Left Atrial Appendage Closure Guidance in Medical Imaging

Abstract

The present invention relates to left atrial appendage closure guidance in medical imaging. An ultrasound imager (20) provides LAA closure guidance. Modeling (32) is performed using ultrasound imaging to allow for time (e.g., throughout the heart cycle). The anatomical model of the LAA over time is used to create a biomechanical model personalized for the patient. The personalized model and models of one or more closure devices are used to select (40) a closure device for the patient that is appropriate for the entire heart cycle and to guide (44) the placement of the selected closure device during implantation.

Description

[0001] This application is a divisional application of the invention patent application with national application number 201811130230.4, the application date of which is September 27, 2018, and the name of the invention is “Left Atrial Appendage Closure Guidance in Medical Imaging”. Background Art

[0002] The present embodiment relates to left atrial appendage (LAA) closure. The LAA is an anatomical structure that may contribute to the formation of life-threatening blood clots. Surgical resection of the LAA structure can treat or prevent the condition. Closure devices can be implanted via a catheter in a manner that is much less invasive to the patient. These closure devices behave like a plug that covers the LAA and prevents the accumulation of blood and the formation of blood clots. Key aspects of the procedure are device selection based on anatomical considerations and planning of the device landing area. Computed tomography (CT) data or three-dimensional transesophageal echocardiography (3D TEE) data can be used to select the appropriate closure device.

[0003] In clinical practice, LAA closure planning and guidance uses manual interaction of physicians with ultrasound and CT datasets. Accurate quantification and characterization of the LAA from medical images has been possible so far on CT or DynaCT images. This limited availability is largely due to the challenging and widely varying shapes of the LAA across patients. Machine learning can combine shape modeling and image processing to provide automatic segmentation with high accuracy from CT and DynaCT data. Once an anatomical model is available, measurements can be automatically derived. However, this approach may not take into account changes due to the heart cycle.

[0004] Most LAA procedures are performed under three-dimensional TEE or volumetric intracardiac echocardiography (volumetric ICE) imaging guidance. ICE imaging is performed by an interventional cardiologist, giving the same person control over both X-ray (e.g., fluoroscopy) and ultrasound imaging, as well as catheter deployment. In the case of ICE imaging in particular, 3D ultrasound imaging is navigated using table-side controls, limiting the complexity of the inputs. As a result, manual measurements can become particularly cumbersome and affect reproducibility. Summary of the invention

[0005] By way of introduction, the preferred embodiments described below include methods, computer readable media, and systems for guiding LAA closure using an ultrasound imager. Ultrasound imaging is used to allow for modeling over time (e.g., throughout the cardiac cycle). The anatomical model of the LAA over time is used to create a biomechanical model personalized for the patient. The personalized model and the model of one or more closure devices are used to select a closure device suitable for the patient throughout the cardiac cycle and to guide the placement of the selected closure device during implantation.

[0006] In a first aspect, a method for guiding left atrial appendage closure using an ultrasound imager is provided. A three-dimensional anatomical model of the patient's left atrial appendage is generated from ultrasound data representing the patient's cardiac volume. An image of the ostium of the left atrial appendage is displayed from the ultrasound data. The position of the ostium is determined by the anatomical model. The placement of a closure device relative to the anatomical model is modeled. The interaction of the closure device with the anatomical model is simulated using a biomechanical model derived from the anatomical model. Three-dimensional blood flow is calculated based on the simulated interaction. A first closure device among the closure devices is selected as a function of the blood flow. An image of a representation of the first closure device overlaid with the modeled placement is output.

[0007] In a second aspect, a method for guiding left atrial appendage closure using an ultrasound imager is provided. Ultrasound is used to perform a three-dimensional scan of a patient's left atrial appendage over at least a cardiac cycle. The geometry of the left atrial appendage over the cardiac cycle is characterized from ultrasound data acquired in the scan. A biomechanical model derived from the geometry is used to model placement of a closure device over the cardiac cycle using the geometric model of the left atrial appendage. Blood flow is calculated over the cardiac cycle based on the modeling. A guidance image shows the placement as optimized based on the modeling and blood flow.

[0008] In a third aspect, a system for left atrial appendage guidance is provided. An ultrasound scanner is configured to acquire ultrasound data representing a three-dimensional region of a patient over time. A fluoroscopic scanner is configured to acquire x-ray data representing the three-dimensional region including a transducer of the ultrasound scanner. An image processor is configured to: (1) model the left atrial appendage over time based on the ultrasound data; (2) align coordinates of the ultrasound scanner with coordinates of the fluoroscopic scanner based on transducer position and orientation as represented in the x-ray data; and (3) generate an image showing placement of a closure device for the left atrial appendage overlaid on the x-ray data based on the model and the aligned coordinates.

[0009] The present invention is defined by the appended claims, and nothing in this section should be taken as limiting those claims.Further aspects and advantages of the present invention are discussed below in conjunction with the preferred embodiments, and may be claimed later, either independently or in combination. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The components and drawings are not necessarily drawn to scale, emphasis instead being placed upon illustrating the principles of the present invention. Moreover, in the drawings, like reference numerals designate corresponding parts throughout the different views.

[0011] Figure 1 is a flow chart of one embodiment of a method for guiding LAA closure using an ultrasound imager;

[0012] Figure 2 Three example images of the displayed ostium of the LAA are shown;

[0013] Figure 3 Four example images tracing the ostium of the LAA are shown;

[0014] Figure 4 Example images for measuring characteristics of the LAA are shown;

[0015] Figure 5 shows example images for multiplanar reconstruction (MPR) from the centerline of the LAA; and

[0016] Figure 6 is a block diagram of one embodiment of a system for LAA closure guidance. DETAILED DESCRIPTION

[0017] LAA closure guidance is provided with an ultrasound-based system. Ultrasound scanning provides effective LAA closure planning and guidance from anatomical quantification, device selection, path planning, and treatment delivery using fast, reproducible measurements and quantification tools. Ultrasound-based systems enable efficient and robust characterization and quantification of LAA anatomy. Ultrasound-based systems automatically provide intuitive rendering of the LAA for planning and guidance, key measurements, and a simulation system for device selection. An integrated multimodal system can combine ultrasound and fluoroscopic imaging.

[0018] Anatomical models enable therapy planning through virtual device placement and deployment. Computational models of tissue biomechanics coupled with organ / device interaction models have the potential to simulate device deployment and analyze the resulting tissue deformations and stresses, providing measures of anchoring stability, side effects, and guidance.

[0019] In one embodiment, a comprehensive LAA guidance system uses ultrasound scanning for closure device selection and guidance. A three-dimensional (3D) data set of the LAA is acquired using 3D TEE or volumetric ICE. The LAA is detected and segmented from the 3D data set, thereby generating a 3D model. Without loss of generality, 3D+time can be processed. Anatomy-specific renderings of the acquired data set are exported and displayed. The export and display can be done offline (inspection mode) or in real time during image acquisition. Key anatomical measurements and their changes over time are calculated based on the 3D model. Virtual placement and visualization of deployed devices are performed based on the 3D model. The optimal device can be suggested based on a database of anatomical structures (shape, size) and / or clinical reports (e.g., organ shape, success rate). The device position can be dynamically updated throughout the frames of the image sequence by taking into account the movement of the LAA wall (e.g., by a motion model based on center of gravity mapping or machine learning). LAA and / or device metrics and implantation roadmaps or paths including optimal transseptal puncture locations can be automatically calculated. Device deployment is simulated using biomechanical modeling of the device and soft tissue to model organ-device interactions and predict therapy outcomes. Mechanical parameters of the anatomical model (e.g., using LA / LAA motion, atrial pressure from Doppler measurements, etc.) are personalized for simulation. 3D blood flow is calculated before and after virtual device deployment to assess the risk of thrombosis and the amount of leakage around the device. For multimodal fusion, anatomical structures, virtually placed devices, implantation paths, and / or roadmaps (e.g., including distance-to-target measurements) can be overlaid on fluoroscopic images. Device appearance and anatomical structures can be enhanced in ultrasound images, such as by using implant locations tracked over time in ultrasound and constrained by X-ray implant detection to optimize beamforming and image reconstruction and reduce noise and artifacts.

[0020] Figure 1 A method for LAA closure guidance using an ultrasound imager is shown. Ultrasound data representing the LAA over time is used to assist in closure device selection and guidance for placement of the selected closure device. The assistance is personalized to the specific patient's LAA. Rather than automating what is done by a physician, the method improves on imaging for LAA closure through a combination of actions implemented by a dedicated image processor.

[0021] Implement in the order shown (e.g., top to bottom or by numerical value) or in a different order Figure 1method. For example, act 40 can be performed based on the geometric metrics calculated in act 30, so act 40 is performed before any of acts 32-38. As another example, act 40 is performed at any time before act 32, where the user manually selects a closed device and uses the selection for modeling. In another example, the implantation plan of act 34 is generated after the simulation of act 36, the selection of act 40, and / or the placement optimization of act 42. These acts can be performed sequentially within minutes or hours of each other. Alternatively, one or more acts are performed days or weeks before other acts.

[0022] Additional, different, or fewer behaviors may be performed. For example, behavior 28 is not performed. As another example, one or more behaviors 38, 40, and / or 42 are not performed. Simulation of behavior 36 may not be performed, such as where geometry is used solely for selection in behavior 40 and / or placement optimization in behavior 42.

[0023] The method is implemented by a medical diagnostic imaging system, a review station, a workstation, a computer, a picture archiving and communication system (PACS) station, a server, a mobile device, on the cloud, a combination thereof, or other devices for image processing of medical data. For example, Figure 6 The system shown in embodies the method, but other systems may be used. A hardware image processor of any type of system interacting with a memory (e.g., a PACS database or cloud storage), a user input device, a display device, a medical imaging system, and / or other components may perform these actions. In one embodiment, an ultrasound scanner performs action 22, and an image processor performs the other actions. The image processor may be part of the ultrasound scanner or separate from the ultrasound scanner.

[0024] These actions may be performed automatically by the image processor. The user causes the patient to be scanned or obtains scan data for the patient from a previous scan. The user may activate the process. Once activated, the other actions are automatically performed without further user input. User input of the location of any anatomical structure in the scan data may be avoided. Some user input may be provided, such as for changing measurement locations, changing placement and / or marking landmarks or other locations.

[0025] In act 22, an ultrasound scanner scans the patient's LAA. 3D TEE, volume ICE, or other scans may be used. The detection is B-mode, Doppler (eg, flow mode), and / or other modes of ultrasound imaging.

[0026] Scanning occurs for one phase of the cardiac cycle. Scanning may be repeated for other phases to acquire a sequence of ultrasound data sets representing all or part of the LAA over a portion of a cardiac cycle, a cardiac cycle, or multiple cardiac cycles. The ultrasound data may be data from a processing path derived prior to generating an image or data as an image for display.

[0027] The scan is three-dimensional, such as using a rocker array, a two-dimensional array, and / or parallel beamforming. Each ultrasound data set represents the LAA in three dimensions. The scan data set may be in a spherical coordinate format or a polar coordinate format for the scan, or may be interpolated to a regular 3D grid.

[0028] The scanned volume includes the LAA. Other parts of the heart or other tissues may or may not be included. The LAA includes tissue, fluid, and / or other anatomical structures.

[0029] Ultrasound data representing the LAA over time allows for automatic quantification of LAA shape for a particular patient. By automatically quantifying LAA shape from volumetric ICE images, a closure device to be used can be selected and guidance for deployment (including placement) can be provided.

[0030] In act 24, the geometry of the LAA is characterized over part, all, or more of a cardiac cycle. Ultrasound data is used to characterize the geometry of the LAA. The geometry is characterized by any measure or parameter of shape and / or size. The geometry may additionally or alternatively be characterized by an anatomical model representing the patient's LAA.

[0031] In one approach, boundary detection is used. In other approaches, an anatomical model (e.g., a template) is fitted to the ultrasound data set. In yet another approach, a machine learning classifier detects the LAA. Any segmentation of the LAA can be used. The segmentation can provide an anatomical model, and other metrics or parameters are calculated based on the segmentation as a representation of the anatomical structure.

[0032] The detection is repeated for different phases or data sets. Alternatively, the fitted anatomical model (e.g., a mesh representing the detected or segmented surface) from one phase is then tracked through the other phases. The geometry is tracked using correlation, speckle tracking, minimum sum of absolute differences, or other measures of fitting the anatomical model from one phase to the ultrasound data of another phase. In an alternative embodiment, the detection is performed for one phase without a sequence of data sets.

[0033] Act 26 represents an example embodiment for characterizing the geometry of the LAA. A 3D anatomical model of the patient's LAA is generated from ultrasound data representing the volume of the heart. The geometry is a mesh, surface, or other representation of the size and shape of the patient's LAA.

[0034] In one embodiment, an anatomical model is formed based on the identified landmarks. The image processor automatically identifies one or more anatomical landmarks (3D points) from which measurements are derived. These landmarks can be the center of the LAA orifice, four landmarks at the borders of the LAA orifice, the tip (s) of the LAA, the LAA valve, the bifurcation (s) and / or any other landmark of the LAA. Landmarks at adjacent structures (e.g., a cross-section of the circumflex artery for indicating the level of the LAA ostium measurement) can be detected alternatively or additionally.

[0035] The landmark is detected using template matching, machine learning classification, or another object detection. Any now known or later developed landmark detection may be used.

[0036] For more accurate detection in noisy data, edge space learning can be applied. A bounding box for LAA is positioned in the data set, and then, landmarks are detected within the bounding box to limit false positives. Bounding box positioning and detection use a machine learning scheme. Any machine learning, such as a probability boosting tree, can be used. In one example, for the bounding box, edge space learning or edge space deep learning (MSDL) is used to sequentially estimate the LAA position, orientation, and scale. The algorithm learns position, orientation, and scale parameters from a large annotation database in an iterative manner by subsequently sampling the parameter space of increasing dimensions. Using deep learning techniques to learn features or filter kernels that distinguish between position, orientation, and / or scale can improve the robustness and accuracy of the scheme. At the end of the process, an oriented bounding box around the LAA is provided.

[0037] Landmarks are detected within a bounding box. For example, a deep learning scheme uses learned features from ultrasound data to create a machine learning classifier to detect one or more landmarks. One or more such classifiers are applied to detect landmarks of interest. In another example, a deep image-to-image network is used to detect landmarks. In yet another example, deep reinforcement learning trains a classifier that is applied to search through or follow learned behaviors to find landmarks. The input is constrained to a bounding box or a region surrounding the box.

[0038] The output is one or more landmarks. When machine learning detection is applied, various possible locations for the landmarks can be identified. A heat map or probability distribution is output. The 3D point with the maximum probability or a 3D point based on a distribution of adjacent higher probabilities can be selected as a landmark. In a semi-automatic solution, a heat map can be displayed with or without an image of the region in the bounding box for user selection of the landmark.

[0039] Once the one or more markers are located, the 3D anatomical model is estimated at least in part based on the markers. Fitting with ultrasound data can be used alternatively or additionally. 3D segmentation utilizes markers. For example, the anatomical model is a template (e.g., an annotated mesh), so the marker positions in the anatomical model are aligned or registered with the marker positions as detected for the patient. Link or align markers over time. In one implementation, the rest of the anatomical model can be interpolated or distorted based on the marker positions and / or fitting with ultrasound data. In another implementation, each point of the 3D mesh is fitted to the data using a constrained fitting statistical shape model to provide a smooth and natural shape. The anatomical model is a statistical shape model based on the distribution of markers across multiple patients. Population analytics is used to fit the anatomical model to the detected markers for the patient. Algorithms for virtual therapy planning can be used to fit models of devices or implants. The dynamics of the anatomical model can be used to account for the spatial displacement of the device or implant through the heartbeat cycle.

[0040] In act 28, the ultrasound scanner and / or image processor uses a display device to display an image of the ostium of the LAA. The anatomical model includes one or more portions of the LAA or the entire LAA. Where the anatomical model includes annotations or where the ostium is positioned as a landmark to fit an anatomical model, the location of the ostium at one or more stages is known. The ostium location is used to generate the image. For example, the ostium is an ellipse or other closed shape in a plane or generally planar region (e.g., within a plate). The plane or plate (plural adjacent planes) that fits or is most closely aligned with the ostium is determined. A long axis view may alternatively or additionally be displayed (see Figure 4 and 5 ).

[0041] Ultrasound data representing a plane or slab for the ostium is used to generate an image. A planar image (e.g., MPR) is reconstructed from the ultrasound data for the plane. Alternatively, a volumetric slab including the plane with a finite depth (e.g., 1 cm) is volume rendered. In yet another embodiment, the volumetric data set is cropped based on the plane, and the remaining data (e.g., the volume outside the plane - the volume extending into the LAA) is volume rendered. The crop plane rendering is displayed.

[0042] The displayed image may include annotations for the heart gates. Figure 2Three images are shown with the ostium outline as annotation. The detected position of the ostium of the LAA is highlighted using graphics or coloring.

[0043] In a semi-automatic or manual approach, the user participates in annotating the displayed image of the ostium or other parts of the LAA. In the case of potential segmentation errors, the user may want to improve the automatic segmentation results. Convenient user interaction is important, especially considering the desk-side operation of the system. A pipeline editing workflow using intelligent mesh manipulation and image intensity-based editing can be used. For example, a circular cursor or a cursor defined for pressing or pulling on the ostium contour is provided. The user positions the cursor relative to the contour to move the contour (for example, to expand or contract the contour and indicate the size of the affected part by its size). Alternatively, control points are added by the user, and the contour or contour line is recalculated by fitting a piecewise parametric curve (e.g., a spline curve) to the control points and ultrasound data. For example, Figure 2 The three images of show the successive user addition of control points from left to right and the corresponding refitted contour lines. In another embodiment, contour corrections are tracked using image-based techniques. By searching for the higher image intensity closest to the mouse pointer on the image, the user typically tracks the blood-tissue interface to improve the segmentation results. Figure 3 The user is shown manually tracing the ostium from left to right. The image processor then modifies the general manual tracing to better fit the image based on the gradient or intensity at the threshold closest to the contour line or mouse pointer. To avoid tracing the image for every phase of the heart cycle, the contour line is found in the image for one phase and then tracked through the data set for other phases. The results of manual or automatic detection of the ostium are propagated to other phases of the heart cycle through tracking.

[0044] In act 30, the image processor calculates a geometric metric of the ostium. The metric can be used to select one or more closure devices for further testing. The metric may be diagnostically useful for planning interventions by a physician.

[0045] Any metric can be calculated. For example, the long axis of the ostium, the short axis of the ostium, the depth of the LAA, the number of flaps of the LAA, and / or the position of the flaps relative to the ostium can be calculated based on the anatomical model and / or the detected ostium. Other metrics showing the shape of the LAA ostium can be used. The metric can be at any ostial location, such as at the level of the circumflex artery.

[0046] In one embodiment, an ellipse is fitted to the LAA ostium. Major axis measurements and minor axis measurements are derived from the fitted ellipse. An irregularity index can be calculated. The irregularity index indicates how far the actual shape of the ostium differs from a perfect ellipse. Other measures of eccentricity, irregularity (e.g., standard or mean deviation from an ellipse), and / or deviation can be used. The area and / or diameter of the ostium contour can be used to measure irregularity or other distortion relative to a circle, ellipse, and / or symmetry. Irregularity can be related to the amount of residual leakage for a circular LAA closure device. The depth of the LAA is calculated, such as the maximum distance between the center of the region of the LAA ostium and the deepest structure of the LAA. Figure 4 This depth measurement is shown on the right. Figure 1 The left side shows the measurement result of the diameter of the ostium. This distance is calculated based on automatic quantification.

[0047] Images from MPR, volume rendering, and / or pertaining to anatomical models can assist in planning and / or guidance. Figure 4 An ultrasound image such as that shown shows a line from the center of the LAA ostium to the deepest structure of the LAA. Additional display capabilities allow the user to visualize the MPR planes rotated 360° around this line. Figure 2 Two ultrasound images are shown for perpendicular planes along the centerline of the LAA (rotated about the centerline shown by the arrow).

[0048] Some LAA structures contain more than one hollow structure. LAA flaps are smaller hollow structures that branch off from the main hollow structure of the LAA. In order to successfully deploy a closure device, the number of flaps is determined. The flaps can be detected and visualized. The closure device is to be positioned so that the closure device covers the ostia of these flaps. Metrics can be calculated for the flaps. The flaps can be detected from the annotations of the fitted model by machine learning classification and / or by template matching. The flaps can be visualized in different colors to make anatomical understanding easier for the user.

[0049] Any of the metrics are calculated for a given phase. The same metrics may be calculated for other phases. The resulting values ​​as a function of time may be displayed to the user in a table or as a graph. The depth axis may be displayed as a moving axis as a function of time in a 3D rendering of the LAA.

[0050] In act 32, the image processor models the placement of one or more closure devices in the LAA. The physician can select the closure device to be tested. Alternatively, the image processor makes the selection based on metrics. The shape and / or size of the LAA and / or the ostium of the LAA are used to select the closure device. More than one closure device can be selected for testing. The placement of the closure device relative to the anatomical model is modeled. The geometry of the closure device is modeled relative to the geometry of the anatomical model.

[0051] The placement is modeled for one phase. Alternatively, the placement is modeled in different (such as all) phases. The placement is modeled within a cardiac cycle. Since the anatomical structure and corresponding anatomical model change within a cardiac cycle, the placement is modeled with respect to different phases. The image processor updates the placement through a sequence or within a cardiac cycle.

[0052] The fit of the closure device to the LAA can be measured for any given placement. Any fit measure can be used, such as the minimum sum of absolute or standard deviations. The fit can be measured for different stages and provided as a graph or average. The image processor calculates the fit of each selected closure device to the anatomical model, such as a geometric fit.

[0053] The placement is modeled using geometry. The modeling is based on the geometry of the closure device and the geometry of the LAA. In an alternative or additional embodiment, the placement is modeled using a biomechanical model for the LAA, allowing for simulation of interactions that distort the geometry of the closure device and / or the anatomical model. Behavior 36 deals with the simulation.

[0054] In one embodiment of geometric modeling, geometric fitting is used to make selections and plan device delivery by optimizing placement. Modeling provides a virtual LAA closure. The geometric model of the closure device is positioned relative to the geometry or anatomical model of the LAA. Based on the anatomical model, a default position is automatically calculated, such as orienting the center of the closure device at the centerline of the LAA through the center of the ostium. Alternatively or additionally, the user can interactively place or adjust the position of the device. The adjustment exists in translation and / or rotation along any axis. The position of the closure device model is then dynamically updated throughout the frames of the image sequence by taking into account the movement of the LAA wall and assuming a constant device-LAA positional relationship (e.g., by center of gravity mapping), or in a more advanced setting by applying the learned movement from a database of patients who have been implanted.

[0055] From the device position in one or more phases, the image processor automatically calculates the fit metric of the LAA to the device. Example fit metrics include oversize or undersize (e.g., mean difference in size). To find the fit for different closure devices, different closure device geometric models are used. These closure device models can be imported models from device manufacturers or created from 3D scans.

[0056] Different positioning of the closure device to the LAA can be tested for geometric fit. The position with the best or optimal geometric fit through the stages can be identified. This optimal position of the virtual device can be used during deployment as an image overlay for guidance on an ultrasound image or fluoroscopic system.

[0057] LAA shape is subject to considerable morphological variation (e.g., "cactus," "chicken wing," "windsock," "cauliflower"), which has potential implications for device selection and placement. The image processor can suggest devices to use based on optimal placement for each device. Alternatively or additionally, morphological similarity can be used to suggest devices based on a database of anatomy (shape, size) and / or clinical reports. Correlations can be used to rank similar patient cases, such as by applying a machine learning classifier. Similar cases can be presented to physicians based on device, LAA geometric similarity, and outcomes. Any outcome information can be used, such as mean residual leak and / or years until subsequent intervention.

[0058] In act 34, the image processor generates an implantation plan. The implantation plan includes placement of the closure device relative to the patient's LAA. Other information can be included in the plan, such as the location of a transseptal puncture for accessing the LAA. Included in the plan is a path along which the catheter travels and the puncture location. A set of different plans can be provided, and one of them can be selected based on the location of the ostium relative to other parts of the heart (e.g., the segmented left and right atria) and / or based on the placement location of the closure device relative to the LAA. Alternatively, the path is calculated based on geometric information. A roadmap for optimal device deployment including an optimal transseptal puncture is generated to guide deployment.

[0059] In act 36, the image processor simulates the interaction of one or more of the closure devices with the anatomical model. The interaction provides more than just geometric fit or geometric information. The interaction uses a biomechanical model derived from the anatomical model. The biomechanical model uses material properties and physics to model the effects of forces and / or distortions caused by the interaction between the closure device and the LAA.

[0060] The simulation was performed in multiple phases. The LAA was affected by cardiac motion forces that varied according to phase. This variation was accounted for in the biomechanical model to determine the interactions and corresponding effects within the various phases of the closure device.

[0061] To personalize the biomechanical model, anatomical models, ultrasound measures of tissue or flow (eg, perfusion and / or elasticity), and / or other information may be incorporated. The biomechanical model may use finite element methods, wavelets, or other representations that relate various properties to each other.

[0062] The simulation can be used to improve or be incorporated into placement determination, generation of implant plans, selection of closure devices, diagnostic information, outcome predictions, and / or other uses. While geometric fit of the deployed device model can provide insights, simulations using the biomechanical model provide predictions of the resulting changes in the LAA and / or left atrial anatomy, along with predictions of potential changes in blood flow patterns. To provide this information, the biomechanical model is a predictive computational model of the anatomical function and interaction of the LAA with the device.

[0063] In one embodiment, ultrasound data from a patient is used to personalize a biomechanical model for that patient. The pathophysiology specific to that patient is captured. The biomechanical model represents the LAA with or without left atrial tissue. The biomechanical model is derived from an anatomical model. The atrial wall and LAA are estimated directly from the ultrasound data (e.g., from an anatomical model) or by fitting a template (e.g., an atlas model) or a statistical shape model via varying thickness and extrusion. The thin volume surface is then meshed with volume elements (e.g., tetrahedrons or hexahedrons).

[0064] The meshing of this anatomical model can be used with a biomechanical solver to compute the displacements of the atrial wall and LAA given internal and external forces. Internal forces account for tissue elasticity (transversely isotropic and nearly incompressible) and active stresses (to capture active contraction of the atrial muscle). External forces account for boundary conditions (e.g., tethers to vessels and left ventricle, pericardial bag, etc.) and hemodynamics (left atrial pressure, wall shear stress, etc.). These forces are treated using Newton's equations for soft tissue mechanics.

[0065] Further personalization is performed for some or all of the force calculations. For example, ultrasound data over time provides tissue motion, such as motion of an anatomical model over time. Based on the observed motion of the left atrium and LAA and given a set of boundary conditions and pressures, the elasticity of the tissue can be estimated globally and regionally, such as from the motion or by applying a machine learning classifier to solve the elasticity as an inverse problem. Alternatively, ultrasound scanning is used to measure elasticity. If atrial pressure is available (e.g., by echo Doppler or by direct catheter measurement), estimates of active and passive stresses become possible. Alternatively, the pressure is derived from the cuff-measured pressure. The availability of magnetic resonance imaging allows regional modification of tissue parameters to capture changes in the tissue matrix, such as fibrosis level. Elasticity scanning can be used to determine the fibrosis level assumed to be associated with stiffness. Alternatively, the fibrosis level is assumed or based on patient averages.

[0066] To simulate the interaction, the closed device is modeled. Computational models of the device mechanics and dynamics are used. For computational mechanics, the finite element method is used to solve for elasticity. More efficient solvers such as the mass-spring method can be used. The parameters of the device (e.g., Young's modulus, shape memory) can be directly obtained from the specifications (e.g., Young's modulus given known materials) based on material composition and dimensions. Alternatively, an indirect evaluation is performed. For example, the dimensions and shape are obtained by scanning the deployed device with a CT imaging scanner. Bench tests are used to estimate the mechanical parameters by measuring the tension given a certain stress and known boundary conditions.

[0067] After fitting or personalizing the biomechanical model of the LAA based on the mechanical parameters and the anatomical model, the image processor simulates the interaction between the biomechanical model and the computational model of the device. The interaction between the LAA and the device is modeled using contact forces. To ensure coupling, the equations of the model are solved simultaneously, such as using an iterative solution. For each time step, the contact is determined and the force is applied to the biomechanical model of the LAA and the computational model of the device at the contact point following the action-reaction principle. The force is modeled as a spring. Other force representations can be used, such as taking into account friction. The solution provides the force and the resulting shape or geometric distortion of the LAA and the closure device.

[0068] In act 38, the image processor calculates three-dimensional blood flow based on the simulated interaction. The blood flow at a given stage can be calculated based on the distorted geometry and any boundary conditions (e.g., pressure and flow). In other embodiments, blood flow is calculated using hemodynamic equations. Changes in position, size, shape, or other changes over time are included in determining the effects on blood flow. The calculation solves for blood flow at different stages (such as all or at least a portion of a heartbeat cycle).

[0069] For hemodynamics, computational fluid dynamics (CFD) is used to estimate the blood flow passing through or through the LAA. The CFD equations are solved based on the Lattice-Boltzmann method (LBM). Other solutions may be used. Given the anatomical structure, dynamics, and biomechanics, the resulting blood flow in the left atrium and / or LAA is calculated. In order to perform patient-specific simulations, the geometry and velocity at each node of the model are passed to the flow solver as boundary conditions. In a bidirectional fluid-structure interaction setting, the resulting fluid forces (wall stresses) are passed back to the biomechanics solver, previously presented as boundary conditions for wall and device mechanics. Flow calculations and interaction simulations are performed simultaneously or as part of an iterative solution for both blood flow and interaction. Alternatively, blood flow calculations are based on already simulated interactions.

[0070] Flow can be calculated before virtual implantation to assess potential low turbulence areas, thrombotic risk, or other information. This calculation does not include interaction with the closure device. Flow can be calculated for a virtual implant afterwards by including a model of the closure device. Flow with respect to the closure device can be used to assess the risk of leakage around the device or other information. The change in flow from before and after placement of the closure device can be calculated.

[0071] In act 40, the image processor selects a closure device. Closure devices of the same type or brand but different sizes may be used. Closure devices of different types and / or brands but the same size may be used. Closure devices of different shapes and / or materials may be used. Several devices are adapted to different anatomical shapes of the LAA (e.g., Watchman from Boston Scientific, Amplatzer / Amulet from St. Jude Medical). The selection is one of these closure devices compared to the other closure devices.

[0072] Geometric fit, leakage around the device, or other information is used to select one closure device compared to other closure devices. The selection can be prior to modeling or simulation, or can be derived from the results of modeling or simulation. The selection can be based on the results of testing one device (e.g., comparing the value of a metric to a threshold). Alternatively, the selection is based on a comparison of modeling or simulation of multiple devices. For example, blood flow (e.g., leakage) is calculated for each of multiple different closure devices. The closure device with the least leakage (e.g., due to size and / or shape) is selected.

[0073] The selection is based on information for a given phase. Since ultrasound scanning provides data sets for different phases, the selection of a closure device can be based on information within these phases. The average or maximum leakage or other blood flow characteristics can be used.

[0074] Instead of or in addition to blood flow, other criteria besides blood flow may be used. The type, material, shape, and / or size of the closure device may be selected based on elasticity in a biomechanical model, the location of the closure device with the resulting leakage, the wall displacements in an anatomical model through the cardiac cycle, and / or the stress at the anchoring location of the closure device.

[0075] In one embodiment, the resulting deformation due to the interaction and the elasticity of the wall are used along with the leakage. The goal is to provide very little leakage at minimal deformation given the elasticity used to select the device and size. The apparent orifice diameter may not be sufficient to ensure proper anchoring. Modeling can also provide other information (such as the displacement of the wall during a complete heartbeat cycle and the interaction with adjacent structures and / or stresses at the anchoring location) to assess the potential risk of device embolism.

[0076] Rather than the image processor making the selection, information may be provided to the user. Leakage, interaction, and / or other information for different closure devices (e.g., different sizes, shapes, and / or configurations) may be presented. Images of interaction and / or blood flow over time, leakage over time, and / or other information may assist the user in selecting a closure device.

[0077] In act 42, the image processor optimizes placement of the closure device. The placement is performed automatically, but semi-automatic placement based on user adjustments may be used.

[0078] Optimal placement is determined based on geometry, metrics from geometry, modeling of geometric fit, and / or simulated interactions. Placement can take into account tissue properties for anchoring and / or the amount of leakage over time. Differences in forces resulting from different placements and / or the amount of tissue or device distortion can be used to optimize placement.

[0079] One criterion or multiple criteria are used to optimize placement. Different placements can be used to minimize leakage, maximize geometric fit, or optimize placement in other ways. In the case of multiple considerations, a weighted combination can be used to determine the optimal placement.

[0080] In act 44, the image processor outputs the image. The image is used as part of modeling, simulation, characterizing geometry, calculating geometric metrics, or other actions. Alternatively or additionally, the image is provided to guide intervention to implant a device in a patient.

[0081] The image is an ultrasound image. Other types of imaging may be used, such as fluoroscopic images. Images of only planes (eg, MPR) or volume renderings of a volume, cropped volumes, or thin slabs (eg, 10 planes or less) may be generated.

[0082] The image is overlaid with a representation of a closure device, such as a selected closure device in an optimized position. The image can be used to guide placement of the actual closure device during implantation. For example, a fluoroscopic or ultrasound image includes a graphic representing the desired placement of the closure device relative to the anatomy. As the image is updated due to scanning, the graphic is maintained relative to the anatomy to indicate the location of deployment. The path and / or puncture site may also be represented. Other information, such as the distance of the catheter tip or closure device from the puncture site and / or deployment location, may be provided.

[0083] Multimodal guidance can be provided. Multimodal imaging guidance is desirable for enhanced workflow and outcomes. In the case of LAA closure procedures, fluoroscopy is used to guide delivery catheter and device deployment, while 3D TEE or volumetric ICE is used to guide transseptal puncture and positioning of the device or delivery catheter prior to device deployment in the LAA. Fluoroscopy is a 2D projection that provides good visualization of the delivery catheter and device; however, it does not provide information about the 3D shape and hemodynamics of the LAA. This information can be obtained using 3D TEE or volumetric ICE.

[0084] A 3D anatomical model of the LAA or a live 3D image of the LAA can be registered with fluoroscopy and overlaid on the fluoroscopy 2D projection image. The user can determine and mark the optimal location for the puncture site on the septum for the delivery catheter. The location is overlaid as a graphic or highlight (e.g., adding color) on the imaging. A transseptal puncture is then performed using fluoroscopy, 3D TEE or volumetric ICE guidance, or a combination of fluoroscopy and 3D TEE or volumetric ICE. After performing the transseptal puncture, the delivery catheter is advanced into the left atrium, pointing toward the LAA. Fluoroscopy with or without ultrasound imaging is used to guide the positioning of the delivery catheter.

[0085] In the next step, the closure device is inserted and advanced to the tip of the delivery catheter. By displaying a perspective image (live or previously acquired) registered with the live 3D TEE or volume ICE image, the actual closure device is positioned at the optimal position in the LAA. This registration allows the calculated optimal position to be displayed (e.g., displayed as a mark or drawing of the device) on either type or both types of imaging. Virtual placement is shown on one or more images to guide the placement of the actual closure device. Once the actual device is close enough to the optimal position, the system shows a corresponding message (e.g., a green flag or distance measurement). Perspective or a combination of perspective and 3D TEE or volume ICE can then be used to fully deploy the actual device.

[0086] After deployment, a drag test is performed to ensure that the closure device is adequately and correctly anchored in the tissue. After a successful drag test, the closure device is released (e.g., detached from the delivery catheter). After release, it is again possible to visualize the actual closure device position together with the calculated optimal position in a 3D rendering view (3D TEE or volumetric ICE).

[0087] The coordinate system of the ultrasound scanner or transducer is spatially registered with the coordinate system of the fluoroscopic scanner. The spatial transformation resulting from the registration causes the position in one coordinate system to be known in the other coordinate system, thereby allowing the anatomical structure and / or device position in one coordinate system to be transferred to the other coordinate system. Any registration can be used. In one embodiment, the position and orientation of the transducer (e.g., TEE probe or ICE catheter) is detected in the fluoroscopic image. Machine learning classification, template matching, landmark detection, fiducial position or other detection can be used. The ultrasound coordinates are based on the transducer, therefore, the position and orientation of the transducer in the fluoroscopic image provides a spatial transformation related to the coordinate system. Echo-fluoroscopic registration provides a transformation matrix that places all two imaging modalities in the same coordinate system.

[0088] Use ultrasound to find placement, implantation paths, and anatomical models. Placement, implantation paths, and anatomical models can be added, overlaid on locations on ultrasound and fluoroscopic images, or used to highlight or mark locations on ultrasound and fluoroscopic images. Similarly, the information provided by the fluoroscopic images can be used to enhance ultrasound images. For example, fluoroscopy can be used to track devices detected in fluoroscopic images over time. Such devices include catheters, implanted objects, or other X-ray opaque structures. The detection uses any of the detections discussed above for catheters. Covers or other markings of devices detected in fluoroscopy can be added to ultrasound images.

[0089] The same object may be detected in both imaging modes. The detection in both imaging modes may be used to refine the position of the object, such as by averaging the positions.

[0090] Co-registration and ultrasound enhancement can be further exploited during ultrasound image formation. For example, the position and orientation of the transducer is detected in fluoroscopy. This position and orientation relative to the LAA can be used to control beamforming (e.g., scanning from a more normal direction relative to the LAA or to avoid shadows from acoustically opaque tissue or bone), for noise or artifact reduction (e.g., motion correction for transducer motion), and / or for image reconstruction (e.g., more or less temporal filtering depending on the amount of movement).

[0091] The use of a method for planning LAA closure and / or for guiding LAA closure is described above. The method or portions of the method can be used for closure device design. During development, different shapes, sizes, materials, structures, or other design aspects are varied and tested through geometric fitting, placement determination, and / or simulated interaction to confirm design, durability, and / or ease of use.

[0092] Figure 6 A system 10 for guiding LAA closure is shown. The system implements Figure 1 or another method. To plan the guidance, the system personalizes a model (e.g., an anatomical model and a biomechanical model) for the patient and determines the closure device to be used and the optimal placement of the closure device. During implantation of the selected closure device, the system guides the placement of the closure device in the patient.

[0093] The system includes an image processor 12, a memory 14, a display 16, an ultrasound scanner 20, and a fluoroscopic scanner 22. The image processor 12, the memory 14, and the display 16 are shown as being separate from the imaging modalities 20, 22 (such as as part of a workstation, computer, or server). In alternative embodiments, the image processor 12, the memory 14, and / or the display 16 are part of one or more of the imaging modalities 20, 22. In yet other embodiments, the system 10 does not include the imaging modalities 20, 22. Additional, different, or fewer components may be used.

[0094] The ultrasound scanner 20 is a medical diagnostic ultrasound scanner. The ultrasound scanner 20 includes a transducer (e.g., a rocker arm or a two-dimensional array) and a parallel receive beamformer for real-time 3D scanning. The ultrasound scanner 20 includes a detector or other image processor to generate an image from the acoustic echoes received by the transducer. By repeating the scan, a sequence of ultrasound data sets is provided. The sequence represents different stages through one or more heartbeat cycles.

[0095] The fluoroscopic scanner 22 is an X-ray system, such as a C-arm X-ray system, an angiography system, or a dedicated fluoroscopic system. An X-ray source is mounted opposite a detector with an intervening patient space. X-rays from the source pass through the patient and impinge on the detector, thereby providing a projected fluoroscopic image. The fluoroscopic image is acquired at any frequency and represents a 3D region of the patient projected onto a plane, which includes the ultrasound transducer.

[0096] The memory 14 is a buffer, cache, RAM, removable media, hard drive, magnetic, optical, database, or other memory now known or later developed. The memory 14 is a single device or a group of two or more devices. The memory 14 is shown as being associated with or part of the image processor 12, but can be external to the system 10 or remote from other components of the system 10. For example, the memory 14 is a PACS database storing scan data from the modalities 20, 22.

[0097] The memory 14 stores scan data, models, metric values ​​and / or information used in image processing, simulation interaction, determining geometric fit, fitting anatomical models, modeling closed devices or other behaviors. Additionally or alternatively, the memory 14 is a non-temporary computer-readable storage medium with processing instructions. The memory 14 stores data representing instructions that can be executed by the programmed image processor 12. Instructions for implementing the processes, methods and / or techniques discussed herein are provided on a computer-readable storage medium or memory (such as a cache, buffer, RAM, removable media, hard drive or other computer-readable storage medium). Computer-readable storage media include various types of volatile and non-volatile storage media. The functions, behaviors or tasks illustrated in the drawings or described herein are performed in response to one or more instruction sets stored in or on a computer-readable storage medium. The function, behavior or task is independent of a specific type of instruction set, storage medium, processor or processing strategy, and can be performed by software, hardware, integrated circuits, firmware, microcode, etc. operating alone or in combination. Similarly, processing strategies can include multi-processing, multi-tasking, parallel processing, etc. In one embodiment, the instructions are stored on a removable media device for reading by a local or remote system. In other embodiments, the instructions are stored in a remote location for transmission over a computer network or on a telephone line. In yet other embodiments, the instructions are stored within a given computer, CPU, GPU, or system.

[0098] The image processor 12 is a general processor, a digital signal processor, a three-dimensional data processor, a graphics processing unit, an application-specific integrated circuit, a field programmable gate array, a digital circuit, an analog circuit, a quantum computer, a combination thereof, or other devices now known or later developed for LAA closure guidance. The image processor 12 is a single device, a plurality of devices, or a network. For more than one device, parallel or sequential division of processing can be used. The different devices that make up the image processor 12 can perform different functions, such as an automated anatomical detector and a separate device for modeling geometric fitting and simulating interaction. In one embodiment, the image processor 12 is a control processor or other processor of a medical diagnostic imaging system (such as an ultrasound scanner 20 or a fluoroscopic scanner 22). The image processor 12 is a hardware device that is configured by stored instructions, designs (e.g., application-specific integrated circuits) or hardware or operates in accordance with stored instructions, designs or hardware to perform various behaviors described herein.

[0099] The image processor 12 is configured to model the LAA over time based on the ultrasound data, register the coordinates of the ultrasound scanner with the coordinates of the fluoroscopic scanner based on the transducer position and orientation as represented in the X-ray data, and generate an image showing the placement of a closure device for the LAA based on the model and the registered coordinates. The image processor 12 can generate images for placement based on the following operations: performing characterization of the LAA geometry, ostium, or other geometric metrics; modeling device placement; determining an implantation plan; simulating the interaction of the device with the LAA using computational and / or biomechanical modeling; calculating blood flow based on hemodynamics; selecting a closure device; and / or optimizing placement.

[0100] The display 16 is a CRT, LCD, plasma, projector, printer, or other output device for showing images or other outputs of the image processor 12 and / or the modalities 20, 22. The display 16 displays images of detected anatomy, geometric measurements, simulated interactions, closure device placement, and / or other guidance information.

[0101] Although the present invention has been described above by reference to various embodiments, it should be understood that many changes and modifications can be made without departing from the scope of the present invention. It is therefore intended that the above detailed description should be regarded as illustrative and not limiting, and it should be understood that it is intended to limit the spirit and scope of the present invention to the appended claims including all equivalents.

Claims

1. A system for guiding left atrial appendage closure using an ultrasound imager, the system comprising: at least one image processor; A memory storing instructions, the image processor executing the instructions to: Performing a three-dimensional scan of the patient's left atrial appendage using ultrasound during at least one heartbeat cycle; characterizing the geometry of the left atrial appendage during the cardiac cycle based on ultrasound data acquired from the scan, including: generating a three-dimensional anatomical model of the patient's left atrial appendage as a geometric shape, the three-dimensional anatomical model generated based on ultrasound data representing a volume of the patient's heart, generating the three-dimensional anatomical model of the left atrial appendage from the ultrasound data, the three-dimensional anatomical model generated based on landmark detection, including detecting a left atrial appendage center, an ostium boundary, a left atrial appendage tip, a left atrial appendage valve, and a bifurcation; modeling placement of a closure device using the geometry of the left atrial appendage during the cardiac cycle using a biomechanical model derived from the geometry; Calculating blood flow during the heartbeat cycle based on the modeling; Generating a guidance image to display the optimized placement based on the modeling and the blood flow, wherein generating the guidance image includes outputting a guidance image displaying the patient's anatomy, the guidance image including a first image of the ostium of the left atrial appendage with a representation of the closure device at the modeled placement superimposed on the anatomy.

2. The system of claim 1, wherein: The characterization includes fitting an anatomical model at one phase of the cardiac cycle, the geometric shape including the fitted anatomical model at the one phase; and tracking the geometric shape at other phases of the cardiac cycle.

3. The system of claim 1, the image processor further executing instructions to optimize the placement based on metrics calculated from the modeling and geometry.

4. The system of claim 2, wherein the image processor further executes instructions to simulate interaction of the closure device with the anatomical model using the biomechanical model.

5. The system of claim 4, wherein: The calculation includes: calculating three-dimensional blood flow according to the simulated interaction.

6. The system of claim 5, wherein: The calculating includes performing calculations using the geometry of the left atrial appendage and the biomechanical model during at least a portion of the heart cycle.

7. The system of claim 6, wherein: The calculating includes calculating the risk of thrombus in the left atrial appendage based on blood flow in the absence of the closure device and calculating leakage based on the simulated interaction.

8. The system of claim 1, wherein the image processor further executes instructions to display an image of the ostium of the left atrial appendage based on the ultrasound data, wherein the position of the ostium is determined based on the geometric shape.

9. The system of claim 4, the image processor further executing instructions to select an occlusive device from a plurality of occlusive devices as a function of the blood flow, and to output an image of the modeled placement with a representation of the selected occlusive device superimposed thereon.

10. The system of claim 9, wherein: Selecting a closure device from a plurality of closure devices includes selecting a type and / or size based on elasticity in the biomechanical model, position of the closure device with resulting leakage, wall displacements in the anatomical model through a cardiac cycle, and stresses at an anchoring location of the closure device.

Images