Computational localization of vibration sources

Abstract

A system for computational localization of a fibrillation source is provided. In some implementations, the system performs operations including generating a representation of electrical activation of a patient's heart and comparing the generated representation to one or more stored representations of the heart based on correlation to identify at least one matching representation of the heart. The operations may also include generating a computational model for the patient's heart based on the at least one matching representation, wherein the computational model includes a representation of one or more fibrillation sources in the patient's heart. Additionally, the operations may include displaying at least a portion of the computational model via a user interface. Related systems, methods, and articles of manufacture are also described.

Description

[0001] This application is a divisional application of Chinese application CN201680075885.9, which was filed on

[0002] On December 22, 2016, the invention was named “Systems and methods for locating the source of fibrillation”.

[0003] CROSS-REFERENCE TO RELATED APPLICATIONS

[0004] This application claims priority to U.S. Provisional Patent Application No. 62 / 271,113, filed on December 22, 2015, entitled “COMPUTATIONAL LOCALIZATION OF VENTRICULAR FIBRILLATION SOURCES,” and also claims priority to U.S. Non-Provisional Patent Application No. 15 / 389,245, filed on December 22, 2016, entitled “COMPUTATIONAL LOCALIZATION OF FIBRILLATION SOURCES,” the disclosures of which are incorporated herein by reference. Technical Field

[0005] The subject matter described herein relates to computational localization of fibrillation sources, and more particularly to identification of ventricular fibrillation (VF) and / or atrial fibrillation (AF) sources. Background Art

[0006] Ventricular fibrillation (VF) and atrial fibrillation (AF) can cause symptoms, morbidity (syncope or stroke), and mortality. Stable electrical rotors, recurring electrical foci, and other mechanisms are important drivers of persistent and / or clinically significant episodes of VF / AF. To treat VF / AF, therapeutic ablation, radiofrequency (RF), cryotherapy, ultrasound, and / or external radiation sources can be used to target and / or eliminate these mechanisms. The ability to map the perpetuating mechanisms of VF and / or AF using noninvasive methods would provide significant benefits in the management of this arrhythmia.

[0007] Current methods for reliably identifying the location of VF / AF sustaining mechanisms (e.g., rotors or focal sources) are currently suboptimal. They typically require invasive surgery, insertion of an expensive 64-electrode basket catheter, and / or use of expensive, difficult to obtain / manufacture, and bulky body surface vests to map arrhythmias, which may interfere with the placement of defibrillator pads. As a result, such procedures are expensive, time consuming, and potentially dangerous to the patient. Therefore, methods for identifying the presence and / or location of a fibrillation source with a lower cost and / or less invasive procedure could provide significant clinical benefits. Summary of the invention

[0008] In some aspects, a method, computer program product, and system are provided. In one implementation, a system for computational localization of a fibrillation source is provided. The system may include (or otherwise utilize) at least one processor and / or memory, which may be configured to perform operations including generating a representation of electrical activation of a patient's heart and comparing the generated representation to one or more stored representations of the heart based on a correlation to identify at least one matching representation of the heart. The operations may also include generating a computational model for the patient's heart based on the at least one matching representation, wherein the computational model includes a representation of one or more fibrillation sources in the patient's heart. Additionally, the operations may include displaying at least a portion of the computational model via a user interface.

[0009] In some aspects, the generated representation and / or one or more stored representations may include three-dimensional data (e.g., may be a 3D model). In some aspects, the generated representation and / or one or more stored representations may include a vector cardiogram. Comparing the representations may include determining, for each of the one or more stored representations, a correlation factor between the generated representation and the stored representations, and at least one matching representation may be identified as the stored representation having the highest correlation factor.

[0010] In some variations, the operations may further include generating a plurality of computational heart models having different shapes, geometries, fiber orientations, scars, fibrillation source types, and fibrillation source locations. The operations may also include filtering the plurality of computational heart models based on the shape or scarring of the patient's heart to identify a filtered set, and / or selecting one or more stored representations for comparison based on the filtered set. In various implementations, the filtering may be based on computerized tomography imaging data, magnetic resonance imaging data, echocardiography data, x-ray data, fluoroscopy data, and / or the like.

[0011] In some variations, the computational model includes one or more of the left atrium, right atrium, left ventricle, and right ventricle, and / or one or more fibrillation sources may be mapped to one or more of the left atrium, right atrium, left ventricle, and right ventricle. The one or more fibrillation sources may include rotors or focal sources, or some other ventricular fibrillation source or atrial fibrillation source. In some variations, the computational model includes a three-dimensional mesh in the shape of the heart and / or a finite state fibrillation source map of cardiac electrical activation mapped to the mesh.

[0012] In some variations, the operations may further include generating an electrocardiogram plot based on the patient's heart, and / or generating a vectorcardiogram based on the electrocardiogram plot. In some variations, the operations may further include generating a second computational model for the patient's heart based on the (original) computational model, wherein the second computational model is generated to include a number of fibrillation sources less than the one or more fibrillation sources. A side-by-side comparison of the computational model and the second computational model may be displayed via a user interface. In some aspects, the second computational model may be generated by removing one of the one or more fibrillation sources from the computational model. In some variations, the operations may further include determining a fibrillation change between the computational model and the second computational model.

[0013] Implementations of the current subject matter may include systems and methods consistent with this specification, including one or more features as described and articles including tangibly implemented machine-readable media, which are operable to cause one or more machines (e.g., computers, etc.) to cause the operations described herein. Similarly, a computer system is also described, which may include one or more processors and one or more memories coupled to the one or more processors. The memory, which may include a computer-readable storage medium, may include one or more programs, encode them, store them, etc., which cause one or more processors to perform one or more of the operations described herein. Computer-implemented methods consistent with one or more implementations of the current subject matter may be implemented by one or more data processors residing in a single computing system or multiple computing systems. Such multiple computing systems may be connected and may exchange data and / or commands or other instructions, etc., via one or more connections, including but not limited to connections through a network (e.g., the Internet, a wireless wide area network, a local area network, a wide area network, a wired network, etc.), via direct connections between one or more of the multiple computing systems, and the like.

[0014] The details of one or more variations of the subject matter described herein are set forth in the accompanying drawings and the following description.While certain features of the presently disclosed subject matter are described for illustrative purposes in relation to enterprise resource software systems or other business software solutions or architectures, it should be readily understood that these features are not intended to be limiting. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings, which are incorporated into and constitute a part of this specification, illustrate certain aspects of the subject matter disclosed herein and, together with the description, help explain some principles associated with the disclosed implementations.

[0016] Figure 1 depicts a block diagram of a system for computational localization of a fibrillation source, according to some exemplary implementations;

[0017] Figure 2A A graph depicting activation time in the right atrium of a patient according to some exemplary implementations;

[0018] Figure 2B A graph depicting activation time in the left atrium of a patient according to some exemplary implementations;

[0019] Figure 2C depicts fibrillation source mapping for VF according to some exemplary implementations;

[0020] Figure 2D depicts fibrillation source mapping for AF according to some exemplary implementations;

[0021] Figure 3A depicts a biventricular computational model of a left ventricle and a right ventricle according to some exemplary implementations;

[0022] Figure 3B depicts a biventricular computational model with identified rotors according to some exemplary implementations;

[0023] Figure 4A depicts a graph of electrocardiogram (EKG) data calculated according to a computational model according to some exemplary implementations;

[0024] Figure 4B depicts a vector cardiogram (VCG) calculated according to a computational model according to some exemplary implementations;

[0025] Figure 5A depicts an endocardial isochronal diagram of cardiac activation during left ventricular VF of a patient according to some exemplary implementations;

[0026] Figure 5B depicts VF source locations within a patient's anatomy as a function of endocardial voltage according to some exemplary implementations;

[0027] Fig. 6A depicts a graph of EKG readings during VF in a human according to some exemplary implementations;

[0028] Figure 6B depicts a VCG derived from recorded EKG data according to some exemplary implementations;

[0029] Fig. 7A depicts a biventricular computational model with VF at different time points according to some exemplary implementations;

[0030] Figure 7B depicts a graph of EKG traces calculated over time intervals according to some exemplary implementations;

[0031] Figure 7C depicts VCG traces computed over time intervals according to some exemplary implementations;

[0032] Fig. 8A depicts a computational model of the left and right atria according to some exemplary implementations;

[0033] Figure 8B depicts a computational model of right and left atria with simulated diseased substrates maintaining identified rotors according to some exemplary implementations;

[0034] Fig. 9 depicts a block diagram of an example computing device according to some example implementations;

[0035] Fig.10 An example of a method for computational localization of a fibrillation source according to some exemplary implementations is depicted;

[0036] Fig.11 An example of a method for calculating the effect of ablation and subsequent risk of cardiac fibrillation according to some exemplary implementations is depicted; and

[0037] Fig.12 Depicted are simulated VCGs and measured human VCGs of the rotor at the same anatomical location, according to some exemplary implementations.

[0038] Indeed, similar reference numerals indicate similar structures, features or elements. DETAILED DESCRIPTION

[0039] As described above, it may be desirable to identify the presence and / or location of a fibrillation source with a lower cost and / or less invasive procedure / method. Accordingly, non-invasive systems and methods for identifying the presence and / or location of rotors or focal sources (collectively referred to herein as "fibrillation mechanisms") in patients suffering from ventricular fibrillation (VF) or atrial fibrillation (AF) are described. In some implementations, a readily available twelve-lead electrocardiogram (EKG) sensor device may be applied to the surface of the patient's skin, rather than a very specialized and expensive disposable 64-electrode basket catheter requiring surgical implantation of twenty or more electrodes and / or an external electrocardiogram ("EKG" or "ECG") vest.

[0040] Figure 1A functional block diagram of a system 100 is shown, wherein features consistent with the described subject matter can be implemented. As shown, the system 100 may include a computing system 110 capable of communicating with one or more user access devices 140 and / or one or more sensor devices 150A-D (collectively referred to as sensor devices 150). In some aspects, the computing system 100 may communicate using one or more interfaces 118. Communication between devices in the system 100 may be by using direct communication, such as by using wireless connections such as Bluetooth, near field communication (NFC), ZigBee, WiFi, some combination thereof, etc. Additionally or alternatively, communication between devices in the system 100 may be by using hardwired connections such as a universal serial bus (USB) and / or the like. Communication may additionally or alternatively occur by indirect communication, such as by a network 160, which may include a local area network, a wide area network, a wireless network, the Internet, some combination thereof, etc.

[0041] Communications over the network 160 may utilize network access devices 165 such as base stations, node Bs, evolved node Bs (eNBs), access nodes (ANs), hotspots, and the like. In some aspects, any of the user access devices 140 may include a personal computer, a desktop computer, a laptop computer, a workstation, a cellular phone, a digital media device, a smart phone, a smart watch, a PDA (personal digital assistant), a tablet computer, a hardware / software server, a sensor, a sensor device, a terminal, an access terminal (AT), a mobile station, a user equipment (UE), a subscriber unit, and the like. In some aspects, any of the sensor devices 150 may include an EKG sensor / device, a vector cardiogram (VCG) sensor / device, a cardiac imaging device, and / or the like. In some implementations, the cardiac imaging device may include one or more of a computed tomography (CT or CAT) scanning device, a magnetic resonance imaging (MRI) scanning device, a Scopulariopter scanning device, a thallium scanning device, a multi-gated acquisition scanning device, an X-ray device, an echocardiography device, a fluoroscopy device, and / or the like. In various implementations, data (e.g., cardiac imaging data) may be provided and / or corresponding devices (e.g., cardiac imaging devices) may not be present. Wired or wireless communications between the computing system 110, the user access device 140, and / or the sensor device 150 may occur according to various protocols and / or access technologies (e.g., Global System for Mobile Communications (GSM), Universal Mobile Telecommunications System (UMTS), technologies developed by IEEE such as WiFi and / or Bluetooth, technologies developed by the Third Generation Partnership Project (3GPP) or 3GPP2 such as Long Term Evolution (LTE) and / or CDMA2000, etc.).

[0042] At least a portion of the illustrated system 100 may include hardware and / or software that interacts with stored data, models, algorithms, etc. and / or receives, sends, defines, creates, and / or updates data. As shown, the computing system 110 may include a processor 190 that may be used to manage / control the operation of the computing system 110. As further shown, the computing system 110 may include core software 112 and / or one or more software modules 114. The core software 112 may provide one or more features of a high-level programming software system. The software module 114 may provide more specialized functions. For example, the core software 112 and / or the software module 114 may include sensor management and / or data processing functions. In some aspects, the core software 112 or other similar software / hardware may access a database layer such as a database 120. The database 120 may store any kind of data, including, but not limited to, data retrieved from the sensor device 150, a computational model, EKG data, VCG data, a machine learning algorithm, a data conversion algorithm, and the like.

[0043] For example, as shown, the database 120 may include a model library 192, a patient-specific model 194, a patient database 195, a VCG library 196, a data transformation algorithm 197, and / or a machine learning algorithm 198. The model library 192 may contain multiple computational models of the heart, a portion of the heart, other organs, etc. The computational models in the model library 192 may be high-resolution (e.g., greater than about 500,000 degrees of freedom) finite element models. At least a portion of the computational model may be generated by incorporating a Fenton-Karma, Bueno-Cherry-Fenton, ten Tussher-Noble, or similar, detailed ionic model of the human ventricular or atrial action potential during simulation of VF or AF, respectively.

[0044] The patient-specific models 194 may include computational models similar to the models in the model library 192, but the models in the patient-specific models 194 may be generated based on data from an actual patient. For example, as shown, an endocardial sensor device 150A may be applied to record data from the inside of the heart of the patient 130A. At the same (or approximately the same) time, an EKG sensor 150B may be applied to the outside of the patient 130A to record an EKG reading. Based on a combination of these readings, one or more patient-specific models 194 may be generated and stored. In some implementations, the patient-specific models 194 may be additionally or alternatively based on CT scan data, MRI scan data, Scopularioplast scan data, thallium scan data, multi-gated acquisition scan data, fluoroscopic data, x-ray data, echocardiogram data, and / or other cardiac imaging data, which may be used to identify the shape, scars, etc., of the heart of the patient 130N.

[0045] The VCG library 196 may include VCG data for the computational models of the model library 192 and / or the model 194 of a specific patient. For example, the VCG model may be simulated based on the computational models in the model library 192 (e.g., based on the EKG data associated with each model). In some implementations, the VCG model may include a three-dimensional tracing of electrical activity in the heart or some part thereof. In some aspects, the VCG data may include spatiotemporal VCG data. The VCG data in the VCG library 196 may serve as a diagnostic template for which a VCG constructed from patient data may be matched. For example, the VCG library data may be compared with the patient's VCG data to identify the location of a VF / AF source in the patient's body. For example, as shown, the EKG sensor device 150D may obtain EKG data from the patient 130N. The EKG data may be used to generate VCG data, and the generated VCG data may be compared with the VCG data from the VCG library 196. Based on the level of correlation between these data sets, the VF / AF source may be identified as described herein. In some implementations, VCG data from VCG library 196 may be filtered for comparison based on physical characteristics (e.g., shape, scarring, etc.) of the heart of patient 130N, which may be determined based on data obtained by cardiac imaging device 150C, which may include a CT scanning device, an MRI scanning device, an echocardiography device, etc. In some implementations, data within VCG library 196 may be stored in addition to / instead within model library 192. For example, each computational heart model may include or be associated with VCG data.

[0046] In some implementations, the computational models within model library 192 may be generated and / or validated based on patient-specific model 194. For example, the VCGs in VCG library 196 may be validated by comparing them to human VCGs, which may be obtained from patient-specific model 194. Fig.12 A comparison of VCGs 1220, 1260 generated from the same ventricular location based on a human rotor 1210 and a simulated rotor 1250 is shown. In some aspects, the human rotor 1210 can be a source of VF in a patient. As shown, the rotor can be in a post-ventricular position within the left ventricle and / or rotate in a counterclockwise direction when viewed from the endocardium. The corresponding first VCG 1220 shows a clinical VCG cycle for a particular period of VF for that patient.

[0047] In some aspects, the second VCG 1260 shows a simulated VCG cycle that closely approximates the clinical VCG 1220 (e.g., for the same period of VF). The second VCG 1260 can be stored in and / or retrieved from the VCG library 196. The corresponding computer simulated rotor 1250 can simulate a VF source indicated by data from the VCG 1260. As shown, the position of the simulated VF source rotor 1250 can also be located in a posterior central ventricular position within the left ventricle and / or rotate in a counterclockwise direction when viewed from the endocardium. As shown, the VCGs 1220, 1260 are colored based on the timing information.

[0048] Back to reference Figure 1 In some aspects, the VCG in the patient-specific model 194 can be constructed from an existing human dataset of surface EKGs, which can be stored in and / or provided from the patient database 195, with known rotors and lesion source locations identified using concurrent electrode invasive endocardial recordings taken from patients with inducible VF.

[0049] In some aspects, the robustness of the patient-specific model 194 can increase the accuracy and / or value of the model library 192, which can increase the likelihood of identifying a VF / AF source within the patient 130 and / or the accuracy of determining the specific location of a VF / AF source within the heart of the patient 130. The machine learning algorithm 198 can be trained based on the patient-specific model 194 to generate an algorithm for detecting VF / AF in the patient based on received EKG sensor data, CT scan data, VCG, and / or the like. For example, the relevant EKG data and VCG can be used to train the machine learning algorithm to identify VF / AF mechanisms and loci. In some example embodiments, an algorithm for deriving diagnostic criteria to predict the presence and location of VF / AF rotors can be provided. To this end, an automatic diagnostic tool can be provided to automatically compare the VCG calculated from the surface EKG of a patient with VF / AF with a diagnostic template (e.g., the VCG library 196) to identify the VF / AF mechanism and location using statistical classification and / or machine learning techniques.

[0050] In some implementations, data conversion algorithm 197 can be used to normalize EKG and / or VCG data measured from patient 130A to account for differences between patients in chest size, lung geometry, degree of musculature, body fat composition, and / or other factors that may affect surface electrode measurements using statistical classification and / or machine learning techniques.

[0051] In some implementations, the model library 192, the patient-specific model 194, the VCG library 196, the machine learning algorithm 198, the data conversion algorithm 197, and / or the patient database 195 can be specific to a VF / AF mechanism. For example, the model library 192 and the patient-specific model 194 can include a model for rotors and a separate model for lesion sources. Thus, the VCG library 196 can include a rotor-based VCG and a separate VCG based on lesion sources. Since rotors and lesion sources are implemented in different ways in the VCG, using a separate model / VCG set can be used to identify the mechanism, as well as its location.

[0052] In some aspects, the core software 112 may be configured to load information from the database 120 into the memory 116 (e.g., main memory) in response to receiving instructions, data, or queries initiated by a user or a computer system via one or more sensor devices 150, user access device 140, etc. Although the database 120 is shown as being located within the computing system, in various implementations, at least a portion of the database 120 may be separate from the computing system 110.

[0053] In some aspects, one or more of the software modules 114 may be configured to utilize data stored in the memory 116, data stored in the database 120, and / or data accessible to the computing system 110. In some aspects, the computing system 110 is able to utilize external software that may provide additional functionality or services that may not be available at the computing system 110. In some aspects, the external software may include a cloud service. In some aspects, the computing system 110 may aggregate or otherwise provide a gateway through which a user may access functionality provided by the external software. In some implementations, the database 120 and / or its contents may be located on one or more servers, and / or communications between the computing system 110, the user access device 140, and / or the sensor device 150 may occur over the network 160.

[0054] In some aspects, database 120 may be physically stored in a hardware server or physically stored across multiple hardware servers.In some aspects, system 100 may be implemented as a cloud-based system and / or data processing system.

[0055] Figure 2AA fibrillation source activation map 200 is depicted in the right atrium 210 of a patient according to some exemplary implementations (in milliseconds according to the shaded scale). As shown, longer and / or more sustained activations of the right atrium 210 may rotate around a source 220. The source 220 may be considered a source of AF, and more specifically, a rotor. In some aspects, the rotor may be considered a tissue center for AF that is at least partially rotating in nature (e.g., the area where the duration of activation time is the longest). Although the right atrium 210 is shown and described, rotors may be present in any chamber of the heart.

[0056] Figure 2B A fibrillation source activation graph 250 is depicted in the left atrium 260 of a patient according to some exemplary implementations (in milliseconds according to the shaded scale). As shown, longer and / or more sustained activations of the left atrium 260 can emanate from a source 270. The source 270 can be considered a source of AF, and more specifically a focal source. In some aspects, a focal source can be considered to be a tissue center that generates AF distally (e.g., an area with the longest duration of activation time). Although the left atrium 260 is shown and described, a focal source can exist in any chamber of the heart.

[0057] Figure 2C A fibrillation source map 280 for VF is depicted according to some exemplary implementations. In some aspects, the fibrillation source map 280 can be the product of a computer algorithm described herein. As shown, regions within the heart that maintain VF are shaded according to the percentage of VF cycles in which a VF source 285 is located at a particular region of the ventricle.

[0058] Figure 2D A fibrillation source map 290 for AF is depicted according to some exemplary implementations. In some aspects, the fibrillation source map 290 can be the product of a computer algorithm as described herein. As shown, regions within the heart that sustain AF are shaded according to the percentage of AF cycles in which AF sources 295 are located at specific regions of the atrium.

[0059] Identifying the locations of the sources 220, 270, 285, and 295 of VF / AF may be beneficial because knowledge of their locations can help guide surgical procedures and minimize the amount of guesswork required of medical professionals. For example, in a patient with AF, a single rotor (e.g., Figure 2A as shown) or lesion sources (e.g. Figure 2B In addition, based on the percentage of cycles / time emitted from these sources (e.g., Figure 2C and Figure 2D220, 270 can be specifically targeted. In addition, there is a lower risk of injury to other organs, nerves, bones, muscles, etc. during the procedure, with a reduced recovery time and / or minimal scarring.

[0060] Heart failure is a complex disease that may involve and / or be based on changes in ventricular shape, fiber orientation, ion channel expression remodeling, and / or other pathologies / changes. Figure 1 The computing system 110 may generate, modify, display, or otherwise utilize a computational model of the heart to identify the source / location of VF / AF based on a particular patient's condition / changes.

[0061] A computational model of the heart (or portion thereof) may be generated based on imaging information representative of the heart (e.g., three-dimensional) and / or electrical data representative of the electrophysiology of the heart. At least some of the computational models may be patient-specific, where the data used to generate the computational model and the parameters / metrics used in conjunction with the computational model may be specific to a given patient, taking into account conditions / variations. However, the time, effort, cost, and invasive procedures required to generate a model for a specific patient may be prohibitive. For example, positioning of an endocardial recording catheter or device may be required in order to measure electrical properties from within the heart. Thus, at least some of the computational models may be general models in that they are not representations of a specific patient's heart. These computational models may be manually generated based on patient-specific models or other patient-specific data, based on measured / known properties of the human heart, and / or the like. The generated model may then be "morphed" into a more specific patient configuration, for example, as Figure 1 As indicated in the computing system 110 .

[0062] The general model can be modified to generate additional general models. For example, based on a general model that has been verified to be sufficiently accurate (e.g., by comparison with a model for a specific patient or other data obtained from the patient), the shape of the heart and / or the location of a scar within the heart can be changed. Because the shape and / or scar of the heart may affect VF / AF, adding models with different hearing shapes / scars can provide additional accuracy in locating the source of VF / AF. In some implementations, thousands of computational models can be generated and / or stored within the model library 192. Any of these generation techniques can be manual, at least partially automated, and / or based on machine learning. In some aspects, the computational model can be viewed as a finite element computational model of the arrhythmia.

[0063] Measurements of cardiac electrical properties of a patient may be generated and / or received. For example, the computing system 100 may receive and / or record EKG data of a patient. In some implementations, the EKG data may be obtained from an EKG sensor device (e.g., a 12-lead EKG) that records continuous, dynamic signals of cardiac electrical function from multiple body locations of the patient (e.g., on the surface of the chest, arms, legs, head, etc.). Additionally, the computing system 100 may receive and / or record endocardial data of the patient. In some embodiments, endocardial data can be obtained from a controllable mapping and / or ablation catheter located within at least one chamber of the heart (e.g., an ablation catheter located in the left ventricle and / or right ventricle for VF, or an ablation catheter located in the left atrium and / or right atrium for atrial fibrillation), or from one or more basket catheters, such as a 64-electrode basket catheter located within at least one chamber of the heart (e.g., one in the left ventricle and one in the right ventricle for VF, or one in the left atrium and one in the right atrium for AF), which records continuous, dynamic signals of cardiac electrical function within the patient's heart.

[0064] In some aspects, the endocardial data can be matched with the EKG data (e.g., in time) so that the computing system 100 can access data representing the electrical function of the heart from inside and outside the patient's heart. Based on the relationship between the endocardial and EKG data, patterns can be identified and / or correlations can be defined to identify VF / AF sources (e.g., rotors or lesion sources). In some implementations, the endocardial and EKG data can be measured when the patient's heart is excited and / or when the patient's heart is in a relaxed state. In an example implementation, the electrical activity of the heart can be recorded at one or more locations according to conventional pacing in the heart to establish a relationship between cardiac excitation and surface EKG recordings and / or calculated VCGs. In another implementation, VF / AF can be induced in the patient, and electrical data during this period can be measured / recorded. Since the data shows the behavior of the patient's heart during VF / AF, it can be compared with baseline data and data from other patients to identify VF / AF sources.

[0065] Imaging data of a patient's heart (e.g., the left atrium, right atrium, left ventricle, and / or right ventricle) may be received. For example, the computing system may receive and / or record image data including an image of the heart (or a portion thereof) obtained from a clinical cardiac CT scanning device, a 2D or 3D ultrasound cardiography device, a myocardial perfusion scanning device, an MRI device, a positron scanning device, an X-ray device, a fluoroscopic device, and / or other device capable of generating or providing an image of the heart (and / or a portion thereof). In order to combine the 3D anatomical model of the heart with the endocardium and / or EKG data, the data processor may register (e.g., align) the data so that the endocardium and / or EKG data is aligned with the appropriate orientation of the heart (e.g., a ventricular or atrial region thereof).

[0066] A four-dimensional (4D) patient-specific computational model with data regarding the patient's electrical activity can be generated. For example, the computing system 110 can generate a 4D model based on 3D data received with additional dynamics from the electrical activity data. The 4D computational model can provide a 3D representation of the morphology and anatomy of the heart (or portion thereof) over time, and can provide time-varying electrical dynamics of the heart (or portion thereof), such as EKG and / or endocardial data that varies over time. The electrical dynamics can include activation patterns and / or electrical trajectories of activation through the myocardium. The electrical dynamics can include patterns (e.g., sequences) of electrical repolarization / recovery. The model may also include additional / optional aspects, such as regional distributions of perfusion or infarction that may be measured or simulated in individual patients.

[0067] In some implementations, the computational model may include EKG data overlaid and / or registered on the 3D biventricular geometry of the patient's heart, the human fiber structure of the heart, areas of heterogeneous conductivity caused by the presence of myocardial ischemia, infarction, anatomical (and / or functional) conductive defects (e.g., partial and / or complete bundle branch block), and / or the like. The model may be generated using a finite element mesh. A patient-specific finite element mesh of the heart (which may include its ventricular anatomy) may be generated based on image data such as clinical CT data, perfusion images, MRI data, and / or other types of image data.

[0068] The computational model may also include the fiber architecture of the heart. For example, the fiber architecture of the heart may be empirically estimated using a logarithmic-Euclidean interpolation framework for registering DT-MR measurements to an anatomical model. The reconstructed diffusion tensor may be fitted to a field of logarithmic transformed components in the corresponding anatomical grid to interpolate local fibers, sheets, and sheet-perpendicular axes. The fiber orientations in the resulting model may be mapped to the patient via large deformation differential homeomorphism mapping and reoriented based on the 3D deformation gradient between the template and target patient ventricular geometry to account for the effects of ventricular shape differences on fiber orientation. The resulting fiber sheet model forms a local basis for transversely isotropic or orthotropic ventricular conductivity (which may have, for example, a fiber sheet anisotropy ratio of approximately 7:1).

[0069] The computational model may also include myocardial ischemic areas, infarcted areas, and / or other similar areas. In this case, myocardial ischemic or infarcted areas may be identified from perfusion images and / or sicogibi perfusion images obtained, for example, during stress and rest. Myocardial ischemic or infarcted boundary areas may be demarcated on the generated anatomical grid of the heart. For example, a patient may have a posterior space infarct and may have a poor infarct. These areas may be registered in the computational model as binary fields of normal and infarcted tissue.

[0070] The computational model may also include myocardial conductivity properties, such as the conductivity and lateral orientation of the left and right ventricular endocardium or most myocardial tissue in the myofiber and the conductivity in the border zone and / or infarcted or ischemic zone. The potential can be described by a human ventricular myocyte model modified to accommodate changes in channel dynamics that occur during heart failure. Action potential propagation can be modeled in a single-domain or dual-domain reaction-diffusion mathematical framework. The conductivity of the ventricular region can be divided into the left and right ventricular subendocardial regions (e.g., ~3mm transmural adjacent ventricular cavity), the infarcted region, and the remaining majority of myocardium. If not explicitly modeled, the conductivity in the endocardial region can be allowed to vary up to about 10 times the conductivity of a large amount of myocardium to explain the rapid conduction of the Purkinje system. In the infarcted or ischemic region, the conductivity can be isotropic, and the conductivity can be allowed to vary between about 10%-90% of the conductivity in most myocardium.

[0071] Figure 3AA biventricular computational model 300 of a left ventricle 310 and a right ventricle 310 is depicted in accordance with some exemplary implementations. The computational model 300 is shown as a series of finite states of the left ventricle 310 with simulated myocardial voltage maps 320a-d (collectively, fibrillation maps 320) of myocardial voltage potentials in different locations. However, as described herein, the computational model 300 may be a three-dimensional model of the left ventricle 310 and the right ventricle 310 with a fibrillation source map 320 (e.g., as a fourth dimension) moving around / through the surface of the left ventricle 310, depending on the pattern indicating VF. In some aspects, the fibrillation source map 320 may represent endocardial, EKG, and / or VCG data.

[0072] In some aspects, the source of VF may be identified based on the fibrillation source map 320. For example, based on the fibrillation map 320, the computing system 110 may determine that the voltage indicative of VF rotates around a specific point / region of the anterior arrhythmia base that may be identified as a rotor in this case (e.g., optionally as a focal activation among others), such as by Figure 3B , as indicated by the highlighted source 330 shown in white in FIG. In some aspects, the source 330 may be the location of the base of the diseased heart.

[0073] In some implementations, the computational model 300 can be displayed via a user interface and can include animations (e.g., showing movement of the fibrillation map 320 and / or intensity / percentage of VF / AF cycles / sizes of the sources 330). In some aspects, the computational model 300 can be a 4D model and / or the fibrillation source map 320 can be considered a fourth dimension. Although only four states of the left ventricle 310 and the fibrillation map 320 are shown, any number of states is possible (e.g., up to infinity). In some implementations, information about the sources 330 (e.g., location, frequency / percentage of activation at a particular location, mechanism, etc.) can be stored and / or displayed to the user.

[0074] Different levels of granularity of determining the location of a fibrillation source are possible. For example, in some implementations, a fibrillation mechanism (e.g., a rotor or focal source), a specific chamber of the heart (e.g., left atrium or right atrium or ventricle), and / or a region of a fibrillation source (e.g., anterior LV) may be identified. However, in some implementations, a specific location may be identified, which may be estimated based on where the fibrillation source is observed to spend a certain amount of time or from where it emanates. The accuracy of identifying the location of a source of VF / AF may be based on the sensitivity of the observed VCG patterns, the robustness of the computational model from which the VCG pattern library is generated, and / or the capabilities of the machine learning method.

[0075] As described above, the computational model may be based on and / or include EKG data. Figure 4AA graph 410 of EKG data calculated from a computational model according to some exemplary implementations is depicted. In some aspects, the EKG data may be generated using a 12-lead EKG sensor device. Although the examples herein relate to endocardial and / or EKG data, the data may additionally or alternatively include other types of electrical data, such as VCG data. For example, Figure 4B Depicted is a VCG 420 model with a VCG trace from a side viewpoint and a VCG model 430 with a trace from a superior viewpoint according to some exemplary implementations. In some aspects, the black heart model located inside the VCG trace is used only for perspective and may not form part of the VCG traces 420, 430. In some implementations, the VCG models 420, 430 may be generated based on EKG data.

[0076] As described above, the VCG models 420, 430 can be spatiotemporal VCG models constructed from a large number of realistic biventricular computational models of VF / AF (e.g., computational models in the model library 192 and / or the patient-specific model 194). In some implementations, the model library 192 can include models that simulate rotor / lesion sources from different anatomical heart segments. More or fewer locations are possible, and in some aspects, the simulated locations can contain most potential source locations.

[0077] In some aspects, the coordinate axes of the VCG data and / or the coordinate axes of the corresponding computational model may be rotated until they are properly aligned with each other, as described above with respect to Figure 1 As described in the data transformation algorithm 197.

[0078] Figure 5A Depicted is a fibrillation map 500 in a patient's left ventricle 510 during VF, according to some exemplary implementations. Figure 2A Fibrillation source map 200, fibrillation map 500 may indicate VF rotors at source 520. For example, Figure 5B A fibrillation source map 550 of a rotor is depicted according to some exemplary implementations. In some aspects, the map 550 can be a posterior view of a patient 560. As shown in the cross-sectional view 570 of the fibrillation source map 550, there can be areas of scar tissue 580, border zone tissue 585, and / or normal tissue 590 in the left ventricle of the patient. In some aspects, the myocardial scar / border zone tissue 580, 585 can sustain VF and therefore should be identified for ablation and / or external radiation therapy in order to reduce the likelihood of future arrhythmias in the patient 560. In some aspects, the scar tissue 580 and / or border zone tissue 585 can be considered a cardiac substrate for arrhythmias.

[0079] To identify the source of VF / AF in a patient, the computational model described herein may be compared to data obtained from the patient. In some aspects, the patient may know that they have VF / AF, but not the location of the source of the VF / AF. At this point, the patient may wish to understand the exact source of their VF / AF (e.g., how bad the patient's VF / AF is and / or their risk of cardiac death), whether an implantable cardioverter defibrillator (ICD) is beneficial, whether ablation or surgery is beneficial for preventing future arrhythmias, what the risks of surgery are, and so on. Using the subject matter described herein, a patient may be able to see a doctor who performs non-invasive measurements on the patient, provide the measurement data to a computing device, and receive data to better address the patient's questions and concerns.

[0080] In some implementations, non-invasive measurements may include obtaining EKG data from the patient, such as by using a 12-lead EKG sensor. Fig. 6A A graph 610 depicts EKG readings during VF according to some exemplary implementations. In some aspects, the EKG data is obtained during induced VF / AF. Based on the EKG readings, a VCG model and / or trace may be generated for the patient. For example, Figure 6B Depicted are a VCG model 620 from a side viewpoint and a VCG model 630 from a superior viewpoint according to some exemplary implementations. In some aspects, the black heart model is provided for orientation only and does not constitute part of the VCG models 620, 630 or the VCG traces. In some implementations, the VCG models 620, 630 can be generated without using EKG readings.

[0081] For example, once the VCG models 620, 630 are obtained for a patient, they can be compared with the VCG models stored in the VCG library 196. Based on the calculated similarity between the patient-derived VCG models 620, 630 and the simulated / stored VCG models, VF / AF sources can be identified. In some aspects, the similarity can be measured in one or more (e.g., two) orthogonal planes. In some implementations, the VCGs can be compared using a Pearson correlation or similar comparison algorithm. Based on the correlation factor, VF / AF sources can be identified in the patient.

[0082] Since a patient may have more than one VF / AF source, multiple comparisons may be performed. For example, VCGs may be obtained from a patient over a series of time intervals. In some aspects, VCGs may be generated at each time interval of 1 ms duration for each VF cycle (which may be approximately 200 ms in duration). Each VF cycle VCG may be compared with a stored VF cycle VCG to identify which VCG is most closely related to the patient. A corresponding computational model from the model library 192 may be identified based on which VCG has the highest correlation. Data associated with the identified computational model (including but not limited to the type of VF source, location, direction of rotation (if any), etc. as described herein) may be considered to represent the patient's heart during a specific time interval. Based on at least a portion of the identified computational model, an estimated computational model may be generated for the patient. In some implementations, a subset of computational models of interest from the model library 192 may be identified based on the shape, scar location, conduction characteristics, etc. of the patient's heart (e.g., as determined by CT, MRI, echo, X-ray, fluoroscopy, or other imaging data, etc.). The identified set of computational models may be associated with a VCG from VCG library 196, which may be the VCG selected for comparison with the patient.

[0083] Different rotor positions can create significantly different VCG loops. Therefore, comparison of the generation of patient-derived VCGs with known / estimated patterns of VF / AF sources can serve as an accurate method for identifying VF / AF sources in patients.

[0084] In some implementations, a method such as Figure 2C The fibrillation source map 280 and / or Figure 2D A fibrillation source map, such as fibrillation source map 290, can be used to illustrate the location of the VF / AF source. For example, based on comparing VCGs, it can be determined how long the VF / AF source spends at one or more locations. This time / percentage information can be used to generate a fibrillation source map, and / or the map can be overlaid on the cardiac imaging data. Utilizing a fibrillation source map to display cardiac imaging data (e.g., a 3D model) can allow a medical professional to identify additional characteristics about the VF / AF source, and / or target the source location for radiofrequency sources, cryogenics, ultrasound or laser ablation, external beam radiation, revascularization, gene transfer therapy, or other interventions to reduce future arrhythmia burden.

[0085] In some implementations, additional computational models can be generated based on a comparison of a patient-derived VCG with a stored VCG. For example, one or more models that are most similar to a VF / AF source identified within a patient can be identified from the model library 192. These one or more models can be combined to generate a composite model, which can include data representing all VF / AF sources in the patient. Using the composite model, one VF / AF source can be removed (e.g., simulated removal) to generate a new model, which can be compared to the composite model to determine whether the removal of each VF / AF source is beneficial to the patient. If the VF / AF change seen between the composite model and the new model is below a threshold / percentage, it can be determined that the removal of the VF / AF source may be detrimental to the patient. The threshold / percentage can vary depending on the patient, medical professional, and / or other factors.

[0086] Although the above measurements / comparisons are with respect to the patient's left ventricle, the measurements / comparisons may be performed in a similar manner, but with respect to the patient's right ventricle. For example, Fig. 7A Depicted is a computational model 700 of a right ventricle 710 according to some exemplary implementations. As shown, the computational model 700 can include a plurality of myocardial voltage maps 710a-d, which are shown as a series of finite states that can form a map of cardiac electrical activation in the right ventricle 710.

[0087] The computational model 700 can be based on EKG data and / or can be used to generate EKG data. For example, Figure 7B A graph 730 of EKG readings generated from a computational model according to some exemplary implementations is depicted. A process for creating an EKG reading may be performed by first calculating the VCG from the electrical dipoles of the computational model for each time period. Then an electrocardiogram trace may be calculated. For example, Figure 7C Depicted are VCG models 740, 750 generated from a heart model according to some exemplary implementations. One or more VCG models are calculated according to the computational model, such as a VCG model 740 from a lateral viewpoint and a VCG model 750 from a vertical viewpoint. An electrocardiogram trace can then be derived from these data. Importantly, as described herein, the VCG models 740, 750 allow comparison and matching with VCG data from a patient. As shown, the VCG models 740, 750 are colored based on timing information.

[0088] While the above measurements / comparisons are of the left and right ventricles of a patient in VF, measurements / comparisons may be made in a similar manner, but with respect to the right or left atrium of a patient during AF. For example, Fig. 8ADepicted is a computational model 800 of the left atrium and right atrium 810 according to some exemplary implementations. As shown, the computational model 800 can include myocardial voltage maps 820a-d, which can represent and / or be based on endocardial and / or EKG data. From the computational model 800, the location of the source of AF (white dots) can be identified. For example, Figure 8B Depicted is a computational model 800 of a right atrium 810 with an identified rotor base 830 (represented by a white dot), according to some exemplary implementations.

[0089] Fig. 9 An example computing device 900 is shown that can be used to implement one or more of the described devices and / or components according to some exemplary implementations. For example, at least a portion of the computing device 900 can be used to implement at least a portion of the computing device 110, a device providing a database 120, a device providing external software 130, one or more of the user access devices 140, one or more of the sensor devices 150, and / or the access device 165. The computing device 900 can perform one or more of the processes described herein.

[0090] As shown, computing device 900 may include one or more processors, such as processor 910, to execute instructions that may implement operations consistent with those described herein. Device 900 may include memory 920 for storing executable instructions and / or information. Memory 920 may include solid-state memory, a solid-state disk drive, a disk drive, or any other information storage device. In some aspects, memory 920 may provide storage for at least a portion of a database (e.g., database 120 or some other data organization). Device 900 may include access to a wired network or a wireless network (e.g., Figure 1 The network interface 940 may include a network interface 940 for a wireless network 160). The wireless network may include WiFi, WiMax, and cellular networks (2G / 3G / 4G / 5G), and / or any other wireless network. To enable wireless communication, the network interface 940 may, for example, utilize one or more antennas, such as antenna 980.

[0091] The apparatus 900 may include one or more user interfaces, such as a user interface 950. The user interface 950 may include a hardware or software interface, such as a keyboard, a mouse, or other interface, some of which may include a touch screen integrated with the display 930. The display 930 may be used to display information such as promotional offers or current inventory, provide prompts to the user, receive user input, and / or the like. In various implementations, the user interface 950 may include one or more peripheral devices and / or the user interface 950 may be configured to communicate with these peripheral devices.

[0092] In some aspects, the user interface 950 may include one or more of the sensors described herein, and / or may include an interface to one or more sensors described herein. The operation of these sensors may be controlled at least in part by the sensor module 960. The device 900 may also include and input and output filter 970, which may filter information received from a sensor or other user interface, received and / or sent through a network interface and / or the like. For example, a signal detected by a sensor may pass through a filter 970 for appropriate signal conditioning, and then the filtered data may be passed to a microcontroller sensor module 960 and / or a processor 910 for verification and processing (e.g., before sending a result or indication via a network interface 940). The device 900 may be powered by using one or more power supplies (e.g., a power supply 990). As shown, one or more of the components of the device 900 may transmit and / or receive power through a system bus 999. In some aspects, the device 900 may be considered to be at least a portion of a server or device utilizing at least a portion of a server.

[0093] Fig.10 A flow chart of a method for computational localization of a fibrillation source according to some exemplary implementations is shown. In various implementations, method 1000 (or at least a portion thereof) may be performed by one or more of computing device 110, an apparatus providing database 120, an apparatus providing external software 130, one or more of user access devices 140, one or more of sensor devices 150, access device 165, computing device 900, other related devices, and / or a portion thereof.

[0094] Method 1000 may begin at operation block 1010, where apparatus 900 may, for example, generate a heart model with different shapes, scars, and fibrillation sources. Other aspects such as geometry, fiber orientation, fibrillation source type, fibrillation source location, etc. may additionally or alternatively be used.

[0095] Method 1000 may proceed to operation block 1020, where apparatus 900 may generate a vector cardiogram plot, for example, based on a heart model. For example, a heart model may be used to calculate a model VCG and / or generate a comprehensive VCG library.

[0096] Method 1000 can proceed to operation block 1030, where apparatus 900 can filter the heart model, for example, based on the patient's heart shape and / or scar. In some aspects, filtering can be based on computed tomography imaging data, magnetic resonance imaging data, echocardiography data, X-ray data, fluoroscopy data, etc.

[0097] The method 1000 may proceed to operation block 1040 where the apparatus 900 may generate a vectorcardiogram plot, for example, based on the patient's heart.

[0098] The method 1000 may proceed to operation block 1050, where the apparatus 900 may, for example, compare the vector cardiogram drawing of the patient to a heart model vector cardiogram drawing. The vector cardiogram drawing may be retrieved from a VCG library. In some implementations, a subset of the vector cardiogram drawings in the VCG library may be selected based on the filtered heart model. The comparison may include determining a correlation factor between the generated VCG and the stored VCG for each stored VCG, and at least one matching VCG may be identified as the stored VCG having the highest correlation factor.

[0099] The method 1000 may proceed to operation block 1060, where the apparatus 900 may, for example, generate a map of fibrillation sources in the patient's heart. In some implementations, the computational model may include a three-dimensional mesh in the shape of the heart and a finite state fibrillation source map of cardiac electrical activation mapped to the mesh. In some implementations, the method 1000 may involve generating an electrocardiogram plot based on the patient's heart and / or generating a representation of electrical activation of the patient's heart based on the electrocardiogram plot.

[0100] The method 1000 may proceed to operation block 1070, where the apparatus 900 may display a map, for example, via a user interface (e.g., user interface 950). The map may be displayed using a heart model, which may include one or more of a left atrium, a right atrium, a ventricle, or a right ventricle. The model may be two-dimensional, three-dimensional, four-dimensional, and / or may be animated. In some aspects, the display may enable the use of radiofrequency, cryogenics, ultrasound, laser, external beam radiation, etc. to guide an ablation procedure.

[0101] In some implementations, method 1000 can involve generating a plurality of vectorcardiogram plots based on one or more stored computational models, and / or generating one or more electrocardiogram plots based on a plurality of vectorcardiogram plots.

[0102] Execution of method 1000 and / or a portion thereof may allow for identification of VF / AF sources with greater accuracy, reduced invasiveness, and / or reduced cost.

[0103] Fig.11 According to some exemplary implementations (e.g., similar to Fig.121000 (or at least a portion thereof) is a flowchart of a method for computationally locating a fibrillation source. In various implementations, method 1000 (or at least a portion thereof) may be performed by one or more of computing device 110, an apparatus providing database 120, an apparatus providing external software 130, one or more of user access devices 140, one or more of sensor devices 150, access device 165, computing device 900, other related devices, and / or a portion thereof.

[0104] Method 1100 may begin at operation block 1110, where apparatus 900, for example, may determine a source of fibrillation in a patient's heart. In some implementations, identification of the source of fibrillation may be combined with information about Fig.10 The identification procedure discussed in method 1000 is similar.

[0105] The method 1100 may proceed to operation block 1120 , where the apparatus 900 may generate a heart model based on the patient's heart shape, scar, and / or fibrillation source, for example.

[0106] Method 1100 may proceed to operation block 1130, where apparatus 900 may, for example, generate another cardiac model with at least one fibrillation source removed. In some aspects, the computational model may be generated by removing one of the one or more fibrillation sources from the original computational model.

[0107] The method 1100 may proceed to operation block 1140 where the apparatus 900 may determine changes in fibrillation in another heart model, for example.

[0108] If there are additional sources to evaluate, the method 1100 can return to operation block 1130. For example, if three VF / AF sources are identified, the method 1100 can evaluate the removal of each of the sources individually or in combination with any number of two additional sources (e.g., seven unique evaluations). Through this process, a user (e.g., a medical professional) can identify the best course of action in removing the VF / AF sources.

[0109] The method 1100 may proceed to operation block 1150, where the device 900 may display at least a portion of a cardiac model, for example, via a user interface (e.g., user interface 950). The cardiac model may include one or more of a left atrium, a right atrium, a ventricle, or a right ventricle. The model may be two-dimensional, three-dimensional, four-dimensional, and / or may be animated. In some aspects, a side-by-side comparison of models with and without VF / AF sources may be provided. In some aspects, the method 1100 may include determining fibrillation changes between the new computational model and the original computational model. In some aspects, displaying a side-by-side comparison may enable targeted ventricular fibrillation or ventricular fibrillation ablation therapy by providing predicted ablation outcome data and / or selection of an optimized ablation strategy.

[0110] Execution of method 1100 and / or a portion thereof may allow for a determination of whether elimination of each individual VF / AF source may benefit the patient. For example, if the change in VF / AF in the model in which a particular source is removed does not decrease by a threshold amount (e.g., a percentage), an invasive procedure to remove (or reduce the strength of a particular source) may be avoided.

[0111] Although several aspects are described herein with respect to VF and AF, other embodiments are possible. For example, other arrhythmias or patterns in humans and / or animals may be observed and / or their sources may be identified. If the source of the abnormal pattern in a patient is identified, it may be displayed (e.g., modeled in two or three dimensions) to a user (e.g., a medical professional) so that the user can see inside the patient's body without making an incision.

[0112] One or more aspects or features of the subject matter described herein may be implemented in digital electronic circuits, integrated circuits, specially designed application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) computer hardware, firmware, software, or a combination thereof. These various aspects or features may be implemented in one or more computer programs executable and / or interpretable on a programmable system, the programmable system including at least one programmable processor, which may be special or general, coupled to receive data and instructions from a storage system, at least one input device, and at least one output device and to send data and instructions to the storage system, at least one input device, and at least one output device. A programmable system or computing system may include a client and a server. The client and the server are usually remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by a computer program running on each computer, and has a client-server relationship with each other.

[0113] These computer programs, which may also be referred to as programs, software, software applications, applications, components or codes, include machine instructions for programmable processors and may be implemented in high-level procedural and / or object-oriented programming languages ​​and / or assembly / machine languages. As used herein, the term "machine-readable medium" refers to any computer program product, device and / or apparatus for providing machine instructions and / or data to a programmable processor, such as, for example, a disk, an optical disk, a memory and a programmable logic device (PLD), including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor. A machine-readable medium may store such machine instructions non-temporarily, such as, for example, a non-temporary solid-state memory or a magnetic hard drive or any equivalent storage medium. A machine-readable medium may optionally or additionally store such machine instructions in a temporary manner, such as, for example, a processor cache or other random access memory associated with one or more physical processor cores.

[0114] In order to provide interaction with a user, one or more aspects or features of the subject matter described herein can be implemented on a computer with a display device, such as a cathode ray tube (CRT) or a liquid crystal display (LCD) or a light emitting diode (LED) monitor for displaying information to a user, and a keyboard and a pointing device (such as a mouse or trackball) through which a user can provide input to the computer. Other types of devices can also be used to provide interaction with a user. For example, the feedback provided to the user can be any form of sensory feedback, such as visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic input, voice input, tactile input, and / or the like. Other possible input devices include touch screens or other touch-sensitive devices, such as single-point or multi-point resistive or capacitive touch pads, voice recognition hardware and software, optical scanners, optical pointers, digital image capture devices, and associated interpretation software, etc.

[0115] Depending on the desired configuration, the subject matter described herein may be embodied in systems, devices, methods and / or articles. The implementations set forth in the foregoing description do not represent all embodiments consistent with the subject matter described herein. On the contrary, they are merely some examples of aspects consistent with the described subject matter. Although some variations have been described in detail above, other modifications or additions are also possible. In particular, in addition to those features and / or variants set forth herein, other features and / or variants may also be provided. For example, the embodiments described above may be directed to various combinations and sub-combinations of the disclosed features and / or combinations and sub-combinations of several other features disclosed above.

[0116] In this article, phrases such as "at least one" or "one or more" may appear before a joint list of elements or features. The term "and / or" may also appear in a list of two or more elements or features. Unless otherwise implicitly or explicitly contradicted by the context in which it is used, such phrases are intended to represent any one of the elements or features listed separately or any one of the elements or features listed with any one of the other listed elements or features. For example, the phrases "at least one of A and B", "one or more of A and B" and "A and / or B" are each intended to represent "A alone, B alone or A and B together". Similar explanations also apply to lists including three or more items. For example, the phrases "at least one of A, B and C;", "one or more of A, B and C" and "A, B and / or C" are each intended to represent "A alone, B alone, C alone, A and B together, A and C together, B and C together or A and B and C together". The term "based on" used in this article is intended to represent "based at least in part", so that unlisted features or elements are also allowed.

[0117] The method shown is exemplary only. Although the method is shown as having a specific operation flow, two or more operations may be combined into a single operation, a single operation may be performed in two or more separate operations, one or more of the operations shown may not exist in various implementations, and / or additional operations not shown may be part of the method. In addition, the logic flow described in the drawings and / or described herein does not necessarily require the specific order shown or the sequential order to achieve the desired result.

Claims

1. A method performed by one or more computing systems, the method comprising: accessing a library comprising a plurality of computational models, each of the plurality of computational models being associated with a set of cardiac electrical activations mapped to a three-dimensional finite element cardiac mesh, each of the plurality of computational models being further associated with a set of cardiac characteristics including source locations of a source of an arrhythmia, at least two sets of electrical activations being associated with different source locations; generating a first simulated cardiogram based on at least a first set of electrical activations associated with a first computational model in the library, the first simulated cardiogram being associated with a first source location of the first computational model; generating a second simulated cardiogram associated with a second source location of a second computational model in the library based on at least a second set of electrical activations associated with the second computational model, the second computational model being associated with a different heart shape and / or a different scar location when compared to the first computational model; training a machine learning algorithm based on at least the first simulated cardiogram and the second simulated cardiogram, the trained machine learning algorithm configured to receive an input comprising a patient cardiogram collected from a patient and to generate an output comprising a patient source location of a patient source of an arrhythmia associated with the patient cardiogram; and A treatment of the patient is determined based at least on the output of the trained machine learning algorithm, the treatment comprising ablation, and the patient source location output by the machine learning algorithm is a target source location for the ablation.

2. The method according to claim 1, further comprising: generating the plurality of computational models based on a plurality of sets of cardiac characteristics including two or more different source locations of arrhythmias; and The plurality of computational models are stored in the library.

3. The method of claim 1 , wherein the machine learning algorithm is further trained based on non-simulated echocardiograms generated by a model of a specific patient associated with a set of specific patient characteristics.

4. The method of claim 1, wherein the plurality of computational models are associated with non-patient-specific characteristics.

5. The method of claim 1 further comprising applying a trained machine learning algorithm to the patient electrocardiogram to identify the patient source location of the patient source of the arrhythmia as the target source location for the ablation. The method according to claim 1 , wherein the first simulated cardiogram and / or the second simulated cardiogram comprises an electrocardiogram.

7. The method of claim 1, wherein the first simulated cardiogram and / or the second simulated cardiogram comprises a vector cardiogram.

8. The method of claim 1, wherein the plurality of computational models are associated with specific arrhythmia sources including rotor or focal sources.

9. The method of claim 1, wherein the machine learning algorithm is trained to identify one or more source locations of specific arrhythmia sources including rotors or focal sources.

10. The method of claim 1, wherein the machine learning algorithm is trained using cycles derived from the cardiograms, and wherein the cycles derived from the cardiograms are associated with source locations associated with the corresponding cardiograms from which the cycles were derived.

11. One or more computing systems, including: one or more processors; and One or more memories storing instructions that, when executed by the one or more processors, result in operations including: accessing a plurality of computational models derived from electrical activation of the heart, each of the plurality of computational models being associated with a characteristic of a pattern of electrical activity of the heart, the characteristic being a source location of a source of the pattern of electrical activity of the heart, the electrical activation being mapped to a three-dimensional finite element mesh representing the heart, at least two of the electrical activations being associated with different source locations; generating a first simulated representation of the cardiac electrical activation based on at least a first electrical activation associated with a first computational model; generating a second simulated representation of the cardiac electrical activation based at least on second electrical activation associated with a second computational model, the second computational model being associated with a different heart shape and / or a different scar location when compared to the first computational model; training a machine learning algorithm based on at least the first simulated representation and the second simulated representation, the trained machine learning algorithm configured to receive an input including a patient representation of electrical activation of the patient's heart and generate an output identifying a patient source location of a patient source of a pattern of electrical activity of the patient's heart; and determining a treatment for the patient based on the output of the machine learning algorithm, Wherein, the electrically activated first simulated representation, the electrically activated second simulated representation and the electrically activated patient representation are echocardiograms.

12. The one or more computing systems of claim 11, wherein the instructions further cause operations comprising: Based on each of the plurality of sets of characteristics of the heart, a corresponding one of a plurality of computational models of cardiac electrical activation is generated.

13. The one or more computing systems of claim 12, wherein each of the plurality of computing models is associated with a source type of the source.

14. The one or more computing systems of claim 11, wherein the first simulated representation and the second simulated representation are each associated with a characteristic that is a source type of a pattern of cardiac electrical activity.

15. One or more computing systems according to claim 11, wherein the electrical activity pattern is an arrhythmia, and wherein the machine learning algorithm is trained to identify one or more source locations of a particular electrical activity pattern source including a rotor or a focal source.

16. A method for identifying a source location of a source of a cardiac arrhythmia in a patient, performed by one or more computing systems, the method comprising: collecting a patient echocardiogram from the patient; applying a machine learning algorithm to the patient's electrocardiogram to identify source locations of sources of arrhythmias in the patient, the machine learning algorithm being trained based on a plurality of training electrocardiograms, the plurality of training electrocardiograms comprising a first simulated electrocardiogram and a second simulated electrocardiogram, the first simulated electrocardiogram being generated based on a first computational model, the second simulated electrocardiogram being generated based on a second computational model, each of the first computational model and the second computational model being associated with a set of characteristics of the heart including one or more source locations of a plurality of sources of arrhythmias, the second computational model being associated with a different heart shape and / or a different scar location when compared to the first computational model, and each of the first computational model and the second computational model being further associated with a set of cardiac electrical activations mapped to a three-dimensional finite element heart mesh; and Treatment for the patient is determined based at least on an output of the machine learning algorithm that identifies a source location of a source of an arrhythmia in the patient.

17. The method of claim 16, wherein the plurality of training cardiograms further comprises non-simulated cardiograms generated based on a patient-specific model associated with a set of patient-specific characteristics.

18. The method of claim 16, wherein the machine learning algorithm is trained to identify one or more source locations of specific arrhythmia sources including rotors or focal sources.

19. The method of claim 16, wherein the machine learning algorithm is stored in a cloud-based system and the patient electrocardiogram is provided via a gateway to the cloud-based system to identify a source location of a source of the patient's arrhythmia.

20. The method of claim 16, further comprising outputting an indication of the identified source location to guide an ablation procedure on the patient's heart.

21. One or more computing systems comprising: one or more processors; and One or more memories storing instructions that, when executed by the one or more processors, result in operations including: collecting a patient representation of the patient's cardiac electrical activation; applying a machine learning algorithm to the patient representation of the electrical activation to identify a characteristic of the patient's cardiac electrical activity pattern, the characteristic being a source location of the cardiac electrical activity pattern, training the machine learning algorithm based on a plurality of training representations of the electrical activation, the plurality of training representations of the electrical activation comprising a first simulated representation of the electrical activation and a second simulated representation of the electrical activation, each of the first simulated representation and the second simulated representation being associated with a respective source location of the electrical activity pattern, each of the one or more simulated representations of the electrical activation generated based on a computational model comprising the electrical activation mapped to a three-dimensional finite element mesh representing the heart, the first simulated representation and the second simulated representation being generated based on the computational model associated with different heart shapes and / or different scar locations; and determining a treatment for the patient based at least on an output of the machine learning algorithm that identifies a source location of a pattern of cardiac electrical activity in the patient, Wherein, the plurality of electrically activated training representations and the electrically activated patient representation are electrocardiograms.

22. The one or more computing systems of claim 21, wherein the plurality of trained representations of electrical activations further include non-simulated representations of electrical activations generated based on a patient-specific model associated with a set of patient-specific characteristics.

23. One or more computing systems according to claim 21, wherein the pattern of electrical activity is an arrhythmia, and wherein the machine learning algorithm is trained to identify one or more source locations of specific arrhythmia sources including rotors or focal sources.

24. One or more computing systems according to claim 21, wherein the machine learning algorithm is stored in a cloud-based system and provides an electrically activated patient representation via a gateway of the cloud-based system to identify a source location of a source of an electrical activity pattern of the patient's heart.

25. The one or more computing systems of claim 24, wherein the instructions further cause operations to output an indication of the identified source location to guide an ablation procedure on the patient's heart.

26. One or more computing systems comprising: one or more processors; and One or more memories storing instructions that, when executed by the one or more processors, result in operations including: collecting a patient representation of the patient's cardiac electrical activation; applying a machine learning algorithm to a patient representation of the electrical activations to identify a characteristic of a pattern of cardiac electrical activity of the patient, the characteristic being a source location of the cardiac electrical activity pattern, the machine learning algorithm being trained based on a plurality of training representations of the electrical activations, the plurality of training representations of the electrical activations comprising a first simulated representation of the electrical activations and a second simulated representation of the electrical activations, each of the first simulated representation and the second simulated representation being associated with a respective source location of the pattern of electrical activity, each of the one or more simulated representations of the electrical activations generated based on a computational model comprising the electrical activations mapped to a three-dimensional finite element mesh representing the heart, the first simulated representation and the second simulated representation being generated based on the computational model associated with different heart shapes and / or different scar locations; determining a treatment for the patient based at least on an output of the machine learning algorithm that identifies a source location of a pattern of cardiac electrical activity of the patient; and A graphic representing a three-dimensional representation of a heart is displayed, the graphic including an indication of the identified source location, wherein the plurality of electrically activated training representations and the electrically activated patient representation are an echocardiogram.

27. The one or more computing systems of claim 26, wherein the operations further comprise displaying, via a user interface, an animation showing movement of a source map of a computational model.

Images