System and method for identifying regions of heterogeneous cardiac electrical propagation

Abstract

A computer-implemented method (e.g., via an electroanatomical mapping system) of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy includes receiving a cardiac activation map, generating an activation direction entropy map from the activation map, and generating an output based on the activation direction entropy map, wherein the output includes an identification of one or more regions of heterogeneous cardiac electrical propagation. The activation direction entropy values may be based upon a distribution of activation directions within a sub-region of the patient's heart and may further be normalized to an entropy score between 0 and 1. The output may be generated using a volumetric convolutional neural network algorithm or a particle swarm optimization algorithm.

Description

Attorney Docket No. 15908WOO 1 / 82410.1297SYSTEM AND METHOD FOR IDENTIFYING REGIONS OF HETEROGENEOUS CARDIAC ELECTRICAL PROPAGATIONCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of United States provisional application no.63 / 748,224, filed 22 January 2025, which is hereby incorporated by reference as though fully set forth herein.FIELD

[0002] The present disclosure relates generally to electrophysiology procedures. In particular, the present disclosure relates to systems, apparatuses, and methods for identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via ablation therapy.BACKGROUND

[0003] Electrophysiological mapping, and more particularly electrocardiographic mapping, is a part of numerous cardiac diagnostic and therapeutic procedures. Such mapping includes measuring the electrical activity of the heart over time using surface leads and / or intracardiac measurement electrodes.

[0004] In conjunction with electrophysiological mapping for diagnosis, ablation therapy may be used to treat various conditions afflicting the human anatomy. For instance, ablation therapy is often used in the treatment of various cardiac arrhythmias. In this regard, it is known that regions of heterogeneous cardiac electrical propagation may be desirable targets for the application of ablation therapy for arrhythmia termination. Thus, electrophysiological mapping may be used to identify regions of heterogeneous cardiac electrical propagation, and ablation therapy may be applied at the identified site(s) to terminate an arrhythmia.BRIEF SUMMARY

[0005] The instant disclosure provides a computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy. The method includes: receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system; generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart fromAttorney Docket No. 15908WOO 1 / 82410.1297the cardiac activation map; and generating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit comprising a classifier trained with data from a plurality of subjects that received ablation therapy for treatment of arrhythmia, wherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

[0006] The activation direction entropy map may include a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart. Each activation direction entropy value of the plurality of activation direction entropy values may be based upon a distribution of activation directions for a respective subregion of the plurality of sub-regions within the portion of the patient’s heart. For instance, each activation direction entropy value of the plurality of activation direction entropy values may be computed according to an equationwhere H is the activation direction entropy value for the respective sub-region and Pi is a proportion of activation directions for the respective sub-region that fall within a bin z of a preset number N of directional bins. It is also contemplated that each activation direction entropy value of the plurality of activation direction entropy values may further be normalized to an entropy score between 0 and 1 according to an equation Score = ( — ) ■ The preset number TV may be user-defined via a user input received at the electroanatomical mapping system.

[0007] The sub-region size may also be user-defined via a user input received at the electroanatomical mapping system.

[0008] The irregular propagation processing unit may include a volumetric convolutional neural network algorithm.

[0009] The method optionally includes a step of the electroanatomical mapping system downsampling the cardiac activation map, such as by voxelizing the cardiac activation map, prior to generating the activation direction entropy map.

[0010] Also disclosed herein is a computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application ofAttorney Docket No. 15908WOO 1 / 82410.1297ablation therapy, including the following steps: receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system; generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; and generating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit, wherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

[0011] The activation direction entropy map may include a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart. Each activation direction entropy value of the plurality of activation direction entropy values may be based upon a distribution of activation directions for a respective subregion of the plurality of sub-regions within the portion of the patient’s heart. For instance, each activation direction entropy value of the plurality of activation direction entropy values may be computed according to an equationwhere H is the activation direction entropy value for the respective sub-region and Pt is a proportion of activation directions for the respective sub-region that fall within a bin z of a preset number N of directional bins. It is also contemplated that each activation direction entropy value of the plurality of activation direction entropy values may further be normalized to an entropy / Hscore between 0 and 1 according to an equation Score = ( — ) ■ Optionally, the electroanatomical mapping system may receive a user input to define the preset number / .

[0012] The electroanatomical mapping system may also receive a user input to define a subregion size.

[0013] The irregular propagation processing unit may include a volumetric convolutional neural network algorithm. Alternatively, the irregular propagation processing unit may include a particle swarm optimization algorithm.

[0014] The instant disclosure also provides a computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia viaAttorney Docket No. 15908WOO 1 / 82410.1297application of ablation therapy. The method includes: receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system; generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; and generating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit comprising a particle swarm optimization algorithm, wherein each particle evaluated by the particle swarm optimization algorithm includes an activation direction entropy value associated with a sub-region of the portion of the patient’s heart and a position of the subregion of the portion of the patient’s heart, and wherein the output includes an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

[0015] The activation direction entropy value may be based upon a distribution of activation directions for the sub-region of the portion of the patient’s heart. For instance, the activation direction entropy value may be computed according to an equationwhere H is the activation direction entropy value and Pi is a proportion of activation directions for the sub-region of the portion of the patient’s heart that fall within the a bin i of a preset number N of directional bins. The activation direction entropy value may further be normalizedto an entropy score between 0 and 1 according to an equation Scor

[0016] Another aspect of the instant disclosure relates to an electroanatomical mapping system for identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy. The electroanatomical mapping system includes an irregular propagation processing unit configured to: receive a cardiac activation map of a portion of a patient’s heart; generate an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; and generate an output, based on the activation direction entropy map for the portion of the patient’s heart, wherein the output includes an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.Attorney Docket No. 15908WOO 1 / 82410.1297

[0017] The activation direction entropy map typically includes a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart. Each activation direction entropy value of the plurality of activation direction entropy values may be based upon a distribution of activation directions for a respective subregion of the plurality of sub-regions within the portion of the patient’s heart. For instance, each activation direction entropy value of the plurality of activation direction entropy values may be computed according to an equationwhere H is the activation direction entropy value and Pi is a proportion of activation directions for the sub-region of the portion of the patient’s heart that fall within the a bin i of a preset number N of directional bins. Each activation direction entropy value of the plurality of activation direction entropy values may further be normalized to an entropy score between 0 and1 according to an equation

[0018] The irregular propagation processing unit may include classifier trained with data from a plurality of subjects that received ablation therapy for treatment of arrhythmia. The irregular propagation processing unit may also include a volumetric convolutional neural network algorithm.

[0019] Alternatively, the irregular propagation processing unit may include a particle swarm optimization algorithm.

[0020] The foregoing and other aspects, features, details, utilities, and advantages of the present invention will be apparent from reading the following description and claims, and from reviewing the accompanying drawings.BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure l is a schematic diagram of an exemplary electroanatomical mapping system.

[0022] Figure 2 is a flowchart of representative steps that can be carried out according to aspects of the instant disclosure.

[0023] Figure 3 is a representative activation direction map.Attorney Docket No. 15908WOO 1 / 82410.1297

[0024] Figure 4A is a simulated activation direction map for a sub-region that exhibits perfectly homogenous electrical propagation.

[0025] Figure 4B is a polar histogram of the distribution of activation directions of the simulated activation direction map of Figure 4A.

[0026] Figure 5A is a simulated activation direction map for a sub-region that exhibits semiuniform activation.

[0027] Figure 5B is a polar histogram of the distribution of activation directions of the simulated activation direction map of Figure 5 A.

[0028] Figure 6A is a simulated activation direction map for a sub-region that exhibits focal breakout.

[0029] Figure 6B is a polar histogram of the distribution of activation directions of the simulated activation direction map of Figure 6A.

[0030] Figure 7 is a voxelized (downsampled) version of the representative activation direction map of Figure 3.

[0031] Figure 8 is a representative volumetric convolutional neural network architecture.

[0032] While multiple embodiments are disclosed, still other embodiments of the present disclosure will become apparent to those skilled in the art from the following detailed description, which shows and describes illustrative embodiments. Accordingly, the drawings and detailed description are to be regarded as illustrative in nature and not restrictive.DETAILED DESCRIPTION

[0033] The instant disclosure provides systems, apparatuses, and methods for identifying regions of heterogeneous cardiac electrical propagation. For purposes of illustration, aspects of the disclosure will be described with reference to the use of a suitably-programmed electroanatomical mapping system, such as the EnSite Precision™ cardiac mapping system (Abbott Laboratories, Abbott Park, IL), to identify regions of heterogeneous cardiac electrical propagation for the termination of arrythmia via application of ablation therapy. Those of ordinary skill in the art will understand, however, how to apply the teachings herein to good advantage in other contexts and / or with respect to other devices and cardiac rhythm disorders.Attorney Docket No. 15908WOO 1 / 82410.1297

[0034] Figure 1 shows a schematic diagram of an exemplary electroanatomical mapping system 8 for conducting cardiac electrophysiology studies by navigating a cardiac catheter and measuring electrical activity occurring in a heart 10 of a patient 11 and mapping the electrical activity and / or information related to or representative of the electrical activity so measured. System 8 can be used, for example, to create an anatomical model of the patient’s heart 10 using one or more electrodes. System 8 can also be used to measure electrophysiology data at a plurality of points along a cardiac surface and store the measured data in association with location information for each measurement point at which the electrophysiology data was measured, for example to create a diagnostic data map of the patient’s heart 10.

[0035] As one of ordinary skill in the art will recognize, system 8 determines the location, and in some aspects the orientation, of objects, typically within a three-dimensional space, and expresses those locations as position information determined relative to at least one reference. This is referred to herein as “localization.”

[0036] For simplicity of illustration, the patient 11 is depicted schematically as an oval. In the embodiment shown in Figure 1, surface electrodes (e.g., patch electrodes) 12, 14, 16, 18, 19, and 22 are shown applied to a surface of the patient 11, pairwise defining three generally orthogonal axes, referred to herein as an x-axis (12, 14), a y-axis (18, 19), and a z-axis (16, 22). In other embodiments, the surface electrodes could be positioned in other arrangements, for example multiple electrodes on a particular body surface. As a further alternative, the electrodes do not need to be on the body surface, but could be positioned internally to the body or on an external frame.

[0037] In Figure 1, the x-axis surface electrodes 12, 14 are applied to the patient along a first axis, such as on the lateral sides of the thorax region of the patient (e.g., applied to the patient’s skin underneath each arm) and may be referred to as the Left and Right electrodes. The y-axis electrodes 16, 22 are applied to the patient along a second axis generally orthogonal to the x-axis, along the sternum and spine of the patient in the thorax region, and may be referred to as the Chest and Back electrodes. The z-axis electrodes 18, 19 are applied along a third axis generally orthogonal to both the x-axis and the y-axis, such as along the inner thigh and neck regions of the patient, and may be referred to as the Left Leg and Neck electrodes. The heart 10 lies between these pairs of surface electrodes 12 / 14, 16 / 22, and 18 / 19.Attorney Docket No. 15908WOO 1 / 82410.1297

[0038] Each surface electrode can measure multiple signals. For example, in embodiments of the disclosure, each surface electrode can measure three resistance (impedance) signals and three reactance signals. These signals can, in turn, be grouped into three resistance / reactance signal pairs. One resistance / reactance signal pair can reflect driven values, while the other two resistance / reactance signal pairs can reflect non-driven values (e.g., measurements of the electric field generated by other driven pairs in a manner similar to that described below for electrodes 17).

[0039] An additional surface reference electrode (e.g., a “belly patch”) 21 provides a reference and / or ground electrode for the system 8. The belly patch electrode 21 may be an alternative to a fixed intra-cardiac electrode 31, described in further detail below. In alternative embodiments where system 8 is capable of magnetic field-based localization instead of or in addition to impedance-based localization, the surface electrode 21 can alternatively or additionally include a magnetic patient reference sensor - anterior (“PRS-A”) positioned on the patient’s chest.

[0040] It should be appreciated that patient 11 may also have most or all of the conventional electrocardiogram (“ECG” or “EKG”) system leads in place. In certain embodiments, for example, a standard set of 12 ECG leads may be utilized for sensing electrocardiograms on the patient’s heart 10. This ECG information is available to system 8 (e.g., it can be provided as input to computer system 20). Insofar as ECG leads are well understood, and for the sake of clarity in the figures, only a single lead 6 and its connection to computer 20 is illustrated in Figure 1.

[0041] A representative catheter 13 having at least one electrode 17 is also shown. This representative catheter electrode 17 is referred to as the “roving electrode,” “moving electrode,” or “measurement electrode” throughout the specification. Typically, multiple electrodes 17 on catheter 13, or on multiple such catheters, will be used. In one embodiment, for example, the system 8 may comprise sixty-four electrodes on twelve catheters disposed within the heart and / or vasculature of the patient. In other embodiments, system 8 may utilize a single catheter that includes multiple splines, each of which in turn includes multiple electrodes, such as the Advisor™ HD Grid Mapping Catheter, Sensor Enabled™ (Abbott Laboratories; Abbott Park, Illinois). These embodiments are merely exemplary, however, and any number of electrodes and / or catheters may be used.Attorney Docket No. 15908WOO 1 / 82410.1297

[0042] As the ordinarily-skilled artisan will appreciate, electrodes 17 carried in the distal portion of catheter 13 can be used to measure unipolar electrograms (e.g., between any electrode 17 and a patch electrode 12, 14, 16, 18, 19, 21, or 22). Unipolar electrograms have the advantage of being orientation-independent (that is, a given electrode 17 will measure substantially the same unipolar electrogram regardless of the orientation of catheter 13 relative to the cardiac surface). On the other hand, unipolar electrograms often include not only the component of interest (e.g., a near-field cardiac activation component), but also various far-field noise components. These far-field noise components include, but are not limited to, far-field cardiac activation components, powerline noise components, patient respiration components, patient motion components, and cardiac motion components. Unipolar electrograms also ignore local directional information in the form of the ion currents that produce electrograms.

[0043] It may also be desirable to measure bipolar electrograms, which are less susceptible to far-field noise than unipolar electrograms and incorporate directional information along their axes. As those of ordinary skill in the art will recognize, any two neighboring electrodes 17 on catheter 13 define a bipole. Any bipole can, in turn, be used to generate a bipolar electrogram according to techniques that will be familiar to those of ordinary skill in the art. Those of ordinary skill in the art will recognize, however, that bipolar electrograms are orientationdependent (that is, they will change as the orientation of catheter 13 relative to the cardiac surface changes).

[0044] There are also extant techniques that allow bipolar electrograms to be combined to generate electrograms for any orientation of catheter 13 relative to the cardiac surface without physically changing the orientation of catheter 13. These orientation-independent techniques are often referred to as “omnipolar” techniques. In turn, the computed electrogram signals that result from the application of such techniques can be referred to as “omnipolar electrograms” or “virtual bipolar electrograms.” These omnipolar electrograms can be thought of as the bipolar electrogram that would be seen by an “omnipole” or “virtual bipole” having its “omnipole orientation” or “virtual bipole orientation” at a particular angle relative to the cardiac anatomy (e.g., the cardiac surface).

[0045] In embodiments of the disclosure, omnipolar techniques can be applied to the unipolar and / or bipolar electrograms that can be measured by a clique of three or more electrodes 17Attorney Docket No. 15908WOO 1 / 82410.1297carried by catheter 13. As used herein, the term “clique” refers to a group or cluster of neighboring or closely-spaced electrodes 17 on catheter 13.

[0046] For instance, each individual electrode 17 within a clique can be used to measure a unipolar electrogram; those of ordinary skill in the art will be familiar with various techniques suitable for defining a representative unipolar electrogram for the clique as a whole (e.g., an average of three individual unipolar electrograms measured by the three electrodes of the clique). Likewise, each pair of electrodes 17 within a clique can be used to measure a bipolar electrogram. Details of computing E-field loops from these various unipolar and bipolar electrograms, and for generating omnipolar electrograms therefrom, are described in United States patent nos. 10,758,137 and 10,194,994; international patent application publication no. WO 2015 / 130824; and Deno et al., Orientation-Independent Catheter-Based Characterization of Myocardial Activation, IEEE Transactions on Biomedical Engineering, Vol. 64, No. 5, 1067- 1077 (May 2017) (“Deno”). Each of the foregoing is hereby incorporated by reference as though fully set forth herein.

[0047] As will be apparent from the foregoing description, catheter 13 can be used to simultaneously collect a plurality of electrophysiology data points for the various unipoles and bipoles defined by electrodes 17 thereon. Each such electrophysiology data point includes both localization information (e.g., position of a unipole; position and orientation of a selected bipole or electrode clique) and corresponding electrogram signals (e.g., unipolar, bipolar, and / or omnipolar electrograms).

[0048] Catheter 13 (or multiple such catheters) are typically introduced into the heart and / or vasculature of the patient via one or more introducers and using familiar procedures. Indeed, various approaches to introduce catheter 13 into a patient’s heart, such as transseptal approaches, will be familiar to those of ordinary skill in the art, and therefore need not be further described herein.

[0049] Since each electrode 17 lies within the patient, location data may be collected simultaneously for each electrode 17 by system 8. Similarly, each electrode 17 can be used to gather electrophysiological data from the cardiac surface (e.g., endocardial electrograms). The ordinarily skilled artisan will be familiar with various modalities for the acquisition and processing of electrophysiology data points (including, for example, both contact and non-Attorney Docket No. 15908WOO 1 / 82410.1297contact electrophysiological mapping), such that further discussion thereof is not necessary to the understanding of the techniques disclosed herein. Likewise, various techniques familiar in the art can be used to generate graphical representations of cardiac geometry (e.g., cardiac surface models) and / or cardiac electrical activity (e.g, electrophysiology maps) from the plurality of electrophysiology data points. Moreover, insofar as the ordinarily skilled artisan will appreciate how to create electrophysiology maps from electrophysiology data points, the aspects thereof will only be described herein to the extent necessary to understand the present disclosure.

[0050] Returning now to Figure 1, in some embodiments, an optional fixed reference electrode 31 (e.g, attached to a wall of the heart 10) is shown on a second catheter 29. For calibration purposes, this electrode 31 may be stationary (e.g., attached to or near the wall of the heart) or disposed in a fixed spatial relationship with the roving electrodes (e.g, electrodes 17), and thus may be referred to as a “navigational reference” or “local reference.” The fixed reference electrode 31 may be used in addition or alternatively to the surface reference electrode 21 described above. In many instances, a coronary sinus electrode or other fixed electrode in the heart 10 can be used as a reference for measuring voltages and displacements; that is, as described below, fixed reference electrode 31 may define the origin of a coordinate system.

[0051] Each surface electrode is coupled to a multiplex switch 24, and the pairs of surface electrodes are selected by software running on a computer 20, which couples the surface electrodes to a signal generator 25. Alternately, switch 24 may be eliminated and multiple (e.g., three) instances of signal generator 25 may be provided, one for each measurement axis (that is, each surface electrode pairing).

[0052] The computer 20 may comprise, for example, a conventional general-purpose computer, a special-purpose computer, a distributed computer, or any other type of computer. The computer 20 may comprise one or more processors 28, such as a single central processing unit (“CPU”), or a plurality of processing units, commonly referred to as a parallel processing environment, which may execute instructions to practice the various aspects described herein.

[0053] Generally, three nominally orthogonal electric fields are generated by a series of driven and sensed electric dipoles (e.g, surface electrode pairs 12 / 14, 16 / 22, and 18 / 19) in order to realize catheter navigation in a biological conductor. Alternatively, these orthogonal fields can be decomposed and any pairs of surface electrodes can be driven as dipoles to provide effectiveAttorney Docket No. 15908WOO 1 / 82410.1297electrode triangulation. Likewise, the electrodes 12, 14, 18, 19, 16, and 22 (or any number of electrodes) could be positioned in any other effective arrangement for driving a current to or sensing a current from an electrode in the heart. For example, multiple electrodes could be placed on the back, sides, and / or belly of patient 11. Additionally, such non-orthogonal methodologies add to the flexibility of the system. For any desired axis, the potentials measured across the roving electrodes resulting from a predetermined set of drive (source-sink) configurations may be combined algebraically to yield the same effective potential as would be obtained by simply driving a uniform current along the orthogonal axes.

[0054] Thus, any two of the surface electrodes 12, 14, 16, 18, 19, 22 may be selected as a dipole source and drain with respect to a ground reference, such as belly patch 21, while the unexcited electrodes measure voltage with respect to the ground reference. The roving electrodes 17 placed in the heart 10 are exposed to the field from navigational currents and are measured with respect to ground, such as belly patch 21. In practice the catheters within the heart 10 may contain multiple electrodes and each electrode potential may be measured. As previously noted, at least one electrode may be fixed to the interior surface of the heart to form a fixed reference electrode 31, which is also measured with respect to ground, such as belly patch 21, and which may be defined as the origin of the coordinate system relative to which system 8 measures positions. Data from the surface electrodes and / or the internal electrodes may be used to determine the location of the roving electrodes 17 within heart 10.

[0055] The measured voltages may be used by system 8 to determine the location in three-dimensional space of the electrodes inside the heart, such as roving electrodes 17 relative to a reference location, such as reference electrode 31. That is, the voltages measured at reference electrode 31 may be used to define the origin of a coordinate system, while the voltages measured at roving electrodes 17 may be used to express the location of roving electrodes 17 relative to the origin. In some embodiments, the coordinate system is a three-dimensional (x, y, z) Cartesian coordinate system, although other coordinate systems, such as polar, spherical, and cylindrical coordinate systems, are contemplated.

[0056] As should be clear from the foregoing discussion, the data used to determine the location of the electrode(s) within the heart is measured while the surface electrode pairs impress an electric field on the heart. The electrode data may also be used to create a respirationAttorney Docket No. 15908WOO 1 / 82410.1297compensation value used to improve the raw location data for the electrode locations as described, for example, in United States Patent No. 7,263,397, which is hereby incorporated herein by reference in its entirety. The electrode data may also be used to compensate for changes in the impedance of the body of the patient as described, for example, in United States Patent No. 7,885,707, which is also incorporated herein by reference in its entirety.

[0057] Therefore, in one representative embodiment, system 8 first selects a set of surface electrodes and then drives them with current signals. While the current pulses are being delivered, electrical activity, such as the voltages measured with at least one of the remaining surface electrodes and in vivo electrodes, is measured and stored. Compensation for artifacts, such as respiration and / or impedance shifting, may be performed as indicated above.

[0058] In aspects of the disclosure, system 8 can be a hybrid system that incorporates both impedance-based (e.g., as described above) and magnetic-based localization capabilities. Thus, for example, system 8 can also include a source 30 coupled to one or more magnetic field generators. In the interest of clarity, only two magnetic field generators 32 and 33 are depicted in Figure 1, but it should be understood that additional magnetic field generators (e.g., a total of six magnetic field generators, defining three generally orthogonal axes analogous to those defined by patch electrodes 12, 14, 16, 18, 19, and 22) can be used without departing from the scope of the present teachings. Likewise, those of ordinary skill in the art will appreciate that, for purposes of localizing catheter 13 within the magnetic fields so generated, can include one or more magnetic localization sensors (e.g., coils).

[0059] In some embodiments, system 8 is the EnSite™ X, EnSite™ Velocity™, or EnSite Precision™ electrophysiological mapping and visualization system of Abbott Laboratories (Abbott Park, Illinois). Other localization systems, however, may be used in connection with the present teachings, including for example the RHYTHMIA HDX™ mapping system of Boston Scientific Corporation (Marlborough, Massachusetts), the CARTO navigation and location system of Biosense Webster, Inc. (Irvine, California), the AURORA® system of Northern Digital Inc. (Waterloo, Ontario), Stereotaxis, Inc.’s NIOBE® Magnetic Navigation System (St. Louis, Missouri), the Affera™ Mapping and Ablation System of Medtronic pic (Minneapolis, Minnesota), as well as MediGuide™ Technology from Abbott Laboratories.Attorney Docket No. 15908WOO 1 / 82410.1297

[0060] The localization and mapping systems described in the following patents (all of which are hereby incorporated by reference in their entireties) can also be used with the present invention: United States Patent Nos. 6,990,370; 6,978,168; 6,947,785; 6,939,309; 6,728,562; 6,640,119; 5,983,126; and 5,697,377.

[0061] Aspects of the disclosure relate to identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy. System 8 (e.g., computer system 20) can therefore include a hardware- and / or software-based irregular propagation processing unit 58 as further described below.

[0062] Exemplary methods according to aspects of the instant disclosure will be explained with reference to the flowchart 200 of representative steps presented as Figure 2. In some embodiments, for example, flowchart 200 may represent several exemplary steps that can be carried out by electroanatomical mapping system 8 of Figure 1 (e.g., by processor 28 and / or irregular propagation processing unit 58). It should be understood that the representative steps described below can be hardware-implemented (e.g, as an application-specific integrated circuit (ASIC) or field-programmable gate array (FPGA)), software-implemented (e.g., as a series of programming instructions executed on one or more processing units), or implemented in a combination of hardware and software.

[0063] As described above, system 8 can localize electrodes 17 and / or multi -el ectrode catheter 13 using techniques that will be familiar to those of ordinary skill in the art. Thus, electrophysiological data (e.g, unipolar electrograms, bipolar electrograms, and / or omnipolar electrograms) can be associated with the cardiac location at which they were measured and / or computed. The association of electrophysiological data (e.g, electrogram signals) with the cardiac location at which such data were gathered (or computed) is commonly referred to as an “electrophysiology data point” (or “EP data point”). A collection or set of EP data points can be used to define an “electrophysiology map,” which, in turn, can be output to display 23 of system 8.

[0064] Various electrophysiology maps will be known to those of skill in the art. In embodiments of the instant disclosure, however, system 8 receives a cardiac activation map of a portion of the patient’s heart in block 202. Insofar as cardiac activation maps will be familiar to those of ordinary skill in the art, there is no need for a detailed discussion thereof in the instantAttorney Docket No. 15908WOO 1 / 82410.1297disclosure. To aid in understanding, and for purposes of illustration only, Figure 3 depicts a representative activation direction map 300, which is one example of a cardiac activation map to which the teachings herein may be applied.

[0065] In block 204, system 8 uses the cardiac activation map received in block 202 to generate an activation direction entropy map for the portion of the patient’s heart. As used herein, the term “activation direction entropy map” includes a plurality of activation direction entropy values, each associated with a corresponding sub-region of the portion of the patient’s heart.

[0066] The size of the sub-region may be user selected or may be determined based upon the size of catheter 13. For instance, where catheter 13 is Abbott’s Advisor™ FID Grid Mapping Catheter, Sensor Enabled™, each sub-region may have dimensions of about 10 mm by about 10 mm, though these dimensions are only illustrative and should not be regarded as limiting the present disclosure.

[0067] The activation direction entropy values may be based upon a distribution of activation directions for a respective sub-region. Such activation direction entropy values can be understood with reference to Figures 4A, 4B, 5 A, 5B, 6A, and 6B.

[0068] Figure 4A is a simulated activation direction map for a sub-region that exhibits perfectly homogenous electrical propagation. Figure 4B is a polar histogram of the activation directions of Figure 4A.

[0069] Figure 5A is a simulated activation direction map for a sub-region that exhibits semiuniform activation. Figure 5B is a polar histogram of the activation directions of Figure 5 A.

[0070] Figure 6A is a simulated activation direction map for a sub-region that exhibits focal breakout. Figure 6B is a polar histogram of the activation directions of Figure 6A.

[0071] As illustrated in Figures 4B, 5B, and 6B, it is contemplated that the activation directions for a given sub-region may be discretized into A bins, each of which may have a bin width of 360 degrees / N (e.g., N bins of uniform width and equal spacing). N may be preset or user-defined. For instance, in Figures 4B, 5B, and 6B, N= 36, such that each bin has a width of 10 degrees (c. ., 335 degrees to 5 degrees; 5 degrees to 15 degrees; 15 degrees to 25 degrees; and so on), though this arrangement is merely illustrative rather than limiting.

[0072] The radial heights of the bins in the polar histograms represent the proportion of activation directions for the sub-region that fall within the corresponding bin (that is, within theAttorney Docket No. 15908WOO 1 / 82410.1297range of activation directions for the bin). For example, in Figure 4B, 100% of the activation directions for the sub-region fall within the bin that extends from 335 degrees to 5 degrees; in Figure 5B, 15% of the activation directions for the sub-region fall within the bin that extends from 335 degrees to 5 degrees; and in Figure 6B, about 4% of the activation directions for the sub-region fall within the bin that extends from 335 degrees to 5 degrees.

[0073] Once the activation directions for the sub-region have been discretized into bins as described above, an activation direction entropy value H for the sub-region can be computed according to an equationwhere Pi is the proportion of activation directions for the sub-region that fall within the itflbin of the A bins.

[0074] The equation above yields a minimum activation direction entropy value (all activation directions in the same bin) of 0 and a maximum activation direction entropy value (activation directions uniformly distributed across all bins) of ln(N). Referring again to Figures 4A, 4B, 5A, 5B, 6A, and 6B, the activation direction entropy value of the simulated activation direction map of Figure 4A may be about 0 (perfectly homogeneous electrical propagation); the activation direction entropy value of the simulated activation direction map of Figure 5 A may be about 2.13; and the activation direction entropy value of the simulated activation direction map of Figure 6A may be about 3.46.

[0075] To further aid a practitioner in rapidly understanding the meaning of a given activation direction entropy value, the activation direction entropy value may be normalized to an entropy score between 0 and 1 according to an equation Score = Thus, for example, the entropyscore of the simulated activation direction map of Figure 4A may be about 0 (perfectly homogenous electrical propagation); the entropy score of the simulated activation direction map of Figure 5A may be about 0.35; and the entropy score of the simulated activation direction map of Figure 6A may be about 0.93 (focal breakout).

[0076] Those of ordinary skill in the art will appreciate that the cardiac activation map received in block 202 may include an extremely large number (e.g., hundreds of millions) of EP dataAttorney Docket No. 15908WOO 1 / 82410.1297points. Generating an activation direction map from this many EP data points may be computationally expensive and / or time consuming to an extent that real-time analysis according to the instant teachings is not feasible.

[0077] Thus, in an optional block 206 that precedes block 204, system 8 may downsample the cardiac activation map. One suitable downsampling algorithm is a voxelization algorithm that may reduce the cardiac activation map to a 32 by 32 by 32 voxel grid (e.g., about 32,000 points) that can be analyzed in real-time or near real-time. Figure 7 illustrates the voxelization 700 of cardiac activation map 300 in Figure 3.

[0078] Returning to flowchart 200, and in particular to block 208, system 8 may, using the activation direction entropy map from block 204, generate an output that identifies one or more regions of heterogeneous electrical propagation in the portion of the patient’s heart. Thus, rather than requiring a practitioner to manually identify regions of heterogeneous electrical propagation via visual interpretation of cardiac activation maps, system 8 can algorithmically identify such locations for the practitioner’s consideration via operation of processor 28 and irregular propagation processing unit 58 as described in further detail below.

[0079] According to aspects of the disclosure, block 210 may use a trained machine learning model applied to the activation direction entropy map to automatically identify regions of heterogeneous electrical propagation. This model is referred to herein as a “classifier.”

[0080] In addition to identifying regions of heterogeneous electrical propagation, the classifier can also output a probability that ablation at an identified location will terminate an arrythmia. This probability is referred to herein as an “arrythmia termination confidence value.” In some embodiments of the disclosure, the classifier can identify a region as exhibiting heterogeneous electrical propagation when the region has an arrhythmia termination confidence value above a preset threshold, such as about 80%, though it is contemplated that this threshold may be varied at a practitioner’s option.

[0081] The classifier can employ a volumetric convolutional neural network algorithm trained with data from a plurality of subjects that received ablation therapy for treatment of arrhythmia. That is, a training data set for the classifier can include cardiac activation maps from such subjects with actual ablation locations annotated, as well as corresponding therapeutic outcomesAttorney Docket No. 15908WOO 1 / 82410.1297(e.g., whether the arrhythmia was successfully terminated via ablation at the annotated locations).

[0082] A representative volumetric convolutional neural network architecture 800 is shown in Figure 8. Representative architecture 800 includes a pretrained volumetric convolutional network 802, such as 3D ResNet (https: / / pytorch.org / hub / facebookresearch__pytorchvideo__resoet / ) (accessed November 6, 2024).

[0083] As depicted, architecture 800 includes three convolution layers interleaved by rectified linear unit (ReLU) activation layers. It should be understood that more (or fewer) convolution layers and / or ReLU layers could be employed without departing from the scope of the present teachings.

[0084] To identify regions of heterogeneous electrical propagation, the final output layer may be a segmentation network. Suitable segmentation network architectures include U-Net and other Fully Convolutional Networks (FCN), which will be familiar to those off ordinary skill in the art as networks that use convolutional neural networks to transform image pixels to pixel classes.

[0085] The fully connected layer in the pre-trained network (e.g., 3D ResNet) of representative architecture 800 can be removed and a 1 x 1 convolutional layer can be added to produce a binary segmentation output. An FCN transforms the height and width of intermediate feature maps back to those of the input image via a transposed convolutional layer. Thus, the model output from representative architecture 800 will have the same height and width as the input.

[0086] Cross-entropy loss may be used as the loss function for training the volumetric convolutional neural network model because it is well-suited to measure the performance of a model where the output is a probability (e.g., a probability that ablation at a particular location will terminate arrhythmia). Adam (a method for stochastic optimization) can be used as the optimization algorithm to train the volumetric convolutional neural network.

[0087] After each evolution, the volumetric convolutional neural network model can be validated using a validation data set. Similar to the training data set, the validation data set can include cardiac activation maps from subjects that received ablation therapy for treatment of arrhythmia, with ablation locations annotated, as well as associated therapeutic outcomes (e.g., ground-truth data regarding whether the arrhythmia was successfully terminated via ablation at the annotated locations).Attorney Docket No. 15908WOO 1 / 82410.1297

[0088] The volumetric convolutional neural network model with the best validation accuracy can be selected as the classifier used by system 8 (e.g., processor 28 and / or irregular propagation processing unit 58) in block 208. In turn, the output of block 208 - that is, the identification of one or more regions of heterogeneous electrical propagation, and optionally their associated arrhythmia termination confidence values - may be output to display 23 (e.g., overlaid on a three-dimensional geometric model of the heart) in block 210. The ultimate decision of whether to ablate at any such identified location, however, is committed to the practitioner’s expertise.

[0089] As an alternative to the trained classifier described above, block 208 can utilize a particle swarm optimization (PSO) algorithm to identify one or more regions of heterogeneous electrical propagation in the portion of the patient’s heart. As those of ordinary skill in the art will appreciate, PSO algorithms converge on global optima by iteratively evaluating a population of candidate solutions, termed particles, with each particle being influenced not only by its own best solution but also by the best solution of the entire swarm.

[0090] By way of illustration, in an exemplary PSO algorithm, each particle can include a candidate solution (e.g., an activation direction entropy value and / or entropy score, as described above) and a corresponding geometric position within the activation direction entropy map. The position X) of each particle (z) can be iteratively updated according to an equationis the particle velocity, given by an equationwhere co is the inertial weight, ci is the self-learning factor, cz is the social learning factor, Pt, best is the position associated with the particle’s best solution, Gbest is the position associated with the swarm’s best solution, and ranch and ranch are bounded random variables to prevent rapid convergence to a local optimum.

[0091] The exit criteria for the PSO algorithm described above can be the sooner of Gbest convergence e.g., Gbest remains substantially constant over consecutive iterations, where the criteria for “substantially constant” may be determined empirically by system 8 or selected by a practitioner) or a maximum number of iterations is reached.Attorney Docket No. 15908WOO 1 / 82410.1297

[0092] For example, a maximum number of iterations may be defined according to the largest number of iterations that can be executed before real-time performance of the PSO algorithm degrades below a level that is acceptable to a practitioner. Quantitatively, this may be expressed according to an equation imax= * tn, where imax is the maximum number of iterations,(total is the total PSO algorithm computation time (e.g., the total computation time beyond which real-time performance will no longer be acceptable to a practitioner, the value of which may be selected by the practitioner), ti is the computation time for a single iteration, and m is a safety margin (e.g., between about 0.75 and about 0.99).

[0093] Once again, the output of block 208 - that is, the identification of one or more regions of heterogeneous electrical propagation - may be output to display 23 (e.g., overlaid on a three-dimensional geometric model of the heart) in block 210. The ultimate decision of whether to ablate at any such identified location, however, is committed to the practitioner’s expertise.

[0094] Although several embodiments have been described above with a certain degree of particularity, those skilled in the art could make numerous alterations to the disclosed embodiments without departing from the spirit or scope of this invention.

[0095] For example, the teachings herein can be applied in real time (e.g., during an electrophysiology study) or during post-processing (e.g., to electrophysiology signals collected during an electrophysiology study performed at an earlier time).

[0096] All directional references (e.g., upper, lower, upward, downward, left, right, leftward, rightward, top, bottom, above, below, vertical, horizontal, clockwise, and counterclockwise) are only used for identification purposes to aid the reader’s understanding of the present invention, and do not create limitations, particularly as to the position, orientation, or use of the invention. Joinder references (e.g., attached, coupled, connected, and the like) are to be construed broadly and may include intermediate members between a connection of elements and relative movement between elements. As such, joinder references do not necessarily infer that two elements are directly connected and in fixed relation to each other.

[0097] It is intended that all matter contained in the above description or shown in the accompanying drawings shall be interpreted as illustrative only and not limiting. Changes inAttorney Docket No. 15908WOO 1 / 82410.1297detail or structure may be made without departing from the spirit of the invention as defined in the appended claims.

[0098] Aspects of the disclosure are also set out in the following numbered clauses:

[0099] Clause 1: A computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy, the method comprising:receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system;generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; andgenerating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit, wherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

[0100] Clause 2: The method according to clause 1, wherein the activation direction entropy map comprises a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart.

[0101] Clause 3: The method according to clause 2, wherein each activation direction entropy value of the plurality of activation direction entropy values is based upon a distribution of activation directions for a respective sub-region of the plurality of sub-regions within the portion of the patient’s heart.

[0102] Clause 4: The method according to clause 3, wherein each activation direction entropy value of the plurality of activation direction entropy values is computed according to an equationAttorney Docket No. 15908WOO 1 / 82410.1297where H is the activation direction entropy value for the respective sub-region and Pi is a proportion of activation directions for the respective sub-region that fall within a bin z of a preset number N of directional bins.

[0103] Clause 5: The method according to clause 4, wherein each activation direction entropy value of the plurality of activation direction entropy values is further normalized to an / \ 2entropy score between 0 and 1 according to an equation Score = ( — )

[0104] Clause 6: The method according to clause 4 or 5, further comprising the electroanatomical mapping system receiving a user input to define the preset number N.

[0105] Clause 7: The method according to any of clauses 1 to 6, further comprising the electroanatomical mapping system receiving a user input to define a sub-region size.

[0106] Clause 8: The system according to clause 1 to 7, wherein the irregular propagation processing unit comprises a classifier trained with data from a plurality of subjects that received ablation therapy for treatment of arrhythmia.

[0107] Clause 9: The method according to any of clauses 1 to 8, wherein the irregular propagation processing unit comprises a volumetric convolutional neural network algorithm.

[0108] Clause 10: The method according to claim any of clauses 1 to 7, wherein the irregular propagation processing unit comprises a particle swarm optimization algorithm.

[0109] Clause 11 : A computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy, the method comprising:receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system;generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; andAttorney Docket No. 15908WOO 1 / 82410.1297generating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit comprising a classifier trained with data from a plurality of subjects that received ablation therapy for treatment of arrhythmia,wherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

[0110] Clause 12: The method according to clause 11, wherein the activation direction entropy map comprises a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart.

[0111] Clause 13: The method according to clause 12, wherein each activation direction entropy value of the plurality of activation direction entropy values is based upon a distribution of activation directions for a respective sub-region of the plurality of sub-regions within the portion of the patient’s heart.

[0112] Clause 14: The method according to clause 13, wherein each activation direction entropy value of the plurality of activation direction entropy values is computed according to an equationwhere H is the activation direction entropy value for the respective sub-region and is a proportion of activation directions for the respective sub-region that fall within a bin i of a preset number Afof directional bins.

[0113] Clause 15: The method according to clause 14, wherein each activation direction entropy value of the plurality of activation direction entropy values is further normalized to an entropy score between 0 and 1 according to an equation Score = ( — J .

[0114] Clause 16: The method according to clause 14 or 15, further comprising the electroanatomical mapping system receiving a user input to define the preset number N.Attorney Docket No. 15908WOO 1 / 82410.1297

[0115] Clause 17: The method according to any of clauses 11 to 16, further comprising the electroanatomical mapping system receiving a user input to define a sub-region size.

[0116] Clause 18: The method according to any of clauses 11 to 17, wherein the irregular propagation processing unit comprises a volumetric convolutional neural network algorithm.

[0117] Clause 19: The method according to clause 18, further comprising the electroanatomical mapping system downsampling the cardiac activation map prior to generating the activation direction entropy map, optionally, wherein downsampling the cardiac activation map comprises voxelizing the cardiac activation map.

[0118] Clause 20: A computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy, the method comprising:receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system;generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; andgenerating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit comprising a particle swarm optimization algorithm,wherein each particle evaluated by the particle swarm optimization algorithm comprises an activation direction entropy value associated with a sub-region of the portion of the patient’s heart and a position of the sub-region of the portion of the patient’s heart, andwherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.Attorney Docket No. 15908WOO 1 / 82410.1297

[0119] Clause 21 : The method according to clause 20, wherein the activation direction entropy value is based upon a distribution of activation directions for the sub-region of the portion of the patient’s heart.

[0120] Clause 22: The method according to clause 21, wherein the activation direction entropy value is computed according to an equationwhere H is the activation direction entropy value and Pi is a proportion of activation directions for the sub-region of the portion of the patient’s heart that fall within the a bin z of a preset number N of directional bins.

[0121] Clause 23 : The method according to clause 22, wherein the activation direction entropy value is further normalized to an entropy score between 0 and 1 according to an equation

[0122] Clause 24: An electroanatomical mapping system for identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy, the electroanatomical mapping system comprising:an irregular propagation processing unit configured to:receive a cardiac activation map of a portion of a patient’s heart;generate an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; andgenerate an output, based on the activation direction entropy map for the portion of the patient’s heart, wherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.Attorney Docket No. 15908WOO 1 / 82410.1297

[0123] Clause 25: The system according to clause 24, wherein the activation direction entropy map comprises a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart.

[0124] Clause 26: The system according to clause 25, wherein each activation direction entropy value of the plurality of activation direction entropy values is based upon a distribution of activation directions for a respective sub-region of the plurality of sub-regions within the portion of the patient’s heart.

[0125] Clause 27: The system according to clause 26, wherein each activation direction entropy value of the plurality of activation direction entropy values is computed according to an equationwhere H is the activation direction entropy value and Pi is a proportion of activation directions for the sub-region of the portion of the patient’s heart that fall within the a bin i of a preset number N of directional bins.

[0126] Clause 28: The system according to clause 27, wherein each activation direction entropy value of the plurality of activation direction entropy values is further normalized to an entropy score between 0 and 1 according to an equation Score = ( — ) •

[0127] Clause 29: The system according to any of clauses 24 to 28, wherein the irregular propagation processing unit comprises a classifier trained with data from a plurality of subjects that received ablation therapy for treatment of arrhythmia.

[0128] Clause 30: The system according to clause 29, wherein the irregular propagation processing unit comprises a volumetric convolutional neural network algorithm.

[0129] Clause 31 : The system according to any of clauses 24 to clause 28, wherein the irregular propagation processing unit comprises a particle swarm optimization algorithm.

Claims

Attorney Docket No. 15908WOO 1 / 82410.1297CLAIMSWhat is claimed is:

1. A computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy, the method comprising:receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system;generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; andgenerating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit comprising a classifier trained with data from a plurality of subjects that received ablation therapy for treatment of arrhythmia, wherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

2. The method according to claim 1, wherein the activation direction entropy map comprises a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart.

3. The method according to claim 2, wherein each activation direction entropy value of the plurality of activation direction entropy values is based upon a distribution of activation directions for a respective sub-region of the plurality of sub-regions within the portion of the patient’s heart.

4. The method according to claim 3, wherein each activation direction entropy value of the plurality of activation direction entropy values is computed according to an equationwhere H is the activation direction entropy value for the respective sub-region and Pi is aAttorney Docket No. 15908WOO 1 / 82410.1297proportion of activation directions for the respective sub-region that fall within a bin z of a preset number N of directional bins.

5. The method according to claim 4, wherein each activation direction entropy value of the plurality of activation direction entropy values is further normalized to an entropy score between0 and 1 according to an equation6. The method according to claim 4, further comprising the electroanatomical mapping system receiving a user input to define the preset number N.

7. The method according to claim 2, further comprising the electroanatomical mapping system receiving a user input to define a sub-region size.

8. The method according to claim 1, wherein the irregular propagation processing unit comprises a volumetric convolutional neural network algorithm.

9. The method according to claim 8, further comprising the electroanatomical mapping system downsampling the cardiac activation map prior to generating the activation direction entropy map.

10. The method according to claim 9, wherein downsampling the cardiac activation map comprises voxelizing the cardiac activation map.

11. A computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy, the method comprising:receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system;generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; andgenerating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit, wherein the output comprises anAttorney Docket No. 15908WOO 1 / 82410.1297identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

12. The method according to claim 11, wherein the activation direction entropy map comprises a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart.

13. The method according to claim 12, wherein each activation direction entropy value of the plurality of activation direction entropy values is based upon a distribution of activation directions for a respective sub-region of the plurality of sub-regions within the portion of the patient’s heart.

14. The method according to claim 13, wherein each activation direction entropy value of the plurality of activation direction entropy values is computed according to an equationwhere His the activation direction entropy value for the respective sub-region and Pi is a proportion of activation directions for the respective sub-region that fall within a bin i of a preset number N of directional bins.

15. The method according to claim 14, wherein each activation direction entropy value of the plurality of activation direction entropy values is further normalized to an entropy score between0 and 1 according to an equation16. The method according to claim 14, further comprising the electroanatomical mapping system receiving a user input to define the preset number N.

17. The method according to claim 12, further comprising the electroanatomical mapping system receiving a user input to define a sub-region size.

18. The method according to claim 11, wherein the irregular propagation processing unit comprises a volumetric convolutional neural network algorithm.Attorney Docket No. 15908WOO 1 / 82410.129719. The method according to claim 11, wherein the irregular propagation processing unit comprises a particle swarm optimization algorithm.

20. A computer-implemented method of identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy, the method comprising:receiving a cardiac activation map of a portion of a patient’s heart at an electroanatomical mapping system;generating, using the electroanatomical mapping system, an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; andgenerating an output, based on the activation direction entropy map for the portion of the patient’s heart, using an irregular propagation processing unit comprising a particle swarm optimization algorithm,wherein each particle evaluated by the particle swarm optimization algorithm comprises an activation direction entropy value associated with a sub-region of the portion of the patient’s heart and a position of the sub-region of the portion of the patient’s heart, andwherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

21. The method according to claim 20, wherein the activation direction entropy value is based upon a distribution of activation directions for the sub-region of the portion of the patient’s heart.

22. The method according to claim 21, wherein the activation direction entropy value is computed according to an equationwhere H is the activation direction entropy value and Pi is a proportion of activation directions for the sub-region of the portion of the patient’s heart that fall within the a bin z of a preset number A of directional bins.Attorney Docket No. 15908WOO 1 / 82410.129723. The method according to claim 22, wherein the activation direction entropy value is further normalized to an entropy score between 0 and 1 according to an equation24. An electroanatomical mapping system for identifying regions of heterogeneous cardiac electrical propagation for treatment of arrhythmia via application of ablation therapy, the electroanatomical mapping system comprising:an irregular propagation processing unit configured to:receive a cardiac activation map of a portion of a patient’s heart;generate an activation direction entropy map for the portion of the patient’s heart from the cardiac activation map; andgenerate an output, based on the activation direction entropy map for the portion of the patient’s heart, wherein the output comprises an identification of one or more regions of heterogeneous cardiac electrical propagation in the portion of the patient’s heart.

25. The system according to claim 24, wherein the activation direction entropy map comprises a plurality of activation direction entropy values associated with a corresponding plurality of sub-regions within the portion of the patient’s heart.

26. The system according to claim 25, wherein each activation direction entropy value of the plurality of activation direction entropy values is based upon a distribution of activation directions for a respective sub-region of the plurality of sub-regions within the portion of the patient’s heart.

27. The system according to claim 26, wherein each activation direction entropy value of the plurality of activation direction entropy values is computed according to an equationwhere His the activation direction entropy value and Pi is a proportion of activation directionsAttorney Docket No. 15908WOO 1 / 82410.1297for the sub-region of the portion of the patient’s heart that fall within the a bin z of a preset number N of directional bins.

28. The system according to claim 27, wherein each activation direction entropy value of the plurality of activation direction entropy values is further normalized to an entropy score between0 and 1 according to an equation29. The system according to claim 24, wherein the irregular propagation processing unit comprises a classifier trained with data from a plurality of subjects that received ablation therapy for treatment of arrhythmia.

30. The system according to claim 29, wherein the irregular propagation processing unit comprises a volumetric convolutional neural network algorithm.

31. The system according to claim 24, wherein the irregular propagation processing unit comprises a particle swarm optimization algorithm.

Images