Arrythmia detection with feature delineation and machine learning

Abstract

To provide a technology for using both feature description and machine learning for detecting cardiac arrhythmia.SOLUTION: A computing device receives cardiac electrogram data of a patient sensed by a medical device. The computing device acquires a first classification of arrhythmia in the patient via feature-based delineation of the cardiac electrogram data. The computing device applies a machine learning model to the received cardiac electrogram data to acquire a second classification of arrhythmia in the patient. As an example, the computing device determines whether an episode of arrhythmia has occurred in the patient by using the first classification and the second classification. As another example, the computing device verifies the first classification of arrhythmia in the patient by using the second classification. The computing device outputs a report indicating that the episode of arrhythmia has occurred and one or more cardiac features that coincide with the episode of arrhythmia.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates generally to medical devices, and more particularly to implantable medical devices. [Background technology]

[0002] Malignant tachyarrhythmias, such as ventricular fibrillation, are uncoordinated contractions of the myocardium of the heart's ventricles and are the most commonly identified arrhythmias in cardiac arrest patients. If this arrhythmia continues for more than a few seconds, it can lead to cardiogenic shock and cessation of effective blood circulation. As a result, sudden cardiac death (SCD) can occur in a matter of minutes.

[0003] In patients at high risk for ventricular fibrillation, the use of implantable medical devices (IMDs), such as implantable cardioverter defibrillators (ICDs), has been shown to be beneficial in preventing SCD. ICDs are battery-powered electrical shock devices that may include an electrical housing electrode (sometimes referred to as a can electrode) and are typically coupled to one or more electrical leads placed within the heart. If an arrhythmia is sensed, the ICD can transmit a pulse using the electrical leads to shock the heart and restore normal rhythm. Some ICDs are configured to terminate a detected tachyarrhythmia by delivery of anti-tachycardia pacing (ATP) prior to delivery of the shock. In addition, ICDs are configured to deliver relatively high amplitude post-shock pacing after successful termination of a shockable tachyarrhythmia to support the heart as it recovers from the shock. Some ICDs also deliver bradycardia pacing, cardiac resynchronization therapy (CRT), or other forms of pacing.

[0004] Other types of medical devices may be used for diagnostic purposes. For example, an implantable or non-implantable medical device may monitor a patient's heart. A user, such as a physician, may review data generated by the medical device for occurrences of cardiac arrhythmias, such as atrial or ventricular tachyarrhythmias, or asystole. The user may diagnose a patient's medical condition based on the occurrence of identified cardiac arrhythmias. Summary of the Invention

[0005] In accordance with the techniques of the present disclosure, a medical device system is described herein that uses both characterization and machine learning to detect and classify a patient's cardiac arrhythmia. For example, a computing device receives a patient's electrocardiogram data sensed by an implantable medical device. The computing device obtains a first classification of the patient's arrhythmia via the feature-based characterization of the electrocardiogram data. The computing device applies a machine learning model to the received electrocardiogram data to obtain a second classification of the patient's arrhythmia. As an example, the computing device uses the first and second classifications to determine whether an arrhythmia episode has occurred in the patient. As another example, the computing device uses the second classification of the arrhythmia obtained from the machine learning model to verify the first classification of the patient's arrhythmia obtained from the feature-based characterization.

[0006] In response to determining that an arrhythmia episode has occurred in the patient, the computing device outputs a report indicating that an arrhythmia episode has occurred and one or more cardiac characteristics consistent with the arrhythmia episode. In response to the report, the computing device may receive one or more adjustments to one or more parameters used by the implantable medical device to sense the patient's electrocardiogram data and to perform such adjustments to the implantable medical device.

[0007] Further, the medical device system described herein may classify the arrhythmia according to an arrhythmia dictionary. For example, the computing device may determine that an arrhythmia episode has occurred in a patient via a feature-based representation of the patient's electrocardiogram data. The computing device may apply a machine learning model to compare cardiac features consistent with the arrhythmia episode with cardiac features of the patient's past arrhythmia episodes to classify the arrhythmia episode as a particular type of arrhythmia episode.

[0008] The techniques of the present disclosure may provide specific improvements in the field of cardiac arrhythmia detection and classification. For example, the use of both characterization and machine learning in combination with one another may improve the accuracy of detection of a patient's arrhythmia over the use of characterization or machine learning individually. Furthermore, the medical device system described herein may enable an implantable medical device of the medical device system to act as a low-granularity filter for detecting arrhythmias in a patient while offloading the power-intensive and computationally complex verification of arrhythmia detection to an external computing device. Thus, a system as described herein may provide high accuracy in arrhythmia detection and classification while reducing power usage and improving the battery life of a device implanted within a patient. Such improvements may also be achieved in a patient monitor in the form of a low-power external device, such as a wearable patch, watch, necklace, or other device worn by the patient, that can detect arrhythmias based on the electrical signals of the heart.

[0009] In one example, the disclosure describes a method including: receiving, by a computing device including a processing circuit and a storage medium, electrocardiogram data of a patient sensed by a medical device; applying, by the computing device, a machine learning model trained using electrocardiogram data of a plurality of patients to the received electrocardiogram data to determine, based on the machine learning model, that an arrhythmia episode has occurred in the patient; performing, by the computing device, feature-based representation of the received electrocardiogram data to obtain cardiac features present in the electrocardiogram data; in response to determining that an arrhythmia episode has occurred in the patient, generating, by the computing device, a report including an indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode; and outputting, by the computing device, the report for display including the indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode.

[0010] In another example, the disclosure describes a method including receiving, by a computing device including a processing circuit and a storage medium, electrocardiogram data of a patient sensed by a medical device; obtaining, by the computing device, a first classification of the patient's arrhythmia determined by a feature-based representation of the received electrocardiogram data, where the feature-based representation identifies cardiac features present in the electrocardiogram data; applying, by the computing device, a machine learning model trained using electrocardiogram data of a plurality of patients to the received electrocardiogram data to determine a second classification of the patient's arrhythmia based on the machine learning model; determining, by the computing device, that an arrhythmia episode has occurred in the patient based on the first classification and the second classification; in response to determining that an arrhythmia episode has occurred in the patient, generating, by the computing device, a report including an indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode; and outputting, by the computing device, for display, the report including the indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode.

[0011] In another example, the disclosure provides a method for determining, by a computing device including a processing circuit and a storage medium, a first classification of arrhythmia for the patient as determined by a feature based representation of the received electrocardiogram data, the feature based representation identifying a first cardiac feature present in the electrocardiogram data that is consistent with the first classification of arrhythmia in the patient; determining, by the computing device, that one or more episodes of arrhythmia of the first classification have previously occurred in the patient; and in response to determining that one or more episodes of arrhythmia of the first classification have previously occurred in the patient, applying, by the computing device, a machine learning model trained using electrocardiogram data of a plurality of patients to the received electrocardiogram data and the first cardiac feature of the electrocardiogram data to determine, based on the machine learning model, A method is described that includes determining that a first cardiac feature is similar to cardiac features consistent with one or more episodes of a first class of arrhythmia that have previously occurred in the patient; determining, by a computing device, in response to determining that the first cardiac feature is similar to cardiac features consistent with one or more episodes of a first class of arrhythmia that have previously occurred in the patient, that an episode of a first class of arrhythmia has occurred in the patient; in response to determining that an episode of a first class of arrhythmia has occurred in the patient, generating, by the computing device, a report including an indication that an episode of a first class of arrhythmia has occurred in the patient and one or more of the cardiac features consistent with the arrhythmia episode; and outputting, by the computing device, the report including an indication that an episode of an arrhythmia has occurred in the patient and one or more of the cardiac features consistent with the arrhythmia episode for display.

[0012] This summary is intended to provide an overview of the subject matter described in the present disclosure. It is not intended to provide an exclusive or comprehensive description of the apparatus and methods described in detail in the accompanying drawings and description below. Further details of one or more examples are set forth in the accompanying drawings and the description below. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 is a conceptual diagram illustrating an example of a medical device system for predicting cardiac arrhythmias including a leadless implantable medical device and a patient-associated external device in accordance with the techniques of the present disclosure. [Diagram 2] FIG. 2 is a block diagram illustrating an example of a leadless implantable medical device of FIG. 1. [Diagram 3] FIG. 2 is a block diagram illustrating another example of the leadless implantable medical device of FIG. 1 . [Figure 4] FIG. 1 is a block diagram illustrating an example computing device that operates in accordance with one or more techniques of the present disclosure. [Diagram 5] 4 is a flowchart illustrating an example operation in accordance with the techniques of this disclosure. [Figure 6] 2 is a chart showing an example of an electrocardiogram obtained from the patient of FIG. 1. [Figure 7] 4 is a flowchart illustrating an example operation in accordance with the techniques of this disclosure. [Figure 8] 4 is a flowchart illustrating an example operation in accordance with the techniques of this disclosure. [Figure 9] 4 is a flowchart illustrating an example operation in accordance with the techniques of this disclosure. [Figure 10] 4 is a flowchart illustrating an example operation in accordance with the techniques of this disclosure.

[0014] Like reference characters refer to like elements throughout the figures and description. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] State-of-the-art signal processing algorithms that perform characterization of electrocardiogram data and machine learning models that process patient data, such as machine learning systems and / or artificial intelligence (AI) algorithms that analyze single-channel and multi-channel patient data Techniques are disclosed for combining multiple decision mechanisms to perform arrhythmia detection and classification. Such patient data may include, for example, electrocardiogram or electrocardiogram (ECG) data.

[0016] As described herein, characterization refers to the use of features obtained through signal processing for use in detecting or classifying episodic cardiac arrhythmias. Typically, characterization involves the use of designed rules to identify or extract features of electrocardiogram data, measure properties of such features, and use the measurements to detect or classify arrhythmias. For example, characterization may be used to identify features such as R waves, QRS complexes, P waves, T waves, the rate of such features, the intervals between such features, the morphology of features, the width or amplitude of such features, or other or other types of cardiac features or properties of features not explicitly described herein. Characterization may include feature extraction, signal filtering, peak detection, refractory analysis, or other types of signal processing, feature engineering, or detection rule development. Characterization algorithms may be optimized for real-time, embedded, and low-power applications, such as for use in implantable medical devices. However, characterization algorithms may require expert design and feature engineering to accurately detect arrhythmias in patients.

[0017] In contrast to characterization techniques for cardiac arrhythmia detection and classification, machine learning techniques may be used for cardiac arrhythmia detection and classification. As described herein, machine learning refers to the use of a machine learning model, such as a neural network or a deep learning model, which is trained on a training data set to detect cardiac arrhythmia from electrocardiogram data. Machine learning techniques contrast with characterization in that characterization relies on signal processing, and a machine learning system may "learn" the underlying features present in electrocardiogram data that indicate arrhythmia episodes without requiring knowledge or understanding of the relationship between the features and arrhythmia episodes on behalf of the system designer.

[0018] Machine learning and AI approaches for arrhythmia detection may provide a flexible platform for developing arrhythmia detection and classification algorithms for a variety of purposes (e.g., detecting atrial fibrillation (AF), ruling out cardiac episodes that do not exhibit arrhythmias, etc.) without the need for expert design or feature engineering required by characterization algorithms. As described in detail herein, techniques, methods, systems, and devices are disclosed that combine characterization and machine learning to detect and classify a patient's cardiac arrhythmias in a manner that improves accuracy and robustness over the use of characterization alone, and also reduces power consumption by an implantable device over the use of machine learning alone.

[0019] FIG. 1 illustrates an example medical device system 2 environment in relation to a patient 4 and a heart 6, in accordance with certain example apparatus and methods described herein. The example technology may be used with an IMD 10, which may be leadless and may communicate wirelessly with an external device 12, as shown in FIG. 1. In some examples, the IMD 10 may be coupled to one or more leads. In some examples, the IMD 10 may be implanted outside the thoracic cavity of the patient 4 (e.g., subcutaneously in the chest as shown in FIG. 1). The IMD 10 may be positioned near the level of the heart 6 and / or just below the sternum.

[0020] In some examples, the IMD 10 may take the form of both a Reveal LINQ™ Insertable Cardiac Monitor (ICM) or a Holter Cardiac Monitor, available from Medtronic plc, Dublin, Ireland. The external device 12 may be a computing device configured for use in settings such as a home, clinic, or hospital, and may be further configured to communicate with the IMD 10 via wireless telemetry. For example, the external device 12 may be coupled to a computing system 24 via a network 25. The computing system 24 may include a remote patient monitoring system, such as Carelink® available from Medtronic plc, Dublin, Ireland. The external device 12 may include, in some examples, a communications device, such as a programmer, an external monitor, or a mobile device, such as a mobile phone, a "smart" phone, a laptop, a tablet computer, a personal digital assistant (PDA), etc.

[0021] In some examples, the example techniques and systems described herein may be used with an external medical device in addition to or instead of the IMD 10. In some examples, the external medical device is a wearable electronic device, such as the SEEQ™ Mobile Cardiac Telemetry (MCT) System available from Medtronic plc, Dublin, Ireland, or other types of “smart” electronic apparel, such as a “smart” watch, a “smart” patch, or “smart” glasses. Such an external medical device may be located external to the patient 4 (e.g., may be located on the skin of the patient 4) and may perform any or all of the functions described herein with respect to the IMD 10.

[0022] In some examples, a user, such as a doctor, technician, surgeon, electrophysiologist, or other clinician, may interact with the external device 12 to retrieve physiological or diagnostic information from the IMD 10. In some examples, a user, such as a patient 4 or a clinician as described above, may also interact with the external device 12 to program the IMD 10, e.g., to select or adjust values ​​of operating parameters of the IMD 10. In some examples, the external device 12 serves as an access point to facilitate communication with the IMD 10 over a network 25, e.g., by a computing system 24. The computing system 24 may include a computing device configured to enable a user to interact with the IMD 10 over the network 25.

[0023] In some examples, the computing system 24 may include at least one of a handheld computing device, a computer workstation, a server or other networked computing device, a smartphone, a tablet, or an external programmer including a user interface for presenting information to a user and receiving input from a user. In some examples, the computing system 24 may include one or more devices implementing a machine learning system 150, such as a neural network, a deep learning system, or other type of predictive analytics system. A user, such as a doctor, technician, surgeon, electrophysiologist, or other clinician, may interact with the computing system 24 to retrieve physiological or diagnostic information from the IMD 10. The user may also interact with the computing system 24 to program the IMD 10, for example, to select values ​​for operating parameters of the IMD. The computing system 24 may include a processor configured to evaluate EGMs and / or other sensed signals transmitted from the IMD 10 to the computing system 24.

[0024] Network 25 may include one or more computing devices (not shown), such as one or more non-edge switches, routers, hubs, gateways, security devices such as firewalls, intrusion detection, and / or intrusion prevention devices, servers, computer terminals, laptops, printers, databases, wireless mobile devices such as cell phones or personal digital assistants, wireless access points, bridges, cable modems, application accelerators, or other network devices. Network 25 may include one or more networks managed by a service provider and thus form part of a larger public network infrastructure such as the Internet. Network 25 may provide access to the Internet for computing devices such as computing system 24 and IMD 10, The network 25 may provide a communications framework that allows the computing devices to communicate with each other. In some examples, the network 25 allows the computing system 24, the IMD 10, and / or the external device 12 to communicate with each other, but separates one or more of the computing system 24, the IMD 10, or the external device 12 from devices outside the network 25 for security purposes. In some examples, communications between the computing system 24, the IMD 10, and the external device 12 are encrypted.

[0025] The external device 12 and the computing system 24 may communicate using wireless communication over the network 25 using any technique known in the art. In some examples, the computing system 24 is a remote device that communicates with the external device 12 through an intermediate device located in the network 25, such as a local access point, a wireless router, or a gateway. In the example of FIG. 1, the external device 12 and the computing system 24 communicate over the network 25, but in some examples, the external device 12 and the computing system 24 communicate directly with each other. Examples of communication techniques may include, for example, communication according to the Bluetooth® or BLE protocols. Other communication techniques are contemplated. The computing system 24 may also communicate with one or more other external devices using a number of known communication techniques, both wired and wireless.

[0026] In any such example, the processing circuitry of the medical device system 2 may transmit patient data, including electrocardiogram data of the patient 4, to a remote computer (e.g., external device 12). In some examples, the processing circuitry of the medical device system 2 may transmit a determination that the patient 4 is experiencing an episode of cardiac arrhythmia, such as an episode of bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block.

[0027] The external device 12 may be a computing device (e.g., used in a home, outpatient, clinic, or hospital setting) for communicating with the IMD 10 via wireless telemetry. The external device 12 may include or be coupled to a remote patient monitoring system, such as Carelink® available from Medtronic plc of Dublin, Ireland. In some examples, the external device 12 may receive data, alerts, patient physiological information, or other information from the IMD 10.

[0028] The external device 12 may be used to program commands or operating parameters into the IMD 10 to control its functions (e.g., when configured as a programmer for the IMD 10). In some examples, the external device 12 may be used to interrogate the IMD 10 to retrieve data, including operational data of the device as well as physiological data stored in the IMD memory. Such interrogations may occur automatically according to a schedule or in response to a remote or local user command. Programmers, external monitors, and consumer devices are examples of external devices 12 that may be used to interrogate the IMD 10. Examples of communication technologies used by the IMD 10 and the external device 12 include radio frequency (RF) telemetry, which may be an RF link established via Bluetooth, WiFi, or Medical Implant Communication Service (MICS). In some examples, the external device 12 may include a user interface configured to allow the patient 4, a clinician, or another user to remotely interact with the IMD 10. In some such examples, the external device 12, and / or other devices of the medical device system 2 may be wearable devices (e.g., in the form of a watch, necklace, or other wearable item).

[0029] The medical device system 2 is configured to perform cardiac arrhythmia detection, verification, and reporting. 4 is an example of a medical device system configured to perform machine learning arrhythmia detection and characterization to detect and classify cardiac arrhythmias in patient 4. In accordance with the techniques of the present disclosure, medical device system 2 performs machine learning arrhythmia detection and characterization to detect and classify cardiac arrhythmias in patient 4. Additional examples of the one or more other implantable or external devices may include an implantable, multi-channel cardiac pacemaker, ICD, IPG, leadless (e.g., intracardiac) pacemaker, extravascular pacemaker and / or ICD, or other IMD, or a combination of such IMDs configured to deliver CRT to heart 6, an external therapy delivery device such as an external monitor, an external pacing or electrical stimulation device, or a drug pump.

[0030] Communications circuitry in each device (e.g., IMD 10 and external device 12) of medical device system 2 may enable the devices to communicate with one another. Additionally, while one or more sensors (e.g., electrodes) are described herein as being disposed on the housing of IMD 10, in other examples, such sensors may be disposed on the housing of another device implanted inside or outside patient 4. In such examples, one or more of the other devices may include processing circuitry configured to receive signals from the electrodes or other sensors on the respective device, and / or communications circuitry configured to transmit signals from the electrodes or other sensors to another device (e.g., external device 12) or a server.

[0031] In accordance with the techniques of the present disclosure, the medical device system 2 detects and classifies cardiac arrhythmias of the patient 4 using both the characterization and machine learning. For example, the computing system 24 receives electrocardiogram data of the patient 4 sensed by the implantable medical device 10. The computing system 24 obtains a first classification of arrhythmia of the patient 4 via feature-based characterization of the electrocardiogram data. In some examples, the feature-based characterization of the electrocardiogram data to determine the first classification of arrhythmia of the patient 4 is performed by any one of the IMD 10, the external device 12, or the computing system 24. The machine learning system 150 applies a machine learning model to the received electrocardiogram data to obtain a second classification of arrhythmia of the patient 4. In one example, the machine learning model is a deep learning model. As an example, the computing system 24 uses the first and second classifications to determine whether an arrhythmia episode has occurred in the patient 4. As another example, the computing system 24 uses the second classification of arrhythmia obtained from the machine learning system 150 to verify the first classification of arrhythmia of the patient 4 obtained from the feature-based characterization.

[0032] In response to determining that an arrhythmia episode has occurred in patient 4, computing system 24 outputs a report indicating that an arrhythmia episode has occurred and one or more cardiac characteristics consistent with an arrhythmia episode. In response to the report, computing system 24 may receive one or more adjustments to one or more parameters used by implantable medical device 10 to sense cardiac electrogram data of patient 4, and implement such adjustments to implantable medical device 10 for subsequent sensing.

[0033] Additionally, the medical device system 2 may classify the arrhythmia according to an arrhythmia dictionary. As described in more detail below, the computing system 24 determines that an arrhythmia episode has occurred in the patient 4 via feature-based depiction of the patient's 4 electrocardiogram data. The machine learning system 150 applies a machine learning model to compare cardiac features consistent with the arrhythmia episode with cardiac features of the patient 4's past arrhythmia episodes to classify the arrhythmia episode as a particular type of arrhythmia episode.

[0034] The techniques of the present disclosure may provide specific improvements in the field of cardiac arrhythmia detection and classification. For example, the use of both characterization and machine learning in combination with one another may provide greater accuracy in detecting arrhythmias in patient 4 than either the use of characterization or the use of machine learning individually. Additionally, the medical device system 2 described herein may enable the implantable medical device 10 to act as a low-granularity filter for detecting arrhythmias in the patient 4 while offloading the power-intensive and computationally complex verification of arrhythmia detection to an external device, such as the external device 12 or the computing system 24. Thus, as described herein, the system 2 may provide high accuracy in detecting and classifying arrhythmias in the patient 4 while reducing power usage and improving battery life of the IMD 10.

[0035] FIGURE 2 is a block diagram illustrating an example of the leadless implantable medical device of FIGURE 1. As shown in FIGURE 2, the IMD 10 includes processing circuitry 50, sensing circuitry 52, communication circuitry 54, memory 56, sensor 58, switching circuitry 60, and electrodes 16A, 16B (hereinafter "electrodes 16"), one or more of which may be disposed within a housing of the IMD 10. In some examples, the memory 56 includes computer-readable instructions that, when executed by the processing circuitry 50, cause the IMD 10 and processing circuitry 50 to perform various functions attributed to the IMD 10 and processing circuitry 50 herein. The memory 56 may include any volatile, nonvolatile, magnetic, optical, or electrical medium, such as random access memory (RAM), read-only memory (ROM), nonvolatile RAM (NVRAM), electrically erasable programmable ROM (EEPROM), flash memory, or any other digital medium.

[0036] The processing circuitry 50 may include fixed function circuits and / or programmable processing circuits. The processing circuitry 50 may include any one or more of a microprocessor, a controller, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or equivalent discrete or analog logic circuitry. In some examples, the processing circuitry 50 may include multiple components, such as any combination of one or more microprocessors, one or more controllers, one or more DSPs, one or more ASICs, or one or more FPGAs, as well as other discrete or integrated logic circuits. The functionality attributed to the processing circuitry 50 herein may be embodied as software, firmware, hardware, or any combination thereof.

[0037] The sensing circuit 52 and the communication circuit 54 may be selectively coupled to the electrodes 16A, 16B via the switching circuit 60 as controlled by the processing circuit 50. The sensing circuit 52 may monitor the electrical activity of the heart of the patient 4 of FIG. 1 and monitor signals from the electrodes 16A, 16B to generate electrocardiogram data for the patient 4. In some examples, the processing circuit 50 may perform characterization of the sensed electrocardiogram data and detect an episode of cardiac arrhythmia of the patient 4. In some examples, the processing circuit 50 transmits the electrocardiogram data of the patient 4 to an external device, such as the external device 12 of FIG. 1, via the communication circuit 54. For example, the IMD 10 transmits the digitized electrocardiogram data to the network 25 for processing by the machine learning system 150 of FIG. 1. In some examples, the IMD 10 transmits one or more segments of the electrocardiogram data in response to detecting an episode of arrhythmia via the characterization. In another example, IMD 10 transmits one or more segments of electrocardiogram data in response to instructions from external device 12 (e.g., when patient 4 experiences one or more symptoms of an arrhythmia and enters a command into external device 12 instructing IMD 10 to upload the electrocardiogram data for analysis by a monitoring center or clinician). The electrocardiogram data may be processed by machine learning system 150 to detect and classify cardiac arrhythmias, as described below.

[0038] In some examples, the IMD 10 performs characterization of the sensed electrocardiogram data, as described in more detail below. In some examples, the characterization performed by the IMD 10 is reduced in complexity to conserve power of the IMD 10. This is The IMD 10 may enable the IMD 10 to perform early or preliminary detection of cardiac arrhythmia. As described in detail below, the computing system 24 may further perform characterization of the electrocardiogram data sensed by the IMD 10 and apply a machine learning system 150 to the electrocardiogram data. The computing system 24 may have more computational resources and less power limitations than the IMD 10, thereby enabling the computing system 24 to perform a more comprehensive and detailed analysis of the electrocardiogram data to more accurately detect cardiac arrhythmia. By shifting the computational burden from the IMD 10 to the computing system 24, the techniques of the present disclosure may help reduce the power consumption of the IMD 10 while increasing the accuracy of arrhythmia detection.

[0039] In some examples, the IMD 10 includes one or more sensors 58, such as one or more accelerometers, microphones, and / or pressure sensors. The sensing circuitry 52 may monitor signals from the sensors 58 and transmit patient data obtained from the sensors 58 to an external device, such as the external device 12 of FIG. 1, for analysis. In some examples, the sensing circuitry 52 may include one or more filters and amplifiers for filtering and amplifying signals received from one or more of the electrodes 16A, 16B and / or other sensors 58. In some examples, the sensing circuitry 52 and / or the processing circuitry 50 may include a rectifier, a filter and / or an amplifier, a sense amplifier, a comparator, and / or an analog-to-digital converter.

[0040] The communications circuitry 54 may include any suitable hardware, firmware, software, or any combination thereof for communicating with another device, such as the external device 12, or another medical device or sensor, such as a pressure sensing device. Under the control of the processing circuitry 50, the communications circuitry 54 may receive downlink telemetry and transmit uplink telemetry from the external device 12 or another device with the aid of an internal or external antenna, such as the antenna 26. In some examples, the communications circuitry 54 may communicate with the external device 12. In addition, the processing circuitry 50 may communicate with external devices (e.g., the external device 12) and networked computing devices via a computer network, such as the Medtronic CareLink® network developed by Medtronic, plc of Dublin, Ireland.

[0041] A clinician or other user may retrieve data from IMD 10 using external device 12 or by using another local or networked computing device configured to communicate with processing circuit 50 via communications circuitry 54. A clinician may also program parameters of IMD 10 using external device 12 or another local or networked computing device. In some examples, a clinician may select one or more parameters that define how IMD 10 senses electrocardiogram data of patient 4.

[0042] One or more components of IMD 10 may be coupled to a power source (not shown in FIG. 2), which may include a rechargeable or non-rechargeable battery disposed within the housing of IMD 10. The non-rechargeable battery may be selected to last for several years, while the rechargeable battery may be inductively charged from an external device, for example, on a daily or weekly basis.

[0043] In accordance with the techniques of the present disclosure, processing circuitry 50 senses electrocardiogram data of patient 4 via electrodes 16 using sensing circuitry 52. ​​In some examples, the electrocardiogram data is an ECG of patient 4. Processing circuitry 50 runs the electrocardiogram data through characterization to obtain one or more cardiac features present in the electrocardiogram data. In some examples, the characterization includes one or more of QRS detection, refractory processing, noise processing, or characterization of the electrocardiogram data. For example, processing circuitry 50 may receive electrocardiogram data from sensing circuitry 52 and / or via sensor 58. and extracting one or more cardiac features from the raw signals. In some examples, processing circuitry 50 identifies one or more cardiac features, such as, for example, one or more of the patient's average heart rate, the patient's minimum heart rate, the patient's maximum heart rate, the patient's heart's PR interval, the patient's heart rate variability, one or more amplitudes of one or more features of the patient's electrocardiogram (ECG), or the intervals between one or more features of the patient's ECG, T-wave alternans, QRS morphology, or other types of cardiac features not expressly described herein.

[0044] As an example, the processing circuit 50 identifies one or more features of a T wave of an electrocardiogram of the patient 4 and applies the model to the one or more identified features to detect an episode of cardiac arrhythmia in the patient 4. In some examples, the one or more identified features are one or more amplitudes of the T wave. In some examples, the one or more identified features are a frequency of the T wave. In some examples, the one or more identified features include at least a T wave amplitude and a T wave frequency. In some examples, the processing circuit 50 identifies one or more relative changes in the one or more identified features indicative of a subsequent episode of cardiac arrhythmia in the patient 4. In some examples, the processing circuit 50 identifies one or more interactions between a plurality of identified features indicative of an episode of cardiac arrhythmia in the patient 4. In some examples, the processing circuit 50 analyzes patient data representing one or more values ​​averaged over a short-term period (e.g., about 30 minutes to about 60 minutes). For example, the patient data may include one or more of an average frequency or an average amplitude of the T wave of an electrocardiogram of the patient 4 to detect an episode of cardiac arrhythmia.

[0045] Processing circuitry 50 may further apply such characterization to determine that one or more cardiac features are indicative of an episode of cardiac arrhythmia. Processing circuitry 50 may further apply the characterization to classify the detected cardiac arrhythmia episode as a particular type of cardiac arrhythmia episode (e.g., bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block). Processing circuitry 50 transmits, via communication circuitry 54, to external device 12 the one or more cardiac electrogram data, the one or more cardiac features present in the cardiac electrogram data, an indication of the detected episode of cardiac arrhythmia, or an indication of a classification of the detected episode of cardiac arrhythmia.

[0046] Although described herein in the context of an exemplary IMD 10 sensing cardiac electrogram data of a patient 4, the techniques for cardiac arrhythmia detection disclosed herein may be used with other types of devices. For example, the techniques may be implemented in an extracardiac defibrillator coupled to electrodes outside the cardiovascular system, a transcatheter pacemaker configured to be implanted within the heart, such as, for example, the Micra™ Transcatheter Pacing System available from Medtronic PLC of Dublin, Ireland, an insertable cardiac monitor, such as the Reveal LINQ™ ICM, also available from Medtronic PLC, a neurostimulator device, a drug delivery device, a medical device external to the patient 4, a wearable device, such as a wearable defibrillator, a fitness tracker, or other wearable device, a mobile device, such as a mobile phone, a “smart” phone, a laptop, a tablet computer, a personal digital assistant (PDA), or “smart” apparel, such as “smart” glasses, a “smart” patch, or a “smart” watch.

[0047] FIGURE 3 is a block diagram illustrating another example of the leadless implantable medical device of FIGURE 1. Components in FIGURE 3 are not necessarily drawn to scale, and instead may be enlarged to show detail. Specifically, FIGURE 3 is a block diagram of a top view of an example configuration of IMD 10 of FIGURE 1.

[0048] Figure 3 is a conceptual diagram illustrating an example IMD 10 that may include components substantially similar to the IMD 10 of Figure 1. In addition to the components shown in Figures 1 and 2, the example IMD 10 shown in Figure 3 also includes a housing 14 and processing circuitry 50 that passes between electrodes 16A, 16B. The IMD 10 may include a wafer-scale insulating cover 74 that may help isolate electrical signals. In some examples, the insulating cover 74 may be placed over the open housing 14 to form a housing for the components of the IMD 10B. One or more components of the IMD 10B (e.g., the antenna 26, the processing circuitry 50, the sensing circuitry 52, the communication circuitry 54, and / or the switching circuitry 60) may be formed on the underside of the insulating cover 74, such as by using flip-chip techniques. The insulating cover 74 may be inverted onto the housing 14. When placed upside down on the housing 14, the components of the IMD 10 formed on the underside of the insulating cover 74 may be disposed within a gap 78 defined by the housing 14. The housing 14 may be formed from titanium or any other suitable material (e.g., a biocompatible material) and may have a thickness of about 200 micrometers to about 500 micrometers. These materials and dimensions are merely examples, and other materials and other thicknesses are possible for the devices of the present disclosure.

[0049] In some examples, the IMD 10 collects patient data for the patient 4, including electrocardiogram data, via the sensing circuitry 50 and / or the sensor 58. The sensor 58 may include one or more sensors, such as one or more accelerometers, pressure sensors, optical sensors for O2 saturation, and the like. In some examples, the patient data includes one or more of the patient's activity level, the patient's heart rate, the patient's posture, the patient's electrocardiogram, the patient's blood pressure, the patient's accelerometer data, or other types of patient data. The IMD 10 uploads the patient data to the external device 12 via the communication circuitry 54, which may then upload such data to the computing system 24 via the network 25. In some examples, the IMD 10 uploads the patient data to the computing system 24 daily. In some examples, the patient data includes one or more values ​​representing average measurements of the patient 4 over a long-term period (e.g., about 24 hours to about 48 hours). In this example, the IMD 10 uploads the patient data to the computing system 24 as well as performing short-term monitoring of the patient 4 (as described below). However, in other examples, the medical device that processes the patient data to detect and / or classify arrhythmias in patient 4 is different from the medical device that performs short-term monitoring of patient 4 .

[0050] FIG. 4 is a block diagram illustrating an example computing device 400 that operates according to one or more techniques of the present disclosure. In one example, the computing device 400 is an example implementation of the computing system 24 of FIG. 1. In one example, the computing device 400 includes a processing circuit 402 for executing an application 424, including a machine learning system 450 or other application described herein. For illustrative purposes, while shown in FIG. 4 as a standalone computing device 400, the computing device 400 may be any component or system that includes a processing circuit or other suitable computing environment for executing software instructions, and does not necessarily need to include one or more elements shown in FIG. 4 (e.g., components such as the input device 404, the communication circuit 406, the user interface device 410, or the output device 412, and in some examples, the storage device 408 may not be located in the same location or in the same housing as other components). In some examples, the computing device 400 may be a cloud computing system distributed across multiple devices.

[0051] 4, the computing device 400 includes a processing circuit 402, one or more input devices 404, a communication circuit 406, one or more storage devices 408, a user interface (UI) device 410, and one or more output devices 412. In one example, the computing device 400 further includes one or more applications 424, such as a machine learning system 450 and an operating system 416, that are executable by the computing device 400. Each of 406, 408, 410, and 412 is coupled (physically, communicatively, and / or operatively) for inter-component communication. In some examples, communication channel 414 may include a system bus, a network connection, an inter-process communication data structure, or any other method for communicating data. As an example, components 402, 404, 406, 408, 410, and 412 may be coupled by one or more communication channels 414.

[0052] In one example, the processing circuitry 402 is configured to implement functions and / or process instructions for execution within the computing device 400. For example, the processing circuitry 402 may be capable of processing instructions stored in the storage device 408. Examples of the processing circuitry 402 may include any one or more of a microprocessor, a controller, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or equivalent discrete or integrated logic circuitry.

[0053] One or more storage devices 408 may be configured to store information within the computing device 400 during operation. In some examples, the storage device 408 is described as a computer-readable storage medium. In some examples, the storage device 408 is a temporary memory, meaning that the primary purpose of the storage device 408 is not long-term storage. In some examples, the storage device 408 is described as a volatile memory, meaning that the storage device 408 does not retain the stored contents when the computer is turned off. Examples of volatile memory include random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), and other forms of volatile memory known in the art. In some examples, the storage device 408 is used to store program instructions for execution by the processing circuit 402. In one example, the storage device 408 is used by software or applications 424 running on the computing device 400 to temporarily store information during program execution.

[0054] In some examples, the storage device 408 also includes one or more computer-readable storage media. The storage device 408 may be configured to store a larger amount of information than volatile memory. The storage device 408 may further be configured for long-term storage of information. In some examples, the storage device 408 includes a non-volatile storage element. Examples of such non-volatile storage elements include magnetic hard disks, optical disks, floppy disks, flash memory, or forms of electrically programmable memory (EPROM) or electrically erasable programmable (EEPROM) memory.

[0055] The computing device 400 also includes communications circuitry 406, in some examples. In one example, the computing device 400 utilizes the communications circuitry 406 to communicate with external devices, such as the IMD 10 and external device 12 of FIG. 1. The communications circuitry 406 may include a network interface card, such as an Ethernet card, an optical transceiver, a radio frequency transceiver, or any other type of device capable of transmitting and receiving information. Other examples of such network interfaces may include 3G and WiFi radios.

[0056] In one example, the computing device 400 also includes one or more user interface devices 410. In some examples, the user interface devices 410 are configured to receive input from a user through tactile, audio, or video feedback. Examples of user interface devices 410 include a presence sensitive display, a mouse, a keyboard, a voice response system, a video camera, a microphone, or any other type of device for detecting commands from a user. In some examples, the presence sensitive display includes a touch screen.

[0057] One or more output devices 412 may also be included in the computing device 400. The output device 412 is configured to provide output to a user using tactile, audio, or video stimuli, in some examples. The output device 412 includes, in one example, a presence-sensitive display, a sound card, a video graphics adapter card, or any other type of device for converting signals into an appropriate form that is human or machine understandable. Additional examples of the output device 412 include speakers, a cathode ray tube (CRT) monitor, a liquid crystal display (LCD), or any other type of device capable of generating output that is understandable to a user.

[0058] The computing device 400 may include an operating system 416, which in some examples controls the operation of components of the computing device 400. For example, the operating system 416, in one example, facilitates communication between one or more applications 424 and a long-term prediction module 450 and the processing circuitry 402, the communication circuitry 406, the storage device 408, the input device 404, the user interface device 410, and the output device 412.

[0059] The applications 422 may also include program instructions and / or data that are executable by the computing device 400. An exemplary application 422 executable by the computing device 400 may include a machine learning system 450. Other additional applications not shown may alternatively or additionally be included to provide other functionality described herein, but are not shown for simplicity.

[0060] In accordance with the techniques of this disclosure, computing device 400 applies machine learning models of machine learning system 450 to patient data sensed by IMD 10 to detect and classify arrhythmia episodes occurring in patient 10. In some examples, machine learning system 450 is an example of machine learning system 150 of FIG.

[0061] In some examples, the machine learning model implemented by the machine learning system 450 is trained with training data including electrocardiogram data of multiple patients labeled with descriptive metadata. For example, during the training phase, the machine learning system 450 processes multiple ECG waveforms. Typically, the multiple ECG waveforms are from multiple different patients. Each ECG waveform is labeled with one or more episodes of one or more types of arrhythmia. For example, the training ECG waveform may include multiple segments, each segment labeled with a descriptor that specifies the absence of arrhythmia or the presence of a particular classification of arrhythmia (e.g., bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block). In some examples, a clinician manually labels the presence of arrhythmia in each ECG waveform. In some examples, the presence of arrhythmia in each ECG waveform is labeled according to a classification by a characterization algorithm. The machine learning system 450 can operate to convert the training data into vectors and tensors (e.g., multidimensional arrays), on which the machine learning system 450 can apply mathematical operations such as linear algebra, nonlinear, or alternative computational operations. The machine learning system 450 uses the training data 104 to teach the machine learning model to weigh different features present in the electrocardiogram data. In some examples, the machine learning system 450 uses the electrocardiogram data to teach the machine learning model to apply different coefficients that represent one or more features of the electrocardiogram as being more or less important with respect to the occurrence of a particular class of cardiac arrhythmia. By processing a large number of such ECG waveforms labeled with arrhythmia episodes, the machine learning system 450 can 1 , that have not been previously analyzed by machine learning system 450, and further processes such electrocardiogram data to detect the presence or absence of various classes of arrhythmias in the patient with high accuracy. Typically, the greater the amount of electrocardiogram data that machine learning system 450 is trained on, the more accurate the machine learning model will be in detecting or classifying cardiac arrhythmias in new electrocardiogram data.

[0062] After the machine learning system 450 trains the machine learning model, the machine learning system 450 may receive patient data, such as electrocardiogram data, of a particular patient, such as patient 4. The machine learning system 450 applies the trained machine learning model to the patient data to detect occurrences of cardiac arrhythmia episodes in the patient 4. Further, the machine learning system 450 applies the trained machine learning model to the patient data to classify the patient's cardiac arrhythmia episodes as indicative of a particular type of arrhythmia. In some examples, the machine learning system 450 may output a preliminary determination that the cardiac arrhythmia episode is indicative of a particular type of arrhythmia, as well as an estimate of the certainty in the determination. In response to determining that the estimate of the certainty in the determination is greater than a predetermined threshold (e.g., 50%, 75%, 90%, 95%, 99%), the computing device 400 may classify the cardiac arrhythmia episode as a particular type of arrhythmia.

[0063] In some examples, the machine learning system may process one or more cardiac features of the electrocardiogram data instead of the raw electrocardiogram data itself. The one or more cardiac features may be obtained via characterization performed by the IMD 10, as described above. The cardiac features may include, for example, one or more of the patient's average heart rate, the patient's minimum heart rate, the patient's maximum heart rate, the patient's heart's PR interval, the patient's heart rate variability, one or more amplitudes of one or more features of the patient's electrocardiogram (ECG), or intervals between one or more features of the patient's ECG, T-wave alternans, QRS morphometry, or other types of cardiac features not explicitly described herein. In such an example implementation, the machine learning system may train the machine learning model via a plurality of training cardiac features labeled with arrhythmia episodes, instead of a plurality of ECG waveforms labeled with arrhythmia episodes as described above.

[0064] In some examples, the machine learning system 450 may apply the machine learning model to other types of data to determine that an arrhythmia episode has occurred in the patient 4. For example, the machine learning system 450 may apply the machine learning model to one or more characteristics of the electrocardiogram data that correlate to the patient's arrhythmia, the activity level of the IMD 10, the input impedance of the IMD 10, or the battery level of the IMD 10.

[0065] In a further example, the processing circuit 402 may generate an intermediate representation of the electrocardiogram data from the electrocardiogram data. For example, the processing circuit 402 may apply one or more signal processing, signal decomposition, wavelet decomposition, filtering, or noise reduction operations to the electrocardiogram data to generate an intermediate representation of the electrocardiogram data. In this example, the machine learning system 450 processes such intermediate representation of the electrocardiogram data to detect and classify arrhythmia episodes of the patient 4. Furthermore, the machine learning system may train a machine learning model via a plurality of training intermediate representations labeled with arrhythmia episodes instead of a plurality of raw ECG waveforms labeled with arrhythmia episodes as described above. The use of such intermediate representations of the electrocardiogram data may enable the machine learning system 450 to train and develop a lighter-weight, less computationally complex machine learning model. Furthermore, the use of such intermediate representations of the electrocardiogram data may require fewer iterations and less training data to build an accurate machine learning model as opposed to the use of raw electrocardiogram data to train the machine learning model.

[0066] In some examples, computing system 24 may use machine learning system 150 to detect other types of arrhythmias other than those detected in the characterization screening analysis. For example, the arrhythmia detection algorithms for performing the characterization implemented by a low-power device such as IMD10 may not be designed to detect infrequent arrhythmias such as atrioventricular block. Machine learning system 150 may train a machine learning model on a large data set where such arrhythmias are available, thereby providing finer granularity and greater accuracy than the characterization performed by IMD10 alone, for example. Thus, the use of machine learning system 150 may extend the arrhythmia diagnosis capabilities of system 2 by enabling IMD10 to implement a general screening algorithm using the characterization, followed by using machine learning system 150 to implement a machine learning model that can provide a broader range of arrhythmia detection. After detecting a type of arrhythmia not detected by the characterization, computing system 24 may nevertheless use the characterization, such as QRS detection, to assist in characterizing and reporting other types of arrhythmias detected by the machine learning model of machine learning system 150.

[0067] In some examples, computing system 24 may tailor machine learning system 150 to a particular use case. For example, machine learning system 150 may implement a machine learning model specific to the detection of AV block and bradycardia if patient 4 is a post-TAVR patient. As another example, machine learning system 150 may implement a machine learning model specific to the detection of PVCs, such that PVC burden can be used to risk stratify patients for whom an ICD may be indicated.

[0068] FIG. 5 is a flow chart illustrating an example operation according to the techniques of this disclosure. For convenience, FIG. 5 is described in conjunction with FIG. 1. In some examples, the operations of FIG. 5 are for detecting and classifying cardiac arrhythmias in patient 4. In the operations of FIG. 5, system 2 combines the capabilities of the machine learning models of machine learning system 150 to learn features and perform classification directly from the input, with the interpretability provided by the characterization algorithm and ECG processing. In the example operations of FIG. 5, system 2 implements the machine learning models of machine learning system 150 in parallel with the characterization algorithm to perform arrhythmia detection and characterization.

[0069] As shown in Figure 5, the IMD 10 senses (502) electrocardiogram data of a patient 4. The electrocardiogram data may be, for example, a transient ECG of the patient 4 or a full disclosure ECG of the patient 4. Furthermore, the electrocardiogram data of the patient 4 may be from a single channel or a multi-channel system. For simplicity, in the example of Figure 5, the electrocardiogram data of the patient 4 is described as single channel transient ECG data.

[0070] The machine learning system 150 of the computing system 24 applies the machine learning model to the sensed electrocardiogram data to detect an arrhythmia episode in the patient 4 (506). In some examples, the machine learning model is trained on a plurality of ECG episodes that have been annotated by a clinician or a monitoring center for several different types of arrhythmias. In one example, the machine learning system 150 applies the machine learning model to one or several sub-segments of the normalized input ECG signal to generate an arrhythmia label and a likelihood of arrhythmia occurrence. In some examples, the machine learning model may be accurate in mapping the input ECG to an output arrhythmia label, but may not provide additional arrhythmia characteristics or identify specific cardiac features, such as average heart rate, maximum heart rate, PR interval characteristics, etc., that are used to determine that an arrhythmia episode has occurred in the patient 4. Additionally, a physician-provided notification or reportable criteria (e.g., a probability of occurrence of an arrhythmia episode in the patient 4) may be generated from the output or intermediate states of the machine learning model that a clinician may utilize to determine that an arrhythmia episode has occurred in the patient 4 for use in providing subsequent treatment to the patient 4. In some cases, the heart rate may not be able to be obtained (four of the 10 heart rates were below 30 beats per minute (BPM)).

[0071] To address this, computing system 24 further applies the characterization to the electrocardiogram data to detect one or more cardiac features (504). In some examples, computing system 24 further applies the characterization to the electrocardiogram data to detect one or more episodes of arrhythmia. For example, computing system 24 may apply a QRS detection characterization and a noise flag (e.g., beat noise or not) to the electrocardiogram data to provide characteristics of the arrhythmia and / or cardiac features of the detected episodes of arrhythmia (e.g., average heart rate during an episode of atrial fibrillation, duration of pauses). Additionally, computing system 24 may apply the characterization to guide notification and reporting criteria for system 2. In the example of FIG. 5, computing system 24 performs the characterization of the electrocardiogram data. However, in other examples of the techniques of this disclosure, other devices, such as IMD 10, external device 12, or another external medical device, may perform the characterization of the electrocardiogram data.

[0072] 5, the computing system applies both the machine learning system 150 and the characterization to determine whether an episode of cardiac arrhythmia is detected in patient 4 (508). If neither the machine learning system 150 nor the characterization detects an episode of cardiac arrhythmia (e.g., "No" at block 508), the computing system may archive the electrocardiogram data for subsequent review by a clinician.

[0073] If the machine learning system 150 or at least one of the characterization operations (504) detects an episode of cardiac arrhythmia (e.g., “Yes” at block 508), the computing system may generate a report of the arrhythmia (512) and output the report to a clinician or monitoring center (514). For example, if the machine learning system 150 detects an episode of bradycardia and the characterization performed on the electrocardiogram data indicates that four out of four noise-free heart beats are less than 30 BPM, the computing system 24 generates a report notifying a physician of the occurrence of an arrhythmia episode.

[0074] In one example, the report includes an indication that an arrhythmia episode occurred in the patient and one or more cardiac features consistent with the arrhythmia episode. In some examples, the report further includes a classification of the arrhythmia episode as a particular type of arrhythmia. In some examples, the report includes a subsection of electrocardiogram data obtained from patient 4 that is consistent with the arrhythmia episode. For example, computing system 24 may identify a subsection of electrocardiogram data of patient 4, where the subsection includes electrocardiogram data for a first period prior to the arrhythmia episode (e.g., typically less than 10 minutes prior to the onset of the arrhythmia episode), a second period during the occurrence of the arrhythmia episode, and a third period after the arrhythmia episode (e.g., typically less than 10 minutes after the cessation of the arrhythmia episode). As an example, the subsection of electrocardiogram data of patient 4 may be approximately 6 seconds in length and includes representative segments before, during, and after the arrhythmia episode (if present in the electrocardiogram data or waveform being analyzed). In some examples, the duration of the episode may vary by type of device and may further depend on the use case of the medical device, one or more settings of the medical device, or the particular type of arrhythmia sensed. For example, some types of arrhythmias are quickly self-terminating (resulting in a shorter duration of the episode) while other types of arrhythmias are sustained, the length of time such that the recorded duration of the episode depends on the designated memory space of the medical device. As an example, in the case of atrial fibrillation (AF), the subsection of the electrocardiogram data of the patient 4 may include electrocardiogram data during the onset period, a segment of maximum AF probability, a segment of fastest AF rate, and an AF offset. Typically, the length of time of the patient's electrocardiogram data is longer than the first, second, and third periods. Additionally, the computing system 24 identifies one or more of the cardiac features consistent with the first, second, and third periods. The computing system 24 includes in the report the subsection of the electrocardiogram data and one or more of the cardiac features consistent with the first, second, and third periods.

[0075] In some examples, computing system 24 may receive one or more adjustments from the clinician to the operation of the feature-based representation of the electrocardiogram data based on the report. Computing device 24 may then perform the feature-based representation of the electrocardiogram data of patient 4 according to the one or more adjustments.

[0076] FIG. 6 is a chart illustrating an exemplary electrocardiogram 602 obtained from the patient 4 of FIG. 1. The electrocardiogram 602 may be sensed, for example, by the sensing circuit 52 of the IMD 10. The machine learning system 150 of FIG. 1 may apply a machine learning model to the electrocardiogram 602 and determine that the electrocardiogram 602 includes a pause 604. The computing system 24 of FIG. 1 or the IMD 10 of FIG. 1 (e.g., as part of the IMD 10 that initially detects the arrhythmia) may perform characterization on the electrocardiogram 602 and determine the length of the pause 604. For the example of FIG. 6, the computing system 24 or the IMD 10 may determine, via the characterization of the electrocardiogram 602, that the pause 604 has a length of 3.061 seconds. In one example, the IMD 10 performs QRS detection from a marker channel on the device. The QRS flagging may be based on a conventional QRS algorithm. The IMD 10 may use a QRS marker and determine that the pause duration is 3.061 seconds.

[0077] Figure 7 is a flow chart illustrating example operations in accordance with the techniques of this disclosure. For convenience, Figure 7 is described with respect to Figure 1. The operations of Figure 7 are for detecting and classifying cardiac arrhythmias in patient 4. Specifically, the operations of Figure 7 illustrate an implementation in which computing system 24 uses machine learning arrhythmia detection and characterization of machine learning system 150 in parallel to perform cardiac arrhythmia detection, validation, and reporting.

[0078] As shown in FIG. 7, the IMD 10 senses electrocardiogram data of the patient 4 (702). The computing system 24 applies the characterization to the electrocardiogram data to detect one or more cardiac features (704). In the example of FIG. 7, the computing system 24 performs the characterization of the electrocardiogram data. However, in other examples of the techniques of this disclosure, another device, such as the IMD 10, the external device 12, or another external medical device, may perform the characterization of the electrocardiogram data. The machine learning system 150 of the computing system 24 applies the machine learning model to the sensed electrocardiogram data to detect an arrhythmia episode in the patient 4 (706). The operations of steps 702, 704, and 706 may occur in a manner substantially similar to steps 502, 504, and 506, respectively, of FIG. 5.

[0079] The computing system 24 determines (708) whether both the machine learning system 150 and the characterization operation of (704) detected an episode of cardiac arrhythmia. For example, the computing system 24 may determine (708) a level of confidence that the determination of arrhythmia by the machine learning system 150 is consistent with the determination of arrhythmia by the characterization operation of 704. For example, if the computing system 24 determines that both the machine learning system 150 and the characterization operation of (704) detected an episode of cardiac arrhythmia (e.g., “Yes” at block 708), the computing system 24 may generate (712) a report of the arrhythmia and output (714) the report to a clinician or monitoring center. For example, the computing system 24 places the detected arrhythmia along with the arrhythmia features in a report and outputs the report to a clinician. The operations of steps 712 and 714 may be performed in a manner substantially similar to steps 512 and 514 of FIG. 5, respectively. This may occur.

[0080] As another example, if the computing system 24 determines that the characterization operations of the machine learning system 150 and (704) are inconsistent as to whether an episode of cardiac arrhythmia is detected (e.g., “no” at block 708), the computing system 24 submits the electrocardiogram data to the monitoring center for reconciliation (710). In other words, the computing system 24 presents the electrocardiogram data for human overview where there is a discrepancy between the two detection methods. Such a workflow may allow for a reduction in the burden of human review for only those arrhythmias that the computing system 24 cannot assess with a high degree of confidence. For example, if an arrhythmia detected via characterization is similar to an arrhythmia independently detected by a machine learning model, the computing system 24 may determine that the arrhythmia detected via characterization is independently verified without the need for human review by an expert. Thus, the techniques of the present disclosure may reduce the amount of review required by clinicians and / or experts, thereby reducing the administrative overhead and costs of cardiac monitoring of the patient 4.

[0081] Figure 8 is a flow chart illustrating example operations in accordance with the techniques of this disclosure. For convenience, Figure 8 is described with respect to Figure 1. The operations of Figure 8 are for detecting and classifying cardiac arrhythmias in patient 4. In particular, the operations of Figure 8 illustrate an implementation in which computing system 24 uses characterization in series with machine learning arrhythmia detection of machine learning system 150 to perform cardiac arrhythmia detection, validation, and reporting.

[0082] As shown in FIG. 8, the IMD 10 senses electrocardiogram data of the patient 4 (802). The operations of step 802 may occur in a manner substantially similar to step 502 of FIG. 5. The computing system 24 applies the characterization to the electrocardiogram data to detect a set of cardiac arrhythmias and one or more cardiac features (804). In some examples, the computing system 24 applies the characterization to detect arrhythmias such as bradycardia, tachycardia, pauses, or atrial fibrillation based on the rate and variability features of the electrocardiogram data. In the example of FIG. 8, the computing system 24 performs the characterization as a screening step before characterizing all arrhythmias (e.g., the computing system 24 may use the characterization to only consider tachyarrhythmias with a heart rate of 120 BPM or more, bradyarrhythmias with a heart rate of 40 BPM or less, or arrhythmias with high RR variability). In other examples, such characterization may be implemented in a low power device, such as the IMD 10, or in other types of devices, such as the external device 12 or another external medical device.

[0083] Upon detecting via the characterization that an episode of cardiac arrhythmia has occurred in patient 4, machine learning system 150 of computing system 24 applies the machine learning model to the sensed electrocardiogram to verify that an arrhythmia episode has occurred (806). In some examples, machine learning system 150 applies the machine learning model to many different types of patient data, such as electrocardiogram data for patient 4, to determine whether an arrhythmia episode has occurred based on the trigger reason that caused the characterization to detect the arrhythmia, one or more types of arrhythmia characterizations that were detected, or device characteristics of IMD 10, such as activity level, input impedance, battery level, etc.

[0084] In the example of Figure 8, computing system 24 determines (808) whether machine learning system 150 validates an arrhythmia trigger of the characterization of step 804. In other words, in response to determining that the characterization of step 804 detects an arrhythmia episode in patient 4, computing system 24 determines whether machine learning system 150 similarly detects an arrhythmia episode in patient 4. Thus, the computing system 24 can verify whether the reason for the detection of the characterization in step 804 was appropriate (e.g., the bradycardia trigger of the characterization truly indicated that an episode of bradycardia occurred in the patient 4). The use of the machine learning system 150 as a validation tool may assist the computing system 24 in providing feedback to the physician for reprogramming the diagnostic device of the patient 4, such as the IMD 10. Additionally, the use of the machine learning system 150 as a validation tool may assist the computing system 24 in automating the reporting of physiological parameters (e.g., if all episodes triggered by AF are appropriate, report the burden of AF detected by the device as is, otherwise consider only the burden for the episodes that were appropriately triggered).

[0085] For example, if the computing system 24 determines that the machine learning system 150 verifies the detection of an episode of cardiac arrhythmia by the characterization operation of 804 (e.g., "yes" at block 808), the computing system 24 may generate a report of the arrhythmia (812) and output the report to a clinician or monitoring center (814). As another example, if the computing system 24 determines that the characterization operations of the machine learning system 150 and 804 disagree as to whether an episode of cardiac arrhythmia is detected (e.g., "no" at block 808), the computing system 24 submits the electrocardiogram data to the monitoring center for reconciliation (810). The operations of steps 810, 812, and 814 may occur in a manner substantially similar to steps 510, 512, and 514, respectively, of FIG. 5.

[0086] Figure 9 is a flow chart illustrating example operations in accordance with the techniques of this disclosure. For convenience, Figure 9 is described with respect to Figure 1. The operations of Figure 9 are for detecting and classifying cardiac arrhythmias in patient 4. Specifically, the operations of Figure 9 illustrate an implementation in which computing system 24 pre-processes electrocardiogram data to generate an intermediate representation of the electrocardiogram data and applies machine learning system 150 to the intermediate representation of the electrocardiogram data to perform cardiac arrhythmia detection, validation, and reporting.

[0087] In the example of FIG. 9, the IMD 10 senses electrocardiogram data of the patient 4 (902). The operations of step 902 may occur in a manner substantially similar to step 502 of FIG. 5. The computing system 24 performs pre-processing of the sensed electrocardiogram data to generate an intermediate representation of the electrocardiogram data (904). For example, the computing system 24 performs QRS detection to detect multiple QRS windows in the sensed electrocardiogram data. In one example, the window around the detected QRS includes 160 milliseconds of data before the detected QRS and 160 milliseconds of data after the detected QRS. In another example, the window around the detected QRS includes a data segment from the T offset of the previous QRS to the T offset of the current QRS. In some examples, the computing system 24 may apply signal processing methods such as bandpass filtering or stationary wavelet decomposition used for QRS detection, flagging, and delineation to the sensed electrocardiogram data. For example, the computing system 24 generates a wavelet decomposition of the electrocardiogram of the patient 4 for the window around the detected QRS.

[0088] Computing system 24 applies the characterization to the intermediate representation of the electrocardiogram data to detect one or more cardiac features (906). For example, computing system 24 applies the characterization to the intermediate representation to detect and depict patient 4's QRS segment (e.g., PR interval) from a window around the detected QRS, as well as a noise flag. In the example of FIG. 9, computing system 24 performs the characterization of the electrocardiogram data. However, in other examples of the techniques of this disclosure, other devices, such as IMD 10, external device 12, or another external medical device, perform the characterization of the electrocardiogram data. It can be done.

[0089] The machine learning system 150 of the computing system 24 applies the machine learning model to the intermediate representation of the sensed electrocardiogram to detect arrhythmia episodes in the patient 4 (908). For example, the machine learning model may receive as input a plurality of electrocardiogram segments, each segment including a window around a detected QRS, a QRS delineation of the segment, and a noise flag for the segment. The machine learning system 150 applies the machine learning model to the received segments to detect arrhythmia episodes in the patient 4.

[0090] In some examples, the machine learning model is tuned to capture segments of interest for each arrhythmia. For example, the machine learning model may process sensed electrocardiograms and capture the onset, offset, maximum heart rate, and minimum heart rate from a segment that includes a window around the detected QRS. In some examples, the computing system 24 uses features derived from characterizations such as QRS detection, such as heart rate values ​​of electrocardiogram segments, to characterize or contextualize the detection of the arrhythmia by the machine learning model.

[0091] The use of signal decomposition to create an intermediate representation of the electrocardiogram may enable the use of existing knowledge about frequency bands of interest for arrhythmia detection. Furthermore, signal decomposition may limit the computational complexity of the machine learning model of the machine learning system 150, such that the machine learning model may learn features for classification only from electrocardiogram subsegments corresponding to the detected QRS. Thus, such techniques may reduce the complexity of the machine learning model, reduce the size of the training set required for the generation of the machine learning model, and increase the accuracy of the machine learning model.

[0092] 5, computing system 24 may use the same signal pre-processing for both the cardiac arrhythmia and / or cardiac feature characterization detection of step 906 and the machine learning model detection of cardiac arrhythmia. Additionally, computing system 24 may use the QRS noise flag and characterization as input for the machine learning model of machine learning system 150. The input cardiac electrogram complexes may be of the same duration (e.g., 320 ms) or different durations (e.g., the segment from the previous T offset to the current T offset).

[0093] Figure 10 is a flow chart illustrating example operations in accordance with the techniques of this disclosure. For convenience, Figure 10 is described with respect to Figure 1. The operations of Figure 10 are for detecting and classifying cardiac arrhythmias in patient 4. In particular, the operations of Figure 10 illustrate an implementation in which computing system 24 uses featurization in series with machine learning arrhythmia detection of machine learning system 150 to build an arrhythmia dictionary for use in detecting, classifying, and reporting cardiac arrhythmias.

[0094] The operations of FIG. 10 monitor electrocardiogram data of patient 4, annotate detected arrhythmias, and report such arrhythmias to a monitoring center. In some examples, the operations of FIG. 10 are performed in a centralized location, such as a monitoring center. As another example, the operations of FIG. 10 may be performed at a clinic on a patient-by-patient basis. As shown in FIG. 10, the IMD 10 senses electrocardiogram data of patient 4 (1002). The computing system 24 further applies the characterization to the electrocardiogram data to detect one or more cardiac features (1004). The operations of steps 1002 and 1004 may occur in a manner substantially similar to steps 502 and 504, respectively, of FIG. 5.

[0095] The computing system 24 further applies the characterization to the electrocardiogram data to One or more episodes of arrhythmia are detected (1006). In some examples, the characterization causes automatic triggering of an electrocardiogram. In the example of FIG. 10, computing system 24 performs the characterization. However, in other examples, arrhythmia detection and electrocardiogram episode automatic triggering may occur at another device, such as the IMD 10, external device 12, or another external medical device, or via post-processing in a Holter-like system.

[0096] If an arrhythmia episode is triggered for the first time from a particular patient, the computing system 24 submits the episode for arrhythmia review so that the episode may be used as a reference episode in a patient-specific "episode dictionary." For example, in response to detecting an arrhythmia episode, the computing system 24 determines whether the arrhythmia episode is a first detected episode. If the arrhythmia episode is a first detected episode (e.g., "Yes" at block 1008), the computing system 24 generates a report of the arrhythmia episode and submits the report to a monitoring center or clinician for evaluation (1010). For example, if the episode is a first AF trigger, the episode is submitted for monitoring center review. As another example, if the episode is a first AF trigger occurring during the night, the episode is submitted for monitoring center review. In one example, the report includes an indication that an arrhythmia episode has occurred for the patient and one or more cardiac features consistent with an arrhythmia episode. The computing system 24 receives instructions from the monitoring center to verify whether the cardiac features included in the report are indicative of an arrhythmia episode. In examples where the cardiac features are indicative of an arrhythmia episode, computing system 24 further receives a classification of the type of arrhythmia indicated by the cardiac features included in the report. Computing system 24 may store an indication of the classification of the type of arrhythmia along with the cardiac features in a database to build a "dictionary" of cardiac arrhythmias.

[0097] In some examples, computing system 24 may detect multiple episodes of arrhythmia with similar arrhythmia content, annotations, and / or cardiac characteristics. For example, for atrial fibrillation (AF) monitoring, most episode triggers have AF. Another example is that the characterization may generate some false triggers of arrhythmia (e.g., PAC with low amplitude P waves) for patient-specific reasons such as location and direction of signal acquisition. For example, computing system 24 may input the subsequently detected episodes into a machine learning model (along with other episode characteristics such as trigger reason, activity level, and date and time). The machine learning model of machine learning system 150 compares the episode characteristics to those in the "episode dictionary" of patient 4. If the machine learning model determines that a similar episode exists in the dictionary with high confidence, the original monitoring center annotations are used as is to report the episode. If no similar episodes are identified, computing system 24 may determine that the episode characteristics are different and therefore present the episode for monitoring center review and reporting. Thus, the operations of FIG. 10 may increase the efficiency of arrhythmia annotation by minimizing redundant annotations in arrhythmia episodes with similar characteristics to reduce the amount of arrhythmia episodes requiring monitoring center review.

[0098] The techniques of the present disclosure may provide the further advantage that the machine learning model of the machine learning system 150 does not need to be tuned to detect a wide variety of arrhythmias. Instead, the machine learning model may be tuned only to accurately identify a new episode as similar or dissimilar to a previous episode. For example, if there is a similarity between two episodes of arrhythmia, the computing system 24 may apply the previous patient-specific findings to the new episode as well. If there is a dissimilarity, the computing system 24 may determine that the episode is an arrhythmia. A machine learning model may require a human expert to make a determination of whether an episode is a cardiac episode and / or the type of arrhythmia presented by the episode. Thus, the machine learning model does not need to identify a particular arrhythmia with a high level of confidence. The machine learning model only needs to be accurate in identifying the difference between two episodes of arrhythmia to accurately present episodes with different cardiac features (e.g., novel or unclassified rhythm content) for human review. Thus, the techniques of the present disclosure may enable the computing system 24 to detect arrhythmia episodes that the machine learning model 150 was not specifically trained to detect. Additionally, the techniques of the present disclosure may reduce the complexity of the machine learning model while maintaining high accuracy in detecting and classifying arrhythmias.

[0099] For example, with respect to the operation of FIG. 10, if the arrhythmia episode is not the first detected episode (e.g., “No” at block 1008), the machine learning system 150 applies a machine learning model to the detected cardiac features to compare the cardiac features to other cardiac features of previous episodes of arrhythmia (1012). For example, the machine learning system 150 may apply a machine learning model to the detected cardiac features to determine whether the cardiac features match other cardiac features of previous episodes of arrhythmia and an estimate of the level of confidence or certainty in the comparison. In some examples, the computing system 24 resets the similarity comparison after a certain period of time (e.g., daily) or on demand (e.g., when a change in the patient's medication occurs). This allows some episodes of arrhythmia to be intermittently reviewed by a monitoring center or clinician to ensure that new or changing arrhythmias are not missed.

[0100] In response to determining that the machine learning model does not have a high confidence level or certainty in the comparison (e.g., “NO” at block 1014), computing system 24 generates a report of the arrhythmia episode and submits the report to a monitoring center or clinician for evaluation (1010). Computing system 24 receives an indication verifying that the cardiac features included in the report are indicative of an arrhythmia episode and a classification of the arrhythmia type, stores the indication of the arrhythmia type classification with the cardiac features in a database, and thereby updates a cardiac arrhythmia dictionary with the detected cardiac features and the classification of the arrhythmia indicated by the detected cardiac features.

[0101] In response to determining that the machine learning model has a high level of confidence or certainty in the comparison (e.g., "YES" at block 1014), computing system 24 may determine that the cardiac feature is indicative of a type of previous episode of arrhythmia. Computing system 24 generates (1016) a report of the arrhythmia and outputs (1018) the report to the monitoring center. The operations of steps 1016 and 1018 may occur in a manner substantially similar to steps 512 and 514, respectively, of FIG. 5.

[0102] The following examples may illustrate one or more aspects of the disclosure.

[0103] Example 1. A method, comprising: receiving, by a computing device including a processing circuit and a storage medium, electrocardiogram data of a patient sensed by a medical device; applying, by the computing device, a machine learning model trained using electrocardiogram data of a plurality of patients to the received electrocardiogram data to determine, based on the machine learning model, that an arrhythmia episode has occurred in the patient; performing, by the computing device, a feature-based representation of the received electrocardiogram data to obtain cardiac features present in the electrocardiogram data; and in response to determining that an arrhythmia episode has occurred in the patient, generating, by the computing device, a report including an indication that an arrhythmia episode has occurred in the patient and one or more of the cardiac features consistent with the arrhythmia episode; and outputting for display a report including an indication that an arrhythmia episode has occurred in the patient by the method and one or more cardiac features consistent with the arrhythmia episode.

[0104] Example 2. The method of example 1, wherein performing feature-based representation of the electrocardiogram data to obtain cardiac features present in the electrocardiogram data includes performing at least one of QRS detection, refractory processing, noise processing, or representation of the electrocardiogram data to obtain cardiac features present in the electrocardiogram data.

[0105] Example 3. The method of Example 1 or 2, wherein applying the machine learning model to determine that an episode of arrhythmia has occurred in the patient comprises applying the machine learning model to determine that an episode of at least one of bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block has occurred in the patient.

[0106] Example 4. The method of any of Examples 1-3, wherein the cardiac features present in the electrocardiogram data are one or more of the patient's average heart rate, the patient's minimum heart rate, the patient's maximum heart rate, the patient's cardiac PR interval, the patient's heart rate variability, one or more amplitudes of one or more features of the patient's electrocardiogram (ECG), or intervals between one or more features of the patient's ECG.

[0107] Example 5. The method of any of Examples 1-4, wherein the machine learning model trained using electrocardiogram data of a plurality of patients comprises a machine learning model trained using a plurality of electrocardiogram (ECG) waveforms, each ECG waveform being labeled with one or more types of arrhythmia episodes in a patient of the plurality of patients.

[0108] Example 6. The method of any of Examples 1-5, wherein applying the machine learning model to the received electrocardiogram data further comprises applying the machine learning model to at least one of one or more characteristics of the received electrocardiogram data that are correlated to arrhythmia in the patient, an activity level of the medical device, an input impedance of the medical device, or a battery level of the medical device.

[0109] Example 7. The method of any of Examples 1-6, further comprising receiving, by the computing device and from a user, adjustments to the feature-based representation of the electrocardiogram data in response to outputting a report including an indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode, and performing the feature-based representation of the electrocardiogram data in accordance with the adjustments to obtain a second cardiac feature present in the electrocardiogram data.

[0110] Example 8. The method of any of Examples 1-7, wherein generating a report, the patient's electrocardiogram data including an electrocardiogram (ECG) of the patient including an indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode, includes identifying a subsection of the patient's ECG, the subsection including ECG data for a first period before the arrhythmia episode, a second period during the arrhythmia episode, and a third period after the arrhythmia episode, the length of time of the patient's ECG being greater than the first, second, and third periods; identifying one or more of the cardiac features consistent with the first, second, and third periods; and including in the report the subsection of the ECG and the one or more of the cardiac features consistent with the first, second, and third periods.

[0111] Example 9. A method for detecting a plurality of patients, further comprising: processing, by a computing device, the received electrocardiogram data to generate an intermediate representation of the received electrocardiogram data. The method of any of Examples 1 to 8, wherein applying a machine learning model trained using the electrocardiogram data of a plurality of patients to the received electrocardiogram data to determine that an arrhythmia episode has occurred in the patient comprises applying a machine learning model trained using intermediate representations of the electrocardiogram data of a plurality of patients to the received intermediate representation of the electrocardiogram data and cardiac features present in the electrocardiogram data to determine that an arrhythmia episode has occurred in the patient based on the machine learning model.

[0112] Example 10. The method of Example 9, wherein processing the received electrocardiogram data to generate an intermediate representation of the received electrocardiogram data includes at least one of applying filtering to the received electrocardiogram data and performing signal decomposition on the received electrocardiogram data.

[0113] Example 11. The method of example 10, wherein performing signal decomposition on the received electrocardiogram data includes performing wavelet decomposition on the received electrocardiogram data.

[0114] Example 12. A method comprising: receiving, by a computing device including a processing circuit and a storage medium, electrocardiogram data of a patient sensed by a medical device; obtaining, by the computing device, a first classification of arrhythmia in the patient determined by a feature-based representation of the received electrocardiogram data, the feature-based representation identifying cardiac features present in the electrocardiogram data; applying, by the computing device, a machine learning model trained using electrocardiogram data of a plurality of patients to the received electrocardiogram data to determine, based on the machine learning model, a second classification of arrhythmia in the patient; determining, by the computing device and based on the first classification and the second classification, that an arrhythmia episode has occurred in the patient; generating, by the computing device, a report including an indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode in response to determining that an arrhythmia episode has occurred in the patient; and outputting, by the computing device, the report including an indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode for display.

[0115] Example 13. The method of Example 12, wherein determining that an arrhythmia episode has occurred in the patient based on the first classification and the second classification includes: determining, by a computing device, a similarity between the first classification and the second classification; and determining, by a computing device, that an arrhythmia episode has occurred in the patient based on the similarity between the first classification and the second classification.

[0116] Example 14. The method of example 12, wherein applying the machine learning model to the received electrocardiogram data to determine a second classification of arrhythmia in the patient includes applying the machine learning model to the received electrocardiogram data and cardiac features identified by the feature-based portrayal of the received electrocardiogram data to determine a second classification of arrhythmia in the patient; determining that an arrhythmia episode has occurred in the patient based on the first classification and the second classification includes determining that the first classification indicates that an arrhythmia episode has occurred in the patient, and in response to determining that the first classification indicates that an arrhythmia episode has occurred in the patient, determining that the second classification verifies that an arrhythmia episode has occurred in the patient; and in response to the second classification verifying that an arrhythmia episode has occurred in the patient, determining that an arrhythmia episode has occurred in the patient.

[0117] Example 15. A method as described in any of Examples 12 to 14, wherein obtaining a first classification of arrhythmia in the patient determined by feature-based representation of the received electrocardiogram data by a computing device includes performing, by the computing device, feature-based representation of the received electrocardiogram data to determine a first classification of arrhythmia in the patient.

[0118] Example 16. A method as described in any of Examples 12 to 15, wherein obtaining a first classification of arrhythmia in the patient determined by feature-based depiction of the received electrocardiogram data by a computing device includes receiving, by the computing device and from the medical device, a first classification of arrhythmia in the patient determined by feature-based depiction of the received electrocardiogram data by the medical device.

[0119] Example 17. A method as described in any of Examples 12 to 16, wherein obtaining a first classification of arrhythmia in the patient determined by feature-based depiction of the received electrocardiogram data includes obtaining a first classification of arrhythmia in the patient determined by at least one of QRS detection, refractory processing, noise processing, or depiction of the electrocardiogram data to obtain cardiac features present in the electrocardiogram data.

[0120] Example 18. The method of any of Examples 12-17, wherein applying the machine learning model to determine a second classification of arrhythmia in the patient includes applying the machine learning model to determine that at least one episode of bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block has occurred in the patient.

[0121] Example 19. The method of any of Examples 12-18, wherein the cardiac features present in the electrocardiogram data are one or more of the patient's average heart rate, the patient's minimum heart rate, the patient's maximum heart rate, the patient's cardiac PR interval, the patient's heart rate variability, one or more amplitudes of one or more features of the patient's electrocardiogram (ECG), or intervals between one or more features of the patient's ECG.

[0122] Example 20. The method of any of Examples 12-19, wherein the machine learning model trained using electrocardiogram data of a plurality of patients comprises a machine learning model trained using a plurality of electrocardiogram (ECG) waveforms, each ECG waveform being labeled with one or more episodes of one or more types of arrhythmia in a patient of the plurality of patients.

[0123] Example 21. The method of any of Examples 12-20, wherein applying the machine learning model to the received electrocardiogram data includes applying the machine learning model to at least one of one or more characteristics of the received electrocardiogram data that correlate to arrhythmia in the patient, an activity level of the medical device, an input impedance of the medical device, or a battery level of the medical device.

[0124] Example 22. The method of any of Examples 12-21, wherein the patient's cardiac electrogram data includes an electrocardiogram (ECG) of the patient, and generating a report including an indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode includes identifying a subsection of the patient's ECG, the subsection including ECG data for a first time period prior to the arrhythmia episode, a second time period during the arrhythmia episode, and a third time period after the arrhythmia episode, the length of the period of the patient's ECG being greater than the first, second, and third time periods, identifying one or more of the cardiac features consistent with the first, second, and third time periods, and including in the report the subsection of the ECG and the one or more cardiac features consistent with the first, second, and third time periods.

[0125] Example 23. A method, comprising: receiving, by a computing device, electrocardiogram data of the patient sensed by the medical device; obtaining, by a computing device, a first classification of arrhythmia for the patient determined by a feature based representation of the received electrocardiogram data, the feature based representation identifying a first cardiac feature present in the electrocardiogram data that is consistent with the first classification of arrhythmia in the patient; determining, by the computing device, that one or more episodes of arrhythmia of the first classification have previously occurred in the patient; and in response to determining that one or more episodes of arrhythmia of the first classification have previously occurred in the patient, applying, by the computing device, a machine learning model trained using electrocardiogram data of the plurality of patients to the received electrocardiogram data and the first cardiac feature in the electrocardiogram data, to determine based on the machine learning model that the first cardiac feature has previously occurred in the patient. determining, by the computing device, that an episode of the first class of arrhythmia has occurred in the patient in response to determining that the first cardiac features are similar to cardiac features consistent with one or more episodes of the first class of arrhythmia that have previously occurred in the patient; generating, by the computing device, in response to determining that an episode of the first class of arrhythmia has occurred in the patient, a report including an indication that an episode of the first class of arrhythmia has occurred in the patient and one or more of the cardiac features consistent with the arrhythmia episode; and outputting, by the computing device, the report including the indication that an episode of arrhythmia has occurred in the patient and the one or more cardiac features consistent with the arrhythmia episode for display.

[0126] Example 24. The method of example 23, comprising: obtaining, by a computing device, a second classification of arrhythmia in the patient as determined by a feature-based representation of the received electrocardiogram data, wherein the feature-based representation identifies a second cardiac feature present in the electrocardiogram data that is consistent with the second classification of arrhythmia in the patient; determining, by the computing device, that one or more episodes of arrhythmia of the second classification have not previously occurred in the patient; outputting, by the computing device, the second cardiac feature and at least a portion of the received electrocardiogram data for display in response to determining that one or more episodes of arrhythmia of the second classification have not previously occurred in the patient; receiving, by the computing device, an indication from the user that the second cardiac feature is indicative of an episode of arrhythmia of the second classification in the patient; and storing, by the computing device, the indication that the second cardiac feature is indicative of an episode of arrhythmia of the second classification in the patient and the second cardiac feature.

[0127] Example 25. Obtaining, by a computing device, a second classification of arrhythmia in the patient as determined by a feature-based representation of received electrocardiogram data, the feature-based representation identifying a third cardiac feature present in the electrocardiogram data that is consistent with the second classification of arrhythmia in the patient; determining, by the computing device, that one or more episodes of arrhythmia of the second classification have previously occurred in the patient; and in response to determining that one or more episodes of arrhythmia of the second classification have previously occurred in the patient, applying, by the computing device, a machine learning model to the received electrocardiogram data and the third cardiac feature in the electrocardiogram data to determine, based on the machine learning model, that the third cardiac feature is similar to the second cardiac feature consistent with one or more episodes of arrhythmia of the second classification that have previously occurred in the patient; and in response to determining that the third cardiac feature is similar to the second cardiac feature consistent with one or more episodes of arrhythmia of the second classification that have previously occurred in the patient, determining, by the computing device, that an episode of arrhythmia of the second classification has occurred in the patient. 25. The method of claim 24, comprising: in response to determining that an episode of a third arrhythmia of the third classification has occurred, generating, by the computing device, a second report including an indication that an episode of a third arrhythmia of the third classification has occurred in the patient and one or more of the third cardiac features consistent with an episode of a third arrhythmia of the third classification; and outputting, by the computing device, the report including an indication that an episode of a third arrhythmia of the third classification has occurred in the patient and one or more of the third cardiac features consistent with an episode of a third arrhythmia of the third classification for display.

[0128] Example 26. A method according to any of Examples 23 to 25, comprising: applying a machine learning model to the received electrocardiogram data and a first cardiac feature present in the electrocardiogram data to determine, based on the machine learning model, that the first cardiac feature is similar to a cardiac feature consistent with one or more episodes of a first classification of arrhythmia that previously occurred in the patient; applying the machine learning model to the first cardiac feature to make a preliminary determination that the first cardiac feature is similar to a cardiac feature consistent with one or more episodes of a first classification of arrhythmia that previously occurred in the patient; and outputting an estimate of the certainty of the preliminary determination; and in response to determining that the estimate of the certainty of the preliminary determination is greater than a predetermined threshold, determining that the first cardiac feature is similar to a cardiac feature consistent with one or more episodes of a first classification of arrhythmia that previously occurred in the patient.

[0129] Example 27. The method of any of Examples 23-26, wherein performing feature-based representation of the electrocardiogram data to obtain cardiac features present in the electrocardiogram data includes performing at least one of QRS detection, refractory processing, noise processing, or representation of the electrocardiogram data to obtain cardiac features present in the electrocardiogram data.

[0130] Example 28. The method of any of Examples 23-27, wherein applying the machine learning model to determine that the first cardiac feature is similar to cardiac features consistent with one or more episodes of arrhythmia of a first classification that have previously occurred to the patient includes applying the machine learning model to determine that the first cardiac feature is indicative of at least one episode of bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block that has previously occurred to the patient.

[0131] Example 29. The method of any of Examples 23-28, wherein the first cardiac feature present in the electrocardiogram data is one or more of the patient's average heart rate, the patient's minimum heart rate, the patient's maximum heart rate, the patient's cardiac PR interval, the patient's heart rate variability, one or more amplitudes of one or more features of the patient's electrocardiogram (ECG), or an interval between one or more features of the patient's ECG.

[0132] Example 30. The method of any of Examples 23-29, wherein the machine learning model trained using electrocardiogram data of a plurality of patients comprises a machine learning model trained using a plurality of electrocardiogram (ECG) waveforms, each ECG waveform being labeled with one or more arrhythmia episodes in a patient of the plurality of patients.

[0133] Example 31. The method of any of Examples 23-30, wherein applying the machine learning model to the received electrocardiogram data includes applying the machine learning model to at least one of one or more characteristics of the received electrocardiogram data that correlate to the patient's arrhythmia, an activity level of the medical device, an input impedance of the medical device, or a battery level of the medical device.

[0134] Example 32. The method of any of Examples 23 to 31, further comprising receiving, by the computing device and from a user, an adjustment to the feature-based representation of the electrocardiogram data in response to outputting a report including an indication that an arrhythmia episode has occurred in the patient and one or more cardiac features consistent with the arrhythmia episode, and performing the feature-based representation of the electrocardiogram data in accordance with the adjustment to obtain a second cardiac feature present in the electrocardiogram data.

[0135] Example 33. The method of any of Examples 23-32, wherein the patient's electrocardiogram data includes an electrocardiogram (ECG) of the patient, and generating a report including an indication that an arrhythmia episode has occurred in the patient and one or more first cardiac features consistent with the arrhythmia episode includes identifying a subsection of the patient's ECG, the subsection including ECG data for a first period prior to the arrhythmia episode, a second period during the arrhythmia episode, and a third period after the arrhythmia episode, the length of the period of the patient's ECG being greater than the first, second, and third periods; identifying one or more of the first cardiac features consistent with the first, second, and third periods; and including in the report the subsection of the ECG and the one or more first cardiac features consistent with the first, second, and third periods.

[0136] Example 34. The method of any of Examples 23 to 33, further comprising processing the received electrocardiogram data by a computing device to generate an intermediate representation of the received electrocardiogram data, and applying a machine learning model trained using the electrocardiogram data of a plurality of patients to the received electrocardiogram data to determine that an arrhythmia episode has occurred in the patient, comprising applying a machine learning model trained using the intermediate representations of the electrocardiogram data of a plurality of patients to the intermediate representation of the received electrocardiogram data and cardiac features present in the electrocardiogram data, and determining, based on the machine learning model, that a similar arrhythmia episode has occurred in the patient.

[0137] Example 35. The method described in Example 34, wherein processing the received electrocardiogram data to generate an intermediate representation of the received electrocardiogram data includes at least one of applying filtering to the received electrocardiogram data and performing signal decomposition on the received electrocardiogram data.

[0138] Example 36. The method of example 35, wherein performing signal decomposition on the received electrocardiogram data includes performing wavelet decomposition on the received electrocardiogram data.

[0139] In some examples, the disclosed techniques include a system including means for performing any of the methods described herein. In some examples, the disclosed techniques include a computer-readable medium including instructions that cause a processing circuit to perform any of the methods described herein.

[0140] It should be understood that the various aspects disclosed herein can be combined in different combinations than those specifically presented in the description and accompanying drawings. It should also be understood that certain acts or events of any of the processes or methods described herein may be performed in different orders, or may be added, combined, or omitted entirely, depending on the embodiment (e.g., not all described acts or events may be necessary to perform the technology). Furthermore, although certain aspects of the disclosure are described for clarity as being performed by a single module, unit, or circuit, it should be understood that the technology of the disclosure may be performed by a combination of units, modules, or circuits associated with, for example, a medical device device.

[0141] In one or more examples, the described techniques may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored as one or more instructions or code on a computer-readable medium and executed by a hardware-based processing unit. The computer-readable medium may be a data storage medium, This may include non-transitory computer-readable media corresponding to tangible media, such as a medium (e.g., RAM, ROM, EEPROM, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer).

[0142] The instructions may be executed by one or more processors, such as one or more digital signal processors (DSPs), general-purpose microprocessors, application specific integrated circuits (ASICs), field programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. Thus, the terms "processor" or "processing circuitry" as used herein may refer to any of the foregoing structures, or any other physical structure suitable for implementing the described techniques. Also, the techniques may be implemented entirely in one or more circuits or logic elements.

Claims

1. 1. A computing device comprising: A storage medium; a processing circuit operably coupled to the storage medium; and receiving electrocardiogram data of a patient sensed by a medical device; applying a machine learning model trained using electrocardiogram data of a plurality of patients to the received electrocardiogram data and determining that an arrhythmia episode has occurred in the patient based on the machine learning model; performing feature based representation of the received electrocardiogram data to obtain cardiac features present in the electrocardiogram data; In response to determining that the arrhythmia episode has occurred in the patient, generating a report including an indication that the arrhythmia episode has occurred in the patient and one or more of the cardiac features consistent with the arrhythmia episode; and outputting, for display, the report including the indication that the arrhythmia episode has occurred in the patient and one or more of the cardiac features consistent with the arrhythmia episode.

2. 2. The computing device of claim 1, wherein the processing circuitry is configured to perform at least one of QRS detection, refractory processing, noise processing, or rendering of the electrocardiogram data to obtain cardiac features present in the electrocardiogram data, in order to perform feature-based rendering of the electrocardiogram data to obtain cardiac features present in the electrocardiogram data.

3. 3. The computing device of claim 1, wherein the processing circuitry is configured to apply the machine learning model to determine that an episode of at least one of bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block has occurred in the patient, in order to apply the machine learning model to determine that an episode of the arrhythmia has occurred in the patient.

4. 4. A computing device according to claim 1, wherein the cardiac features present in the electrocardiogram data are one or more of the patient's average heart rate, the patient's minimum heart rate, the patient's maximum heart rate, the patient's cardiac PR interval, the patient's heart rate variability, one or more amplitudes of one or more features of the patient's electrocardiogram (ECG), or intervals between the one or more features of the patient's ECG.

5. 5. The computing device of claim 1, wherein the machine learning model trained using electrocardiogram data of the plurality of patients comprises a machine learning model trained using a plurality of electrocardiogram (ECG) waveforms, each ECG waveform labeled with one or more episodes of one or more types of arrhythmia in a patient of the plurality of patients.

6. To apply the machine learning model to the received electrocardiogram data, the processing circuitry performs the machine learning model: one or more characteristics of the received electrocardiogram data that are correlated to arrhythmia in the patient; an activity level of the medical device; the input impedance of the medical device; or A computing device according to any preceding claim, configured to apply to at least one of: a.

7. to output the report including the indication that the arrhythmia episode has occurred in the patient and the one or more of the cardiac features consistent with the arrhythmia episode. receiving adjustments to the feature-based representation of the electrocardiogram data from a user; and performing a feature-based representation of the electrocardiogram data in accordance with the adjustment to obtain second cardiac features present in the electrocardiogram data.

8. the electrocardiogram data of the patient includes an electrocardiogram (ECG) of the patient; to generate the report including the indication that the arrhythmia episode has occurred in the patient and the one or more of the cardiac features consistent with the arrhythmia episode, identifying a subsection of the ECG of the patient, the subsection including ECG data for a first time period prior to the arrhythmic episode, a second time period during the arrhythmic episode, and a third time period after the arrhythmic episode, the length of time of the ECG of the patient being greater than the first, second, and third time periods; identifying one or more of the cardiac features consistent with the first, second, and third time periods; and including in the report the subsection of the ECG and the one or more of the cardiac features that coincide with the first, second, and third time periods.

9. the processing circuitry is further configured to process the received electrocardiogram data to generate an intermediate representation of the received electrocardiogram data; 9. The computing device of claim 1, wherein the processing circuitry is configured to: apply a machine learning model, trained using an intermediate representation of the electrocardiogram data of a plurality of patients, to the intermediate representation of the received electrocardiogram data and the cardiac features present in the electrocardiogram data to determine that the arrhythmia episode has occurred in the patient, the processing circuitry being configured to: apply a machine learning model, trained using an intermediate representation of the electrocardiogram data of a plurality of patients, to the intermediate representation of the received electrocardiogram data and the cardiac features present in the electrocardiogram data, and determine based on the machine learning model that the arrhythmia episode has occurred in the patient.

10. the processing circuitry for processing the received electrocardiogram data to generate the intermediate representation of the received electrocardiogram data; applying filtering to the received electrocardiogram data; and performing signal decomposition on the received electrocardiogram data.

11. 11. The computing device of claim 10, wherein to perform signal decomposition on the received electrocardiogram data, the processing circuitry is configured to perform wavelet decomposition on the received electrocardiogram data.

12. 1. A computing device comprising: means for receiving electrocardiogram data of a patient sensed by a medical device; means for applying a machine learning model trained using electrocardiogram data of a plurality of patients to the received electrocardiogram data and determining that the arrhythmia episode has occurred in the patient based on the machine learning model; means for performing feature based representation of the received electrocardiogram data to obtain cardiac features present in the electrocardiogram data; In response to determining that the arrhythmia episode has occurred in the patient, generating a report including an indication that the arrhythmia episode has occurred in the patient and one or more of the cardiac features consistent with the arrhythmia episode; and outputting, for display, the report including the indication that the arrhythmia episode has occurred in the patient and one or more of the cardiac features consistent with the arrhythmia episode.

13. A system comprising the computing device according to any preceding claim.

Images