Feature description and machine learning-based arrhythmia detection

Abstract

Techniques are disclosed for using both characterization and machine learning to detect cardiac arrhythmias. A computing device receives a patient's electrocardiogram data sensed by a medical device. The computing device obtains a first classification of the patient's arrhythmia through 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 one 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 to verify the patient's first classification of the arrhythmia. The computing device outputs a report indicating that an arrhythmia episode has occurred and one or more cardiac features consistent with the arrhythmia episode.

Description

Technical Field

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

Background Art

[0002] Malignant tachyarrhythmias such as ventricular fibrillation are uncoordinated contractions of the myocardium of the ventricles of the heart and are the most commonly identified arrhythmias in cardiac arrest patients. If this arrhythmia persists 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 within just a few minutes.

[0003] In patients at high risk of ventricular fibrillation, the use of implantable medical devices (IMDs) such as implantable cardioverter defibrillators (ICDs) has been shown to be beneficial in preventing SCD. An ICD is a battery-powered electrical shock device that can include an electrical housing electrode (sometimes referred to as a can electrode) and is typically coupled to one or more electrical leads placed within the heart. When an arrhythmia is sensed, the ICD can use the electrical leads to transmit a pulse and deliver a shock to the heart to restore a normal rhythm. Some ICDs are configured to terminate detected tachyarrhythmias by delivering anti-tachycardia pacing (ATP) prior to shock delivery. Additionally, the ICD is configured to deliver relatively high-amplitude post-shock pacing after the normal termination of a tachyarrhythmia with shock 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 can be used for diagnostic purposes. For example, implantable or non-implantable medical devices can monitor a patient's heart. A user, such as a physician, can review data generated by the medical device for, e.g., the occurrence of atrial or ventricular tachyarrhythmias, or arrhythmias of cardiac arrest. The user can diagnose the patient's medical condition based on the occurrence of the identified arrhythmia. [Overview of the project]

[0005] This disclosure describes a medical device system that uses both feature description and machine learning to detect and classify cardiac arrhythmias in a patient, in accordance with the technology of this disclosure. For example, a computing device receives electrocardiogram data from a patient sensed by an implantable medical device. The computing device obtains a first classification of the patient's arrhythmias through a feature-based description of the electrocardiogram data. The computing device obtains a second classification of the patient's arrhythmias by applying a machine learning model to the received electrocardiogram data. As an example, the computing device uses the first and second classifications to determine whether an episode of arrhythmia has occurred in the patient. As another example, the computing device uses the second classification of arrhythmias obtained from the machine learning model to validate the first classification of the patient's arrhythmias obtained from the feature-based description.

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

[0007] Furthermore, the medical device systems described herein may classify arrhythmias according to an arrhythmia dictionary. For example, a computing device may determine that an episode of arrhythmia has occurred in a patient through a feature-based description of the patient's electrocardiogram data. The computing device may apply a machine learning model to classify the episode of arrhythmia as a specific type of arrhythmia by comparing cardiac features that match the episode of arrhythmia with cardiac features of the patient's past episodes of arrhythmia.

[0008] The technologies of this disclosure may provide concrete improvements in the field of cardiac arrhythmia detection and classification. For example, the combined use of both feature description and machine learning may improve the accuracy of detecting patient arrhythmias compared to the use of feature description or machine learning individually. Furthermore, the medical device systems described herein may enable the implantable medical device of the medical device system to function as a low-granularity filter for detecting patient arrhythmias while offloading the power-intensive and computationally complex verification of arrhythmia detection to an external computing device. Thus, systems such as those described herein can reduce power consumption and improve the battery life of implanted devices in the patient while providing high accuracy in arrhythmia detection and classification. Such improvements may also be achieved in low-power external devices that can detect arrhythmias based on the electrical signals of the heart, such as patient monitors in the form of wearable patches, watches, necklaces, or other devices worn by the patient.

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

[0010] In another example, the disclosure describes a method comprising: a computing device including processing circuits and a storage medium receiving patient electrocardiogram data sensed by a medical device; the computing device obtaining a first classification of the patient's arrhythmia determined by a feature-based description of the received electrocardiogram data, wherein the feature-based description identifies and obtains cardiac features present in the electrocardiogram data; the computing device applying a machine learning model trained using electrocardiogram data of multiple patients to the received electrocardiogram data to determine a second classification of the patient's arrhythmia based on the machine learning model; the computing device determining, based on the first and second classifications, that an episode of arrhythmia has occurred in the patient; the computing device generating a report in response to the determination that an episode of arrhythmia has occurred in the patient, including an indicator that an episode of arrhythmia has occurred in the patient and one or more cardiac features that match the episode of arrhythmia; and the computing device outputting the report, including the indicator that an episode of arrhythmia has occurred in the patient and one or more cardiac features that match the episode of arrhythmia, for display.

[0011] In another example, the disclosure relates to a computing device including processing circuits and a storage medium that receives patient electrocardiogram data sensed by a medical device, and the computing device obtains a first classification of the patient's arrhythmias determined by a feature-based description of the received electrocardiogram data, wherein the feature-based description identifies and obtains a first cardiac feature present in the electrocardiogram data that matches the first classification of arrhythmias in the patient, and the computing device determines that one or more episodes of the first classification of arrhythmias have previously occurred in the patient, and in response to the determination that one or more episodes of the first classification of arrhythmias have previously occurred in the patient, the computing device applies a machine learning model trained using electrocardiogram data from multiple patients to the received electrocardiogram data and the first cardiac feature of the electrocardiogram data, based on the machine learning model. The present invention describes a method comprising: determining whether a first cardiac feature is similar to cardiac features that match one or more episodes of a first-class arrhythmia previously occurring in the patient; determining, in response to the determination that the first cardiac feature is similar to cardiac features that match one or more episodes of a first-class arrhythmia previously occurring in the patient, that an episode of a first-class arrhythmia has occurred in the patient; generating a report by the computing device that includes an indicator that an episode of a first-class arrhythmia has occurred in the patient and one or more cardiac features that match the episode of arrhythmia; and outputting the report by the computing device for display, which includes an indicator that an episode of arrhythmia has occurred in the patient and one or more cardiac features that match the episode of arrhythmia.

[0012] This summary is intended to provide an overview of the subject matter described herein. 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 provided in the accompanying drawings and description below. [Brief explanation of the drawing]

[0013] [Figure 1] This 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-related external device, in accordance with the technology of this disclosure. [Figure 2] Figure 1 is a block diagram showing an example of a leadless implantable medical device. [Figure 3] Figure 1 is a block diagram showing another example of a leadless implantable medical device. [Figure 4] A block diagram shows an exemplary computing device operating in accordance with one or more of the technologies of this disclosure. [Figure 5] This flowchart shows an exemplary operation of the technology disclosed herein. [Figure 6] This chart shows an example of an electrocardiogram obtained from the patient in Figure 1. [Figure 7] This flowchart shows an exemplary operation of the technology disclosed herein. [Figure 8] This flowchart shows an exemplary operation of the technology disclosed herein. [Figure 9] This flowchart shows an exemplary operation of the technology disclosed herein. [Figure 10] This flowchart shows an exemplary operation of the technology disclosed herein.

[0014] Similar reference characters refer to the same elements throughout the figure and description. [Modes for carrying out the invention]

[0015] Disclosed are techniques for combining multiple decision mechanisms to perform the detection and classification of patient cardiac arrhythmias, such as machine learning systems and / or artificial intelligence (AI) algorithms that analyze single-channel and multi-channel patient data, including state-of-the-art signal processing algorithms that perform feature description of electrocardiogram data and machine learning models that process patient data. Such patient data may include, for example, electrocardiogram data or electrocardiogram (ECG) data.

[0016] Where used herein, feature depiction refers to the use of features obtained through signal processing for use in detecting or classifying episodic cardiac arrhythmias. Typically, feature depiction involves identifying or extracting features from electrocardiogram data, measuring the properties of such features, and using the measurements to use designed rules for detecting or classifying arrhythmias. For example, feature depiction may be used to identify features such as R waves, QRS complexes, P waves, T waves, the velocity of such features, the interval between such features, the morphology of such features, the width or amplitude of such features, or other or other types of cardiac features or properties of features not expressly described herein. Feature depiction may include feature extraction, signal filtering, peak detection, refractory analysis, or other types of signal processing, feature engineering, or the development of detection rules. Feature depiction algorithms may be optimized for real-time, embedded, and low-power applications, such as use in implantable medical devices. However, feature depiction algorithms may require expert design and feature engineering to accurately detect patient arrhythmias.

[0017] In contrast to feature description techniques for detecting and classifying cardiac arrhythmias, machine learning techniques can be used for the detection and classification of cardiac arrhythmias. As described herein, machine learning refers to the use of machine learning models, such as neural networks or deep learning models, which are trained on training datasets for detecting cardiac arrhythmias from electrocardiogram data. Machine learning techniques are in contrast to feature description in that feature description relies on signal processing, and machine learning systems can "learn" the underlying features present in electrocardiogram data that indicate arrhythmia episodes without requiring knowledge or understanding of the relationships between features and arrhythmia episodes on behalf of the system designer.

[0018] Machine learning and AI techniques for arrhythmia detection can provide a flexible platform for developing arrhythmia detection and classification algorithms for various purposes (e.g., detection of atrial fibrillation (AF), exclusion of cardiac episodes that do not exhibit arrhythmias) without requiring the feature engineering required by expert-designed or feature-drawing algorithms. Disclosed herein are techniques, methods, systems, and devices that combine feature drawing and machine learning to detect and classify patient cardiac arrhythmias in a manner that improves accuracy and robustness compared to the use of feature drawing alone, and further reduces power consumption by implantable devices compared to the use of machine learning alone.

[0019] Figure 1 shows an environment of an exemplary medical device system 2 related to a patient 4 and a heart 6, according to the apparatus and methods of a particular example described herein. The exemplary technique may be used with an IMD 10, which may be leadless as shown in Figure 1 and capable of wireless communication with an external device 12. 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 Figure 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 can take the form of both a Reveal LINQ™ insertable cardiac monitor (ICM) or a Holter cardiac monitor, available from Medtronic plc in Dublin, Ireland. The external device 12 can be a computing device configured to be used in settings such as a home, clinic, or hospital, and can further be configured to communicate with the IMD 10 via wireless telemetry. For example, the external device 12 can be coupled to a computing system 24 via a network 25. The computing system 24 can include a remote patient monitoring system such as Carelink® available from Medtronic plc in Dublin, Ireland. The external device 12 can, in some examples, include a programmer, a communication device such as an external monitor, or a mobile device such as a cellular phone, a "smart" phone, a laptop, a tablet computer, a personal digital assistant (PDA), etc.

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

[0022] In some examples, a user such as a physician, technician, surgeon, electrophysiologist, or other clinician may interact with the external device 12 and retrieve physiological or diagnostic information from the IMD 10. In some examples, a user such as the patient 4 or clinician as described above may also interact with the external device 12 and be able to program the IMD 10, for example, select or adjust values of the operating parameters of the IMD 10. In some examples, the external device 12 functions as an access point to facilitate communication between the IMD 10 via the network 25, for example, by the computing system 24. The computing system 24 may include a computing device configured to enable a user to interact with the IMD 10 via 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 that includes a user interface for presenting information to a user and receiving input from the 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 physician, technician, surgeon, electrophysiologist, or other clinician may interact with the computing system 24 and retrieve physiological or diagnostic information from the IMD 10. The user may also interact with the computing system 24 to be able to program the IMD 10, for example, select values of the 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 non-edge switches, routers, hubs, gateways, firewalls, security devices such as intrusion detection and / or intrusion prevention devices, one or more computing devices (not shown) such as servers, computer terminals, laptops, printers, databases, wireless mobile devices such as mobile 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 may form part of a larger public network infrastructure such as the Internet. Network 25 may provide computing devices such as computing system 24 and IMD 10 with access to the Internet and may provide a communication framework that allows computing devices to communicate with each other. In some examples, network 25 allows computing system 24, IMD 10, and / or external devices 12 to communicate with each other, but for security purposes, one or more of the external devices 12 are isolated from computing system 24, IMD 10, or devices outside network 25. In some examples, communication between computing system 24, IMD 10, and external devices 12 is encrypted.

[0025] The external device 12 and the computing system 24 may communicate wirelessly 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 via an intermediate device located on the network 25, such as a local access point, wireless router, or gateway. In the example in Figure 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 include, for example, communication according to the Bluetooth® or BLE protocol. Other communication techniques are also conceivable. The computing system 24 may also communicate with one or more other external devices using several known communication techniques, both wired and wireless.

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

[0027] The external device 12 may be a computing device for communicating with the IMD 10 via wireless telemetry (for example, used in a home, outpatient, clinic, or hospital setting). The external device 12 may include or be coupled with a remote patient monitoring system such as Carelink®, available from Medtronic plc in 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 (for example, if configured as a programmer for the IMD 10). In some examples, the external device 12 may be used to query the IMD 10 to retrieve data including the device's operating data as well as physiological data stored in the IMD memory. Such queries may occur automatically according to a schedule or in response to remote or local user commands. Programmers, external monitors, and consumer devices are examples of external devices 12 that may be used to query the IMD 10. An example of a communication technology used by the IMD 10 and the external device 12 is radio frequency (RF) telemetry, which may be an RF link established via Bluetooth, WiFi, or Medical Implant Communication Services (MICS). In some examples, the external device 12 may include a user interface configured to allow a patient 4, a clinician, or another user to interact with the IMD 10 remotely. In some such examples, the external device 12, and / or other devices in the medical device system 2, may be wearable devices (for example, in the form of a watch, necklace, or other wearable item).

[0029] Medical device system 2 is an example of a medical device system configured to perform the detection, verification, and reporting of cardiac arrhythmias. In accordance with the art of this disclosure, medical device system 2 performs machine learning arrhythmia detection and feature depiction to detect and classify cardiac arrhythmias in patient 4. Additional examples of one or more other implantable or external devices include implantable, multi-channel cardiac pacemakers, ICDs, IPGs, leadless (e.g., intracardiac) pacemakers, extravascular pacemakers and / or ICDs, or other IMDs, or combinations of such IMDs configured to deliver CRT to the heart 6, external therapeutic delivery devices such as external monitors, external pacing or electrical stimulation devices, or drug pumps.

[0030] Communication circuits in each device of the medical device system 2 (e.g., IMD 10 and external device 12) may enable the devices to communicate with each other. Furthermore, one or more sensors (e.g., electrodes) are described herein as being located on the housing of IMD 10, but in other examples, such sensors may be located 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 circuits configured to receive signals from electrodes or other sensors on each device, and / or communication circuits configured to transmit signals from electrodes or other sensors to another device (e.g., external device 12) or a server.

[0031] According to the technology of this disclosure, a medical device system 2 uses both feature description and machine learning to detect and classify cardiac arrhythmias in patient 4. For example, a computing system 24 receives electrocardiogram data of patient 4 sensed by an implantable medical device 10. The computing system 24 obtains a first classification of patient 4's arrhythmia through a feature-based description of the electrocardiogram data. In some examples, the feature-based description of the electrocardiogram data to determine the first classification of patient 4's arrhythmia is performed by one of the following: IMD 10, an external device 12, or the computing system 24. A machine learning system 150 applies a machine learning model to the received electrocardiogram data to obtain a second classification of patient 4's arrhythmia. In one example, the machine learning model is a deep learning model. In one example, the computing system 24 uses the first and second classifications to determine whether an episode of arrhythmia occurred in patient 4. In another example, the computing system 24 uses the second classification of arrhythmia obtained from the machine learning system 150 to validate the first classification of patient 4's arrhythmia obtained from the feature-based description.

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

[0033] Furthermore, the medical device system 2 can classify arrhythmias according to an arrhythmia dictionary. As will be described in more detail below, the computing system 24 determines that an episode of arrhythmia occurred in patient 4 through a feature-based description of patient 4's electrocardiogram data. The machine learning system 150 applies a machine learning model to classify the episode of arrhythmia as a specific type of arrhythmia by comparing cardiac features that match the episode of arrhythmia with cardiac features of past episodes of arrhythmia in patient 4.

[0034] The technology of this disclosure may provide specific improvements in the field of cardiac arrhythmia detection and classification. For example, using both feature description and machine learning in combination may improve the accuracy of arrhythmia detection in patient 4 compared to using feature description or machine learning individually. Furthermore, the medical device system 2 described herein may allow the implantable medical device 10 to function as a low-granularity filter for detecting arrhythmias in patient 4, while offloading the power-intensive and computationally complex verification of arrhythmia detection to an external device such as an external device 12 or computing system 24. Thus, as described herein, system 2 may provide high accuracy in the detection and classification of arrhythmias in patient 4 while reducing power consumption and improving the battery life of the IMD 10.

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

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

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

[0038] In some examples, the IMD10 performs feature description of the perceived electrocardiogram data, as described in more detail below. In some examples, the feature description performed by the IMD10 is simplified to conserve power on the IMD10. This may allow the IMD10 to perform early or preliminary detection of cardiac arrhythmias. As described in more detail below, the computing system 24 may further perform feature description of the perceived electrocardiogram data by the IMD10 and apply a machine learning system 150 to the electrocardiogram data. The computing system 24 may have more computing resources and fewer power constraints than the IMD10, thereby enabling the computing system 24 to perform a more comprehensive and detailed analysis of the electrocardiogram data to more accurately detect cardiac arrhythmias. By shifting the computing load from the IMD10 to the computing system 24, the techniques of this disclosure may help reduce the power consumption of the IMD10 while improving 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 circuit 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 in Figure 1, for analysis. In some examples, the sensing circuit 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 circuit 52 and / or the processing circuit 50 may include rectifiers, filters and / or amplifiers, sensing amplifiers, comparators, and / or analog-to-digital converters.

[0040] The communication circuit 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 circuit 50, the communication circuit 54 may receive downlink telemetry and transmit uplink telemetry from the external device 12 or another device with the help of an internal or external antenna, e.g., antenna 26. In some examples, the communication circuit 54 may communicate with the external device 12. In addition, the processing circuit 50 may communicate with the external device (e.g., the external device 12) and networked computing devices via a computer network such as the Medtronic CareLink® network developed by Medtronic, plc in Dublin, Ireland.

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

[0042] One or more components of the IMD10 may be coupled to a power source (not shown in Figure 2) which may include a rechargeable or non-rechargeable battery located within the housing of the IMD10. Non-rechargeable batteries can be selected to last for several years, while rechargeable batteries can be inductively charged from an external device, for example, daily or weekly.

[0043] According to the technology of this disclosure, a processing circuit 50 senses electrocardiogram data of patient 4 via electrodes 16 using a sensing circuit 52. In some examples, the electrocardiogram data is patient 4's ECG. The processing circuit 50 processes the electrocardiogram data via feature depiction to obtain one or more cardiac features present in the electrocardiogram data. In some examples, feature depiction includes one or more of QRS detection, refractory processing, noise processing, or depiction of the electrocardiogram data. For example, the processing circuit 50 receives a raw signal from the sensing circuit 50 and / or via sensor 58 and extracts one or more cardiac features from the raw signal. In some examples, the processing circuit 50 identifies one or more cardiac features such as, for example, the patient's mean 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, the amplitude of one or more features of the patient's electrocardiogram (ECG), or the interval between one or more features of the patient's ECG, T-wave alternation, QRS morphometry, or one or more other types of cardiac features not expressly described herein.

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

[0045] The processing circuit 50 may further apply such feature descriptions and determine that one or more cardiac features indicate an episode of cardiac arrhythmia. The processing circuit 50 further applies feature descriptions to classify the detected episode of cardiac arrhythmia as a specific type of episode (e.g., bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block). The processing circuit 50 transmits one or more electrocardiogram data, one or more cardiac features present in the electrocardiogram data, an index of the detected episode of cardiac arrhythmia, or an index of the classification of the detected episode of cardiac arrhythmia to the external device 12 via the communication circuit 54.

[0046] Although described herein in the context of an exemplary IMD 10 sensing electrocardiogram data of patient 4, the techniques for detecting cardiac arrhythmias disclosed herein may be used in conjunction with other types of devices. For example, the techniques may be implemented with an extracardiac defibrillator coupled to an electrode outside the cardiovascular system, a transcatheter pacemaker configured to be implanted in the heart, such as the Micra® transcatheter pacing system commercially available from Medtronic PLC in Dublin, Ireland, an implantable cardiac monitor such as the Reveal LINQ® ICM, also commercially available from Medtronic PLC, a nerve stimulation device, a drug delivery device, a medical device located outside 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 showing another example of the leadless implantable medical device shown in Figure 1. The components in Figure 3 do not necessarily have to be drawn to a fixed scale; instead, they may be enlarged to show details. Specifically, Figure 3 is a block diagram of a top view of an example configuration of the IMD10 shown in Figure 1.

[0048] Figure 3 is a conceptual diagram showing an exemplary IMD10 which may include components substantially similar to those of the IMD10 in Figure 1. In addition to the components shown in Figures 1 and 2, the example of the IMD10 shown in Figure 3 may also include a wafer-scale insulating cover 74 which can help insulate electrical signals passing between electrodes 16A, 16B on the housing 14 and processing circuit 50. In some examples, the insulating cover 74 may be placed on top of the open housing 14 to form a housing for the components of the IMD10B. One or more components of the IMD10B (e.g., antenna 26, processing circuit 50, sensing circuit 52, communication circuit 54, and / or switching circuit 60) may be formed on the underside of the insulating cover 74, for example, by using flip-chip technology. The insulating cover 74 may be flipped over onto the housing 14. When placed flipped over onto the housing 14, the components of the IMD10 formed on the underside of the insulating cover 74 may be placed within a gap 78 defined by the housing 14. The housing 14 can 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 this disclosure.

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

[0050] Figure 4 is a block diagram showing an exemplary computing device 400 operating in accordance with one or more of the techniques of this disclosure. In one example, the computing device 400 is an exemplary implementation of the computing system 24 of Figure 1. In one example, the computing device 400 includes processing circuitry 402 for running an application 424, including a machine learning system 450 or other applications described herein. For illustrative purposes, Figure 4 shows the computing device 400 as a standalone device, but the computing device 400 may be any component or system including processing circuitry or other preferred computing environment for executing software instructions, and does not necessarily have to include one or more elements shown in Figure 4 (for example, components such as an input device 404, a communication circuit 406, a user interface device 410, or an output device 412, and in some examples, a storage device 408, may not be located in the same place as the other components or in the same enclosure). In some examples, the computing device 400 may be a cloud computing system distributed across multiple devices.

[0051] As shown in the example in Figure 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 a machine learning system 450 that can be run by the computing device 400, and one or more applications 424 such as an operating system 416. Each of the components 402, 404, 406, 408, 410, and 412 is coupled (physically, communicatively, and / or operationally) for communication between components. In some examples, the 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, the components 402, 404, 406, 408, 410, and 412 may be coupled by one or more communication channels 414.

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

[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 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 volatile memory, meaning that when the computer is turned off, the storage device 408 does not retain the contents it has stored. 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 an application 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 larger amounts of information than volatile memory. The storage device 408 may be further configured for long-term storage of information. In some examples, the storage device 408 includes non-volatile storage elements. Examples of such non-volatile storage elements include magnetic hard disks, optical disks, floppy disks, flash memory, or electrically programmable memory (EPROM) or electrically erasable programmable memory (EEPROM).

[0055] In some examples, the computing device 400 also includes a communication circuit 406. In one example, the computing device 400 uses the communication circuit 406 to communicate with an external device such as the IMD 10 and external device 12 in Figure 1. The communication circuit 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 sending and receiving information. Other examples of such network interfaces may include 3G and WiFi radio.

[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 the user through haptic, audio, or visual feedback. Examples of user interface devices 410 include presence-sensitive displays, mice, keyboards, voice response systems, video cameras, microphones, or any other type of device for detecting commands from the user. In some examples, the presence-sensitive display includes a touchscreen.

[0057] One or more output devices 412 may also be included in the computing device 400. In some examples, the output devices 412 are configured to provide output to the user using tactile, audio, or visual stimuli. Output devices 412 include, 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 a suitable format that can be understood by humans or machines. Additional examples of output devices 412 include speakers, cathode ray tube (CRT) monitors, liquid crystal displays (LCDs), or any other type of device capable of producing user-understandable output.

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

[0059] Application 422 may also include program instructions and / or data that can be executed by the computing device 400. An exemplary application 422 that can be executed by the computing device 400 may include a machine learning system 450. Other additional applications not shown herein may be included alternatively or additionally to provide other functions described herein, but are not illustrated for the sake of simplification.

[0060] In accordance with the technology of this disclosure, the computing device 400 applies the machine learning model of the machine learning system 450 to patient data sensed by IMD 10 to detect and classify episodes of arrhythmia occurring in patient 10. In some examples, the machine learning system 450 is an example of the machine learning system 150 shown in Figure 1.

[0061] In some examples, the machine learning model implemented by the machine learning system 450 is trained on training data that includes electrocardiogram data from 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 arrhythmias. For example, a training ECG waveform may contain multiple segments, each segment labeled with a descriptor specifying the absence of arrhythmias or the presence of arrhythmias of a particular classification (e.g., bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block). In some examples, a clinician manually labels the presence of arrhythmias in each ECG waveform. In some examples, the presence of arrhythmias in each ECG waveform is labeled according to a classification by a feature description 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 compare and contrast different features shown 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 in relation to the occurrence of cardiac arrhythmias of a particular classification. By processing a large number of such ECG waveforms labeled with episodes of arrhythmia, the machine learning system 450 can build and train a machine learning model to receive electrocardiogram data from patients such as patient 4 in Figure 1, which the machine learning system 450 had not previously analyzed, and further process such electrocardiogram data to detect with high accuracy the presence or absence of arrhythmias of various classifications in the patient. Generally, the larger the amount of electrocardiogram data to which the machine learning system 450 is trained, the higher the accuracy of the machine learning model in detecting or classifying cardiac arrhythmias with new electrocardiogram data.

[0062] After the machine learning system 450 has trained a machine learning model, it may receive patient data, such as electrocardiogram data from a specific patient, such as patient 4. The machine learning system 450 applies the trained machine learning model to the patient data to detect the occurrence of cardiac arrhythmia episodes in patient 4. Furthermore, the machine learning system 450 applies the trained machine learning model to the patient data to classify the patient's cardiac arrhythmia episodes as indicating a specific type of arrhythmia. In some examples, the machine learning system 450 may output a preliminary determination that the cardiac arrhythmia episode indicates a specific type of arrhythmia, as well as an estimate of the certainty in the determination. Depending on whether the estimate of 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 specific type of arrhythmia.

[0063] In some examples, a machine learning system may process one or more cardiac features of electrocardiogram data instead of the raw electrocardiogram data itself. One or more cardiac features may be obtained through feature depiction performed by IMD10 as described above. Cardiac features may include, for example, one or more of the following: the patient's mean 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, the amplitude of one or more features of the patient's electrocardiogram (ECG), or the interval between one or more features of the patient's ECG, T-wave alternation, QRS morphometry, or one or more of other types of cardiac features not expressly described herein. In such exemplary implementations, a machine learning system may train a machine learning model through multiple training cardiac features labeled with episodes of arrhythmia, instead of multiple ECG waveforms labeled with episodes of arrhythmia as described above.

[0064] In some examples, the machine learning system 450 may apply machine learning models to other types of data to determine that an episode of arrhythmia occurred in patient 4. For example, the machine learning system 450 may apply machine learning models to one or more characteristics of electrocardiogram data that correlate with the patient's arrhythmia, the activity level of IMD10, the input impedance of IMD10, or the battery level of IMD10.

[0065] In a further example, the processing circuit 402 may generate an intermediate representation of electrocardiogram (ECG) data from the ECG 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 ECG data to generate an intermediate representation of the ECG data. In this example, the machine learning system 450 processes such an intermediate representation of the ECG data to detect and classify episodes of arrhythmia in patient 4. Furthermore, the machine learning system may train a machine learning model through multiple training intermediate representations labeled with episodes of arrhythmia, instead of multiple raw ECG waveforms labeled with episodes of arrhythmia as described above. The use of such intermediate representations of ECG data may enable the machine learning system 450 to train and develop a lighter and less computationally complex machine learning model. Moreover, the use of such intermediate representations of ECG data may require fewer iterations and less training data to build an accurate machine learning model, in contrast to the use of raw ECG data to train a machine learning model.

[0066] In some cases, computing system 24 may use machine learning system 150 to detect other types of arrhythmias not detected by feature description screening analysis. For example, arrhythmia detection algorithms for performing feature description implemented by low-power devices such as IMD10 may not be designed to detect less frequent arrhythmias such as atrioventricular block. Machine learning system 150 can train machine learning models on large datasets in which such arrhythmias are available, thereby providing finer granularity and higher accuracy than, for example, feature description performed by IMD10 alone. Thus, the use of machine learning system 150 can extend the arrhythmia diagnostic capabilities of system 2 by enabling IMD10 to implement general screening algorithms using feature description, and then by using machine learning system 150 to implement machine learning models that can provide a wider range of arrhythmia detection. After detecting types of arrhythmias not detected by feature description, computing system 24 may nevertheless use feature description, such as QRS detection, to assist in the characterization and reporting of other types of arrhythmias detected by the machine learning models of machine learning system 150.

[0067] In some examples, the computing system 24 may adapt the machine learning system 150 to specific use cases. For example, the machine learning system 150 may implement a machine learning model specific to the detection of atrioventricular block and bradycardia if patient 4 is a post-TAVR patient. In another example, the machine learning system 150 may implement a machine learning model specific to the detection of PVCs so that it can be used to risk stratify patients whose PVC burden may be an indication for an ICD.

[0068] Figure 5 is a flowchart illustrating exemplary operation by the technology of this disclosure. For convenience, Figure 5 is described in relation to Figure 1. In some examples, the operation in Figure 5 is an operation for detecting and classifying cardiac arrhythmias in patient 4. In the operation in Figure 5, system 2 combines the capabilities of the machine learning model of machine learning system 150 to learn features, perform classification directly from the input, and have the interpretability provided by the feature description algorithm and ECG processing. In the example operation in Figure 5, system 2 implements the machine learning model of machine learning system 150 in parallel with the feature description algorithm to perform arrhythmia detection and characterization.

[0069] As shown in Figure 5, the IMD10 senses electrocardiogram data from patient 4 (502). The electrocardiogram data may be, for example, a transient ECG of patient 4 or a fully disclosed ECG of patient 4. Furthermore, the electrocardiogram data of patient 4 may be from a single-channel or multi-channel system. For simplicity, in the example in Figure 5, the electrocardiogram data of patient 4 is described as single-channel transient ECG data.

[0070] The machine learning system 150 of the computing system 24 applies a machine learning model to sensed electrocardiogram data to detect episodes of arrhythmia in patient 4 (506). In some examples, the machine learning model is trained on multiple ECG episodes annotated by a clinician or several different types of arrhythmia monitoring centers. In one example, the machine learning system 150 applies the machine learning model to one or several subsegments of a normalized input ECG signal to generate arrhythmia labels and the likelihood of arrhythmia occurrence. In some examples, the machine learning model may be accurate in mapping the input ECG to output arrhythmia labels, but may not be able to provide additional arrhythmia characteristics or identify specific cardiac features such as mean heart rate, maximum heart rate, or PR interval characteristics that can be used to determine that an episode of arrhythmia has occurred in patient 4. Furthermore, it may not be possible to obtain physician-provided notifiable or reportable criteria (e.g., four out of four heartbeats in patient 4 showed a heart rate of less than 30 beats / minute (BPM)) from the output state or intermediate state of the machine learning model that would allow the clinician to utilize the determination that an episode of arrhythmia has occurred in patient 4 for use when providing subsequent treatment to patient 4.

[0071] To address this, the computing system 24 further applies feature descriptions to the electrocardiogram data to detect one or more cardiac features (504). In some examples, the computing system 24 further applies feature descriptions to the electrocardiogram data to detect one or more episodes of arrhythmia. For example, the computing system 24 may apply QRS detection descriptions and noise flags (e.g., whether it is beat noise) to the electrocardiogram data to provide cardiac features of the arrhythmia and / or detected episodes of arrhythmia (e.g., mean heart rate during an episode of atrial fibrillation, duration of pause). Furthermore, the computing system 24 may apply feature descriptions to derive notification and reporting criteria for system 2. In the example of Figure 5, the computing system 24 performs feature descriptions of electrocardiogram data. However, in other examples of the art of this disclosure, other devices such as the IMD 10, external device 12, or another external medical device may perform feature descriptions of electrocardiogram data.

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

[0073] If at least one of the feature description operations of the machine learning system 150 or (504) detects an episode of cardiac arrhythmia (e.g., "yes" in block 508), the computing system may generate an arrhythmia report (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 feature description performed on electrocardiogram data shows that four of the four noiseless heartbeats are less than 30 BPM, the computing system 24 generates a report notifying the physician of the occurrence of an arrhythmia episode.

[0074] In one example, the report includes an indicator that an episode of arrhythmia occurred in the patient, and one or more cardiac features that match the episode of arrhythmia. In some examples, the report further includes a classification of the episode of arrhythmia as a specific type of arrhythmia. In some examples, the report includes a subsection of electrocardiogram data obtained from patient 4 that matches the episode of arrhythmia. For example, computing system 24 may identify a subsection of electrocardiogram data from patient 4, where the subsection includes electrocardiogram data for a first period before the episode of arrhythmia (e.g., typically less than 10 minutes before the onset of the episode of arrhythmia), a second period during the occurrence of the episode of arrhythmia, and a third period after the episode of arrhythmia (e.g., typically less than 10 minutes after the cessation of the episode of arrhythmia). As an example, the subsection of electrocardiogram data from patient 4 may be about 6 seconds long and include representative segments before, during, and after the episode of arrhythmia (if present in the electrocardiogram data or waveform being analyzed). In some cases, the duration of an episode may vary depending on the type of device and may also depend on the usage of the medical device, one or more settings of the medical device, or the specific type of arrhythmia perceived. For example, some types of arrhythmias terminate quickly (resulting in a shorter episode duration), while others persist, with the recorded duration of the episode depending on the length of the medical device's specified memory space. As an example, in the case of atrial fibrillation (AF), a subsection of patient 4's electrocardiogram data 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 that of the first, second, and third periods. Furthermore, the computing system 24 identifies one or more cardiac features that match the first, second, and third periods. The computing system 24 includes in its report the subsection of the electrocardiogram data and one or more cardiac features that match the first, second, and third periods.

[0075] In some examples, the computing system 24 receives one or more adjustments from the clinician to the operation of a feature-based depiction of electrocardiogram data based on the report. Subsequently, the computing device 24 may perform a feature-based depiction of patient 4's electrocardiogram data according to one or more adjustments.

[0076] Figure 6 is a chart showing an exemplary electrocardiogram 602 obtained from patient 4 in Figure 1. The electrocardiogram 602 may be detected, for example, by the sensing circuit 52 of the IMD 10. The machine learning system 150 in Figure 1 may apply a machine learning model to the electrocardiogram 602 and determine that the electrocardiogram 602 contains a pause 604. The computing system 24 in Figure 1 or the IMD 10 in Figure 1 (for example, as part of the IMD 10 that initially detects arrhythmias) may perform feature drawing on the electrocardiogram 602 and determine the length of the pause 604. With respect to the example in Figure 6, the computing system 24 or the IMD 10, through feature drawing of the electrocardiogram 602, determines 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. QRS flagging may be based on conventional QRS algorithms. The IMD 10 may use QRS markers and determine that the pause period is 3.061 seconds.

[0077] Figure 7 is a flowchart illustrating exemplary operation of the technology of this disclosure. For convenience, Figure 7 is illustrated with respect to Figure 1. The operation in Figure 7 is for detecting and classifying cardiac arrhythmias in patient 4. Specifically, the operation in Figure 7 illustrates an implementation in which a computing system 24 uses machine learning arrhythmia detection and feature rendering of a machine learning system 150 in parallel to perform cardiac arrhythmia detection, validation, and reporting.

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

[0079] The computing system 24 determines whether both the machine learning system 150 and the feature description operation (704) have detected an episode of cardiac arrhythmia (708). For example, the computing system 24 may determine the level of confidence that the arrhythmia detection by the machine learning system 150 is consistent with the arrhythmia detection by the feature description operation (704) (708). For example, if the computing system 24 determines that both the machine learning system 150 and the feature description operation (704) have detected an episode of cardiac arrhythmia (e.g., "yes" in block 708), the computing system 24 may generate an arrhythmia report (712) and output the report to a clinician or monitoring center (714). For example, the computing system 24 may place the detected arrhythmias along with the arrhythmia features in the report and output the report to the clinician. The operations in steps 712 and 714 may occur in substantially the same manner as steps 512 and 514 in Figure 5, respectively.

[0080] As another example, if the computing system 24 determines that the feature description operations of the machine learning system 150 and (704) are inconsistent regarding whether an episode of cardiac arrhythmia has been detected (e.g., "No" in block 708), the computing system 24 submits electrocardiogram data to the monitoring center for mediation (710). In other words, the computing system 24 presents electrocardiogram data for a human overview where there is an inconsistency between the two detection methods. Such a workflow may allow for a reduction in the burden of human review only for arrhythmias that the computing system 24 cannot evaluate with high confidence. For example, if an arrhythmia detected via feature description is similar to an arrhythmia independently detected by the machine learning model, the computing system 24 may determine that the arrhythmia detected via feature description is independently validated without requiring a human review by an expert. Thus, the technology of this disclosure may reduce the amount of review required by clinicians and / or experts, thereby reducing the administrative overhead and cost of cardiac monitoring for patient 4.

[0081] Figure 8 is a flowchart illustrating exemplary operation of the technology of this disclosure. For convenience, Figure 8 is illustrated with respect to Figure 1. The operation in Figure 8 is for detecting and classifying cardiac arrhythmias in patient 4. Specifically, the operation in Figure 8 shows an implementation in which a computing system 24 performs cardiac arrhythmia detection, validation, and reporting using feature depiction in series with machine learning arrhythmia detection of a machine learning system 150.

[0082] As shown in Figure 8, the IMD 10 senses electrocardiogram data from patient 4 (802). The operation of step 802 may occur in substantially the same manner as step 502 in Figure 5. The computing system 24 applies a feature description 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 feature description to detect arrhythmias such as bradycardia, tachycardia, pause, or atrial fibrillation based on the rate and variability features of the electrocardiogram data. In the example in Figure 8, the computing system 24 performs the feature description as a screening step before describing all arrhythmias (for example, the computing system 24 may use the feature description to consider only tachyarrhythmias with a heart rate of 120 BPM or higher, bradyarrhythmias with a heart rate of 40 BPM or lower, or arrhythmias with high RR variability). In other examples, such a feature description may be implemented in a low-power device such as the IMD 10, or in other types of devices such as an external device 12 or another external medical device.

[0083] When the machine learning system 150 of the computing system 24 detects that an episode of cardiac arrhythmia has occurred in patient 4 through feature description, it applies a machine learning model to the sensed electrocardiogram to verify that an episode of arrhythmia has occurred (806). In some examples, the machine learning system 150 applies the machine learning model to many different types of patient data, such as electrocardiogram data for patient 4, based on the triggering reason that caused the feature description to detect the arrhythmia, one or more types of arrhythmia feature descriptions detected, or device characteristics of the IMD 10 such as activity level, input impedance, and battery level.

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

[0085] For example, if computing system 24 determines that machine learning system 150 verifies the detection of an episode of cardiac arrhythmia by the feature description operation of 804 (e.g., "yes" in block 808), computing system 24 may generate an arrhythmia report (812) and output the report to a clinician or monitoring center (814). In another example, if computing system 24 determines that the feature description operations of machine learning system 150 and 804 are inconsistent regarding whether an episode of cardiac arrhythmia has been detected (e.g., "no" in block 808), computing system 24 submits electrocardiogram data to the monitoring center for mediation (810). The operations in steps 810, 812, and 814 may occur in substantially the same manner as steps 510, 512, and 514 in Figure 5, respectively.

[0086] Figure 9 is a flowchart illustrating exemplary operation of the technology of this disclosure. For convenience, Figure 9 is illustrated with reference to Figure 1. The operation in Figure 9 is for detecting and classifying cardiac arrhythmias in patient 4. Specifically, the operation in Figure 9 shows an implementation in which a computing system 24 preprocesses electrocardiogram data to generate an intermediate representation of the electrocardiogram data, and a machine learning system 150 applies the intermediate representation of the electrocardiogram data to perform cardiac arrhythmia detection, verification, and reporting.

[0087] In the example in Figure 9, IMD 10 senses electrocardiogram data for patient 4 (902). The operation in step 902 may occur in substantially the same manner as in step 502 in Figure 5. The computing system 24 performs preprocessing 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 a detected QRS includes data for 160 milliseconds before the detected QRS and data for 160 milliseconds after the detected QRS. In another example, the window around a 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 steady wavelet decomposition used for QRS detection, flagging, and drawing to the sensed electrocardiogram data. For example, the computing system 24 generates a wavelet decomposition of patient 4's electrocardiogram for the window around the detected QRS.

[0088] The computing system 24 applies feature depiction to an intermediate representation of electrocardiogram data to detect one or more cardiac features (906). For example, the computing system 24 applies feature depiction to an intermediate representation to detect and depict patient 4's QRS segments (e.g., PR intervals) from a window around the detected QRS, as well as noise flags. In the example of Figure 9, the computing system 24 performs feature depiction of electrocardiogram data. However, in other examples of the art of this disclosure, other devices such as the IMD 10, external device 12, or another external medical device may perform feature depiction of electrocardiogram data.

[0089] The machine learning system 150 of the computing system 24 applies a machine learning model to an intermediate representation of the perceived electrocardiogram to detect episodes of arrhythmia in patient 4 (908). For example, the machine learning model may receive multiple electrocardiogram segments as input, each segment including a window around the detected QRS, a QRS depiction 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 episodes of arrhythmia in patient 4.

[0090] In some examples, the machine learning model is tailored to capture a segment of interest for each arrhythmia. For instance, the machine learning model may process a perceived electrocardiogram and capture the onset, offset, maximum heart rate, and minimum heart rate from a segment containing the window around the detected QRS complex. In some examples, the computing system 24 characterizes or contextualizes the arrhythmia detection by the machine learning model using features derived from feature descriptions such as QRS detection, including heart rate values ​​in the electrocardiogram segment.

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

[0092] In contrast to the operation shown in Figure 5, the computing system 24 may use the same signal preprocessing for both the detection of cardiac arrhythmias and / or cardiac features and the detection of a machine learning model of cardiac arrhythmias in step 906. Furthermore, the computing system 24 may use QRS noise flags and feature descriptions as input to the machine learning model of the machine learning system 150. The input electrocardiogram complexes may be of the same duration (e.g., 320 milliseconds) or different durations (e.g., segments from the previous T offset to the current T offset).

[0093] Figure 10 is a flowchart illustrating exemplary operation by the technology of this disclosure. For convenience, Figure 10 is illustrated with respect to Figure 1. The operation in Figure 10 is for detecting and classifying cardiac arrhythmias in patient 4. Specifically, the operation in Figure 10 shows an implementation in which a computing system 24 constructs an arrhythmia dictionary for use in detecting, classifying, and reporting cardiac arrhythmias using machine learning arrhythmia detection and serial feature depiction of a machine learning system 150.

[0094] The operation in Figure 10 involves monitoring patient 4's electrocardiogram data, annotating detected arrhythmias, and reporting such arrhythmias to a monitoring center. In some examples, the operation in Figure 10 takes place in a centralized location, such as a monitoring center. In another example, the operation in Figure 10 may take place in a clinic for each patient. As shown in Figure 10, IMD 10 senses patient 4's electrocardiogram data (1002). The computing system 24 further applies feature descriptions to the electrocardiogram data to detect one or more cardiac features (1004). The operations in steps 1002 and 1004 may occur in substantially the same manner as steps 502 and 504 in Figure 5, respectively.

[0095] The computing system 24 further applies feature description to the electrocardiogram data to detect one or more episodes of arrhythmia (1006). In some examples, feature description triggers automatic triggering of the electrocardiogram. In the example in Figure 10, the computing system 24 performs feature description. However, in other examples, arrhythmia detection and automatic triggering of electrocardiogram episodes may occur in another device such as the IMD 10, external device 12, or another external medical device, or through post-processing in a system such as Holter.

[0096] When an arrhythmia episode is first triggered in a particular patient, the computing system 24 presents the episode for arrhythmia review so that it can be used as a reference episode in the patient-specific “episode dictionary.” For example, upon detection of an arrhythmia episode, the computing system 24 determines whether the arrhythmia episode is the first detected episode. If the arrhythmia episode is the first detected episode (e.g., “yes” in block 1008), the computing system 24 generates a report on the arrhythmia episode and submits the report to the monitoring center or clinician for evaluation (1010). For example, if the episode is the first AF trigger, the episode is presented for review by the monitoring center. As another example, if the episode is the first AF trigger occurring at night, the episode is presented for review by the monitoring center. In one example, the report includes an indicator that the arrhythmia episode occurred in the patient, and one or more cardiac features that match the arrhythmia episode. The computing system 24 receives instructions from the monitoring center to verify whether the cardiac features included in the report indicate an arrhythmia episode. In cases where cardiac features indicate an episode of arrhythmia, the computing system 24 further receives a classification of the type of arrhythmia indicated by the cardiac features included in the report. The computing system 24 may store the representation of the arrhythmia type classification along with the cardiac features in a database in order to build a "dictionary" of cardiac arrhythmias.

[0097] In some cases, the computing system 24 may detect multiple episodes of arrhythmias that have similar arrhythmia content, annotations, and / or cardiac features. For example, with respect to atrial fibrillation (AF) monitoring, most episode triggers have AF. Another example is that, for patient-specific reasons such as the location and direction of signal acquisition, the feature description may generate some false triggers for arrhythmias (e.g., PAC with low-amplitude P waves). For example, the computing system 24 may input 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 the machine learning system 150 compares the episode features to the episode features in the patient 4's "episode dictionary". If the machine learning model determines with high confidence that similar episodes exist in the dictionary, the episode is reported using the original monitoring center annotations as they are. If no similar episodes are identified, the computing system 24 may determine that the episodes have different characteristics and therefore present the episodes for review and reporting by the monitoring center. Therefore, the operation shown in Figure 10 can improve the efficiency of arrhythmia annotation by minimizing redundant annotations in arrhythmia episodes with similar characteristics, thereby reducing the number of arrhythmia episodes that require review by the monitoring center.

[0098] The technology of this disclosure may offer a further advantage in 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 only need to be tuned to accurately identify new episodes as similar or dissimilar to previous episodes. For example, if there is similarity between two episodes of arrhythmia, the computing system 24 may apply previous patient-specific findings to the new episode as well. If there is dissimilarity, the computing system 24 may ask a human expert to determine whether the episode is an episode of arrhythmia and / or the type of arrhythmia presented by the episode. Thus, the machine learning model does not need to identify specific arrhythmias with a high level of confidence. The machine learning model only needs to be accurate in identifying the differences between two episodes of arrhythmia in order to accurately present episodes with different cardiac features (e.g., novel or unclassified rhythmic content) for human review. Thus, the technology of this disclosure may enable the computing system 24 to detect episodes of arrhythmia that have not been specifically trained for detection by the machine learning model 150. Furthermore, the technology of this disclosure may reduce the complexity of the machine learning model while maintaining high accuracy in arrhythmia detection and classification.

[0099] For example, with respect to the operation in Figure 10, if an episode of arrhythmia is not the first episode detected (e.g., "No" in block 1008), the machine learning system 150 applies a machine learning model to the detected cardiac feature to compare it with other cardiac features from previous episodes of the arrhythmia (1012). For example, the machine learning system 150 may apply a machine learning model to the detected cardiac feature to determine whether the cardiac feature matches other cardiac features from previous episodes of the arrhythmia and whether it matches the confidence level or certainty estimate 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 request (e.g., when a change in the patient's medication occurs). This allows some episodes of arrhythmia to be intermittently confirmed by the monitoring center or clinician, ensuring that new or changing arrhythmias are not missed.

[0100] If the machine learning model determines that it does not have a high level of confidence or certainty in the comparison (for example, "No" in block 1014), the computing system 24 generates a report of the arrhythmia episode and submits the report to the monitoring center or clinician for evaluation (1010). The computing system 24 receives an index that verifies that the cardiac features included in the report indicate an arrhythmia episode and a classification of the arrhythmia type, stores the index for the arrhythmia type classification along with the cardiac features in the database, and updates the cardiac arrhythmia dictionary with the detected cardiac features and the arrhythmia classification indicated by the detected cardiac features.

[0101] Depending on whether the machine learning model determines that it has a high level of confidence or certainty in the comparison (e.g., "yes" in block 1014), the computing system 24 may determine that the cardiac features indicate the type of previous episode of arrhythmia. The computing system 24 generates an arrhythmia report (1016) and outputs the report to the monitoring center (1018). The actions of steps 1016 and 1018 may occur in substantially the same manner as steps 512 and 514 in Figure 5, respectively.

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

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

[0104] Example 2. The method according to Example 1, wherein obtaining cardiac features present in the electrocardiogram data is performed by performing at least one of QRS detection, refractory processing, noise processing, or electrocardiogram data rendering to obtain cardiac features present in the electrocardiogram data.

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

[0106] Example 4. The method according to any one of Examples 1 to 3, wherein the cardiac features present in the electrocardiogram data are one or more of the following: the patient's mean 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, the amplitude of one or more features of the patient's electrocardiogram (ECG), or the interval between one or more features of the patient's ECG.

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

[0108] Example 6. The method according to any one of Examples 1 to 5, further comprising applying the machine learning model to received electrocardiogram data, applying the machine learning model to at least one of the following characteristics of received electrocardiogram data correlated with arrhythmia in a patient: the activity level of a medical device, the input impedance of a medical device, or the battery level of a medical device.

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

[0110] Example 8. The method according to any one of Examples 1 to 7, wherein the method generates a report in which the patient's electrocardiogram data includes the patient's electrocardiogram (ECG), an indicator that an episode of arrhythmia occurred in the patient, and one or more cardiac features consistent with the episode of arrhythmia, the method identifies a subsection of the patient's ECG, the subsection includes ECG data for a first period before the episode of arrhythmia, a second period during the episode of arrhythmia, and a third period after the episode of arrhythmia, and the length of the patient's ECG is longer than the first, second, and third periods, the method identifies one or more cardiac features consistent with the first, second, and third periods, and the report includes the subsection of the ECG and one or more cardiac features consistent with the first, second, and third periods.

[0111] Example 9. The method according to any one of Examples 1 to 8, further comprising processing received electrocardiogram data by a computing device to generate an intermediate representation of the received electrocardiogram data, applying a machine learning model trained using electrocardiogram data from multiple patients to the received electrocardiogram data to determine that an episode of arrhythmia occurred in the patient, or applying a machine learning model trained using intermediate representations of electrocardiogram data from multiple patients to the intermediate representations of the received electrocardiogram data and cardiac features present in the electrocardiogram data to determine, based on the machine learning model, that an episode of arrhythmia occurred in the patient.

[0112] Example 10. The method according to 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 according to Example 10, wherein the received electrocardiogram data is subjected to signal decomposition, and the received electrocardiogram data is subjected to wavelet decomposition.

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

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

[0116] Example 14. The method of Example 12, comprising applying a machine learning model to received electrocardiogram data to determine a second classification of arrhythmias in a patient, applying a machine learning model to received electrocardiogram data and cardiac features identified by a feature-based depiction of the received electrocardiogram data to determine a second classification of arrhythmias in a patient, and determining, based on the first and second classifications, that an episode of arrhythmia has occurred in the patient, comprising determining that the first classification indicates that an episode of arrhythmia has occurred in the patient, and determining that, in response to the determination that the first classification indicates that an episode of arrhythmia has occurred in the patient, the second classification verifies that an episode of arrhythmia has occurred in the patient, and determining that an episode of arrhythmia has occurred in the patient in response to the verification that an episode of adysregulation has occurred in the patient.

[0117] Example 15. The method according to any one of Examples 12 to 14, wherein the computing device performs a feature-based depiction of the received electrocardiogram data to determine the first classification of arrhythmias in the patient.

[0118] Example 16. The method according to any one of Examples 12 to 15, wherein the computing device obtains a first classification of arrhythmias in a patient determined by a feature-based description of received electrocardiogram data, and the computing device receives a first classification of arrhythmias in a patient determined by a feature-based description of received electrocardiogram data by a medical device from a medical device.

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

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

[0121] Example 19. The method according to any one of Examples 12 to 18, wherein the cardiac features present in the electrocardiogram data are one or more of the following: the patient's mean 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, the amplitude of one or more features of the patient's electrocardiogram (ECG), or the interval between one or more features of the patient's ECG.

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

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

[0124] Example 22. The method according to Examples 12 to 21, wherein the patient's electrocardiogram data includes the patient's electrocardiogram (ECG), and a report is generated that includes an indicator that an episode of arrhythmia occurred in the patient and one or more cardiac features that match the episode of adysregulation, wherein the subsection of the patient's ECG includes ECG data for a first period before the episode of arrhythmia, a second period during the episode of adysregulation, and a third period after the episode of arrhythmia, and the length of the patient's ECG period is longer than the first, second, and third periods, and one or more cardiac features that match the first, second, and third periods are identified, and the report includes the subsection of the ECG and one or more cardiac features that match the first, second, and third periods.

[0125] Example 23. A method comprising: receiving patient electrocardiogram data sensed by a medical device using a computing device including a processing circuit and a storage medium; obtaining a first classification of the patient's arrhythmia determined by a feature-based description of the received electrocardiogram data, wherein the feature-based description identifies and obtains a first cardiac feature present in the electrocardiogram data that matches the first classification of arrhythmia in the patient; determining that one or more episodes of the first classification of arrhythmia have previously occurred in the patient; and, in response to determining that one or more episodes of the first classification of arrhythmia have previously occurred in the patient, applying a machine learning model trained using electrocardiogram data from multiple patients to the received electrocardiogram data and the first cardiac feature in the electrocardiogram data, based on the machine learning model. A method comprising: determining that a first cardiac feature is similar to cardiac features that match one or more episodes of a first classification arrhythmia previously occurring in the patient; determining, in response to the determination that a first cardiac feature is similar to cardiac features that match one or more episodes of a first classification arrhythmia previously occurring in the patient, that an episode of a first classification arrhythmia has occurred in the patient; generating a report, in response to the determination that an episode of a first classification arrhythmia has occurred in the patient, that the computing device includes an indicator that an episode of a first classification arrhythmia has occurred in the patient and one or more cardiac features that match the episode of arrhythmia; and outputting, for display, the report, in which the computing device includes an indicator that an episode of arrhythmia has occurred in the patient and one or more cardiac features that match the episode of arrhythmia.

[0126] Example 24. The method according to Example 23, comprising: obtaining a second classification of arrhythmias in a patient determined by a feature-based description of received electrocardiogram data using a computing device, wherein the feature-based description identifies and obtains a second cardiac feature present in the electrocardiogram data that matches the second classification of arrhythmias in the patient; determining, by the computing device, that one or more episodes of arrhythmias 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 the determination that one or more episodes of arrhythmias of the second classification have not previously occurred in the patient; receiving from a user an index that the second cardiac feature reveals an episode of arrhythmia of the second classification in the patient; and storing, by the computing device, the index that reveals the second cardiac feature and the second cardiac feature that reveals an episode of arrhythmia of the second classification in the patient.

[0127] Example 25. Obtaining a second classification of arrhythmias in a patient determined by a feature-based description of received electrocardiogram data using a computing device, wherein the feature-based description identifies and obtains a third cardiac feature present in the electrocardiogram data that matches the second classification of arrhythmias in the patient, the computing device determines that one or more episodes of arrhythmias of the second classification have previously occurred in the patient, and, in response to the determination that one or more episodes of arrhythmias of the second classification have previously occurred in the patient, the computing device applies 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 a second cardiac feature that matches one or more episodes of arrhythmias of the second classification that have previously occurred in the patient. The method according to Example 24, comprising: determining that a third cardiac feature is similar to a second cardiac feature that matches one or more episodes of a second classification arrhythmia that have previously occurred in the patient, and determining that an episode of a second classification arrhythmia has occurred in the patient; generating a second report by the computing device that includes an indicator that an episode of a third classification arrhythmia has occurred in the patient and one or more third cardiac features that match an episode of a third classification arrhythmia, and outputting the report by the computing device for display, which includes an indicator that an episode of a third classification arrhythmia has occurred in the patient and one or more third cardiac features that match an episode of a third classification arrhythmia.

[0128] Example 26. The method according to any one of Examples 23 to 25, wherein applying a machine learning model to received electrocardiogram data and a first cardiac feature present in the electrocardiogram data, and determining based on the machine learning model that the first cardiac feature is similar to a cardiac feature that matches one or more episodes of a first classification arrhythmia previously experienced in the patient, is further comprising applying the machine learning model to the first cardiac feature and outputting a preliminary determination that the first cardiac feature is similar to a cardiac feature that matches one or more episodes of a first classification arrhythmia previously experienced in the patient, an estimate of certainty in the preliminary determination, and determining that the estimate of certainty in the preliminary determination is greater than a predetermined threshold, thereby determining that the first cardiac feature is similar to a cardiac feature that matches one or more episodes of a first classification arrhythmia previously experienced in the patient.

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

[0130] Example 28. The method according to any one of Examples 23 to 27, comprising applying a machine learning model to determine that a first cardiac feature is similar to a cardiac feature that matches one or more episodes of a first classification arrhythmia previously occurring in the patient, or applying a machine learning model to determine that a first cardiac feature represents at least one episode of bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block previously occurring in the patient.

[0131] Example 29. The method according to any one of Examples 23 to 28, wherein the first cardiac feature present in the electrocardiogram data is one or more of the following: the patient's mean 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, the amplitude of one or more features of the patient's electrocardiogram (ECG), or the interval between one or more features of the patient's ECG.

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

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

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

[0135] Example 33. The method according to any one of Examples 23 to 32, wherein the patient's electrocardiogram data includes the patient's electrocardiogram (ECG), and the method generates a report including an indicator that an episode of arrhythmia occurred in the patient and one or more first cardiac features that match the episode of adysregulation, wherein the subsection of the patient's ECG includes ECG data for a first period before the episode of arrhythmia, a second period during the episode of adysregulation, and a third period after the episode of arrhythmia, and the length of the patient's ECG period is longer than the first, second, and third periods, and the method identifies one or more first cardiac features that match the first, second, and third periods, and the report includes the subsection of the ECG and one or more first cardiac features that match the first, second, and third periods.

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

[0137] Example 35. The method according to 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 technology of the present disclosure includes a system that includes means for performing any of the methods described herein. In some examples, the technology of the present disclosure includes a computer-readable medium that includes instructions for causing a processing circuit to perform any of the methods described herein.

[0140] It should be understood that the various embodiments disclosed herein may be combined in combinations other than those specifically presented in the description and accompanying drawings. It should also be understood that any particular action or event among the processes or methods described herein may be performed in a different order depending on the embodiment, and may be added, merged, or omitted entirely (for example, not all described actions or events may be necessary to perform the Art). Furthermore, while certain embodiments of this disclosure are described for clarity as being performed by a single module, unit, or circuit, it should be understood that the Art of this disclosure may be performed, for example, by a combination of units, modules, or circuits related to a medical device.

[0141] In one or more examples, the techniques described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functionality may be stored as one or more instructions or codes on a computer-readable medium and executed by a hardware-based processing unit. The computer-readable medium may include non-temporary computer-readable media corresponding to tangible media, such as data storage media (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] Instructions may be executed by one or more processors, such as digital signal processors (DSPs), general-purpose microprocessors, application-specific integrated circuits (ASICs), field-programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuits. Therefore, the terms “processor” or “processing circuit” as used herein may refer to any of the aforementioned structures or any other physical structure suitable for implementing the described technology. Furthermore, the technology may be fully implemented in one or more circuits or logic elements.

Claims

1. A computing device, Storage media and A processing circuit which is operably coupled to the storage medium, Receiving patient electrocardiogram data detected by a medical device, Performing a feature-based description of the received electrocardiogram data to obtain cardiac features present in the electrocardiogram data, A machine learning model trained using electrocardiogram data from multiple patients is applied to the received electrocardiogram data, and based on the cardiac features obtained by the feature-based description and the machine learning model, it is determined that an episode of arrhythmia occurred in the patient. In response to determining that the aforementioned arrhythmia episode occurred in the patient, To generate a report that includes an indicator that the arrhythmia episode occurred in the patient, and one or more of the cardiac features that are consistent with the arrhythmia episode. A processing circuit is configured to output, for display, a report including the indicator that the arrhythmia episode occurred in the patient, and one or more of the cardiac features that are consistent with the arrhythmia episode. A computing device in which the machine learning model trained using electrocardiogram data of the plurality of patients includes a machine learning model trained using a plurality of electrocardiogram (ECG) waveforms, each ECG waveform being labeled with a plurality of episodes of a plurality of types of arrhythmias in one of the plurality of patients, each ECG waveform comprising a plurality of segments, each segment being labeled with a descriptor specifying the presence of an arrhythmia in a classification including bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block.

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

3. The computing device according to claims 1 to 2, wherein the processing circuit is configured to apply the machine learning model to determine that multiple episodes of bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block have occurred in the patient, in order to determine that the arrhythmia episode has occurred in the patient.

4. The computing device according to any one of claims 1 to 3, wherein the cardiac features present in the electrocardiogram data are one or more of the following: 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, the amplitude of one or more features of the patient's electrocardiogram (ECG), or the interval between one or more features of the patient's ECG.

5. To output the report, which includes the indicator that the arrhythmia episode occurred in the patient, and one or more of the cardiac features that match the arrhythmia episode, the processing circuit: Receiving adjustments from the user to the feature-based depiction of the electrocardiogram data, A computing device according to any one of claims 1 to 4, configured to perform a feature-based description of the electrocardiogram data in accordance with the adjustments described above, to obtain a second cardiac feature present in the electrocardiogram data.

6. The electrocardiogram data of the aforementioned patient includes the patient's electrocardiogram (ECG), To generate the report, which includes the indicator that the arrhythmia episode occurred in the patient, and one or more of the cardiac features that match the arrhythmia episode, the processing circuit: Identifying a subsection of the patient's ECG, wherein the subsection includes ECG data for a first period before the episode of arrhythmia, a second period during the episode of arrhythmia, and a third period after the episode of arrhythmia, and the length of the patient's ECG is longer than the first, second, and third periods. Identifying one or more of the cardiac features that coincide with the first, second, and third periods, A computing device according to any one of claims 1 to 5, configured to include in the report the subsection of the ECG and one or more of the cardiac features that coincide with the first, second, and third periods.

7. The processing circuit is further configured to preprocess the received electrocardiogram data to generate an intermediate representation of the received electrocardiogram data. A computing device according to any one of claims 1 to 6, wherein the processing circuit is configured to apply the machine learning model, which has been trained using electrocardiogram data of the plurality of patients, to the received electrocardiogram data in order to determine that an episode of arrhythmia has occurred in the patient, and to apply the machine learning model, which has been trained using electrocardiogram data of the plurality of patients, to the received electrocardiogram data in order to determine that an episode of arrhythmia has occurred in the patient, and based on the machine learning model, to the received electrocardiogram data.

8. In order to process the received electrocardiogram data and generate the intermediate representation of the received electrocardiogram data, the processing circuit: Applying filtering to the received electrocardiogram data, The computing device according to claim 7, configured to perform at least one of the following: performing signal decomposition on the received electrocardiogram data.

9. The computing device according to claim 8, wherein the processing circuit is configured to perform wavelet decomposition on the received electrocardiogram data in order to perform signal decomposition on the received electrocardiogram data.

10. A computing device, A means for receiving patient electrocardiogram data sensed by a medical device, Means for performing a feature-based description of the received electrocardiogram data to obtain cardiac features present in the electrocardiogram data, A means for applying a machine learning model trained using electrocardiogram data from multiple patients to the received electrocardiogram data, and for determining whether an episode of arrhythmia occurred in the patient based on the cardiac features obtained by the feature-based description and the machine learning model, In determining that the aforementioned arrhythmia episode occurred in the patient, To generate a report that includes an indicator that the arrhythmia episode occurred in the patient, and one or more of the cardiac features that are consistent with the arrhythmia episode. The system includes means for outputting, for display, a report that includes the indicator that the arrhythmia episode occurred in the patient, and one or more of the cardiac features that are consistent with the arrhythmia episode. A computing device in which the machine learning model trained using electrocardiogram data of the plurality of patients includes a machine learning model trained using a plurality of electrocardiogram (ECG) waveforms, each ECG waveform being labeled with a plurality of episodes of a plurality of types of arrhythmias in one of the plurality of patients, each ECG waveform comprising a plurality of segments, each segment being labeled with a descriptor specifying the presence of an arrhythmia in a classification including bradycardia, tachycardia, atrial fibrillation, ventricular fibrillation, or atrioventricular block.

11. A system comprising the computing device according to any one of claims 1 to 10.

Images