System and method for recommending an ablation line

Abstract

A system (10) for improving cardiac ablation procedures includes a recommendation unit (14) configured to provide an initial recommendation of at least one proposed ablation line (4) for an ablation procedure on a patient's anatomical structure. The system displays at least one proposed ablation line on an anatomical map of the anatomical structure. The recommendation unit includes a first trained machine learning model and a second trained machine learning model. Both the first trained machine learning model and the second trained machine learning model have the same structure.

Description

Technical Field

[0001] The present invention relates to a system for the treatment of atrial fibrillation.

Background Art

[0002] Atrial fibrillation (AF) is the most commonly diagnosed sustained arrhythmia, characterized by rapid and irregular activation of the atria. Atrial fibrillation can be paroxysmal, lasting for 7 days or less with or without intervention, or it may persist for more than 7 days (persistent, PS-AF) or more than 12 months (long-standing, LS-PS-AF). Permanent AF is a term used for longstanding persistent AF when attempts to restore sinus rhythm have been abandoned or proven impossible. (See, for example, Kaba, R.A., Momin, A. & Camm, J. Persistent atrial fibrillation: The role of left atrial posterior wall isolation and ablation strategies. J. Clin. Med. 10, (2021)).

[0003] Treatment strategies vary depending on the degree and duration of AF. As an example, paroxysmal AF is typically managed by wide antral circumferential ablation (WACA), i.e., by delivering high-frequency energy in a point-by-point fashion along pre-defined lines in the atria. However, the majority of patients with AF have PS-AF or LS-PS-AF, which are likely more difficult to treat due to arrhythmogenic triggers and more extensive spread of the substrate outside the pulmonary veins. The posterior wall of the left atrium is a common site of these changes and is the target of ablation strategies for treating these more resistant forms of AF. Other common sites are the right and left carinas and the roof.

Summary of the Invention

Means for Solving the Problems

[0004] Accordingly, according to a preferred embodiment of the present invention, a system for improving a cardiac ablation procedure including a recommendation unit is provided. The recommendation unit provides an initial recommendation of at least one proposed ablation line for an ablation procedure on a patient's anatomical structure and displays at least one proposed ablation line on an anatomical map of the anatomical structure. The recommendation unit includes a first trained machine learning model and a second trained machine learning model, both of which have the same structure.

[0005] Furthermore, according to a preferred embodiment of the present invention, the system includes an ablation line editor that enables a physician to modify a selected one of the at least one proposed ablation lines.

[0006] Furthermore, according to a preferred embodiment of the present invention, the system includes a treatment decision maker that determines the recommended energy delivery per segment of the selected ablation line.

[0007] Furthermore, according to a preferred embodiment of the present invention, the system includes a guidance unit that displays the next ablation site on the anatomical structure based at least on the selected ablation line and the catheter position within the anatomical structure.

[0008] Furthermore, according to a preferred embodiment of the present invention, the recommendation unit includes a map segmenter, an ablation line trainer, and an ablation line proposer. The map segmenter segments an anatomical map into output portions of anatomical structures, and the map segmenter utilizes a first trained machine learning model. The ablation line trainer trains a second trained machine learning model to output a proposed ablation line. The ablation line proposer utilizes the second trained machine learning model to propose at least one proposed ablation line for an anatomical structure.

[0009] Furthermore, according to a preferred embodiment of the present invention, the same structure is a graph convolutional neural network (GCN) or a classifier.

[0010] Furthermore, according to a preferred embodiment of the present invention, the recommendation unit recommends which of the at least one proposed ablation lines is most suitable for treatment of a patient.

[0011] Furthermore, according to a preferred embodiment of the present invention, at least one of the proposed ablation lines is a wide area circumferential ablation (WACA) line.

[0012] Furthermore, according to a preferred embodiment of the present invention, the system includes a keypoint filter that compares the actual ablation points of a training case with intersections or lines of adjacent portions on the segmented map of the training case and selects the ablation point closest to the intersection as a keypoint.

[0013] According to a preferred embodiment of the present invention, a computer-implemented method for improving cardiac ablation procedures is also provided. The method includes providing an initial recommendation of at least one proposed ablation line for an ablation procedure on a patient's anatomical structure. Providing includes displaying at least one proposed ablation line on an anatomical map of the anatomical structure. The initial recommendation utilizes both a first trained machine learning model and a second trained machine learning model having the same structure.

[0014] Furthermore, according to a preferred embodiment of the present invention, the method includes enabling a physician to modify a selected one of the at least one proposed ablation lines.

[0015] Furthermore, according to a preferred embodiment of the present invention, the method includes determining a recommended energy delivery per segment of the selected ablation line.

[0016] Furthermore, according to a preferred embodiment of the present invention, the method includes displaying a next ablation site on the anatomical structure based at least on the selected ablation line and a catheter position within the anatomical structure.

[0017] Furthermore, according to a preferred embodiment of the present invention, providing includes segmenting the anatomical map to output portions of the anatomical structure, where segmenting utilizes the first trained machine learning model, training the second trained machine learning model to output a proposed ablation line, proposing at least one proposed ablation line for the anatomical structure, and the proposed ablation line utilizes the second trained machine learning model.

[0018] Furthermore, according to a preferred embodiment of the present invention, the same structure is a graph convolutional neural network (GCN) or a classifier.

[0019] Furthermore, according to a preferred embodiment of the present invention, providing includes recommending which of at least one of the proposed ablation lines is most suitable for treatment of a patient.

[0020] Furthermore, according to a preferred embodiment of the present invention, at least one of the proposed ablation lines is a wide area circumferential ablation (WACA) line.

[0021] Finally, according to a preferred embodiment of the present invention, the method includes comparing the actual ablation points of the training case with the intersections or lines of adjacent portions on the segmentation map of the training case, and selecting the ablation point closest to the intersection as a key point.

Brief Description of the Drawings

[0022] The subject matter regarded as the present invention is particularly pointed out and distinctly claimed in the concluding portion of this specification. However, the present invention, together with its objects, features, and advantages, can be best understood by reference to the following detailed description, taken in conjunction with the accompanying drawings, with regard to both the structure of the mechanism and the method of operation.

Figure 1A

Figure 1B

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

[0023] It will be understood that, for the sake of simplicity and clarity, the elements shown in the figures are not necessarily drawn to scale. For example, the dimensions of some elements may be exaggerated relative to other elements for clarity. Further, reference numbers may be repeated between drawings to indicate corresponding or similar elements where appropriate. DETAILED DESCRIPTION

[0024] The following detailed description sets forth numerous specific details in order to provide a thorough understanding of an example of the present invention. However, it will be understood by those skilled in the art that the present invention may be practiced without following the specific details of the specific exemplary examples provided herein and may be practiced in numerous other examples or embodiments within the scope of the amended claims. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the present invention.

[0025] The applicant recognized that a deep learning system can be used to plan ablation lines, such as a wide area circumferential ablation (WACA) line, and to check how close the lines actually executed on the left atrium are to the planned lines. In addition, the applicant recognized that the problem of defining ablation lines is a segmentation problem. Thus, the input to such a system may be an anatomical map of a patient's heart, and the same type of deep learning structure may be used both to automatically segment the anatomical map and to define the location of the ablation lines.

[0026] Referring now to FIG. 1A, which shows an exemplary left atrium 2 in a posteroanterior (PA) view, and FIG. 1B, which shows an exemplary ablation line 4 marked on the left atrium 2. The left atrium 2 has a plurality of regions such as an upper wall 3 and a posterior wall 5. On its right side are a right superior pulmonary vein (RSPV) 6 and a right inferior PV (RIPV) 7, and on its left side are a left inferior PV (LIPV) 8 and a left superior PV (LSPV) 9.

[0027] Referring now to FIG. 2, a deep learning system 10 for rendering ablation line(s) recommended on an anatomical structure is shown. The deep learning system 10 includes three components: an ablation line trainer 12 that constructs a trained deep learning unit for an ablation line proposer 13, an ablation line recommender 14 that uses the ablation line proposer 13 to generate ablation line recommendations, and an ablation line guider 16 that provides ablation line guidance during a procedure.

[0028] The ablation line trainer 12 may receive an anatomical map of the heart, which may be generated by using anatomical images generated by using a suitable medical imaging system 20, or by using the fast anatomical mapping (FAM) technology available in the CARTO™ system manufactured by Biosense Webster Inc. (Irvine, Calif.), or by using any other suitable technology, or by using any suitable combination of the above. The map / model of the heart on which the recommended ablation line(s) may be rendered may include any suitable type of three-dimensional (3D) anatomical map generated by using any suitable technology. The map can be obtained from intracardiac ECG (electrocardiogram) and can include both the position information of the catheter moving around the atrium and the ECG measurements such as voltage at each position.

[0029] The ablation line trainer 12 may include an automatic map segmenter 22 that segments a 3D map of the heart received from the imaging system 20 into various parts such as the posterior and anterior walls, superior wall, inferior wall, left lateral, septum, mitral valve, left atrial appendage (LAA), and four pulmonary veins (LSPV, LIPV, RSPV, RIPV), and an ablation line neural network trainer 24 that examines thousands of ablation cases on different maps received from the imaging system 20 along with their different ablation strategies to generate a trained deep learning unit for the ablation line proposer 13.

[0030] The ablation line recommender 14 includes an ablation line proposer 13 that provides an initial recommendation of typical ablation lines during a procedure, and the recommended line(s) are rendered on an anatomical structure, such as a model of the heart segmented by an automated map segmenter 22 and / or on an anatomical map, such that a physician can ablate as close as possible to the proposed line, thus resulting in an ablation line that is as close as possible to the ablation line preferred by a key opinion leader (KOL) on site. However, note that the ablation line proposer 13 can display the recommended ablation lines described herein, whether on an anatomical map or on an ultrasonic image, such as a 4D ultrasonic image, or on any digital depiction of an anatomical structure, together with an ablation line display 25. Thus, as used herein, a "map" of an anatomical structure refers to any such digital depiction on which an ablation line can be projected.

[0031] The ablation line proposer 13 may enable a physician to specify the required ablation line, which may be, for example, a known ablation line such as a WACA, roof line, carina line (right, left), posterior line, anterior line, inferior line, mitral valve line, posterior wall de-bulking, anterior wall de-bulking, inferior wall de-bulking, left lateral de-bulking, LAA isolation, etc.

[0032] The ablation line recommender 14 may also include an ablation line editor 27 to enable a physician to review and modify the selected ablation line according to the physician's requirements, preferences, and findings during the case. For example, a physician may want to edit the initial ablation line such that the initial ablation line includes a trigger observed on the posterior wall.

[0033] The ablation line proposer 13 may segment the WACA circle into sections, and the energy delivery per segment, i.e., may be pre-defined for a specific ablation machine such as the CARTO (trademark) machine, and / or may be configured by a physician before performing ablation, and / or may be for each type of ablation line, and may recommend the recommended ablation index (radiofrequency (RF) or irreversible electroporation (IRE)) and / or the recommended distance between adjacent points.

[0034] The ablation line editor 27 may also enable a physician to update the recommended treatment delivery per segment.

[0035] The ablation line recommender 14 may require the physician to approve the plan after the physician reviews all the parameters.

[0036] The ablation line guider 16 provides ablation line guidance during the procedure. It can receive an anatomical map of the heart from the medical imaging system 20 during the procedure, and can include an automatic map segmenter 22 for indicating various elements of the heart during the procedure, an ablation line guidance unit 30, and a performance reporter 32.

[0037] During the procedure, based on the catheter position and the relevant ablation line from the approved plan, the ablation line guidance unit 30 may present a dynamic display of the next ablation site based on the previous site and the recommended ablation line (approved or edited by the physician). The next ablation site may then be displayed in the forward direction of the line, along with the minimum distance from the proposed line, based on a pre-defined distance, e.g., 4 mm from the previous site.

[0038] The purpose of ablation treatment is to form an effective lesion that can block abnormal electrical pathways causing arrhythmia. For this purpose, the ablation line guidance unit 30 can calculate an ablation index indicating the quality of each high-frequency (RF) ablation lesion during treatment. The ablation index is a function of at least contact force, time, and power, and thus increases as the ablation operation continues. The instantaneous ablation index provides real-time feedback to the physician, enabling the physician to monitor and adjust their technique during treatment.

[0039] Since the physician can define a "required ablation index" for a given ablation line and a given segment within the line, the ablation line guidance unit 30 can present both the "required ablation index" and the instantaneous ablation index of the current site. This may enable the physician to achieve the desired ablation index.

[0040] The performance reporter 32 can provide a report to the physician based on the physician's performance during treatment regarding the ablation plan generated by the ablation line recommender 14. The report can include data such as, for example, the maximum and median deviations [mm] from what was planned for each ablation line, the deviation from the planned position, including the number and percentage of ablation points having a distance greater than (or less than) the planned distance, the deviation from the ablation planned for each ablation line, etc. Note: This report can be particularly useful when training less experienced physicians.

[0041] The ablation line recommender 14 may be configured to always draw the recommended WACA (left-right) ablation line.

[0042] As described above, the applicant has recognized that the same type of deep learning structure may be utilized and the same type of input may be provided to both the automatic map segmenter 22 and the ablation line neural network trainer 24. FIG. 3, which is hereby incorporated by reference, shows a deep learning structure that will be described in more detail below and shows that the input is the anatomical map described above and an adjacent map that will be described below. However, FIG. 3 shows two types of outputs, a segmentation code for the automatic map segmenter 22 and an ablation line code for the ablation line neural network trainer 24.

[0043] The segmentation code can be a set of codes indicating parts of the heart or parts of the atria. For example, there may be 13 different parts of the atria. The ablation line code can be a set of codes indicating the type of ablation line and codes for all regions lacking an ablation line. As will be described below, each training case can include an input map and an output map, and each position is labeled with one of the associated code, the segmentation code for training the automatic map segmenter 22, or the ablation line code for training the ablation line neural network trainer 24.

[0044] The initial dataset may be composed of a sufficient number of paroxysmal and persistent AF cases (e.g., 5000 or more cases) having the anatomical map for each case and at least one ablation line performed in that case. The ablation line is typically one of (1) left WACA, (2) right WACA, (3) inlet PVI, (4) roof line, (5) right carina, (6) left carina, (7) posterior line, (8) inferior line, (9) mitral isthmus line, (10) anterior line, (11) inferior vena cava tricuspid annulus isthmus (CTI) line, (12) superior vena cava isolation (SVCI), (13) LAA isolation.

[0045] To prepare an initial dataset to be used as input for training and testing the automatic map segmenter 22 and the ablation line neural network trainer 24, a specialist doctor (e.g., an on-site KOL) may first add manual annotations of ablation points on the anatomical map of a case as generated by the imaging system 20. Each point is associated with at least one of the ablation lines (1)-(13).

[0046] In the next step, data preprocessing may be performed, whereby each ablation point may be rendered on the anatomical map using a specific radius (e.g., a radius of 5 mm in this example). This is shown in Figure 4 and is hereby briefly referred to. Figure 4 shows a posterior view of an exemplary left atrium 40 after projecting ablation points with a 5 mm radius. Figure 4 shows the left WACA line 42 and the right WACA line 44.

[0047] In another step, each map is inspected and maps having gaps / or unclear lines may be excluded from the training set and the test set.

[0048] Subsequently, the following further data preprocessing may be performed. a. The atrium is transformed into a fixed dimension (e.g., by generating a fixed mesh file with a fixed number of points such as values (e.g., values such as 1000, 2000, 4000, 10000, etc., all values are configurable), e.g., 2000 points and an adjacency matrix of 2000×2000). b. Feature extraction for 3D general segmentation: To construct a deep learning framework for 3D segmentation, the graph data is characterized by improving the local features of each node. This can be achieved by utilizing the spatial and structural features of one triangular neighborhood of each mesh surface, as shown in Figure 5 which shows a mesh surface formed from points 46 having a position (x, y, z) and a normal N thereto, as indicated by a vector (N x , N y , N z ).

[0049] In one triangular mesh, each face contains at most three adjacent faces. The 3D mesh structure data is characterized by M = {V, E, F} of vertices, edges and faces.

[0050] Therefore, one triangular mesh is transformed into a graph G = {feature, adjacency}. The faces {F} on the mesh M have three corresponding nodes (vertices) on the graph {G}. The adjacency matrix {adjacency} is a 6000×6000 matrix in the case of the present invention. The adjacency matrix represents the connections between vertices in M based on {F}. To improve the perception of the local region of the graph nodes, spatial and structural geometric features are used based on a single vertex of the simplified mesh (consisting of 6000 vertices in this example). The feature space is a 6000×9 matrix, and for each vertex (a triangle composed of three vertices with three adjacent faces), the preprocessing utilizes the following features. The normalized (x i , y i , z i ) position of vertex Vi in pi - Euclidean space, where i represents the three indices of the vertex. The spherical coordinates i (φ i , θ i , r i ) of the spi - vertex, where φ i is normalized by π and θ is normalized by 2π. The normal of vertex i defined by ni - (xni, yni, zni).

[0051] c. Network architecture: The automatic map segmenter 22 can utilize a machine learning model composed of two layers. The first layer is composed of six graph convolutional neural network units, and the second layer is a decision layer composed of another GCN with a softmax unit. Each GCN unit gives a prediction of a given vertex to be associated with each of the target categories (anatomical structure or ablation line).

[0052] As shown in Figure 3, the anatomical map and the adjacency matrix function as inputs to all units within the network. Two types of GCNs are used: 1) approximation personalized propagation of neural prediction (APPNP) and 2) auto regressive moving average layer (ARMA). The units are very efficient as softmax layers for increasing the overall performance of the network during the training and cross-validation phases, provide a more flexible frequency response, are more robust to noise, and capture the global graph structure better. The personalized graph neural network is described by Gasteiger, J., Bojchevski, A, and Unnemann, S.G., Predict then propogate: Graph neural networks meet personalized PageRank. Int. Conf. Learn. Represent. (2018). The neural network with graph CNN ARMA filters is described by Maria Bianchi, F., Grattarola, D., Livi, L, and Alippi, C., Graph Neural Networks with Convolutional ARMA Filters (2018). The APPNP unit models the relationship between the GCN network and the PageRank algorithm (originally a link analysis algorithm designed by Google that assigns a numerical weight to each node based on the amount and quality of the links pointing to that node) and extends it to personalized PageRank. It uses information from a large adjustable neighborhood to classify each node. The model is computationally efficient and outperforms some state-of-the-art methods for semi-supervised classification on multiple graphs. APPNP has 13 channels with a multi-layer perceptron (MLP) having two hidden layers with 128 and 64 neurons respectively in each layer.

[0053] The network weights are trained using a weighted categorical cross - entropy loss function. The automatic map segmenter 22 can provide the segmentation code to its segmentation loss function, and the ablation line neural network trainer 24 can provide the ablation line code to its ablation line loss function.

[0054] d. In an exemplary embodiment, the training set includes 5000 maps. Each map is represented using three matrices, namely, a 6000×9 feature matrix, a 6000×6000 adjacency matrix, and a 6000×1 target matrix, one for the segmentation code and one for the ablation code. In the case of the ablation code, 0 represents no line, and non - 0 represents the ablation line ID. e. When the neural network of the automatic map segmenter 22 is trained, the automatic map segmenter 22 may segment the anatomical map of each atrium and provide the segmented map to the ablation line display 25. f. From Segmentation to Lines - The ablation line proposer 13 can create a focused line having potential positions for ablation point - by - point on the anatomical map. For example, the map of FIG. 3 is shown in FIG. 6 with point - by - point ablation indicated by points 50. G. The ablation “points” point - by - point may be updated at any time based on the actual ablation performed by the physician in each case. The mechanism for updating the points may involve minimizing the distance between the planned line and the next dot in the forward direction of the ablation line.

[0055] As described, system 10 can use mapping data and automatic segmentation data to recommend which of the ablation lines is more suitable for a given case. The recommendations may be set based on instructions from a particular research hospital or a particular physician (e.g., a KOL in the field of cardiac ablation), or based on data from a number of physicians and / or research hospitals.

[0056] In an alternative embodiment, the structure used for both the automatic map segmenter 22 and the ablation line trainer 24 may include a classifier that operates on intracardiac ECG data. During intracardiac ECG recording, a catheter is inserted into the heart to measure the electrical activity of the heart. The electrical signals for each anatomical location are represented as voltage amplitudes.

[0057] In this alternative embodiment, the classifier is based on the following papers, namely, Breiman L(2001). “Random Forests”. Machine Learning. 45(1):5 - 32. Bibcode:2001, Ho, Tin Kam(1995). Random Decision Forests(PDF), Proceedings of the 3rd International Conference on Document Analysis and Recognition, Montreal, QC, 14 - 16 August 1995.pp.278 - 282. Archived from the original(PDF)on 17 April 2016.Retrieved 5 June 2016, Any suitable classifier, such as a random forest classifier, a support vector machine (SVM) classifier, or a deep learning classifier, as described in Cortes, Corinna, and Vapnik, Vladimir (1995). "Support-vector networks" (PDF), Machine Learning. 20(3): 273-297. CiteSeerX 10.1.1.15.9362. doi:10.1007 / BF00994018. S2CID 206787478, may be used.

[0058] Each case in the training dataset may be preprocessed with an indication of whether one of a given number of ablation lines (e.g., 14) was performed by a key opinion leader (KOL). Thus, the target is a matrix of size N×14, where N represents the number of cases in the training dataset.

[0059] iii. Input data: The ECG data from the intracardiac ECG is stored as an additional feature matrix of 6000×22. The following trigger and substrate maps from the intracardiac ECG data are used as the feature space. 1) Focus source per "finder" vertex (6000×1) represents the number of S waves per second. 2) "Finder" RAP - Each vertex is represented by the number of rotation patterns per second. 3) CL - Cycle length mapping - Each vertex is represented by the cycle length in milliseconds. 4) STD CL - Cycle length standard deviation - Each vertex is represented by the cycle length standard deviation per second. 5) Spatial - temporal dispersion - Each vertex represents the ratio of spatial - temporal dispersion (among 10 - 30 seconds of the recording). 6) Periodic spatial - temporal dispersion - Each vertex represents the ratio of periodic spatial - temporal dispersion (among 10 - 30 seconds of the recording). 7) Voltage map - Each vertex represents the bipolar amplitude in mV. 8) CLM ROI - Each vertex represents the presence of a cycle length region of interest based on a cutoff where cycle length < minimum CL + 15 milliseconds and STD CL < 15 milliseconds. 9) Segmentation one - hot encoding is the output of the automatic map segmenter 22 and is a matrix of size (6000 x K), where K represents the number of anatomical structures (13 for LA) (output from the segmentation algorithm). Ablation line encoding is the output of the ablation line trainer 24. 10) Global voltage per segmentation region (6000 × 1) mV. 11) Low - voltage zone region, which is a region within the left atrium where the recorded voltage amplitude is relatively low compared to the surrounding region. For example, these may be regions where the voltage is between 0.2 - 0.5 mV. Typically, a scar can be defined as a region where the voltage is less than 0.2 mV. It is understood that these thresholds can vary based on the operator's decision in the treatment.

[0060] The classifier can receive as input a feature matrix that can have dimensions of 6000 × M, where 6000 represents the number of vertices on the map and M represents the number of features (such as focus, substrate - related, ripple, frequency, anatomical location, etc.).

[0061] The target vector has dimensions of K x 1, where K indicates the number of ablation approaches or IDs. In this embodiment, the data may be from cases where the treatment was successful, specifically, cases where there has been no arrhythmia for 12 months. For each of these cases, if a specific ablation approach was performed during the treatment, the K x 1 vector is set to 1.

[0062] The neural network of the classifier can utilize a set of dense nets, where the last layer consists of a softmax layer with K neurons in the output layer. This modification enables the prediction of K ablation approaches.

[0063] The systems and methods described herein may be in various local or distributed computers (e.g., cloud-based processing systems) owned by the applicant and incorporated herein by reference, including but not limited to the forms described by U.S. Patent No. 11,198,004 or U.S. Patent Publication No. 20200352652(A1).

[0064] The applicant recognizes that the actual ablation line data may be noisy and the ablation lines they represent may not be ideal. In another embodiment shown in FIG. 7, referred to herein, the system 10' may further comprise a keypoint filter 23 for reviewing noisy ablation line data and, based on the segmentation map received from the automatic map segmenter 22, finding those ablation points within the ablation line data that can be defined as keypoints (i.e., important points along the ablation lines that should not be missed) defined by experts such as KOLs associated therewith. Typically, these keypoints can be points at the intersection of two parts of the atrium. This is shown in FIG. 8, to which reference is now made. The input 100 to the keypoint filter 23 may be an image or mesh of the left atrium having actual ablation lines that may have been segmented (102) by the automatic map segmenter 22. The keypoint filter 23 can detect keypoints 104 by comparing the ablation points for each case with the intersections on the segmentation map 102 for that case and selecting the ablation point closest to the intersection. The keypoint filter 23 can combine keypoints from multiple cases to generate an ablation line 106 obtained as a result for training, which can be thinner and better defined than otherwise. The keypoint filter 23 may provide the resulting ablation line 106 to the ablation line neural network trainer 24 for training.

[0065] Unless otherwise specified, as is apparent from the foregoing description, throughout this specification, descriptions using terms such as "processing," "computing," "calculating," "determining," etc. refer to actions and / or processes of any type of general-purpose computer, such as a client / server system, a mobile computing device, a smart appliance, a cloud computing unit, or a similar electronic computing device that operates on and / or transforms data in a register and / or memory of a computing system into other data in a memory, register, or other such information storage, transmission, or display device of the computing system.

[0066] Embodiments of the present invention can include an apparatus for performing the operations of this specification. This apparatus may be specially configured for a desired purpose or may comprise a computing device or system typically having at least one processor and at least one memory that is selectively activated or reconfigured by a computer program stored in a computer. As a result, the resulting apparatus can transform a general-purpose computer into an inventive element as described herein when instructed by software. The instructions can define an apparatus of the present invention that operates with a desired computer platform. Such a computer program can be stored in any type of computer-readable storage medium, including but not limited to any type of disk such as an optical disk, magneto-optical disk, read-only memory (ROM), volatile and non-volatile memories, random access memory (RAM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic or optical cards, flash memory, disk on key, or any other type of medium suitable for storing electronic instructions and capable of being coupled to a computer system bus. The computer-readable storage medium may also be implemented in cloud storage.

[0067] Some general-purpose computers can be provided with at least one communication element that enables communication with a data network and / or a mobile communication network.

[0068] The processes and displays presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems can be used with programs in accordance with the teachings herein, or it may prove convenient to construct more specialized apparatus to perform the desired method. The desired structure of various of these systems will become apparent from the following description. Further, embodiments of the invention are not described with reference to any particular programming language. It will be understood that various programming languages can be used to implement the teachings of the invention as described herein.

[0069] Although particular features of the invention have been illustrated and described herein, many modifications, substitutions, changes, and equivalents will occur to those skilled in the art. Accordingly, it is to be understood that the appended claims are intended to cover all such modifications and changes that fall within the true spirit and scope of the invention.

[0070] 〔Embodiments〕 (1) A system for improving a cardiac ablation procedure, the system comprising a recommendation unit configured to provide an initial recommendation for at least one proposed ablation line for an ablation procedure on a patient's anatomical structure, the system comprising the recommendation unit that displays the at least one proposed ablation line on an anatomical map of the anatomical structure. The recommendation unit comprises a first trained machine learning model and a second trained machine learning model, and both the first trained machine learning model and the second trained machine learning model have the same structure. (2) The system according to embodiment 1, further comprising an ablation line editor configured to enable a physician to modify a selected one of the at least one proposed ablation line. (3) The system according to embodiment 1, further comprising a treatment decision maker configured to determine a recommended energy delivery per segment of a selected ablation line. (4) The system according to embodiment 1, further comprising a guidance unit configured to display a next ablation site on the anatomical structure based at least on a selected ablation line and a catheter position within the anatomical structure. (5) The recommendation unit is a map segmenter that segments the anatomical map to output parts of the anatomical structure, the map segmenter utilizing the first trained machine learning model, a map segmenter, and an ablation line trainer that trains the second trained machine learning model to output proposed ablation lines, and an ablation line proposer that proposes the at least one proposed ablation line for the anatomical structure using the second trained machine learning model, the system according to embodiment 1.

[0071] (6) The system according to embodiment 5, wherein the same structure is a graph convolutional neural network (GCN). (7) The system according to embodiment 5, wherein the same structure is a classifier. (8) The system according to embodiment 1, wherein the recommendation unit is further configured to recommend which one of the at least one proposed ablation line is most suitable for treatment of the patient. (9) The at least one proposed ablation line of the system according to any one of embodiments 1 to 8 is a wide antral circumferential ablation (WACA) line. (10) The system according to embodiment 5 further comprises a keypoint filter that compares the actual ablation points of a training case with the intersections or lines of adjacent portions on the segmentation map of the training case and selects the ablation point closest to the intersection as a keypoint.

[0072] (11) A computer-implemented method for improving a cardiac ablation procedure, the method comprising: providing an initial recommendation for at least one proposed ablation line for an ablation procedure on a patient's anatomical structure, the providing including displaying the at least one proposed ablation line on an anatomical map of the anatomical structure, wherein the initial recommendation utilizes a first trained machine learning model and a second trained machine learning model, and both the first trained machine learning model and the second trained machine learning model have the same structure. (12) The method according to embodiment 11 further comprising enabling a physician to modify a selected one of the at least one proposed ablation lines. (13) The method according to embodiment 11 further comprising determining a recommended energy delivery per segment of a selected ablation line. (14) The method according to embodiment 11 further comprising displaying a next ablation site on the anatomical structure based at least on the selected ablation line and a catheter position within the anatomical structure. (15) The providing is Segmenting the anatomical map to output portions of the anatomical structure, wherein the segmenting utilizes the first trained machine learning model, and outputting; Training the second trained machine learning model to output a proposed ablation line; Proposing, utilizing the second trained machine learning model, at least one proposed ablation line for the anatomical structure, the method of embodiment 11 comprising:

[0073] (16) The method of embodiment 15, wherein the same structure is a graph convolutional neural network (GCN). (17) The method of embodiment 15, wherein the same structure is a classifier. (18) The method of embodiment 11, wherein providing further comprises recommending which of the at least one proposed ablation line is most suitable for treatment of the patient. (19) The method according to any one of embodiments 11 to 18, wherein the at least one proposed ablation line is a wide area circumferential ablation (WACA) line. (20) The method of embodiment 15, further comprising comparing actual ablation points of a training case with intersections or lines of adjacent portions on a segmented map of the training case, and selecting, as key points, the ablation points closest to the intersections.

Claims

1. A system for improving cardiac ablation procedures, the system comprising a recommendation unit configured to provide an initial recommendation for at least one proposed ablation line for an ablation procedure on a patient's anatomical structure, the system displaying the at least one proposed ablation line on an anatomical map of the anatomical structure, the system comprising a recommendation unit; the recommendation unit comprises a first trained machine learning model and a second trained machine learning model, both the first trained machine learning model and the second trained machine learning model having the same structure, the system.

2. The system according to claim 1, further comprising an ablation line editor configured to enable a physician to modify a selected one of the at least one proposed ablation lines.

3. The system according to claim 1, further comprising a treatment decision maker configured to determine a recommended energy delivery per segment of a selected ablation line.

4. The system according to claim 1, further comprising a guidance unit configured to display a next ablation site on the anatomical structure based at least on a selected ablation line and a catheter position within the anatomical structure.

5. The recommendation unit a map segmenter that segments the anatomical map to output a portion of the anatomical structure, the map segmenter utilizing the first trained machine learning model, the map segmenter; an ablation line trainer that trains the second trained machine learning model to output a proposed ablation line; An ablation line proposer that proposes the at least one proposed ablation line for the anatomical structure by using the second trained machine learning model, and the system according to claim 1, comprising:

6. The system according to claim 5, wherein the same structure is a graph convolutional neural network (GCN).

7. The system according to claim 5, wherein the same structure is a classifier.

8. The system according to claim 1, wherein the recommendation unit is further configured to recommend which of the at least one proposed ablation line is most suitable for treatment of the patient.

9. The system according to any one of claims 1 to 8, wherein the at least one proposed ablation line is a wide area circumferential ablation (WACA) line.

10. The system according to claim 5, further comprising a keypoint filter that compares the actual ablation points of the training case with the intersections or lines of adjacent portions on the segmentation map of the training case, and selects the ablation point closest to the intersection as the keypoint.

11. A computer-implemented method for improving cardiac ablation procedures, the method comprising: Providing an initial recommendation for at least one proposed ablation line for an ablation procedure on a patient's anatomical structure, including displaying the at least one proposed ablation line on an anatomical map of the anatomical structure, the providing including, The initial recommendation uses a first trained machine learning model and a second trained machine learning model, and both the first trained machine learning model and the second trained machine learning model have the same structure, a computer-implemented method.

12. The method according to claim 11, further comprising enabling a physician to modify a selected one of the at least one proposed ablation line.

13. The method according to claim 11, further comprising determining a recommended energy delivery per segment of a selected ablation line.

14. The method according to claim 11, further comprising displaying a next ablation site on the anatomical structure based at least on a selected ablation line and a catheter position within the anatomical structure.

15. said providing is segmenting the anatomical map to output portions of the anatomical structure, said segmenting utilizing the first trained machine learning model to output, training the second trained machine learning model to output proposed ablation lines, proposing at least one proposed ablation line for the anatomical structure, said proposing utilizing the second trained machine learning model, comprising the method according to claim 11.

16. The method according to claim 15, wherein the same structure is a graph convolutional neural network (GCN).

17. The method according to claim 15, wherein the same structure is a classifier.

18. The method according to claim 11, wherein said providing further comprises recommending which of the at least one proposed ablation line is most suitable for treatment of the patient.

19. The method according to any one of claims 11 to 18, wherein the at least one proposed ablation line is a wide area circumferential ablation (WACA) line.

20. The method according to claim 15, further comprising: comparing an actual ablation point of a training case with an intersection or a line of adjacent portions on a segmentation map of the training case; and selecting, as a key point, an ablation point closest to the intersection.

Images