model obtained using the method of the present invention, showing a comparison between the matched stent and a model segmented from a patient's CT scan. [Figure 6C] FIG. 1 shows a sample patient-specific (balloon-expandable) valve model obtained using the method of the present invention, showing the final matched valve model. [Figure 6D] FIG. 1 shows a sample patient-specific (balloon-expandable) valve model obtained using the method of the present invention, showing the final matched valve model. [Figure 7A] FIG. 11 shows a known stent (ground truth) after deployment showing the synthetic error assessment. [Figure 7B] FIG. 11 shows a stent after applying a Gaussian blurring algorithm to mimic blooming artifacts, showing the composite error assessment. [Figure 7C] Figure 1 shows a comparison between the matched stent and the ground truth, showing the composite error score. The matched stent is shown in darker grey. [Figure 7D] 13 is a histogram showing the distribution of deviations between nodes (in mm) illustrating the composite error rating. [Figure 8A] Figure 13. Effect of grid size on mean error (10 cases) and computation time (only one Evolution case) showing composite error assessment and sensitivity analysis. [Figure 8B] FIG. 13 shows the effect of the number of selected landmarks on matching accuracy and computation time for one Evolut case, showing a composite error assessment and sensitivity analysis. [Figure 9]FIG. 4 shows the restored leaflet geometry and stress distribution using FEA simulation. Note the difference when compared to FIG. 4B and FIG. 4C. The scale indicates the stress distribution in MPa. [Figure 10A] Figure 1 shows an FEA simulation to estimate the stress distribution in a stent. The scale shows the stress distribution in MPa. [Figure 10B] Figure 1 shows the FEA simulation to estimate the stress distribution in the stent and the bounding box of the crimped stent. The scale shows the stress distribution in MPa. [Figure 10C] FIG. 10D shows an FEA simulation to estimate the stress distribution in the stent, and the estimated stress distribution is compared to FIG. 10C (ground truth). The scale indicates the stress distribution in MPa. [Figure 10D] Figure 1 shows the FEA simulation to estimate the stress distribution in a stent and the ground truth. The scale indicates the stress distribution in MPa. [Figure 11] FIG. 1 illustrates an example computer architecture of a computer system capable of executing software components that can use output of the example methods described herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] A system and a fast and accurate method for reconstructing patient-specific valve geometry from CT scans. The approach exploits known valve topology. Guided by a few manually selected landmarks, a known valve model is deformed to fit patient-specific data in intensity space using an image-matching based framework. Results show that the constructed patient-specific valve model is highly accurate, with a mean error around 0.1 mm, lower than the typical resolution of cardiac CT scans. The shape and stress distribution of the deformed leaflets can be restored in FEA simulations using boundary conditions derived from the deformed stent. It is also possible to reasonably estimate the stress distribution in the stent using a simplified crimping method. The reconstructed model can be used for high-fidelity pre-operative ViV planning or post-TAVR assessment. It can also serve as a ground truth for validating patient-specific valve deployment simulations. The method can also be used to reconstruct the geometry of other stented medical devices from CT.

[0013] A flow chart describing an exemplary method is shown in Figure 1. It consists of three main parts: obtaining a clean valve model and some easily identifiable landmarks, determining the corresponding landmarks in the patient's CT scan, and applying a multi-pass matching algorithm.

[0014] Figure 1 shows a flowchart of the image matching based method for reconstructing implant device geometry (e.g., TAV / SAV device geometry) from patient-specific CT scans.

[0015] In FIG. 1, the exemplary system and method starts by obtaining a clean model of the implanted device (e.g., SAV / TAV) from the device manufacturer or by reverse engineering from a micro-CT scan. Then, a few corresponding landmarks between this reference model (e.g., valve model) and the patient-specific implanted model (e.g., patient-specific stent) are used as an initial pass of registration. Subsequently, an intensity-based non-rigid B-spline matching may be performed and the clean stent model may be altered to accurately match those in the patient's CT scan. Finally, a transformation may be applied to the valve leaflets and skirt to generate a patient-specific valve model.

[0016] Clean valve model and landmarks If a CAD (Computer-Aided Design) model is not available from the manufacturer, a high-quality valve model can be reconstructed by reverse engineering a micro-CT scan of the valve. The typical resolution of such scans is well below 50 μm. Contrast agent can be sprayed on the leaflets and skirt to increase their signal intensity. The valve model segmented from the scan can be easily used in matching and modeling. However, when high mesh quality is required, a clean CAD model is desirable. To obtain such a model, various software can be used to reverse engineer the scanned data into a defect-free CAD model. As an illustration, Figure 2A shows a reconstructed Evolut valve model. The resolution of the scan was 50 μm, and SolidWorks was used to build the model.

[0017] The following steps determine several easily identifiable landmarks on the valve model. For example, the Evolut valve shown in FIG. 2A has two protruding hooks on the top of the stent, which are used in crimping. Guided by these two hooks, all 15 strut tips on the top of the stent, labeled clockwise from 1 to 15, can be quickly identified (dashed lines in FIG. 2A and FIG. 2B). In addition, the 15 strut tips on the bottom can also be easily identified (FIG. 2C). The absolute coordinates of the 30 landmarks were recorded in the coordinate system of the model. As long as the valve topology does not change, this step only needs to be performed once for one type of valve. The landmarks can be curated by the clinician or obtained by artificial intelligence (AI) and / or machine learning approaches. The number and location of the landmarks are not limited to those described.

[0018] Patient valve model and landmarks A DICOM file from the patient's cardiac CT scan may be read, cropped around the valve, and resampled into a 3D array. A simple thresholding algorithm may then be applied to segment the stent. As shown in Figure 2D, a stent model was obtained from the patient's post-TAVR CT scan using 1000 HU (Hounsfield Units). Note that the stent wireframe is about 5-10 times thicker than the clean model shown in Figure 2A. Following the same procedure, all 30 corresponding landmarks in this patient's valve model can be found (Figures 2E and 2F). The landmark coordinates were recorded in the coordinate system of the 3D array. In general, it is not necessary to use all 30 landmarks in most cases. A total of 10 landmarks (five at the top and five at the bottom) works very well, as shown by the following example. The landmarks can be curated by the clinician or obtained by AI and / or machine learning approaches. The number and location of the landmarks are not limited to those described.

[0019] Multi-pass matching framework An exemplary multi-pass matching framework may be based on the open source Insight Toolkit platform and implemented in Python (Python Software Foundation, Wilmington, DE). Before matching, the stent of the clean valve model may be voxelized and converted into a 3D array. The size of this array may match the size of the trimmed array of the patient around the valve.

[0020] In our example, we employed an array size of 133 × 182 × 134 with a voxel spacing of 0.35 mm. Two sample slices of the voxelized array are shown in Figures 2B and 2C. To start, these two models were roughly aligned using only the 10 landmarks mentioned previously (top: 1, 4, 7, 9, 12; bottom: 1, 4, 7, 10, 13 (Figures 2A-F)). The initial alignment consisted of scaling, rigid body rotation, and translation. Figures 3A-C show the comparison between the stent model and the patient CT scan during various steps of the matching process of the method in Figure 1.

[0021] The initial result is shown in Figure 3A, where the clean stent roughly matches the patient model. A second pass of deformable B-spline transformation was then used to improve the initial alignment (Figure 3B). The method works by dividing the 3D domain into a mesh of uniformly spaced control points. By matching corresponding landmarks, a deformation field controlled by B-splines is computed and applied to the entire 3D volume. In this example, a 4 × 4 × 4 mesh grid was used. Compared to Figure 3B, the alignment of the two models is already good, but there are still discrepancies. A final intensity-based B-spline matching step corrected all the misalignments and produced a highly accurate result (Figure 3C). In this step, the 3D domain was divided into a 6 × 6 × 6 mesh grid, whose deformations were controlled by cubic B-splines. Mutual information of the joint histograms was used as the similarity measure, and a line search steepest descent algorithm was employed as the optimizer.

[0022] Other similarity measures (cross-correlation) and optimizers can be used to achieve similar results. The matching method for the initial alignment is not limited to the method described here, and the number of passes and the order of implementation can be modified. The matching method for the final step is not limited to the B-spline matching described here. Other non-rigid matching methods can be used without changing the framework.

[0023] To validate the accuracy of our method, slices from the patient's CT data are overlaid on the matched stent in Figures 4A and 4B. The darker blurred spots are the stents as they appeared in the CT scan, and the bright spots are the clean stents that were voxelized and matched. Figures 4A and 4B show the comparison between the matched stent model and the patient's CT scan at various cross sections.

[0024] The slices shown in Figure 4A show excellent agreement between the two, with the estimation error being less than 20% of the stent width. Further comparisons of various slices are shown in Figure 4B, and visual inspection shows good agreement in all of them.

[0025] Final Valve Assembly The final step is to apply the transformation functions obtained from the stent matching process to the original geometry file of the valve. Here, the STL format was used for the clean valve model. STL files consist of vertices and faces (connectivity between vertices). Transformations were applied to individual vertices and the vertices were mapped to their final location. Since the connectivity of the vertices does not change during the deformation, a new mesh can also be generated using the mapped vertices and their original connectivity map. The transformations can be applied not only to the stent but also to the leaflets and skirt, and the final patient-specific clean valve model was obtained by assembling all the transformed components (Figure 5A-C).

[0026] 5A-5C show a sample patient-specific (self-expanding) valve model obtained using the method of FIG.

[0027] Note that the leaflet area was not preserved in the process. As shown in the example in Figures 5A-5C, the patient's valve appears elliptical with its sides curved. The final step is to transform the model from the image coordinate system to the patient's physical coordinate system using metadata from the patient's DICOM file. This model can then be used to perform post-TAVR evaluation and ViV pre-procedure assessment.

[0028] Furthermore, the orientation of the valve leaflets was also restored during the process, making it possible to assess the risk of coronary artery occlusion in ViV simulations, a task that was previously very challenging.

[0029] The exemplary algorithm is applicable to other types of valves. Figure 6A shows a balloon-expandable Edwards® Sapien 3 valve model generated by reverse engineering a micro-CT scan. The same procedure was applied and the final results are compared in Figure 6B. Using 12 landmarks (6 on the top and 6 on the bottom), the match between the stent model and the patient geometry is perfect (Figure 6B). Since the stent design is axisymmetric, matching the landmarks does not guarantee correct leaflet orientation. The method may involve aligning at least one commissure point of the leaflets in the patient's CT scan with the commissure of the valve model. The final deformed patient-specific valve model after applying the transformation to the stent, leaflets, and skirt geometry is shown in Figures 6C and 6D. The valve is highly deformed and also elliptical in cross section.

[0030] error rating Although visual inspection already showed a good agreement between the matched valve and the ground truth, a quantitative error assessment was performed using the synthetic CT data. The ground truth, i.e. the previously acquired deformed valve, was voxelized and a blurring algorithm was applied to mimic blooming artifacts in the CT scan. An image matching based process was then performed to reconstruct the stent geometry from this synthetic CT data. Finally, the results were compared with the ground truth. An example is shown in Figure 7A-7D, where in Figure 7A we show the deformed stent. The geometry was converted (voxelized) into a 3D image array with a grid spacing of 0.35 mm, the same resolution as the original CT scan. The blurring algorithm was a Gaussian filter with a kernel width of 9 pixels and a standard deviation of 1. A segmented stent from this synthetic CT data is shown in Figure 7B. The width of the stent is measured at 1.5 mm (4-5 pixels). We chose the same 12 landmarks (as in the Multi-path Matching Framework section) and repeated the entire process using a B-spline grid size of 63 to reconstruct the stent geometry. The results are compared in Figure 7C. The reconstructed stent is shaded. The two models look identical. Both models have the same mesh (same vertices and faces) and are in the same coordinate system, so it is possible to quantify the error using the distance between corresponding vertices. The average distance between the nodes of the two models is 0.106 ± 0.091 mm, which is less than one-third of the stent width of 0.45 mm. The distribution of the errors is shown in Figure 7D. 95% of the errors are less than 0.29 mm.

[0031] Sensitivity analysis The effects of the number of landmarks used, the deformation grid size and artificial input bias were evaluated using the same synthetic CT data.

[0032] Grid size: The sensitivity of our method to grid size in the non-rigid matching step is shown in Figure 8A. Grid sizes were varied between zero (rigid) and 23, 43, 63, and 83, with the same 12 landmarks used for all grid sizes. Mean error plots were obtained from 10 different patient-specific valves, including 8 Evolut and 2 Sapien. In the rigid matching only case, the mean error (0.80 ± 0.32 mm) was far greater than the width of the stent. However, it dropped dramatically to less than 0.2 mm with non-rigid B-spline matching using the 43 grid. As the grid size increased from 43 to 83, the mean error continued to drop to around 0.1 mm. However, the computation time (only for the Evolut case shown in Figures 5A-C on an 8-core desktop computer) shot up from less than 10 minutes for the 23 grid to more than 27 minutes for the 83 grid. The results show that the method works well at a reasonable computational cost when used with a moderately fine grid size (e.g., 43). Using a very fine grid does not improve the results significantly, as the resolution of a typical CT scan is usually greater than 0.3 mm.

[0033] Number of landmarks used: For the single Evolut case, the effect of the number of landmarks used in the matching process is shown in Figure 8B. We started with 30 landmarks as a baseline and reduced the number to 16, 12, 8, 6, and 4. All subsequent non-rigid matchings were performed on a 63 grid. The average error only increased slightly when fewer landmarks were used, from 0.102 mm for the baseline case to 0.112 mm when only four landmarks were used (Figure 8B). This shows that the method is not sensitive to the number of landmarks. However, the computation time increased for fewer landmarks, suggesting that the algorithm takes more time to find a perfect match given less initial information. Overall, these results show that the algorithm can produce accurate results at a reasonable speed even with fewer than 10 landmarks.

[0034] Typing bias: Finally, to assess sensitivity to artificial typing bias, we added random numbers ranging from -1.5mm to 1.5mm (a variation of 9 voxels) to the three coordinates of every landmark to represent artificial typing errors. We performed 15 independent trials on the Evolut case with 63 grid sizes using the same 12 landmarks. The average error from all 15 trials was 0.146 ± 0.017mm. This was higher than the cases with correctly selected landmarks, but the matching error was still within 1 / 3 of the stent width.

[0035] Leaflet deformation and stress calculation Although the method of the present invention provides an estimate of the final leaflet geometry (Figures 5A-5C and 6A-6D), the correct deformation and residual stress distribution on the leaflets are important as input for the flow fluid-structure simulation. One way to obtain such information is to perform a finite element analysis with boundary conditions derived from the matching results. Specifically, the displacements of the individual vertices can be calculated between the original stent mesh and the one deformed after matching. With that information, the deformed stent can be restored by applying the displacement conditions to the original stent in the FEA simulation. Since the stress distribution in the stent is not the focus here, any material model in this simulation can be used. By applying tie constraints between the stent mesh and the leaflet mesh, the leaflets will deform together with the stent when the displacement boundary conditions are applied. Thus, the geometry and stress distribution on the leaflets can be correctly calculated. As an illustration, the deformed leaflets for the Evolut valve (same as in Figures 5A-5C) are shown in Figure 9, shaded by the stress distribution (in MPa) on the biological leaflets. The commissure lines between the leaflets are not straight, unlike those shown in Figures 5B and 5C of the same model.

[0036] Stent Stress Estimation This disclosure describes a simple method for reasonably estimating residual stresses in self-expanding stents.

[0037] To estimate the stress distribution in the restored model, we now introduce the crimped box method. A comparison of the crimping process and stress distribution is shown in Figure 10A-D. Specifically, a bounding box of the stent was constructed and meshed with quad shell elements in Hypermesh. This bounding box (Figure 10A) had exactly the same shape as the undeformed stent, so it should match the shape of the deformed stent by applying the same transformation as the stent. Therefore, rather than specifying displacements for the vertices of the stent, we applied displacements to the bounding box and used it to "crimp" the valve model to its final shape (Figure 10B). The "hard" contact model with a friction coefficient of 0.1 was used in Abaqus to define the contact properties between the box and the stent. The stent material was the same hyperelastic Nitinol. A slight over-crimping followed by a release step was applied to account for the nonlinear behavior of Nitinol. In this way we reduced the local errors and avoided spikes in the stress distribution. The restored stent and stress distribution (Figure 10C) are compared to the ground truth in Figure 10D. The restored stent geometry has a difference of around 1 mm and the calculated stress distribution is slightly higher (around 60 MPa or 15% of the peak stress) compared to the ground truth. Overall, this crimping method produces a similar mesh and stress distribution to the ground truth with minimal effort compared to the process described by Gessat et al.

[0038] The exemplary method is not limited to TAVR, but can also be applied to recover the geometry of surgical bioprosthetic valves or any known implant from CT scans. With the database of patient-specific valve models generated by the method, a deep learning model can be trained to recover stent geometries quickly.

[0039] It will be apparent to those skilled in the art that various modifications and variations can be made in the present disclosure without departing from the scope or spirit of the invention. Other aspects of the present disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the methods disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

[0040] It should be appreciated that the logical operations described above may be implemented as (1) a sequence of computer-implemented operations or program modules executed on a computing system, and / or (2) as interconnected machine logic circuits or circuit modules within a computing system. The implementation is a matter of choice dependent on the performance and other requirements of the computing system. Accordingly, the logical operations described herein are referred to variously as state operations, operations, or modules. These operations, operations, and / or modules may be implemented in software, firmware, special purpose digital logic, hardware, and any combination thereof. It should also be appreciated that more or fewer operations may be performed than illustrated in the figures and described herein. These operations may also be performed in different orders than described herein.

[0041] 11 illustrates an example computer architecture of a computer system 200 capable of executing software components that can use the output of the example methods described herein. The computer architecture illustrated in FIG. 11 illustrates an example computer system configuration in which computer 200 may be utilized to execute any aspect of the components and / or modules presented herein that are described as executing on an analysis system or any components in communication therewith.

[0042] In one aspect, computing device 200 may comprise two or more computers in communication with each other that cooperate to perform a task. For example, but not limited to, an application may be divided in a manner that allows for simultaneous and / or parallel processing of instructions of the application. Alternatively, data processed by an application may be divided in a manner that allows for simultaneous and / or parallel processing of different portions of the data set by two or more computers. In one aspect, virtualization software may be employed by computing device 200 to provide the functionality of multiple servers that are not directly bound to the number of computers in computing device 200. For example, the virtualization software may provide 20 virtual servers on four physical computers. In one aspect, the functionality disclosed above may be provided by running an application and / or multiple applications in a cloud computing environment. Cloud computing may include providing computing services over a network connection using dynamically scalable computing resources. Cloud computing may be supported at least in part by virtualization software. Cloud computing environments may be established by enterprises and / or rented as needed from third party providers. Some cloud computing environments may include cloud computing resources that are owned and operated by the enterprise, as well as cloud computing resources rented and / or leased from third-party providers.

[0043] In its most basic configuration, computing device 200 typically includes at least one processing unit 220 and system memory 230. Depending on the exact configuration and type of computing device, system memory 230 may be volatile (such as random-access memory (RAM)), non-volatile (such as read-only memory (ROM), flash memory, etc.), or some combination of the two.

[0044] This most basic configuration is illustrated in FIG. 11 by dashed line 210. Processing unit 220 may be a standard programmable processor that performs arithmetic and logical operations necessary for the operation of computing device 200. Although one processing unit 220 is shown, there may be multiple processors. As used herein, processing unit and processor refer to physical hardware devices that execute coded instructions to perform functions on inputs and create outputs, including, for example, but not limited to, microprocessors (MCUs), microcontrollers, graphical processing units (GPUs), and application specific circuits (ASICs). Thus, although instructions may be described as being executed by one processor, the instructions may be executed simultaneously, sequentially, or alternatively by one or more processors. Computing device 200 may also include a bus or other communication mechanism for communicating information between various components of computing device 200.

[0045] Computing device 200 may have additional features / functionality. For example, computing device 200 may include additional storage, such as removable storage 240 and non-removable storage 250, including, but not limited to, magnetic or optical disks or tape.

[0046] Computing device 200 may also include network connections 280 that enable the device to communicate with other devices, for example, via communication paths described herein. Network connections 280 may take the form of a modem, a modem bank, an Ethernet card, a universal serial bus (USB) interface card, a serial interface, a token ring card, a fiber distributed data interface (FDDI) card, a wireless local area network (WLAN) card, a wireless transceiver card such as code division multiple access (CDMA), global system for mobile communications (GSM), long-term evolution (LTE), worldwide interoperability for microwave access (WiMAX) and / or other air interface protocol wireless transceiver cards, and other well-known network devices. Computing device 200 may have input devices 270, such as a keyboard, keypad, switches, dials, a mouse, a track ball, a touch screen, a voice recognition device, a card reader, a paper tape reader, or other well-known input devices. It may also include output devices 260, such as a printer, a video monitor, a liquid crystal display (LCD), a touch screen display, a display, speakers, etc. Additional devices may be connected to the bus to facilitate communication of data between the components of computing device 200. All of these devices are well-known in the art and need not be described at length here.

[0047] The processing unit 220 may be configured to execute program code encoded in a tangible computer-readable medium. A tangible computer-readable medium refers to any medium capable of providing data that causes the computing device 200 (i.e., a machine) to operate in a specific manner. A variety of computer-readable media may be utilized to provide instructions to the processing unit 220 for execution. Exemplary tangible computer-readable media may include, but are not limited to, volatile, non-volatile, removable, and non-removable media implemented in any method or technology for storage of information, such as computer-readable instructions, data structures, program modules, or other data. System memory 230, removable storage 240, and non-removable storage 250 are all examples of tangible computer storage media. Example tangible computer-readable recording media include, but are not limited to, integrated circuits (e.g., field programmable gate arrays, or application specific ICs), hard disks, optical disks, magneto-optical disks, floppy disks, magnetic tape, holographic storage media, solid state devices, RAM, ROM, electrically erasable program read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices.

[0048] In view of the above, it should be appreciated that many types of physical transformations may occur in computer architecture 200 to store and execute the software components presented herein. It should also be appreciated that computer architecture 200 may include other types of computing devices, including handheld computers, embedded computer systems, personal digital assistants, and other types of computing devices known to those of skill in the art. It is also contemplated that computer architecture 200 may not include all of the components shown in FIG. 11, may include other components not explicitly shown in the figure, or may utilize a different architecture than that shown in FIG. 11.

[0049] In the illustrated embodiment, processing unit 220 may execute program code stored in system memory 230. For example, a bus may carry data to system memory 230, from which processing unit 220 receives and executes instructions. Data received by system memory 230 may optionally be stored on removable storage 240 or non-removable storage 250 before or after execution by processing unit 220.

[0050] It is to be understood that the various techniques described herein may be implemented in connection with hardware or software or, where appropriate, a combination thereof. Thus, the methods and apparatus of the presently disclosed subject matter, or certain aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in a tangible medium, such as a floppy diskette, CD-ROM, hard drive, or any other machine-readable storage medium, which when loaded into and executed by a machine, such as a computing device, causes the machine to become an apparatus for practicing the presently disclosed subject matter. In the case of program code execution on a programmable computer, the computing device generally includes a processor, a processor-readable storage medium (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. One or more programs may perform or utilize the processes described in connection with the presently disclosed subject matter, for example, through application programming interfaces (APIs), reusable controls, or the like. Such programs may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, the programs may be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language, and combined with hardware implementations.

[0051] Further, the various components may be in communication via wireless and / or wired or other desired available communication means, systems, and hardware. Further, the various components and modules may be replaced by other modules or components providing similar functionality.

[0052] The computer architecture 200 includes the necessary software and / or hardware components and modules to enable the functionality of the modeling, simulations and methods disclosed in this disclosure. In some embodiments, the computer architecture 200 may include artificial intelligence (AI) modules or algorithms and / or machine learning (ML) modules or algorithms (e.g., stored in the system memory 230, the removable storage 240, the non-removable storage 250, and / or a cloud database). The AI ​​and / or ML modules / algorithms may enhance the predictive power of the models, simulations, and / or methods disclosed in this disclosure. For example, by using deep learning, AI and / or ML model training, including patient information and any relevant input data to the computational model, the predictive power of the computational model may be significantly improved. The AI ​​and / or MI modules / algorithms also help improve the sensitivity and specificity of predictions as the database grows. In some aspects, the computer architecture 200 may include virtual reality (VR), augmented reality (AR), and / or mixed reality displays, headsets, glasses, or any other suitable display devices as part of the output devices 260 and / or input devices 270. In some aspects, the display devices may be interactive to allow a user to select from options including with or without AR, with or without VR, or fusion with real-time clinical imaging to help the clinician interact and make decisions.

[0053] Although illustrative aspects of the disclosure are described in detail in certain instances herein, it should be understood that other aspects are contemplated. Thus, the disclosure is not intended to be limited in scope to the details of construction and the arrangement of components set forth in the following detailed description or illustrated in the drawings. The disclosure is capable of other aspects and of being practiced or carried out in various ways.

[0054] As used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from "about" or "approximately" one particular value and / or to "about" or "approximately" another particular value. When such a range is expressed, another exemplary embodiment includes from the one particular value and / or to the other particular value.

[0055] "Comprising," "containing," or "including" means that the named compounds, elements, particles, or method steps are present in a composition, article, or method, but do not exclude the presence of other such compounds, materials, particles, or method steps, even if those other compounds, materials, particles, or method steps have the same function as the one named.

[0056] In describing the exemplary embodiments, technical terms are used for clarity. Each term is intended to have the broadest meaning as understood by a person skilled in the art and to include all technical equivalents that operate in a similar manner to achieve a similar purpose. It should also be understood that the reference to one or more steps of a method does not exclude the presence of additional or intervening method steps between those steps that are explicitly identified. The steps of the method may be performed in a different order than described herein without departing from the scope of the present disclosure. Similarly, it should also be understood that the reference to one or more components in a device or system does not exclude the presence of additional or intervening components between those components that are explicitly identified.

[0057] As used herein, a "subject" may be any applicable human, animal, or other organism, living or dead, or other biological or molecular structure or chemical environment, and may relate to a particular component of the subject, such as a particular tissue or fluid of the subject (e.g., human tissue in a particular area of ​​the body of a living subject), which may be a particular location of the subject, referred to herein as an "area of ​​interest" or "region of interest."

[0058] As described herein, it is to be appreciated that the subject may be a human or any animal. It is to be appreciated that the animal may be of any of a wide variety of applicable types, including, but not limited to, mammals, veterinary animals, livestock animals or pet-type animals, etc. By way of illustration, the animal may be a laboratory animal (e.g., rats, dogs, pigs, monkeys) specifically selected to have certain characteristics similar to humans, etc. It is to be appreciated, for example, that the subject may be any applicable human patient.

[0059] The term "about" as used herein means approximately, in the region of, roughly, or around. When the term "about" is used in conjunction with a numerical range, it modifies the range by extending the boundaries above and below the numerical values ​​set forth. In general, the term "about" is used herein to modify a numerical value above and below the stated value with a variance of 10%. In one embodiment, the term "about" means plus or minus 10% of the numerical value of the number with which it is used. Thus, about 50% means within a range of 45% to 55%. Numerical ranges described herein by endpoints include all numbers and decimals subsumed within that range (e.g., 1 to 5 includes 1, 1.5, 2, 2.75, 3, 3.90, 4, 4.24, and 5).

[0060] Similarly, numerical ranges recited herein by endpoints include the subranges subsumed within that range (e.g., 1 to 5 includes 1 to 1.5, 1.5 to 2, 2 to 2.75, 2.75 to 3, 3 to 3.90, 3.90 to 4, 4 to 4.24, 4.24 to 5, 2 to 5, 3 to 5, 1 to 4, and 2 to 4). It is also to be understood that all numbers and decimals thereof are intended to be modified by the term "about."

Claims

1. 1. A model of an in vivo heart valve device implanted in a patient, comprising: Providing a known model of a pre-implant heart valve device; performing an imaging scan on the patient having an implanted heart valve device to obtain at least one patient-specific landmark; deforming the known model to conform to the at least one patient-specific landmark associated with the implanted heart valve device to obtain a constructed patient-specific valve model; It is generated by performing The constructed patient-specific valve model accurately represents the geometry of the implanted heart valve device in its current in vivo configuration with an error of less than 0.5 mm.

2. 2. The model of claim 1, wherein the constructed patient-specific valve model accurately represents the geometry of the implanted heart valve device in its current in vivo configuration with an error of less than 0.1 mm.

3. The model of claim 1 , wherein the known model of the pre-implanted heart valve device is reconstructed from a micro-CT scan of the pre-implanted valve.

4. 2. The model of claim 1, wherein the heart valve device is a transcatheter aortic valve (TAV), a transcatheter mitral valve (TMV), a surgical valve (SAV), a metallic or radiopaque implant.

5. The model of claim 1 , wherein the constructed patient-specific valve model further comprises a model for patient-specific post-operative evaluation and assessment.

6. 10. The model of claim 1, wherein the constructed patient-specific valve model further comprises modeling for high-fidelity pre-operative assessment or pre-interventional procedure planning required for heart valve device replacement or reconstruction.

7. The model of claim 6 , wherein the high fidelity pre-operative assessment includes developing boundary conditions at valve-in-valve (ViV).

8. The model of claim 1 , wherein the imaging scan is a CT scan or an MRI scan.

9. The model of claim 1 , wherein no more than 12 patient-specific landmarks are obtained from the imaging scan.

10. 2. The model of claim 1, wherein deforming the known model to fit the at least one patient-specific landmark to obtain a constructed patient-specific valve model is performed using a multi-path matching framework.

11. A model of an in vivo device implanted in a patient, comprising: Providing a known model of a pre-implant device; performing an imaging scan on the patient having an implanted device to obtain at least one patient-specific landmark associated with the implanted device; deforming the known model to conform to the at least one patient-specific landmark to obtain a constructed patient-specific model; It is generated by performing The constructed patient-specific model accurately represents the geometry of the implanted device in its current in vivo configuration with an error of less than 0.5 mm.

12. 12. The model of claim 11, wherein the constructed patient-specific model accurately represents the geometry of the implanted device in its current in vivo configuration with an error of less than 0.1 mm.

13. 1. A method for generating a patient-specific heart valve device model, comprising: Providing a known model of a pre-implant heart valve device; performing an imaging scan on a patient having an implanted heart valve device to obtain at least one patient-specific landmark associated with said implanted heart valve device; deforming the known model to conform to the at least one patient-specific landmark to obtain a constructed patient-specific valve model; Including, The method of claim 1, wherein the constructed patient-specific valve model accurately represents a geometry of the patient-specific heart valve device in its current in vivo configuration with an error of less than 0.5 mm.

14. 14. The method of claim 13, wherein the constructed patient-specific valve model accurately represents the geometry of the patient-specific heart valve device in its current in vivo configuration with an error of less than 0.1 mm.

15. The method of claim 13 , wherein the known model of the pre-implanted heart valve device is reconstructed from a micro-CT scan of the pre-implanted valve.

16. 14. The method of claim 13, wherein the heart valve device is a transcatheter aortic valve (TAV), a transcatheter mitral valve (TMV), a surgical valve (SAV), a metallic or radiopaque implant.

17. The method of claim 13 , wherein the constructed patient-specific valve model further comprises a model for patient-specific post-operative evaluation and assessment.

18. 14. The method of claim 13, wherein the constructed patient-specific valve model further comprises modeling for high-fidelity pre-operative assessment or pre-interventional procedure planning required for heart valve device replacement or reconstruction.

19. 20. The method of claim 18, wherein the high fidelity pre-operative assessment includes developing valve-in-valve (ViV) boundary conditions.

20. The method of claim 13, wherein the imaging scan is a CT scan or an MRI scan.

Images