Specification of a 2D-ICE-based left atrial appendage closure device for the treatment of patients with atrial fibrillation.

Abstract

This invention provides a system and method for determining the specifications of a left atrial appendage closure device based on two-dimensional (2D) intracardiac echocardiography (ICE). [Solution] In imaging procedures, 2D-ICE is used to acquire images of the patient's left atrium. A model of the left atrium is generated from the images of the left atrium. The model of the left atrium is used to validate subsequent imaging of the patient's left atrial appendage. Anatomical measurements from the validated subsequent imaging are used to select or identify a left atrial appendage closure device, for example, for the treatment of atrial fibrillation patients.

Description

[Technical Field]

[0001] This disclosure relates to medical imaging. [Background technology]

[0002] The left atrial appendage (LAA) is a small sac extending from the side of the left atrium in the heart, which functions as a decompression chamber when atrial pressure is high. The LAA plays a significant role in the risk of thromboembolism associated with atrial fibrillation and may also induce atrial tachyarrhythmias. Atrial fibrillation (A-fib) is a type of abnormal cardiac rhythm in which the upper ventricles of the heart beat irregularly and rapidly, increasing the risk of blood clot formation within the heart. If a blood clot in the left upper ventricle (left atrium) detaches from the cardiac region, it can travel to the brain and cause a stroke, a leading cause of death. Studies have shown that the LAA is a particularly common site for blood clot formation in people with A-fib.

[0003] Understanding the morphology and function of the laryngeal aorta (LAA) is crucial for the treatment of arteriovenous fibrillation (A-fib). LAA orifice size has generally been measured using cardiac CT or transesophageal echocardiography (TEE). Both CT and TEE have drawbacks. CT is a pre-procedure imaging procedure using ionizing radiation. TEE requires general anesthesia and associated risks, and an anesthesiologist must be present during anesthesia administration. Furthermore, TEE image quality often depends on the operator. [Overview of the Initiative]

[0004] As a preliminary note, the preferred embodiments described below include methods, systems, instructions, and computer-readable media relating to LAA closure device specification determination based on 2D-ICE.

[0005] In a first embodiment, a method for evaluating the left atrial appendage (LAA) is provided. The method includes: acquiring multiple images of the patient's left atrium; generating a model of the left atrium from the multiple images of the left atrium; acquiring multiple images of the patient's LAA; determining, based on the model, whether the multiple images of the LAA are sufficient to accurately estimate the specifications of the LAA closure device; if the multiple images of the LAA are sufficient, calculating and providing anatomical measurements of the LAA; or, if the multiple images of the LAA are insufficient, providing instructions to the operator to acquire additional images of the LAA.

[0006] In a second embodiment, a system for left atrial appendage (LAA) device selection is provided. The system includes a 2D-ICE imaging system configured to acquire left atrial imaging data and left atrial appendage imaging data of the patient's heart, and an image processing processor configured to generate a three-dimensional model of the patient's left atrium from the left atrial imaging data, wherein the image processing processor is further configured to register the left atrial appendage imaging data against the three-dimensional model and to determine that the left atrial appendage is sufficiently visualized in the left atrial appendage imaging data, and the image processing processor is further configured to calculate at least one anatomical measurement of the left atrial appendage and to select a device for left atrial appendage closure based on this at least one anatomical measurement.

[0007] In a third embodiment, a non-temporary computer-executable storage medium is provided that stores machine-readable instructions executable by at least one processor for left atrial appendage (LAA) evaluation. The machine-readable instructions include taking multiple images of the patient's left atrium, generating a model of the left atrium from the multiple images of the left atrium, taking multiple images of the patient's LAA, determining, based on the model, whether the multiple images of the LAA are sufficient to accurately estimate the specifications of an LAA closure device, calculating and providing anatomical measurements of the LAA if the multiple images of the LAA are sufficient, or providing instructions to the operator to take additional images of the LAA if the multiple images are insufficient.

[0008] Any one or more of the above embodiments may be used individually or in combination. These and other embodiments, features, and advantages will become apparent from the following detailed description of preferred embodiments, which should be read in conjunction with the drawings. The present invention is defined by its claims, and nothing in this section should be construed as a limitation to any of those claims. Further embodiments and advantages of the present invention are described below in connection with preferred embodiments and may be claimed individually or in combination thereafter. [Brief explanation of the drawing]

[0009] The components and drawings are not necessarily to scale; rather, the emphasis is on illustrating the principles of the embodiments. Furthermore, in the drawings, the same reference numerals indicate corresponding parts throughout the drawings. [Figure 1] Here is an example of the elements that make up the heart. [Figure 2] An example of ICE images of the patient's LA and LAA is shown. [Figure 3] Examples of different orientations and scanning planes for ICE catheters are shown. [Figure 4] An example of a system for determining LAA closure device specifications based on 2D-ICE according to one embodiment is shown. [Figure 5] An example of a workflow for determining LAA closure device specifications based on 2D-ICE according to one embodiment is shown. [Figure 6] An embodiment of the 2D-ICE scanning plane and a 3D model of the heart are shown. [Figure 7] One embodiment of a method for determining LAA closure device specifications based on 2D-ICE is shown. [Figure 8] An example of an artificial neural network is shown. [Figure 9] An example of a convolutional neural network is shown. [Modes for carrying out the invention]

[0010] The embodiments disclosed herein provide a system and method for LAA closure device sizing based on two-dimensional (2D) intracardiac echocardiography (ICE). In an imaging procedure, 2D-ICE is used to acquire an image of a patient's left atrium. A model of the left atrium is generated from the image. The model is used to validate subsequent imaging of the patient's LAA. Anatomical measurements from the validated subsequent imaging can be used, for example, to select or specify an LAA closure device for treating a patient with atrial fibrillation.

[0011] 2D-ICE is an imaging modality that provides high-resolution real-time visualization of cardiac structures, continuous monitoring of catheter position within the heart, and early recognition of intraoperative complications such as pericardial fluid retention or thrombus formation. ICE has widely replaced transesophageal echocardiogram as an imaging modality for guiding certain procedures such as atrial septal defect closure and catheter ablation of cardiac arrhythmias, and also has a new role in other procedures including mitral valve formation, transcatheter aortic valve replacement, and left atrial appendage closure. The ICE catheter can be enhanced by a tracking device such as a radio frequency transmitter that enables determination of the catheter's position when an image is acquired.

[0012] FIG. 1 shows an example of each element constituting a patient's heart. FIG. 1 shows an example of the left atrium 20 and surrounding structures of interest. The structures include the left atrial appendage (LAA) 26, left inferior pulmonary vein (LIPV) 25, left superior pulmonary vein (LSPV) 28, right inferior pulmonary vein (RIPV) 22, and right superior pulmonary vein (RSPV) 24. When performing an ultrasonic procedure, the left ventricle or other cardiac chamber can be the anatomical structure of interest. The embodiments described herein focus particularly on the LAA, but the methods and systems can be applied to other organs, tissues, or structures that can benefit from real-time segmentation and quantification.

[0013] As shown in FIG. 1, the LAA 26 is located on the side of the pulmonary vein that carries blood from the lungs. The position of the LAA 26 can cause blood to pool there without flowing into the left ventricle of the patient's heart. In patients with a normal heart rhythm, the LAA 26 contracts rhythmically together with the rest of the left atrium. This rhythmic contraction ensures that the blood in the sac is discharged when blood is pumped out of the left atrium into the left ventricle and then sent throughout the body. However, patients with A-fib experience abnormal atrial contractions during an episode. The weak contractions can lead to thrombosis associated with blood pooling, and the thrombus can form within the LAA 26. Anticoagulant therapy is mainly used as the main embolism prevention treatment for patients with atrial fibrillation, but this has a low long-term compliance and potential bleeding complications. Left atrial appendage closure (LAAC) is a minimally invasive cardiac intervention to prevent thrombus formation within the LAA 26 by closing the LAA with a device. This can significantly reduce the risk of stroke in people with A-fib at high risk of thrombus occurrence. This technique functions as an alternative solution to anticoagulant therapy and is recommended for patients with A-fib who cannot take blood thinners.

[0014] In left atrial appendage closure (LAAC) procedures, ICE is increasingly being used to provide guidance and navigation assistance. For example, as an example, the imaging most frequently used for the Watchman FLX implant and other LAA 26 implants is ICE. Using ICE for LAAC has advantages over the TEE-guided approach. This procedure can be performed under local anesthesia with the patient awake and responsive. The risks and postoperative discomfort associated with general anesthesia or deeper sedation and endotracheal intubation can be avoided, and the procedure can be performed in patients with general anesthesia contraindications and patients with gastroesophageal diseases. Anesthesia teams and TEE operators are not required, and the procedure turnover time is reduced.

[0015] However, there are several challenges to using 2D-ICE in procedures such as LAAC. One of the key measurements used when selecting a particular LAAC device is the diameter of the LAA opening. The size of the opening (ostium) varies considerably from patient to patient. Measuring the size is an important task when performing LAAC, for example, to determine which of several variations of LAAC devices to use for a particular patient. This can be challenging because 2D-ICE provides a 2D view of the 3D anatomical structure of the LAA26, and many clinical facilities lack experience working with ICE and obtaining the best images. The LAA opening may be visible in a particular ICE image frame, but this may not represent the full range of the LAA opening diameter. It may not be clear from a single image frame or a series of image frames that the entire diameter is not fully visible.

[0016] Figure 2 shows an example of a 2D-ICE image of LAA26. This 2D image includes both the patient's LA20 and LAA26. While the location of LAA26 can be easily identified, it may not be clear whether the 2D image includes the entire LAA26 or only a portion of it. Figure 3 illustrates how different scanning planes can result in different views and renderings of LAA26, particularly the size of the rendered LAA ostia 205. Figure 3 shows two different scenarios in which an ICE catheter 210 is used to image the LAA ostia 205, which is represented as an ellipse. The ICE image frame is shown in a wedge shape representing an ultrasonic fan of image information. The measured LAA ostia size 215 is rendered on the ICE image frame. The size in Figure 3A underestimates the maximum size of the LAA ostia 205. The bore in Figure 3B is an accurate estimate of the maximum bore size (LAA Ostia Size) 215 of the LAA opening 205. As shown in the figure, if the catheter is not positioned correctly by the operator, the resulting measurement may be significantly off. Due to the limitations of two-dimensional (2D) ICE and the difficulty of manually manipulating the ICE transducer, the LAA opening 205 and other three-dimensional (3D) anatomical structures may not be fully observable in certain views. This presents difficulties for electrophysiologists and echocardiographic image analysis algorithms.

[0017] In one embodiment, a mechanism is provided for verifying or confirming whether the acquired images are sufficient for selecting an LAAC device. In one embodiment, an estimated model of the LA anatomical structure, including LAA26, and a known position of the 2D-ICE frame relative to the model are used to confirm that the entire diameter of the LAA opening 205 is seen. Uncertainty in the model and the ICE image frame position can be used to estimate the probability that the left atrial appendage diameter was measured with a certain degree of confidence. If the diameter of the LAA opening 205 is not seen reliably, the system may, based on the model, suggest in which direction the ICE catheter 210 should be moved to obtain a better estimate of the diameter of the LAA opening 205. For example, shifting the catheter forward or rotating the catheter ("clocking"). Simulating alternative views from the LA model that the ICE catheter 210 can reach may enable simulated calculations of the diameter of the LAA opening 205. Calculations of all such different views are possible, and the best view may be reported to the user in the form of guidance for moving the ICE catheter 210 to obtain the best view.

[0018] Whether an image is sufficient for selecting the appropriate LAAC device size depends on the specific occlusion device used and the manufacturer's recommendations regarding its placement. Device manufacturers may identify several measurements that need to be performed when selecting which occlusion device to use for a patient. For example, a device manufacturer might require that the LAAC orifice diameter be measured with an accuracy of 2 mm. An image showing the LAA diameter within 2 mm of the actual LAA diameter would be sufficient. Because there may be some uncertainty in models recovered from ICE images, the diameter measured from the model may only be known with 90% accuracy. In one example, assuming the diameter seen in the image is within 2 mm of the maximum diameter measured on the model, this system would output that the current image is sufficient with 90% probability.

[0019] Figure 4 shows an ultrasound system 100 for specifying (selecting) an LAAC device using 2D-ICE. The system includes an Image Processing System 100, a Medical Imaging Device 130, and optionally a Server 140. The Server 140 may be configured to perform any task of the Image Processing System 100, including processing and / or storing data and models. The Server 140 may consist of or include a cloud-based platform. The Image Processing System 100 includes a processor 110 (image processing processor 110), memory 120, and a display 115. The Image Processing System 100 may be included in or connected to the Medical Imaging Device 130. The Image Processing System 100 is configured to generate a model of the LA using the acquired 2D-ICE images of the LA. The Image Processing System 100 is configured to use the model to verify that the 2D-ICE images of the LAC are sufficient to identify an LAAC device for the patient. For example, LAC images not acquired using the appropriate orientation / scanning plane may be rejected because these images are not sufficient to accurately measure the LA aperture (ostia). Sufficient might mean, for example, that at least 80, 90, or 95% of the LAA / LA aperture (ostia) is visible in at least one image, so that an accurate assessment can be performed. The image processing system 100 may also be configured to train or store machine learning models for these tasks. Imaging data is acquired in real time from the medical imaging device 130, for example, as an ultrasound sequence. Additional or different components may be provided, or there may be fewer components. For example, a computer network may be included for remote processing of locally acquired ultrasound data by, for example, a server 140. In another example, a user input device (e.g., a keyboard, buttons, sliders, dials, trackballs, mice, or other devices) may be provided for the user to change or position one or more markers.Further examples may include one or more devices or components used in medical procedures such as left atrial appendage closure (LAAC). For example, a device that closes the opening of the left atrial appendage to prevent a thrombus formed within the LAA26 from entering the bloodstream, a device that clamps and closes the base of the LAA26, or, among many medical devices, a device that closes the LAA26 using a band or suture loop.

[0020] The medical imaging device 130 may be an ultrasound system 130 configured to acquire scanning planes of the LA and LAA26. Ultrasound imaging uses sound waves to image internal structures. Various ultrasound procedures include transthoracic echocardiography (TTE), transesophageal echocardiography (TEE), and intracardiac ultrasound (ICE). TTE is a non-invasive procedure in which a transducer (or probe) is placed on the patient's chest. Images are recorded using ultrasound data. In the case of TEE, the probe is inserted through the patient's esophagus to get close to the patient's heart. This probe may have an ultrasound transducer at its tip to provide imaging capabilities. ICE uses an ultrasound-tipped catheter, which is inserted through a vein in the groin and advanced into the heart. ICE can be used to generate detailed images of the size, structure, and function of the heart, as well as detailed images of the heart valves. The usage described herein relates to the treatment of atrial fibrillation under the guidance of ICE imaging, although other procedures and ultrasound techniques may be used. The left atrium may be the most common anatomical structure of interest in this context. The left ventricle or other cardiac chambers may be the anatomical structure of interest in other embodiments. Other ICE imaging locations may result in other anatomical structures of interest, such as arteries or veins. The left atrium and its corresponding anatomical structure are used as examples below. In one embodiment, for LAAC procedures, the catheter is introduced via femoral vein access and positioned in either the right atrium, right ventricular outflow tract, coronary sinus, or left atrium. The medical imaging device 130 is configured to acquire a 2D scanning plane of the left atrium and LAA 26.

[0021] The processor 110 is a general-purpose processor, a digital signal processor, a graphics processing unit, an application-specific integrated circuit, a field-programmable gate array, an artificial intelligence processor, a digital circuit, an analog circuit, a combination thereof, or any other device currently known or to be developed for LAAC device specification from 2D-ICE images, among other processes described below. The processor 110 is one device, multiple devices, or a network. In the case of multiple devices, parallel or sequential partitioning of processing may be used. The various devices constituting the processor 110 may perform different functions. In one embodiment, the processor 110 is a control processor for a medical imaging device 130 or another processor. In other embodiments, the processor 110 is part of a separate workstation or computer. The processor 110 operates according to stored instructions and performs the various processes described herein. The processor 110 consists of software, design, firmware, and / or hardware and is configured to perform any or all of the processes in Figures 5 and 7, and any other arbitrary calculations described herein.

[0022] Figure 5 shows a workflow for LAA closure device specification based on 2D-ICE according to one embodiment. The patient is prepared for the procedure. ICE is performed using an ultrasound-tipped catheter, which is inserted through a vein in the groin and advanced into the heart. ICE overview 310 of the left atrium (ICE scan data of Left Atrium) is acquired. The left atrium is modeled and a Left Atrium Model 320 is generated. ICE imaging of LAA 26 is performed to acquire ICE scan data of LAA 315. ICE measurements from the ICE imaging of LAA 26 are checked against the model. If the measurements are sufficient, an LAAC device is selected / identified based on the measurements and the LAA Device Specification 325, and the procedure is performed. If the measurements are not sufficient, the operator is notified, and additional imaging of LAA 26 is performed.

[0023] The processor 110 is configured to generate a model of the patient's LA from images acquired by the medical imaging device 130. The processor 110 is configured to use the model to determine whether the acquired LAA26 images are sufficient to identify the device for LAAC treatment. The processor 110 may further be configured to use the model to validate measurements from subsequent images of LAA26 acquired by the medical imaging device 130. Validation may include, for example, determining the orientation of the scanning plane and determining whether an acceptable view of the aperture (ostium) exists.

[0024] In one embodiment, the processor 110 is configured to generate a 3D model of the left atrium by combining multiple scanning planes acquired by the medical imaging device 130. In ICE imaging procedures, a sensor provides the position / orientation of the scanning plane. The provided known position is used to fill the volume with data from the scanning plane. The position of the catheter may be identified, for example, from a magnetic position sensor. The position and orientation of the catheter / transducer define the scanning plane. Different scanning planes have different positions / orientations. The detected position is used to assign scalar or other ultrasound data to each voxel in the 3D volume. Alternatively, the detected position is used to align the planar positions represented by the ultrasound data in the volume. The ultrasound data is mapped to three dimensions using the position of the scanning plane.

[0025] Figure 6A shows an example of various scan planes 370 superimposed on a 3D environment. Different scan planes are captured as the catheter / transducer is rotated to different orientations. When data from some or all of the scan planes 370 are combined, a 3D view can be visualized. The 3D view may be sparse, and some data may be missing between the scan planes, for example. However, this 3D view may contain enough data to generate a mesh and / or allow the system to identify the position of each scan plane 370 relative to each other, as well as any landmarks or features of the heart. Figure 6B shows an exemplary model generated by combining each scan plane, segmenting the data, generating a mesh model, and adding some appearance data to flesh out the model. In both Figure 6A and Figure 6B, the scan planes 370 are visible.

[0026] In one embodiment, the processor 110 is configured to generate an LA model by performing segmentation of the acquired image using, for example, a machine learning model. Different types of models or networks may be trained and used for the segmentation task. Segmentation divides an image into regions based on a specified description, such as segmenting body organs / tissues. In one embodiment relating to segmented volumes, the segmented data includes multiple pixels. Each pixel represents a two-dimensional display element. For example, pixels represent a quantity of 2D data. For segmentation, landmark detection, and tracking, the processor 110 may apply one or more machine learning networks or models trained for each task.

[0027] A machine learning network or model may include a neural network defined as a series of sequential feature units or layers. The term "sequential" is used to describe the general flow of output feature values ​​from one layer to the input of the next layer. Information from the next layer is fed to the next layer, and so on, until the final output. Layers can be feedforward (forward only) or bidirectional, involving some feedback to the previous layer. Nodes in each layer or unit may connect to all nodes of the previous and / or subsequent layers or units, or only a subset of nodes. Skip connections are also possible, for example, a layer may output to other layers in addition to sequentially outputting to the next layer. Rather than pre-programming features and attempting to associate them with attributes, deep architectures are defined to learn features at different levels of abstraction based on input data. Features are learned to reconstruct lower-level features (i.e., features at a more abstract or compressed level). Each node in a unit represents a feature. Different units are provided to learn different features. Various units or layers can be used, such as convolution, pooling (e.g., max pooling), deconvolution, fully connected, or other types of layers. Within a unit or layer, any number of nodes are provided. For example, 100 nodes are provided. Subsequent or subsequent units may have more, fewer, or the same number of nodes.

[0028] Segmented data can be used to form a mesh model of the patient's LA. Algorithms can be used to compute a graphical mesh representation from image segmentation. Features such as landmarks / tissues / contours / boundaries can be identified and labeled in the segmented data / 3D model. These features can be used to determine whether the newly acquired LA ultrasound data was acquired at an acceptable angle / orientation.

[0029] Image data, machine-trained networks, training data, computed metrics, and other data may be stored in memory 120. Memory 120 may be, or include, external storage devices, RAM, ROM, databases, and / or local memory (e.g., solid-state drives or hard drives). The same or different non-temporary computer-readable media may be used for instructions and other data. Memory 120 may be implemented using a database management system (DBMS) and reside in memory 120 such as a hard disk, RAM, or removable media. Alternatively, memory 120 is built into processor 110 (e.g., cache). Instructions for performing the processes, methods, and / or techniques described herein are provided to non-temporary computer-readable storage media or memory such as caches, buffers, RAM, removable media, hard drives, or other computer-readable storage media (e.g., memory 120). Instructions are executable by processor 110 or another processor. Computer-readable storage media include various types of volatile and non-volatile storage media. The functions, processes, or tasks shown in the drawings or described herein are executed in response to one or more instruction sets stored in a computer-readable storage medium. The functions, processes, or tasks are independent of the instruction sets, storage medium, processor 110, or processing strategies and may be executed by software, hardware, integrated circuits, firmware, microcode, etc., operating alone or in combination. In one embodiment, the instructions are stored in a removable media device for reading by a local or remote system. In another embodiment, the instructions are stored in a remote location for transfer over a computer network. In yet another embodiment, the instructions are stored in a given computer, CPU, GPU, or system. Since some of the illustrated system components and method steps may be implemented in software, the actual connections between system components (or process steps) may vary depending on how this embodiment is programmed.

[0030] The processor 110 is configured to determine whether the subsequent image data of LAA26 is sufficient to identify the LAAC device. To determine whether the image data of LAA26 is sufficient, the processor 110 may register the image slices of LAA26 to a cardiac model and then determine their orientation. For example, a view of the LA from a particular scanning plane may be acquired at an orientation offset by 5, 10, or 20 degrees from the optimal. The offset angle makes it difficult to properly evaluate the LA aperture (ostia). The processor 110 is configured to determine if the newly acquired image data of the LA is insufficient by matching the features of the newly acquired image to a 3D model and deriving where the scanning plane slices the cardiac region.

[0031] The processor 110 is further configured to check / validate various measurements from ICE imaging of LAA26 against an LA model. In one embodiment, the processor 110 is configured to measure features of LAA26 and validate the measurements, at least in part, based on a cardiac model. The processor 110 may determine and use the orientation of the catheter / scanning plane and an algorithm to determine whether the patient's LA opening measurements fall within a predefined range of estimated LA opening measurements derived from the cardiac model. The algorithm is then used to determine whether the ICE images of LAA26 are sufficient to accurately estimate the need for an LAA closure device for this patient, based on the ICE images of LAA26 and the LA model. If the ICE images of LAA26 are insufficient, the system user may be notified of this and informed how to obtain additional appropriate images to more accurately determine the patient's specific needs for an LAA closure device. If the ICE images are sufficient, the processor 110 may further be configured to select a device for LAAC treatment based on the measurements of LAA26.

[0032] The display 115 is configured to display or otherwise provide patient images and measurements to the user. The display may be configured to display a 3D model, acquired 2D-ICE images, and an overlay of the 2D-ICE images and 3D model of the LAA26. In one example, the scanning plane of the acquired 2D-ICE image of the LAA is displayed to the operator so that the operator can understand why the image is insufficient, for example, how much the scanning plane is offset from the optimal or useful scanning plane. The display 115 is a CRT, LCD, projector, plasma, printer, tablet, smartphone, or other display device currently known or to be developed in the future for displaying the output.

[0033] In one embodiment, the system is configured to measure not only the LAA bore but also any measurements of the size or shape of the LAA26 used when selecting a device for LAA26 closure. Many manufacturers of LAA closure devices offer solutions in different sizes. When the sole selection between devices is based on size, this is referred to as device sizing. When anatomical measurements are used to select between various available devices, the system may be used to assist in sizing or to select between devices from different manufacturers.

[0034] In one embodiment, the system can detect whether LAA26 is visible in the ICE frame. The system further registers the ICE frame to a model of left atrial anatomy and estimates the variability in the alignment of the ICE frame with the anatomical model. The system is further configured to infer based on the uncertainty in the ICE frame alignment and the uncertainty in the anatomical model and to determine the uncertainty in specific measurements, such as the diameter of the LAA opening 205. The system reduces the need for pre-procedure CT imaging, TEE imaging, or other imaging separate from ICE, which can be used as guidance during LAA closure procedures. The system provides a measure of confidence that LAA26 was measured from ICE for sizing of the LAA closure device. Furthermore, the system provides instructions on how to precisely position the ICE catheter 210 to obtain ICE images that better support LAA26 measurements useful for selecting the LAA closure device.

[0035] Figure 7 illustrates an exemplary method for guiding 2D-ICE imaging of the LAA opening 205 using an anatomical estimation model and for optimally selecting a device for LAA closure. Each step is performed by the systems shown in Figures 4, 5, 8, and 9, other systems, workstations, computers, and / or servers. Additional or different steps may be provided, and there may be fewer steps. Each step is performed in the order shown (e.g., from top to bottom) or in any other order. Some steps may be omitted or modified depending on the results of previous steps and the patient's condition.

[0036] In process A110, multiple images of the patient's left atrium are acquired. The ultrasound scanner uses an ICE catheter 210 for imaging. The transducer of the ICE catheter 210 scans a plane. The scanning plane is oriented based on the catheter's position. As the catheter moves (e.g., translation or rotation), different scanning planes are scanned. Each scan generates a frame of data representing the scanning plane at that point in time. The frames of ultrasound data may be scalar values ​​or representation values ​​(e.g., RGB) in polar or Cartesian coordinate format. The frames of ultrasound data may be B-mode, color flow, or other ultrasound images. A sequence of data frames results from the ICE imaging. Since each frame represents a 2D scanning plane, a set of frames representing different 2D scanning planes within and / or around the cardiac volume is acquired.

[0037] In process A120, a model of the left atrium is generated from multiple images of the left atrium. This model is generated by combining multiple slices of the left atrium and registering the slices in a common coordinate system using the known positions of the catheter / transducer. Examples of scanning planes (slice) and how they are combined to generate the model are provided in Figures 6A and 6B. The volume is filled using the known positions represented by the ultrasound data (multiple images). The detected positions are used to assign scalar or other ultrasound data to each voxel. Alternatively, the detected positions are used to align the planar positions represented by the ultrasound data within the volume. The ultrasound data is mapped three-dimensionally using the scanning plane positions. The mapping forms a 3D sparse ICE volume using the positional information associated with each ICE image. A set of 2D-ICE images is input, each containing a portion of the heart within the field of view. The detected 3D positions are used to map all 2D-ICE images into 3D space, thereby forming a sparse ICE volume. The generated 3D sparse ICE volume maintains the spatial relationships between individual ICE views.

[0038] In one embodiment, the model is segmented using a machine learning-based network. 3D segmentation is labeling each anatomical structure by voxel or location. The anatomical structure represented by each location is labeled. Alternatively, segmentation forms a 3D mesh for each anatomical structure. Other segmentation results may be provided. 3D segmentation provides boundaries for one or more structures in 3D. Segmentation is of one or more structures of interest, for example, identifying different subsets or all of the anatomical structures of interest.

[0039] In one embodiment, the processor 110 generates 3D segmentation from the input of the ICE volume to a machine learning-prepared network. 3D segmentation is labeling of each anatomical structure by voxel or location. The anatomical structure represented by each location is labeled. Alternatively, segmentation forms a 3D mesh for each anatomical structure. Other segmentation results may be provided. 3D segmentation provides boundaries for one or more structures in 3D. Segmentation is of one or more structures of interest, for example, identifying the locations of a subset or all of the anatomical structures of interest relating to the left atrium, specifically including the cusps of heart valves (tricuspid valve, aortic valve, mitral valve), segments of each cardiac chamber, and the proximal parts of arteries and veins leading to the heart. Feature data / identified landmarks may be used to register newly acquired images into a 3D cardiac model.

[0040] In one embodiment, 3D segmentation uses a machine learning network. For example, multiple 2D slices, with or without ICE volume and / or other data, are input to the machine learning network, and 3D segmentation is output in response. The machine learning network may be further configured to complement portions of the 3D volume resulting from the absence of data from a limited number of 2D slices. An example is provided in U.S. Patent 11,534,136B2, which is incorporated herein by reference.

[0041] Machine learning for image segmentation can be performed by extracting feature selections from an input image. These features may include, for example, pixel grayscale, pixel location, image moment, and information about the pixel's neighborhood. The vector of image features is then input to a trained classifier, where each pixel in the image is classified into a class. The classifier's parameters are automatically learned by providing the classifier with input images for which the ground truth classification results are known. The model's output is comparable to the ground truth, and the model's parameters are adjusted so that the model's output better fits the ground truth values. This procedure is repeated for a large number of input images, and the trained parameters are generalized to new and unknown examples. The process of adjusting the model parameters is called training. Deep learning can also be used for segmentation, for example, using neural networks. Deep learning-based image segmentation can be performed, for example, using a convolutional neural network (CNN). A convolutional neural network is a layered structure in which a series of convolutions are performed on the input image. The convolution kernels are learned during training. The convolution results are then combined using a pre-trained statistical model that outputs segmented images.

[0042] A machine learning-trained network may be an image-to-image network, such as a fully convolutional U-Net, trained to convert an ICE volume into a 3D segmentation. For example, the network's trained convolutional units, weights, links, and / or other properties are applied to the data and / or derived feature values ​​of the ICE volume, extracting corresponding features through multiple layers and outputting a 3D segmentation. Input features may be extracted from ICE images arranged in 3D. Other, more abstract features may be extracted from the extracted features using this architecture. Depending on the number and / or arrangement of units or layers, other features may be extracted from the input. The network includes an encoder (convolutional) network and a decoder (transposed convolutional) network, forming a "U" shape with connections that pass features from the encoder to the decoder at the most compressed or abstract level. Skip connections may be provided. Any currently known or future U-Net architecture may be used. Other fully convolutional networks may also be used. In one embodiment, the network is a U-Net with one or more skip connections. Skip connections pass features from the encoder to the decoder at different levels of abstraction and resolution, other than the most abstract level (i.e., the bottleneck). Skip connections provide more information to the decoding layer. Fully convolutional layers can be the bottleneck in the network (i.e., between the encoder and decoder at the most abstract level of the layer). Fully connected layers can ensure that as much information as possible is encoded. Batch normalization may be added to stabilize training.

[0043] Various optimization techniques such as Adadelta, SGD, RMSprop, or Adam may be used to train any network. Network weights are initialized randomly, but other initialization methods may be used. End-to-end training is performed, but one or more features may be set. Batch normalization, dropout, and data augmentation are not used, but may be used (e.g., batch normalization and dropout are used). Different discriminant features are learned during optimization. Features that provide indications of anatomical location or missing volume information given an input sparse ICE volume are learned. The network minimizes errors or losses such as mean squared error (MSE), Huber loss, L1 loss, or L2 loss. In one embodiment, machine learning uses a combination of adversarial loss and reconstruction loss (e.g., using a GAN). The GAN classifier provides the adversarial loss for segmentation and / or volume completion. Reconstruction loss is a measure of the difference between the 3D segmentation and the ground truth segmentation, and the difference between the complete volume and the ground truth volume.

[0044] Figure 8 shows one embodiment of the artificial neural network 500 according to one or more embodiments. The term "artificial neural network" can also be referred to as "neural network," "artificial neural network," or "neural network." The artificial neural network 500 may be used, for example, as part of one or more machine learning-based networks used for segmentation, rendering, etc.

[0045] The artificial neural network 500 includes nodes 502-522 and edges 532, 534, ..., 536, where each edge 532, 534, ..., 536 is a directed connection from the first nodes 502-522 to the second nodes 502-522. Generally, the first nodes 502-522 and the second nodes 502-522 are different nodes, but it is also possible for the first nodes 502-522 and the second nodes 502-522 to be the same. In Figure 8, edge 532 is a directed connection from node 502 to node 506, and edge 534 is a directed connection from node 504 to node 506. Edges 532, 534, ..., 536 from the first nodes 502-522 to the second nodes 502-522 are shown as "inflow edges" to the second nodes 502-522 and as "outflow edges" to the first nodes 502-522.

[0046] In this embodiment, the nodes 502-522 of the artificial neural network 500 may be arranged in layers 524-530, which may include a specific order introduced by edges 532, 534, ..., 536 between the nodes 502-522. In particular, the edges 532, 534, ..., 536 may exist only between adjacent layers of nodes. In the embodiment shown in Figure 8, there is an input layer 524 containing only nodes 502 and 504 which have no inflow edges, an output layer 530 containing only node 522 which has no outflow edges, and hidden layers 526, 528 between the input layer 524 and the output layer 530. Generally, the number of hidden layers 526, 528 can be arbitrarily selected. The number of nodes 502 and 504 in the input layer 524 is usually related to the number of input values ​​to the neural network 500, and the number of nodes 522 in the output layer 530 is usually related to the number of output values ​​to the neural network 500.

[0047] Specifically, a certain (real) number can be assigned as a value to all nodes 502-522 of the neural network 500. Here, x (n) iThis shows the values ​​of the i-th node 502-522 in the nth layer 524-530. The values ​​of nodes 502-522 in the input layer 524 are equivalent to the input values ​​of the neural network 500, and the value of node 522 in the output layer 530 is equivalent to the output values ​​of the neural network 500. Furthermore, each edge 532, 534, ..., 536 may contain weights that are real numbers, in particular the weights being real numbers within the interval [-1,1] or the interval [0,1]. Here, w (m,n) i,j This shows the edge weights between the i-th node 502-522 of layer m 524-530 and the j-th node 502-522 of layer n 524-530. Furthermore, the abbreviated form w (n) i,j But, weight lol (n,n+1) i,j It is defined as follows.

[0048] Specifically, input values ​​are propagated through the neural network 500 in order to calculate the output values ​​of the neural network 500. In particular, the values ​​of nodes 502-522 in the (n+1)th layer 524-530 can be calculated based on the values ​​of nodes 502-522 in the nth layer 524-530 using the following formula. JPEG2026071155000002.jpg2295

[0049] Here, the function f is a transfer function (also known as an "activation function"). Known transfer functions include the step function, sigmoid function (e.g., logistic function, generalized logistic function, hyperbolic tangent, arctangent function, error function, smoothstep function), or rectifier function. Transfer functions are primarily used for normalization purposes.

[0050] In a concrete example, values ​​are propagated layer by layer through the neural network, the value of the input layer 524 is given by the input to the neural network 500, the value of the first hidden layer 526 is calculated based on the value of the input layer 524 of the neural network, the value of the second hidden layer 528 can be calculated based on the value of the first hidden layer 526, and so on.

[0051] Edge value w (m,n) i,j In order to set (m,n) and i,j , the neural network 500 must be trained using training data. Specifically, the training data includes training input data and training output data (denoted as t i ). During the training step, the neural network 500 is applied to the training input data to generate the calculated output data. In particular, the learning data and the calculated output data include a number of values, the number of which is equal to the number of nodes in the output layer.

[0052] In a specific example, the comparison between the calculated output data and the training data is used (by the backpropagation algorithm) to recursively adapt the weights within the neural network 500. Specifically, the weights are changed according to the following formula. JPEG2026071155000003.jpg2186

[0053] γ in the formula is the learning rate, and the number δ (n) j can be recursively calculated as follows. JPEG2026071155000004.jpg27145δ (n+1) j Based on (n+1) and j , when the (n + 1)-th layer is not the output layer, and JPEG2026071155000005.jpg21135 when the (n + 1)-th layer is the output layer 530, where f’ is the first derivative of the activation function, and y (n+1) j is the comparison training value of the j-th node in the output layer 530.

[0054] FIG. 9 shows a convolutional neural network 600 according to one or more embodiments. For example, the machine learning networks described herein, such as for segmentation, model generation, rendering, etc., can be implemented using the convolutional neural network 600.

[0055] In the embodiment shown in Figure 9, the convolutional neural network 600 includes an input layer 602, a convolutional layer 604, a pooling layer 606, a fully connected layer 608, and an output layer 610. Alternatively, the convolutional neural network 600 may include other types of layers in addition to multiple convolutional layers 604, multiple pooling layers 606, and multiple fully connected layers 608. The order of the layers is arbitrarily selectable, and typically the fully connected layer 608 is used as the last layer before the output layer 610.

[0056] . In particular, within the convolutional neural network 600, nodes 612-620 of a certain layer 602-610 can be considered to be organized as a d-dimensional matrix or a d-dimensional image. In particular, in the two-dimensional case, the values ​​of nodes 612-620 indexed by i and j in the nth layer 602-610 are x (n) [i,j] This can be expressed as follows. However, the arrangement of nodes 612-620 in layers 602-610 does not affect the calculations themselves performed within the convolutional neural network 600, since these calculations are given solely by the edge structure and weights.

[0057] In a specific example, the convolutional layer 604 is characterized by the fact that the structure and weights of the input edges constitute a convolution operation based on a predetermined number of kernels. In particular, the structure and weights of the input edges are the values ​​x of node 614 of the convolutional layer 604. (n) k However, the value x of node 612 in the preceding layer 602 (n-1) Based on, convolution x (n) k =K k *x (n-1) It is chosen to be calculated as, where convolution* is defined as follows in the case of two dimensions: JPEG2026071155000006.jpg14138

[0058] In this case, the kth kernel K kThe kernel is a d-dimensional matrix (a two-dimensional matrix in this embodiment) and is typically small compared to the number of nodes 612-618 (e.g., a 3x3 or 5x5 matrix). In particular, this means that the weights of the input edges are not independent and are selected to generate the convolution equation. Specifically, if the kernel is a 3x3 matrix, there are only 9 independent weights (each entry in the kernel matrix corresponds to one independent weight) and this is independent of the number of nodes 612-620 in the layers 602-610. In particular, in the convolutional layer 604, the number of nodes 614 in the convolutional layer is equal to the number of nodes 612 in the preceding layer 602 multiplied by the number of kernels.

[0059] If node 612 of the preceding layer 602 is organized as a d-dimensional matrix, using multiple kernels may be interpreted as adding an additional dimension (referred to as the "depth" dimension), and as a result, node 614 of the convolutional layer 604 is organized as a (d+1)-dimensional matrix. If node 612 of the preceding layer 602 is already organized as a (d+1)-dimensional matrix including the depth dimension, using multiple kernels may be interpreted as extending along the depth dimension, and as a result, node 614 of the convolutional layer 604 is also organized as a (d+1)-dimensional matrix, where the size of the (d+1)-dimensional matrix with respect to the depth dimension is larger by a coefficient of the number of kernels than that in the preceding layer 602.

[0060] The advantage of using the convolutional layer 604 is that it allows for the utilization of spatially local correlations of the input data by forcing local connection patterns between nodes in adjacent layers, particularly by ensuring that each node is connected only to small areas of the nodes in the preceding layer.

[0061] In the embodiment shown in Figure 9, the input layer 602 includes 36 nodes 612 organized as a two-dimensional 6x6 matrix. The convolutional layer 604 includes 72 nodes 614 organized as two two-dimensional 6x6 matrices, each of which is the result of convolving the values ​​of the input layer with a kernel. Alternatively, the nodes 614 of the convolutional layer 604 may be interpreted as being organized as a three-dimensional 6x6x2 matrix, where the last dimension is the depth dimension.

[0062] The pooling layer 606 can be characterized by the structure and weights of its input edges, as well as the activation function of its node 616, which constitute a pooling operation based on a nonlinear pooling function f. For example, in the two-dimensional case, the value x of node 616 of the pooling layer 606. (n) This is the value x of node 614 in the preceding layer 604. (n-1) Based on this, it can be calculated as follows: JPEG2026071155000007.jpg12139

[0063] In other words, by using the pooling layer 606, the number of nodes 614,616 can be reduced by replacing the adjacent d1-d2 number of nodes 614 in the preceding layer 604 with a single node 616 calculated in the pooling layer as a function of the values ​​of that number of adjacent nodes. In particular, the pooling function f can be the maximum function, the mean, or the L2 norm. Especially for the pooling layer 606, the input edge weights are fixed and are not changed by training.

[0064] The advantage of using a pooling layer of 606 is that the number of nodes (614,616) and the number of parameters are reduced. This leads to a reduction in computational complexity within the network and helps to suppress overfitting.

[0065] In the embodiment shown in Figure 9, the pooling layer 606 is max pooling, where four adjacent nodes are replaced by one node, the value of which is the maximum value of the four adjacent nodes. Max pooling is applied to each d-dimensional matrix of the previous layer. In this embodiment, max pooling is applied to each of the two 2-dimensional matrices, reducing the number of nodes from 72 to 18.

[0066] The fully connected layer 608 can be characterized by the fact that most of the edges, in particular all of them, exist between the node 616 of the previous layer 606 and the node 618 of the fully connected layer 608, and that the weight of each edge can be adjusted individually.

[0067] In this embodiment, the nodes 616 of the preceding layer 606 of the fully connected layer 608 are represented both as a two-dimensional matrix and as unrelated nodes (shown as a column of nodes, where the number of nodes is reduced for better readability). In this embodiment, the number of nodes 618 of the fully connected layer 608 is equal to the number of nodes 616 of the preceding layer 606. Alternatively, the number of nodes 616,618 may be different.

[0068] Furthermore, in this embodiment, the value of node 620 of the output layer 610 is determined by applying a softmax function to the value of node 618 of the preceding layer 608. By applying the softmax function, the sum of the values ​​of all nodes 620 of the output layer 610 is 1, and all values ​​of all nodes 620 of the output layer are real numbers between 0 and 1.

[0069] The convolutional neural network 600 may also include ReLU (rectified linear units) layers or activation layers with nonlinear transfer functions. Specifically, the number and structure of nodes in a ReLU layer are equal to the number and structure of nodes in the preceding layer. In particular, the value of each node in the ReLU layer is calculated by applying a rectification function to the value of the corresponding node in the preceding layer.

[0070] The inputs and outputs of different convolutional neural network blocks can be wired using summation (residual / dense neural networks), element-wise multiplication (attention), or other differentiable operators. Therefore, if the entire pipeline is differentiable, the architecture of a convolutional neural network can be nested rather than sequential.

[0071] In a concrete example, the convolutional neural network 600 may be trained based on a backpropagation algorithm. To prevent overfitting, regularization methods can be used, such as dropout of nodes 612-620, stochastic pooling, use of artificial data, weight decay based on L1 or L2 norms, or maximum norm constraints. Different loss functions may be combined to train the same neural network and reflect a common training objective. A subset of the neural network's parameters may be excluded from optimization to preserve weights pre-trained on other datasets.

[0072] Any machine training architecture can be used for segmentation. Similarly, different machine training architectures can be used for the other tasks described here. For example, U-Net can be used. A convolution-to-transposed convolution network can be used. One segment of a layer or unit applies convolution to increase the level of abstraction or compression. The most abstract feature values ​​are then output to another segment. The other segment of the layer or unit then applies transposed convolution to decrease the level of abstraction or compression, resulting in the output of positional class belonging indications. This architecture can be a fully convolutional network. Other deep networks can be used.

[0073] Referring back to Figure 7, in process A130, multiple images of the patient's LAA26 are acquired. Similar to process A110, the ultrasound scanner uses the ICE catheter 210 to image the patient's cardiac region. Images of the LAA26 may be acquired. In one embodiment, it is determined whether the LAA26 is visible in the image. If the acquired image does not contain the LAA26, a notification may be provided. For this procedure, the transducer of the ICE catheter 210 scans a plane. The scanning plane is oriented based on the position of the catheter. As the catheter moves (e.g., translation or rotation), different scanning planes are scanned. Each scan generates a frame of data representing the scanning plane at that time. Frames of ultrasound data may be scalar values ​​or representation values ​​(e.g., RGB) in polar or Cartesian coordinate form. Frames of ultrasound data may be B-mode, color flow, or other ultrasound images.

[0074] In process A140, the system determines, based on the model, whether multiple images of LAA26 are sufficient to accurately estimate the dimensions of each LAA closure device. For example, being sufficient might mean that at least 80, 90, or 95% of the LAA / LA opening is visible in at least one image, allowing for an accurate assessment. Whether the images are sufficient to estimate the size of the LAA for the LAAC device can be determined based on the device manufacturer's recommendations regarding device placement. Different devices and device manufacturers may require different measurements or values ​​for anatomical features. For example, a device manufacturer might require that the diameter of the LAAC opening be measured with an accuracy of 2 mm. If the image shows the diameter of the LAA within 2 mm of the actual diameter of the LAA, then the image is sufficient. Because there may be some uncertainty in the model reconstructed from the ICE images, the diameter measured from the model can only be known with 90% accuracy. Therefore, if the aperture visible in the image is within 2 mm of the maximum aperture measured on the model, the system will output that the current image is sufficient with a 90% probability.

[0075] In one embodiment, one or more images of LAA26 are registered to a 3D model to determine the orientation of the scanning plane relative to other features of the 3D cardiac model, such as the LA opening. For example, features that are landmarks identified in the images of LAA26 may be registered / matched to relevant features in the 3D model. The orientation may be derived therefrom. In one embodiment, features are identified by segmenting the images of LAA26 using, for example, a machine learning-prepared network configured for segmentation as described above. In one embodiment, uncertainties in ICE frame alignment and uncertainties in the anatomical model are determined and used to identify uncertainties in specific measurements, such as the diameter of the LAA opening 205. These uncertainties may be provided to the operator so that the operator can determine whether to obtain a different view. The uncertainty in the sizing measurement may be used when determining whether the device is suitable for LAAC treatment.

[0076] In process A150, if multiple images of LAA26 are sufficient, a suitable LAA closure device is identified. In one embodiment, the device may be identified before performing the procedure. Multiple images of LAA26 may be used to determine whether the device is suitable, for example, by comparing estimated anatomical feature values ​​with the device specifications. In one embodiment, multiple measurements of LAA26 are calculated and compared with measurements predicted from the LA model. These measurements may include not only the measurement of the diameter of the LAA opening 205, but also any measurements of the size or shape of LAA26 used when selecting a device for LAA closure. For example, the LAA opening is measured from the pulmonary vein ridge to the junction between the LA and LAA26. The landing zone (the area in LAA26 where the device is seated) is measured 10 mm inward from the opening at an angle perpendicular to the neck axis. Another important distance to be measured is the maximum length of the anchoring lobe to ensure that the lobe has sufficient space to accommodate the selected device. Based on the measurements, different LAAC devices may be used for each patient. Many manufacturers of LAA closure devices offer solutions in different sizes. When the choice between devices is based solely on size, this is called "sizing" the device. However, additional anatomical measurements other than size may be calculated and used to assist in sizing or to select between devices from different manufacturers.

[0077] If multiple images of LAA26 are insufficient, the operator will be instructed to obtain additional images of LAA26. Instructions may be provided on how to better position the ICE catheter 210 to obtain ICE images that better support LAA26 measurements useful for selecting an LAA closure device. For example, instructions may include rotating or adjusting the catheter position so that the transducer orientation provides image slices that fully depict the LA opening.

[0078] In one embodiment, 2D-ICE treatment provides sufficient visualization of the LAA26 and can be used as an alternative to TEE when guiding LAA occlusion treatment. The embodiment provides an anatomical estimation model for guiding 2D-ICE imaging of the LAA opening 205 to optimally select a device for LAA occlusion. Furthermore, the guidance can be used to provide 2D-ICE images that exclude LAA thrombi, select device size by measuring landing zone dimensions, guide transseptal puncture, verify the positioning of the delivery sheath in the LAA26, assist device delivery, confirm stability before and after release, check for peri-device leakage with color Doppler flow imaging, and monitor for complications such as cardiac tamponade.

[0079] It should be understood that the elements and features described in the claims may be combined in different ways to create new claims that similarly fall within the scope of the invention. That is, while a dependent (referenced) claim depends on only one independent (independent) claim or dependent claim, these dependent claims may alternatively depend on any preceding or succeeding claim, whether independent or dependent, and such new combinations should be interpreted as forming part of the disclosure.

[0080] While the present invention has been described above with reference to various embodiments, it should be understood that many changes and modifications can be made to the disclosed embodiments. Therefore, it should be understood that the above description is intended to be illustrative rather than restrictive, and that all equivalents and / or combinations of the embodiments are included in the description. Regardless of grammatical usage, the terms include individuals having a male, female, or other gender identity.

[0081] The following is a list of non-exclusive exemplary embodiments disclosed herein.

[0082] Exemplary Embodiment 1: A method for evaluating the left atrial appendage, To obtain multiple images of the patient's left atrium, To generate a model of the left atrium from the aforementioned multiple images of the left atrium, To obtain multiple images of the left atrial appendage of the aforementioned patient, Based on the aforementioned model, determine whether the multiple images of the left atrial appendage are sufficient to accurately estimate the specifications of an appropriate left atrial appendage closure device. A method comprising: calculating and providing anatomical measurements of the left atrial appendage when the plurality of images of the left atrial appendage are sufficient; or providing instructions to the operator to obtain additional images of the left atrial appendage when the plurality of images of the left atrial appendage are insufficient.

[0083] Exemplary Embodiment 2: A method according to exemplary embodiment 1, A method further comprising identifying the appropriate left atrial appendage closure device based on the anatomical measurements of the left atrial appendage.

[0084] Exemplary Embodiment 3: A method according to exemplary embodiment 2, A method further comprising performing a left atrial appendage closure procedure using the appropriate left atrial appendage closure device.

[0085] Exemplary Embodiment 4: A method relating to any of the exemplary embodiments 1 to 3, The acquisition described above is a method that includes acquiring a two-dimensional image using an ICE ultrasound imaging system.

[0086] Exemplary Embodiment 5: A method relating to any of the exemplary embodiments 1 to 4, The model includes a method comprising three-dimensional segmentation of the left atrium of the patient.

[0087] Exemplary Embodiment 6: A method relating to any of the exemplary embodiments 1 to 5, Generating the aforementioned model means To determine the position of the scanning plane within the patient's cardiac system for the aforementioned multiple images of the left atrium, To construct an ICE volume based on the aforementioned multiple images and positions of the left atrium, A method comprising generating a three-dimensional segmentation model by inputting the aforementioned ICE volume into a machine learning-prepared network.

[0088] Exemplary Embodiment 7: A method according to exemplary embodiment 6, Determining whether the aforementioned multiple images of the left atrial appendage are sufficient is: Registering one or more of the aforementioned images to the three-dimensional segmentation model, A method comprising checking that the entire diameter of the left atrial appendage opening is present in at least one of the multiple images of the left atrial appendage relative to the model.

[0089] Exemplary Embodiment 8: A method relating to any of the exemplary embodiments 1 to 7, The aforementioned determination is, A method comprising providing uncertainty measurements in the model and at the image frame position, and determining an estimate of the probability that the left atrial appendage opening diameter has been measured to a predetermined level of confidence.

[0090] Exemplary Embodiment 9: A method relating to any of the exemplary embodiments 1 to 8, The foregoing provides a method further comprising one or more additional measurements of the size or shape of the left atrial appendage used when selecting a device for left atrial appendage closure.

[0091] Exemplary Embodiment 10: A method relating to any of the exemplary embodiments 1 to 9, The instructions relate to a method based on a simulated alternative view from the model of the left atrium that the ICE catheter can reach, which enables accurate calculation of the left atrial appendage orifice diameter.

[0092] Exemplary Embodiment 11: A system for selecting a left atrial appendage device, A 2D-ICE imaging system configured to acquire left atrial imaging data of a patient's left atrium and left atrial appendage imaging data of the patient's left atrial appendage, Includes an image processing processor configured to generate a three-dimensional model of the patient's left atrium based on the left atrial image data, The image processing processor is further configured to register the left atrial appendage image data with the three-dimensional model and to determine that the left atrial appendage is sufficiently visible in the left atrial appendage image data. The system further comprises an image processing processor configured to calculate at least one anatomical measurement of the left atrial appendage and to select a device for left atrial appendage closure based on the at least one anatomical measurement.

[0093] Exemplary Embodiment 12: A system according to exemplary embodiment 11, The left atrium is sufficiently visualized when the diameter of the left atrial appendage opening is measured to a predetermined level of confidence.

[0094] Exemplary Embodiment 13: A system relating to exemplary embodiment 11 or 12, The image processing processor is further configured to determine that the left atrial appendage is not sufficiently visible in the left atrial appendage image data. The system includes an image processing processor configured to generate instructions for an operator to acquire additional image data of the left atrial appendage.

[0095] Exemplary Embodiment 14: A system relating to any of the exemplary embodiments 11 to 13, The three-dimensional model is generated by segmenting and combining multiple scanning planes from the left atrial image data.

[0096] Exemplary Embodiment 15: A system according to exemplary embodiment 14, Generating the aforementioned three-dimensional model is To determine the position of the left atrial image data within the patient's cardiac system on the scanning plane, To configure the ICE volume based on the scanning plane and the position, A system comprising generating a three-dimensional segmentation model by inputting the aforementioned ICE volume into a machine learning-prepared network.

[0097] Exemplary Embodiment 16: A system relating to any of the exemplary embodiments 11 to 15, A system further comprising a display configured to display the left atrial image data, the left atrial appendage image data, and / or the three-dimensional model of the left atrium.

[0098] Exemplary Embodiment 17: A non-temporary computer executable storage medium storing machine-readable instructions that can be executed by at least one processor for left atrial appendage evaluation, The aforementioned machine-readable instruction is: To obtain multiple images of the patient's left atrium, To generate a model of the left atrium from the aforementioned multiple images of the left atrium, To obtain multiple images of the left atrial appendage of the aforementioned patient, Based on the aforementioned model, determine whether the multiple images of the left atrial appendage are sufficient to accurately estimate the specifications of the left atrial appendage closure device. A non-temporary computer execution storage medium that includes calculating and providing anatomical measurements of the left atrial appendage when the plurality of images of the left atrial appendage are sufficient, or providing the operator with instructions to obtain additional images of the left atrial appendage when the plurality of images of the left atrial appendage are insufficient.

[0099] Exemplary Embodiment 18: A non-temporary computer execution storage medium according to exemplary embodiment 17, The machine-readable instructions for generating the aforementioned model are: To determine the position of the scanning plane within the patient's cardiac system for the aforementioned multiple images of the left atrium, To construct an ICE volume based on the aforementioned multiple images and positions of the left atrium, A non-temporary computer execution storage medium, comprising generating a three-dimensional segmentation model by inputting the aforementioned ICE volume into a machine learning-prepared network.

[0100] Exemplary Embodiment 19: A non-temporary computer execution storage medium according to exemplary embodiment 17 or 18, The machine-readable instruction for making the aforementioned determination is: Registering the plurality of images of the left atrial appendage to the model of the left atrium, A non-temporary computer execution storage medium, which includes determining whether an overall view of the diameter of the left atrial appendage is visualized in the plurality of images of the left atrial appendage.

[0101] Exemplary Embodiment 20: A non-temporary computer execution storage medium according to any of the exemplary embodiments 17 to 19, The acquisition described above is performed using a 2D-ICE imaging system and involves a non-temporary computer execution storage medium.

Claims

1. A method for evaluating the left atrial appendage, To obtain multiple images of the patient's left atrium, To generate a model of the left atrium from the aforementioned multiple images of the left atrium, To obtain multiple images of the left atrial appendage of the aforementioned patient, Based on the aforementioned model, determine whether the multiple images of the left atrial appendage are sufficient to accurately estimate the specifications of an appropriate left atrial appendage closure device. A method comprising: calculating and providing anatomical measurements of the left atrial appendage when the plurality of images of the left atrial appendage are sufficient; or providing instructions to the operator to obtain additional images of the left atrial appendage when the plurality of images of the left atrial appendage are insufficient.

2. The method according to claim 1, further comprising identifying the appropriate left atrial appendage closure device based on the anatomical measurements of the left atrial appendage.

3. The method according to claim 2, further comprising using the appropriate left atrial appendage closure device for left atrial appendage closure procedure.

4. The method according to claim 1, wherein the acquisition includes acquiring a two-dimensional image using an ICE ultrasound imaging system.

5. The method according to claim 1, wherein the model includes three-dimensional segmentation of the left atrium of the patient.

6. Generating the aforementioned model means To determine the position of the scanning plane within the patient's cardiac system for the aforementioned multiple images of the left atrium, To construct an ICE volume based on the aforementioned multiple images and positions of the left atrium, The method according to claim 1, comprising generating a three-dimensional segmentation model by inputting the ICE volume into a machine learning-trained network.

7. Determining whether the aforementioned multiple images of the left atrial appendage are sufficient is: Registering one or more of the aforementioned images to the three-dimensional segmentation model, The method according to claim 6, comprising checking that the entire diameter of the left atrial appendage opening is present in the at least one image by utilizing the position of the image of at least one of the plurality of images of the left atrial appendage relative to the model.

8. The aforementioned determination is, The method according to claim 1, comprising providing uncertainty measurements in the model and at the image frame position, and determining an estimate of the probability that the left atrial appendage opening diameter has been measured to a predetermined level of confidence.

9. The method according to claim 1, further comprising one or more additional measurements of the size or shape of the left atrial appendage used when selecting a device for left atrial appendage closure.

10. The method according to claim 1, wherein the instructions are based on a simulated alternative view from the model of the left atrium that the ICE catheter can reach, which enables accurate calculation of the left atrial appendage orifice diameter.

11. A system for selecting a left atrial appendage device, A 2D-ICE imaging system configured to acquire left atrial image data of a patient's left atrium and left atrial appendage image data of the patient's left atrial appendage, Includes an image processing processor configured to generate a three-dimensional model of the patient's left atrium based on the left atrial image data, The image processing processor is further configured to register the left atrial appendage image data with the three-dimensional model and to determine that the left atrial appendage is sufficiently visible in the left atrial appendage image data. The system further comprises an image processing processor configured to calculate at least one anatomical measurement of the left atrial appendage and to select a device for left atrial appendage closure based on the at least one anatomical measurement.

12. The system according to claim 11, wherein the left atrium is sufficiently visualized when the diameter of the left atrial appendage opening is measured to a predetermined level of confidence.

13. The image processing processor is further configured to determine that the left atrial appendage is not sufficiently visible in the left atrial appendage image data. The system according to claim 11, wherein the image processing processor is configured to generate instructions to an operator for acquiring additional image data of the left atrial appendage.

14. The system according to claim 11, wherein the three-dimensional model is generated by segmenting and combining a plurality of scanning planes from the left atrial image data.

15. Generating the aforementioned three-dimensional model is To determine the position of the left atrial image data within the patient's cardiac system on the scanning plane, To configure an ICE volume based on the scanning plane and the position, The system according to claim 14, comprising generating a three-dimensional segmentation model by inputting the ICE volume into a machine learning-trained network.

16. The system according to claim 11, further comprising a display configured to display the left atrial image data, the left atrial appendage image data, and / or the three-dimensional model of the left atrium.

17. A non-temporary computer executable storage medium storing machine-readable instructions that can be executed by at least one processor for left atrial appendage evaluation, The aforementioned machine-readable instruction is: To obtain multiple images of the patient's left atrium, To generate a model of the left atrium from the aforementioned multiple images of the left atrium, To obtain multiple images of the left atrial appendage of the aforementioned patient, Based on the aforementioned model, determine whether the multiple images of the left atrial appendage are sufficient to accurately estimate the specifications of the left atrial appendage closure device. A non-temporary computer execution storage medium that includes calculating and providing anatomical measurements of the left atrial appendage when the plurality of images of the left atrial appendage are sufficient, or providing the operator with instructions to obtain additional images of the left atrial appendage when the plurality of images of the left atrial appendage are insufficient.

18. The machine-readable instructions for generating the aforementioned model are: To determine the position of the scanning plane within the patient's cardiac system for the aforementioned multiple images of the left atrium, To construct an ICE volume based on the aforementioned multiple images and positions of the left atrium, The non-temporary computer execution storage medium according to claim 17, comprising generating a three-dimensional segmentation model by inputting the ICE volume into a machine learning-prepared network.

19. The machine-readable instruction for making the aforementioned determination is: Registering the plurality of images of the left atrial appendage to the model of the left atrium, The non-temporary computer execution storage medium according to claim 17, comprising determining whether an overall view of the diameter of the left atrial appendage is visualized in the plurality of images of the left atrial appendage.

20. The acquisition described above is performed using a 2D-ICE imaging system, and the non-temporary computer execution storage medium is as described in claim 17.

Images