Medical system configured for discrimination of cardiac events using ensemble learning and morphology comparison

Abstract

A medical system comprises an implantable medical device (IMD) and a computing system. The IMD comprises sensors configured to detect physiological parameters corresponding to cardiac physiology of a patient and sensing circuitry configured to generate patient physiological data based on the physiological parameters. The IMD receives, from the computing system, a set of cardiac morphology templates based on historical patient physiological data and applies a first model to the generated patient physiological data to determine a cardiac event. Responsive to determining a cardiac event, the IMD generates a cardiac morphology template based on at least a portion of the generated patient physiological data, applies a second model to the set of cardiac morphology templates and the generated cardiac morphology template to determine a first instance of a cardiac morphology template match, and transmits, to the computing system, data indicative of the first instance of the cardiac morphology template match.

Description

MEDICAL SYSTEM CONFIGURED FOR DISCRIMINATION OF CARDIAC EVENTS USING ENSEMBLE LEARNING AND MORPHOLOGY COMPARISON

[0001] This application claims the benefit of U.S. Provisional Patent Application Serial No. 63 / 623,501, filed January 22, 2024, the entire content of which is incorporated herein by reference.TECHNICAL FIELD

[0002] The disclosure relates, inter alia, to implantable medical devices and, more particularly, it relates to systems, devices, and methods for detecting, evaluating, and predicting cardiac events.BACKGROUND

[0003] Some medical systems may include an implantable medical device (IMD) to collect various measurements used to assess a patient’s current and historical physiological state and / or predict impending events or conditions. The IMD may monitor various data (e.g., cardiac electrophysiological activities and signals) of a patient to detect or predict changes in patient health. The patient’s heart’s electrical activity can be detected on the patient’s skin or subcutaneously (via an electrocardiogram (ECG)) and inside the heart itself (via a cardiac electrogram (EGM)). In some examples, the IMD may monitor ECG data to detect one or more types of arrhythmia, such as bradycardia, tachycardia, fibrillation, or asystole (e.g., caused by sinus pause or AV block). However, for certain patients, some types of arrhythmia may be less common and more severe, such as non-sustained ventricular tachycardia (NSVT) in hypertrophic cardiomyopathy (HCM) patients. As such, accurately detecting and distinguishing various cardiac events is important for patient treatment recommendations.

[0004] Over time, IMDs have increased in complexity while decreased in size. One hurdle to achieving such small and highly functional devices, however, is computing power. In particular, many IMDs may not have the computing power to execute computationally expensive algorithms, some of which have shown to significantly improve the identification and / or prediction of impending events or conditions.SUMMARY

[0005] Medical systems and techniques as described herein detect and predict changes in a patient’s health (e.g., heart health) based upon data describing that patient’s physiology, includingthe patient’s cardiac physiology. A medical system may include any one or more of a variety of medical devices (e.g., implantable devices, wearable devices, etc.) that detect (e.g., evidence of) a cardiac event (which may also be referred to herein as an “episode”) of a certain type (e.g., an arrhythmia) based on a match of patient physiological data (e.g., patient physiological data comprising electrocardiogram data for the patient) with one or more stored heartbeat morphology templates (e.g., cardiac morphology templates based on historical patient physiological data).

[0006] An example medical device may be implanted into the patient’s body in order to monitor the electrical activity of the patient’s heart during which the implanted device receives, as input, cardiac ECG (or EGM) signals representing such electrical activity. Hereinafter the terms ECG and EGM may refer to either or both of an ECG or EGM. The medical device may pre-loaded with multiple templates representative of morphologies of various cardiac events (e.g., premature ventricular contraction (PVC), NSVT, etc.). A “template”, as described herein, may be considered a “snapshot” or representation (e.g., based on application of a transform) of at least a portion of patient physiological data (e.g., a QRS complex or other portion of ECG signal data) corresponding to one or more cardiac events. The medical device may also receive morphology templates from a computing system (e.g., cloud computing system) that are based on the condition of a patient and / or previous data transmitted to the cloud computing device from the medical device. The morphology templates may include specific patterns and durations of cardiac events (e.g., specific patterns and durations of PVC that trigger NSVT).

[0007] The implantable medical device may be configured to apply a first model to patient physiological data (e.g., cardiac ECG signals) to determine at least one cardiac event. In some examples, the first model includes at least one of a cardiac rate or cardiac interval threshold. In some examples, the first model includes a detection algorithm. Responsive to determining at least one cardiac event, the implantable medical device may be configured to apply a second model (e.g., a template matching algorithm) to the set of cardiac morphology templates and one or more segments of the patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match. As described herein, the implantable medical device may be configured to transmit, to a computing system (e.g., a cloud computing system) in communication with the implantable medical device, data indicative of a first confirmed instance of a particular morphology template match (e.g., a segment of ECG signal corresponding to a particular morphology template match) for further analysis (e.g., determining presence of the cardiac event, severity and / or risk level of the cardiac event, and / or treatment recommendations for the cardiac event).

[0008] The present disclosure includes a number of benefits and advantages that example medical systems and techniques described herein realize by implementing cloud-based machine learning algorithms. Medical systems including medical devices may benefit from increased accuracy with respect to detecting cardiac events that occurred in a patient. As an advantage, such medical systems may improve upon the ability of clinicians to diagnose and treat a patient, and thus improve the patient’s general health, as well as specifically help prevent cardiac events. When the machine learning algorithms described herein are implemented on a separate computing system, medical devices may also realize such improvements in event detection without increased resource utilization, especially in consumption of compute resources (e.g., processing cycles) and / or storage resources (e.g., storage space), overall cost, and power consumption.

[0009] In one example, this disclosure describes a medical system comprising: an implantable medical device comprising: a communication system configured for wireless communication; one or more sensors configured to detect at least one physiological parameter corresponding to cardiac physiology of a patient; sensing circuitry configured to generate patient physiological data based on the at least one physiological parameter; and processing circuitry operatively coupled to the communication system, the one or more sensors, and the sensing circuitry, wherein the processing circuitry is configured to: receive, from a computing system in communication with the implantable medical device via the communication system, a set of cardiac morphology templates based on historical patient physiological data; apply a first model to the generated patient physiological data to determine at least one cardiac event; responsive to determining the at least one cardiac event, apply a second model to the set of cardiac morphology templates and one or more segments of the generated patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match; and transmit, to the computing system, data indicative of the first instance of the cardiac morphology template match.

[0010] In another example, this disclosure describes a method comprising: receiving, by an implantable medical device and from a computing system, a set of cardiac morphology templates based on historical patient physiological data; applying, by the implantable medical device, a first model to generated patient physiological data to determine at least one cardiac event; responsive to determining the at least one cardiac event, applying, by the implantable medical device, a second model to the set of cardiac morphology templates and one or more segments of the generated patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match; and transmitting, by the implantable medicaldevice and to the computing system, data indicative of the first instance of the cardiac morphology template match.

[0011] In another example, this disclosure describes a method comprising: receiving, by a computing system and from an implantable medical device, data indicative of a first instance of a cardiac morphology template match, wherein the data indicative of the first instance of the cardiac morphology template match includes patient physiological data corresponding to at least one cardiac event; applying, by the computing system, a model to the data indicative of the first instance of the cardiac morphology template match; and determining, by the computing system and based on at least one output of the model, whether the at least one cardiac event is a true cardiac event.

[0012] In another example, this disclosure describes a non-transitory computer-readable storage medium comprising instructions that, when executed, cause processing circuitry of an implantable medical device to: receive, from a computing system, a set of cardiac morphology templates based on historical patient physiological data; apply a first model to generated patient physiological data to determine at least one cardiac event; responsive to determining the at least one cardiac event, apply a second model to the set of cardiac morphology templates and one or more segments of the generated patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match; and transmit, to the computing system, data indicative of the first instance of the cardiac morphology template match.

[0013] In another example, this disclosure describes a non-transitory computer-readable storage medium comprising instructions that, when executed, cause processing circuitry of a computing system to: receive, from an implantable medical device, data indicative of a first instance of a cardiac morphology template match, wherein the data indicative of the first instance of the cardiac morphology template match includes patient physiological data corresponding to at least one cardiac event; apply a model to the data indicative of the first instance of the cardiac morphology template match; and determine, based on at least one output of the model, whether the at least one cardiac event is a true cardiac event.

[0014] The disclosure also provides means for performing any of the techniques described herein.

[0015] The summary is intended to provide an overview of the subject matter described in this disclosure. It is not intended to provide an exclusive or exhaustive explanation of the systems, device, and methods described in detail within the accompanying drawings and description below. Further details of one or more examples of this disclosure are set forth in the accompanyingdrawings and in the description below. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims.BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG. 1 illustrates example environment of an example medical system in conjunction with a patient, in accordance with one or more examples of the present disclosure.

[0017] FIG. 2 is a functional block diagram illustrating an example configuration of an implantable medical device, in accordance with one or more examples of the present disclosure.

[0018] FIG. 3 is a conceptual side-view diagram illustrating an example configuration of the implantable medical device of FIGS. 1 and 2, in accordance with one or more examples of the present disclosure.

[0019] FIG. 4Ais a perspective drawing illustrating an insertable cardiac monitor, in accordance with one or more examples of the present disclosure.

[0020] FIG. 4B is a perspective drawing illustrating another insertable cardiac monitor, in accordance with one or more examples of the present disclosure.

[0021] FIG. 5 is a functional block diagram illustrating an example configuration of the computing system of FIG. 1, in accordance with one or more examples of the present disclosure.

[0022] FIG. 6 is a conceptual diagram illustrating an example machine learning model configured to determine an extent to which data of a patient indicates an acute health event.

[0023] FIG. 7 is a conceptual diagram illustrating an example training process for a machine learning model, in accordance with examples of the current disclosure.

[0024] FIG. 8 is a block diagram illustrating an example system that includes an access point, a network, external computing devices, such as a server, and one or more other computing devices, which may be coupled to the medical device and external device of FIGS. 1-4, in accordance with one or more examples of the present disclosure.

[0025] FIG. 9 is a flow diagram illustrating an example operation for determining a cardiac event, in accordance with one or more examples of the present disclosure.

[0026] FIG. 10 is a flow diagram illustrating another example operation for determining a cardiac event, in accordance with one or more examples of the present disclosure.

[0027] FIG. 11 is a flow diagram illustrating another example operation for determining a cardiac event, in accordance with one or more examples of the present disclosure.

[0028] Like reference characters denote like elements throughout the description and figures.DETAILED DESCRIPTION

[0029] Implantable medical devices (IMDs) can sense and monitor signals and use those signals to determine various conditions of a patient and / or provide therapy to the patient. Example IMDs include monitors, such as the Reveal LINQ™ or LINQ II™ Insertable Cardiac Monitor, available from Medtronic, Inc., of Minneapolis, Minnesota. Such IMDs may facilitate relatively longer-term monitoring of patients during normal daily activities, and may periodically transmit collected data to a network service, such as the Medtronic CareLink® Network, developed by Medtronic, or some other network linking patient 4 to a clinician. In other examples, IMDs may also deliver therapy based on the sensed and collected data. In some examples the delivered therapy may include electrical stimulation therapy or drug or other fluid delivery.

[0030] In general, medical systems according to the present disclosure implement techniques for detecting and predicting changes in a patient’s health (e.g., heart health) based upon data describing the patient’s physiology, including the patient’s cardiac physiology. Cardiac events (e.g., arrhythmias) and other maladies negatively affect heart health in general and may be potentially fatal to the patient. Furthermore, while certain cardiac events may be common within the general population, they may be less common and more severe in certain patients. For instance, while PVCs are relatively common within the general population, NSVT (or three or more consecutive ventricular depolarizations that occur at a rate greater than 100 beats-per-minute) are less common and more severe in HCM patients. As such, accurately detecting and distinguishing various cardiac events is important for the identification, prediction, and treatment of impending events or conditions.

[0031] An example medical device may receive, e.g., from a computing system in communication with the medical device via a communication system, a set of cardiac morphology templates based on historical patient physiological data. The medical device may detect at least one physiological parameter corresponding to cardiac physiology of a patient and generate patient physiological data based on the at least one physiological parameter to monitor, evaluate, and / or determine at least one cardiac event. The example medical device may apply a first model to the patient physiological data to generate a prediction value indicating whether or not the patient physiological data depicts a particular cardiac event; for example, the first model may include a cardiac rate or cardiac interval threshold, or a detection algorithm that analyzes deviations from R- R interval and cardioacceleration baselines or patterns to determine the presence of a cardiac event. The first model may determine an extent to which patient parameter data is indicative of an acute health event, such as a ventricular tachyarrhythmia or SC A. Responsive to determining thepresence of at least one cardiac event, the example medical device may apply a second model to one or more segments of the patient physiological data corresponding to the at least one cardiac event and the stored set of cardiac morphology templates to determine similarity, and transmit, to the computing system, data indicative of a first confirmed instance of a particular morphology template match for further analysis (e.g., the computing system may apply third model, such as a segmentation neural network that further classifies the patient physiological data).

[0032] Logical circuitry or circuitries (and, possibly, other hardware) enable the example medical systems and techniques described herein including the above medical device that collects and evaluates the at least one physiological parameter corresponding to cardiac physiology of a patient. Examples of the medical device include wearable devices, implantable devices, and / or any other medical device in communication with one or more sensors (e.g., an electrode) continuously measuring a patient’s heart’s electrical activities and recording such measurements as the patient’s cardiac physiology data.

[0033] An example computing system may be configured with programmable logic configured to perform an algorithm (e.g., a machine learning algorithm) and may include processing circuitry that, by executing the programmable logic, performs operations in accordance with such an algorithm. Machine learning generally refers to algorithms (e.g., learning) and structures (e.g., machine learning models) enabling functionality (e.g., classification) for making predictions or automating decisions without user direction or explicit rules, which may or may not be hard-coded or otherwise programmed to do so. As instructed by an example machine learning algorithm, the computing system may build a model based on foundational concepts and some training data (e.g., samples of sensor data including patient cardiac EGM data or ECG data, sample cardiac morphology templates, etc.), test that model using additional training data and validation data and various metrics, and if the model satisfies the metrics, deploy the model to a medical system that is configured to monitor the patient’s cardiac physiology for cardiac events. The medical device may upload samples of cardiac EGM data or ECG data to serve as additional training data for testing models.

[0034] The medical systems and devices described herein include techniques to detect cardiac events and / or predict cardiac event risk in accordance with one or more algorithms, including machine learning algorithms. By implementing cloud-based machine learning algorithms, not only can on-device detection logic be simplified, but operational resource requirements and overall resource utilization may be lowered. Conserving resource capacities may also result in time and capital savings. Furthermore, a lower resource footprint may enable smaller and less complexembodiments of these techniques (e.g., implantable or wearable embodiments). A medical device equipped with substantial resource capacities is no longer needed; instead, a medical device with fewer capacities of processing, network, and storage resources can be configured to detect cardiac events, in accordance with any technique described herein.

[0035] In view of the above, the present disclosure describes a technological improvement in a technical field or a technical solution that is integrated into a practical application. Medical systems and devices benefit from reductions in false determinations and an improved accuracy of cardiac event detections. Provided that cloud computing systems can store up to exabytes of data, any highly complex or computationally expensive algorithms can be employed to more accurately detect and / or predict cardiac events.

[0036] An example computing system (e.g., a cloud-based system) may support the example medical devices with data and / or application services. The example computing system may receive data from the example medical devices and perform additional analysis to determine cardiac events. Additionally, the computing system may configure or provide updates to the example medical devices to improve the accuracy and / or detection parameters of the example medical devices. The computing system may also provide information or reports regarding the patient’s cardiac activities and health to external devices operated by the patient or a clinician.

[0037] In this manner, the techniques of this disclosure may advantageously enable improved accuracy in the detection of changes in patient health and, consequently, better evaluation of the condition of the patient.

[0038] FIG. 1 illustrates the environment of an example medical system 2 in conjunction with a patient 4, in accordance with one or more techniques of this disclosure. Patient 4 ordinarily, but not necessarily, will be a human. For example, patient 4 may be an animal needing ongoing monitoring for cardiac conditions. System 2 includes IMD 10. IMD 10 may include one or more electrodes on a housing of IMD 10, or may be coupled to one or more leads that carry one or more electrodes (not shown in FIG. 1). System 2 may also include external device 12 and / or computing system 6.

[0039] The example techniques may be used with an IMD 10, which may be configured to be in wireless communication with at least one of external device 12, an external device in communication with computing system 6, and other devices not pictured in FIG. 1. As shown in the example of FIG. 1, computing system 6 and external device 12 may be in communication via network 92. Network 92, which may be described in more detail with respect to FIG. 8, may comprise a local area network, wide area network, or global network, such as the Internet. In someexamples, IMD 10 may transmit data to external device 12 via a wireless connection, and external device 12 may transmit the data to computing system 6 via network 92.

[0040] In some examples, IMD 10 may be implanted within patient 4. For example, IMD 10 may be implanted outside of a thoracic cavity of patient 4 (e.g., pectoral location illustrated in FIG. 1). In some examples, IMD 10 may be positioned near the sternum near or just below the level of the heart of patient 4, e.g., at least partially within the cardiac silhouette. In some examples, IMD 10 may be implanted in patients who have experienced prior sudden cardiac arrest (SC A), sustained ventricular tachycardia (SVT), a ventricular fibrillation (VF) event, family history of sudden cardiac death (SCD), massive left ventricular hypertrophy, unexplained syncope, apical aneurysm, or ejection fraction < 50%. IMD 10 includes a plurality of electrodes (not shown in FIG. 1), and is configured to sense electrical activity of the heart via the plurality of electrodes. The sensed electrical activity may be herein referred to as an ECG or a cardiac EGM. IMD 10 may record and / or evaluate cardiac physiological data for patient 4 in the form of a cardiac EGM or ECG. IMD 10 may be a type of cardiac monitoring device. In some examples, IMD 10 takes the form of the Reveal LINQ™ or LINQ II™ Insertable Cardiac Monitor (ICM), or another ICM similar to, e.g., a version or modification of, the LINQ™ ICM, available from Medtronic, Inc. Such IMDs may facilitate relatively longer-term monitoring of patients during normal daily activities, and may periodically transmit collected data to a network service, such as the Medtronic CareLink® Network. Example medical system 2 may be adapted for other devices that are capable of monitoring patient 4’s cardiac activity, including smart / personal EKG devices and ECG devices, surface ECG devices, and other heart monitoring devices.

[0041] In some examples, IMD 10 may sense ECG signals via the plurality of electrodes and / or operate as a therapy delivery device. For example, IMD 10 may operate as a therapy delivery device to deliver electrical signals to the heart of patient 4, such as an implantable pacemaker, a cardioverter, and / or defibrillator, a drug delivery device that delivers therapeutic substances to patient 4 via one or more catheters, or as a combination therapy device that delivers both electrical signals and therapeutic substances.

[0042] In some examples, system 2 may include any suitable number of leads coupled to IMD 10, and each of the leads may extend to any location within or proximate to a heart or in the chest of patient 4. For example, other examples therapy systems may include three transvenous leads and an additional lead located within or proximate to a left atrium of a heart. As other examples, a therapy system may include a single lead that extends from IMD 10 into a right atrium or right ventricle, or two leads that extend into a respective one of a right ventricle and a right atrium.

[0043] As described herein, IMD 10 may be configured to continuously (e.g., on a periodic or triggered basis without human intervention) sense cardiac signals (EGM or ECG) while subcutaneously implanted in patient 4 over months or years and perform numerous operations per second on patient data. As such, the systems described herein may be enabled to monitor a patient's health throughout a portion of their lifetime, which may improve upon the ability of clinicians to accurately diagnose and treat the patient, and furthermore implement proactive measures to prevent cardiac events that the patient may be predicted to be at risk for. As one example, continuous monitoring of ECG signals to detect events such as NSVTs may enable a risk score that is a more complete representation of a patient’s risk, e.g., of more lethal arrhythmias.

[0044] Computing system 6 herein refers to a remote computing system that provides, over a network (e.g., network 92), IMD 10, external device 12, and / or other devices not shown in FIG. 1 with data and / or application services. In some examples of computing system 6, a cloud service provider maintains a set of computing devices operating as data / appli cation servers that are configured to support IMD 10 functionality, for instance, with updates to one or more algorithms in use by IMD 10 and / or data stored by IMD 10.

[0045] As described herein, IMD 10 may be configured to receive, from computing system 6, patient history, data corresponding to previously detected cardiac events (e.g., NSVT episodes, PVC episodes, etc.), and / or patient daily activity. While the techniques described herein may be directed towards discrimination of PVC from NSVT, IMD 10 and / or computing system 6 may be configured to detect and analyze other cardiac events, such as, but not limited to, myocardial infarction (MI), atrial fibrillation (AF), atrial flutter, ventricular tachycardia (VT), supraventricular tachycardia (SVT), ventricular fibrillation (VF), atrioventricular nodal reentrant tachycardia (AVNRT), atrioventricular reentrant tachycardia (AVRT), bradycardia, pause, premature atrial contractions (PAC), long QT syndrome, Wolff-Parkinson-White Syndrome (WPW), Sick Sinus Syndrome, and the like. IMD 10 may receive data pertaining to a patient and / or one or more cardiac events from computing system 6 continuously or periodically.

[0046] IMD 10 may detect a potential cardiac event based on a comparison of continuous ECG data to an established threshold (e.g., a cardiac rate or cardiac interval threshold), implementing detection algorithms with parameters based on the patient data, and / or by implementing cardiac morphology template matching. In some examples, IMD 10 may implement one or more rulebased systems and / or one or more machine learning models, e.g., to detect a cardiac event, determine an extent to which patient parameter data is indicative of an acute health event and / or to confirm a detected cardiac event. IMD 10 may then confirm a potential event using, for example,AF burden, NSVT burden, patient activity, and / or changes in ECG. Other variables of interest for detecting and / or confirming an event may include, but are not limited to, the number of nonsustained arrhythmias occurring during a specified period of time, the duration of the non-sustained arrhythmias, the atrial and / or ventricular intervals during the non-sustained arrhythmias, characteristics of the ECG morphology during non-sustained arrhythmias, characteristics of the ECG morphology triggering the non-sustained arrhythmias, changes in the frequency or duration of non-sustained arrhythmias, frequency of non-sustained arrhythmia episodes determined as the number of episodes occurring within a predetermined amount of time, average duration of a given number of non-sustained episodes or the average of a number (such as 20, 30, or some other number of all non-sustained episodes occurring within a predetermined amount of time), an NSVT index defined as the product of the number of NSVT episodes / day times the mean number of beats per episode (i.e., total NSVT beats per day (total beats / day)), the average of the stored episode durations or the total number of non-sustained arrhythmia intervals occurring during a timer interval calculated as the sum of all the stored episode durations, and the product of the number of episodes detected and the average of all episode durations. In some examples, IMD 10 may determine one or more operating parameters, thresholds for detection, or other metrics based on the data received from computing system 6, such as one or more cardiac morphology templates.

[0047] IMD 10 may upload or transmit detected event data to computing system 6. Computing system 6 may further analyze the confirmed event data and determine if IMD 10 detected a true event. Computing system 6 may employ, as an example, one or more machine learning models (e.g., neural network) to further determine or classify cardiac events detected by IMD 10. For example, the one or more machine learning models may determine an extent to which patient parameter data is indicative of an acute health event, such as a ventricular tachyarrhythmia or SCA. If computing system 6 determines that IMD 10 has detected a false event, computing system 6 may send data back to IMD 10 for further adjustment of the detection algorithm parameters or thresholds for detection. If computing system 6 determines that IMD 10 has detected a true event, computing system 6 may send data back to IMD 10 to keep the detection algorithm parameters or thresholds for detection and transmit counters corresponding to the detected events instead of the entire ECG repeatedly; this can help save battery life by limiting repeated transmissions of similar ECGs, while getting diagnostically relevant parameters in the form of event counters.

[0048] As such, the techniques described herein may enable efficient hardware resource utilization by IMD 10. In one example, IMD 10 may make efficient use of battery power and memory space by only recording patient physiological data corresponding to determined cardiacevents. Additionally, IMD 10 may only record and / or transmit data indicative of a first instance of a cardiac event. Furthermore, by employing cloud computing systems such as, in some examples, computing system 6, system 2 may employ more advanced machine learning techniques or algorithms that are typically computationally expensive to execute. Additionally, the use of more powerful algorithms may aid in improving the accuracy of cardiac event detection (e.g., reduce false positives) and risk prediction.

[0049] As described above, system 2 may also provide a feedback loop that aids in improving the accuracy of IMD 10. Computing system 6 may confirm or reject initial detections of cardiac events detected by IMD 10, and IMD 10 may use that confirmation or rejection to determine how much and / or which of patient 4’s cardiac activity to record.

[0050] As shown in the example of FIG. 1, external device 12 may be configured to communicate wirelessly with IMD 10 and communicate over network 92 with computing system 6. System 2 may be configured to facilitate data transfers between external device 12, IMD 10, computing system 6, and / or any other networked computing devices. In some examples, external device 12 is a server of computing system 6. In some examples, external device 12 may be used to retrieve data from IMD 10 and / or computing system 6. The retrieved data may include patient physiological data (e.g., ECG signal data) recorded by IMD 10, e.g., due to IMD 10 and / or computing system 6 determining that a cardiac event occurred, or in response to a request to record the data from patient 4 or another user.

[0051] In other examples, the user may also use external device 12 to retrieve information from IMD 10 regarding other sensed physiological parameters of patient 4, such as activity, temperature, tissue impedance, intrathoracic impedance, or posture. Additionally, one or more remote computing devices may interact with IMD 10 in a manner similar to external device 12, e.g., to program IMD 10 and / or retrieve data from IMD 10, via network 92.

[0052] Processing circuitry of medical system 2, e.g., of IMD 10, external device 12, and / or computing system 6, may be configured to perform techniques for detecting changes in patient health, including the example techniques of this disclosure for detecting occurrences of cardiac events in patient 4. Processing circuitry of IMD 10 may be communicably coupled to one or more sensors, each being configured to sense physiological parameters in some form (e.g., cardiac physiology parameter(s) corresponding to electrical activity of patient 4’s heart), and sensing circuitry configured to generate patient physiological data based on the sensed physiological parameters (e.g., data corresponding to the sensed electrical activity). In some examples, processing circuitry of IMD 10 and / or processing circuitry of computing system 6 may executelogic to detect the above-mentioned cardiac events (e.g., arrhythmias), in which the logic is configured to implement one or more algorithms (e.g., a detection algorithm and a template matching algorithm) that produce a prediction value (e.g., probability, similarity score, etc.) indicative of a likelihood that a cardiac event has occurred. In some examples, IMD 10 may be configured with high sensitivity and low specificity, and thus may be configured to detect many events, including false events. IMD 10 may be configured with a threshold for detection that is set very low, and thus may implement algorithms that are sensitive to a wide range of morphologies and possible cardiac events. In some examples, processing circuitry of IMD 10 may apply false detection criteria configured to detect one or more indicators of noise and / or amplitude variations in the sensed physiological parameters. In some examples, IMD 10 may perform some or all of the techniques described herein in the same manner described herein with respect to external device 12 and / or computing system 6.

[0053] In some examples, processing circuitry in a wearable device may execute the same or similar logic as the logic executed by processing circuitry of IMD 10 and / or other processing circuitry as described herein. In this manner, a wearable device or other device may perform some or all of the techniques described herein in the same manner described herein with respect to IMD 10. In some examples, the wearable device operates with IMD 10 and / or external device 12 as potential providers of computing / storage resources and sensors for monitoring patient 4’s cardiac physiology, including electrical activity and / or other parameters. For example, the wearable device may communicate patient physiological data to external device 12 and / or computing system 6 for storage in non-volatile memory and / or for computing prediction values, in accordance with the one or more algorithms described herein. Additionally, external device 12 and / or computing system 6 may perform some or all of the techniques described herein in the same manner described herein with respect to IMD 10. Additionally, in some examples, IMD 10 may perform some or all of the techniques described herein in the same manner described herein with respect to external device 12 and / or computing system 6.

[0054] FIG. 2 is a functional block diagram illustrating an example configuration of IMD 10 of FIG. 1 in accordance with one or more techniques described herein. In the illustrated example, IMD 10 includes electrodes 16A - 16N (collectively “electrodes 16”), processing circuitry 50, sensing circuitry 52, communication system 56, switching circuitry 58, storage device 60, sensors 62, and power source 70. Storage device 60 further includes templates data storage 63, patient data storage 64, detection logic 66, template matching 67, and detected event data storage 69. Themodules and components described with respect to FIG. 2 may be referred herein collectively as “components 50-70”.

[0055] Power source 70 may be any type of device that is configured to hold a charge to operate the circuitry of IMD 10. Power source 70 may be provided as a rechargeable or non- rechargeable battery. In other examples, power source 70 may incorporate an energy scavenging system that stores electrical energy from movement of IMD 10 within patient 4.

[0056] Processing circuitry 50 may include fixed function circuitry and / or programmable processing circuitry. Processing circuitry 50 may include any one or more of a microprocessor, a controller, a graphics processing unit (GPU), a tensor processing unit (TPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or equivalent discrete or analog logic circuitry. In some examples, processing circuitry 50 may include multiple components, such as any combination of one or more microprocessors, one or more controllers, one or more GPUs, one or more TPUs, one or more DSPs, one or more ASICs, or one or more FPGAs, as well as other discrete or integrated logic circuitry. The functions attributed to processing circuitry 50 herein may be embodied as software, firmware, hardware, or any combination thereof. In some examples, storage device 60 includes computer-readable instructions that, when executed by processing circuitry 50, cause IMD 10 and processing circuitry 50 to perform various functions attributed herein to IMD 10 and processing circuitry 50. Storage device 60 may include any volatile, non-volatile, magnetic, optical, or electrical media, such as a random-access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically-erasable programmable ROM (EEPROM), flash memory, or any other digital media. Storage device 60 may store, as examples, programmed values for one or more operational parameters of IMD 10 and / or data collected by IMD 10 for transmission to another device using communication system 56. Data stored by storage device 60 and transmitted by communication system 56 to one or more other devices may include patient data indicative of a cardiac event.

[0057] Processing circuitry 50 may employ communication system 56 to communicate with a remote computing system such as computing system 6 of which external device 12 may or may not be a server. Communication system 56 may include any suitable hardware, firmware, software or any combination thereof for communicating with another device, such as external device 12 and / or computing system 6, another networked computing device, or another IMD or sensor. Under the control of processing circuitry 50, communication system 56 may receive downlink telemetry from, as well as send uplink telemetry to, external device 12 and / or computing system 6 or another device with the aid of an internal or external antenna. In addition, processing circuitry 50 maycommunicate with a networked computing device via an external device (e.g., external device 12 and / or computing system 6) and a computer network, such as the Medtronic CareLink® Network. Communication system 56 may be configured to transmit and / or receive signals via inductive coupling, electromagnetic coupling, Near Field Communication (NFC), Radio Frequency (RF) communication, Bluetooth, WiFi, or other proprietary or non-proprietary wireless communication schemes.

[0058] As described herein, IMD 10 may receive, from a computing system (e.g., remote computing system 6) in communication with IMD 10 via communication system 56, a set of cardiac morphology templates based on historical patient physiological data. In some examples, IMD 10 may receive previous data transmitted to computing system 6 and / or one or more external devices. As such, IMD 10 may be pre-loaded with data, such as multiple heartbeat templates of various cardiac event morphologies. The set of cardiac morphology templates may include specific patterns and durations cardiac events. The set of cardiac morphology templates may include various onset and offset patterns. As an example, the set of cardiac morphology templates may include various patterns and durations of PVC that trigger NSVT. IMD 10 may store the set of cardiac morphology templates received from computing system 6 and / or one or more external devices 12 in templates data storage 63 of storage device 60. Templates data storage 63 may include any volatile, non-volatile, magnetic, optical, or electrical media, such as a random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically-erasable programmable ROM (EEPROM), flash memory, or any other digital media. In general, templates data storage 63 may be configured as a database, flat file, table, or other data structure stored within storage device 60.

[0059] Processing circuitry 50 may control sensing circuitry 52 to sense at least one physiological parameter corresponding cardiac physiology of a patient, e.g., by sensing electrical activity of the patient’s heart via a plurality of electrodes, such as electrodes 16. Sensing circuitry 52 may be selectively coupled to electrodes 16 via switching circuitry 58, e.g., to sense electrical signals of the heart of patient 4, for example by selecting the electrodes 16 and polarity, referred to as the sensing vector, used to sense cardiac activity, as controlled by processing circuitry 50. As such, sensing circuitry 52 may monitor signals from sensors 62, which may include one or more accelerometers, pressure sensors, and / or optical sensors, as examples. One or more sensors 62 may be configured to detect at least one physiological parameter corresponding to cardiac physiology of patient 4. Sensing circuitry 52 may be configured to generate patient physiological data based on the at least one physiological parameter, or generate sensor data from sensor signals received fromsensor(s) 62 that encode patient physiological parameters. For example, sensing circuitry 52 may convert signals corresponding to the sensed electrical activity to digital form (e.g., ECG) and provide the digitized signals to processing circuitry 50 for detection analysis. As such, the generated patient physiological data may comprise recorded digitized signals representing the electrical activity. In some examples, sensing circuitry 52 may include one or more filters and amplifiers for filtering and amplifying signals received from electrodes 16 and / or sensors 62. Sensing circuitry 52 and / or processing circuitry 50 may store some or all of the sensor data or patient physiological data as a portion of patient data storage 64 in storage device 60.

[0060] In some examples, patient data storage 64 is shared between various components executing at IMD 10 (e.g., between one or more of components 50-70 or other components not shown in FIG. 2). In some examples, IMD 10 may receive and store information from a patient (e.g., sensed data) and / or external computing device or system over a specified period of time. Patient data storage 64 may include any volatile, non-volatile, magnetic, optical, or electrical media, such as a random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically-erasable programmable ROM (EEPROM), flash memory, or any other digital media. In general, patient data storage 64 may be configured as a database, flat file, table, or other data structure stored within storage device 60. In some examples, patient data storage 64 may operate, at least in part, as a cache or temporary storage for patient data received by IMD 10. In various examples, patient data storage 64 may store, at least temporarily, patient physiological data generated by sensing circuitry 52 and / or other data or contextual information received by IMD 10 that corresponds with the physiology of patient 4. Detection logic 66 may query, either continuously or periodically, patient data storage 64 to determine whether at least a portion of the patient physiological data indicates at least one cardiac event. In some examples, responsive to detection logic 66 determining a portion of the patient physiological data does not indicate at least one cardiac event, patient data storage 64 may delete or otherwise erase the portion of the patient physiological data. As such, the memory storage of IMD 10 may be managed.

[0061] Patient data storage 64 may store patient physiological data corresponding to one or more physiological parameters (e.g., ECG signal data corresponding to sensed electrical activity), which may encompass temporal data or a period of time (e.g., during which a cardiac event may have occurred) and, in some examples, include a sequence of values representing waveforms (e.g., ECG waves, and / or ECG-type waveforms) of a cardiac rhythm. It should be noted that the data corresponding to typical sensed electrical activity records a series of samples representing points on waves (e.g., the P wave, Q wave, R wave, S wave, T wave and U wave), intervals (e.g., PRinterval, QRS interval (also called QRS duration), QT interval or R-R interval), segments (e.g., PR segment, ST segment or TP segment), complex(es) (e.g., QRS complex), and other components. As described herein, in some examples, IMD 10 may be configured to monitor and / or analyze continuous ECG signal data generated by sensing circuitry 52, which may be stored, at least temporarily, in patient data storage 64.

[0062] In some examples, IMD 10 may be configured to implement two stages of a three-stage (or more than three-stage) algorithm to detect cardiac events. Detection logic 66 may be configured to apply a first stage model or algorithm (e.g., logic, one or more criteria, a machine learning model, etc.) to the generated patient physiological data (e.g., ECG signal) to determine at least one cardiac event. Specifically, processing circuitry 50 may execute detection logic 66 to perform a detection analysis (e.g., apply a cardiac rate or cardiac interval threshold and / or a detection algorithm) on generated patient physiological data to identify cardiac events, including arrhythmias, that are likely to cause a change (e.g., a decline) in patient health or, otherwise, negatively affect the patient. As described herein, the detection analysis may be performed on data stored in patient data storage 64 and / or any other sensor data. In some examples, the first stage algorithm may run continuously and develop rate-based features after performing peak detections.

[0063] As such, detection logic 66 may be configured to identify one or more waveforms from the generated patient physiological data (e.g., ECG signals) (or portions thereof) and / or other sensed signals from any electrodes 16 connected to the patient’s heart. In some examples, the first stage algorithm may be configured to compare the generated patient physiological data to at least one of a cardiac rate or cardiac interval threshold. In some examples, the first stage algorithm may be a tachyarrhythmia detection algorithm that uses R-R intervals and cardioacceleration, e.g., to detect NSVT. In some examples, detection logic may detect NSVT based on identifying a threshold number of R-R intervals less than or equal to a threshold length (or greater than or equal to a threshold rate). The threshold number of intervals may be consecutive or X of the Y intervals meeting the length threshold. In some examples, the one or more waveforms may indicate an initial detection of a cardiac event. Detection logic 66 may apply the first model or algorithm to the one or more waveforms, and based on one or more prediction values generated by the first model or algorithm, determine whether the one or more waveforms indicate at least one cardiac event.

[0064] In some examples, a baseline level or certain threshold may be established for each variable that contributes to a certain cardiac event. For example, in the case of detecting NSVT episodes, a baseline for the patient physiological data may be based on the AHA guidelines relating to NSVT burden that specify a frequency > 3, a duration > 10 beats, and a rate > 200 bpmoccurring over 24 to 48 hours of monitoring is significant. For pediatric patients, a VT rate that exceeds the baseline sinus rate by >20% may be considered significant. Detection logic 66 may be configured to analyze the patient physiological data and determine any deviations from the baseline or if at least a portion of the data has met the threshold. Detection logic 66 may determine that any data that exceeds a baseline or threshold is indicative of a cardiac event.

[0065] In some examples, IMD 10 may determine one or more operating parameters, thresholds for detection, or other metrics based on the data received from computing system 6, such as one or more cardiac morphology templates. In some examples, IMD 10 may establish a baseline or threshold for ECG data based on patient history, previously detected events, and / or patient daily activity received from computing system 6. Detection logic 66 may compare continuous ECG data to the baseline to determine whether an event has occurred. As described herein, detection logic 66 may be configured to implement threshold analysis and / or detection algorithms that analyze the continuous ECG data to detect events. Furthermore, based on patient physiological data or patient history, detection logic 66 may change the sensitivity or parameters of the thresholds and / or detection algorithms.

[0066] For example, for post myocardial infarction (MI) patients, an increased sensitivity to NSVT may be desired, and ST elevation or other ST deviations may be more significant. Based on this post MI patient history, IMD 10 may be programmed with a lower detection threshold for NSVT and a greater weight assigned to ST elevation. In another example, for heart failure (HF) patients, night or resting heart rate and potential for increased prevalence of wide complex tachycardia (WCT) may be more significant. Based on this HF patient history, IMD 10 may be programmed with greater weights assigned to resting heart rate and WCT occurrence. In some examples, detection logic 66 may change the sensitivity or parameters of the detection algorithms based on data or feedback received from computing system 6 and / or external device 12. In some examples, detection logic 66 may apply one or more signal processing algorithms, such as filtering algorithms and feature extraction algorithms.

[0067] In some examples, detection logic 66 may apply one or more signal processing techniques, such as filtering and feature extraction. In some examples, detection logic 66 may apply a pattern recognition technique to interpret electrical vectors recorded in corresponding ECG data as one or more of the above wave components. In some examples, detection logic 66 may implement a Hidden Markov Model or Dynamic Time Warping. In some examples, detection logic 66 may apply one or more classification algorithms, such as decision trees or k-nearest neighbors. In some examples, detection logic 66 may apply a rule-based system. In some examples, detectionlogic 66 may apply a time-frequency analysis, such as wavelet transform. In some examples, detection logic 66 may be configured to additionally or alternatively apply one or more machine learning algorithms to the one or more waveforms, and based on one or more prediction values of the machine learning model, determine whether the one or more waveforms indicate at least one cardiac event. Detection logic 66 may apply one or more supervised machine learning algorithms (e.g., support vector machines, random forest, neural networks, logistic regression, etc.) and / or one or more unsupervised machine learning algorithms (k-means clustering, hierarchal clustering, isolation forests, etc.). Some other non-limiting examples of machine learning techniques include K-Nearest Neighbor algorithm and Multi-layer Perceptron. In some examples, detection logic 66 may apply a deep learning algorithm. For example, a CNN model of ResNet-18 may be used. Some non-limiting examples of models that may be used for transfer learning include AlexNet, VGGNet, GoogleNet, ResNet50, or DenseNet, etc.

[0068] In some examples, IMD 10 may confirm a potential cardiac event using, for example, AF burden, NSVT burden, patient activity, changes in ECG (e.g., changes in ST elevation and QRS width, amplitude, and rise rate), any of the variables detailed above, or any of the algorithms listed above.

[0069] Responsive to detection logic 66 determining at least one cardiac event, processing circuitry 50 may implement template matching unit 67 to apply a second model or second stage algorithm to one or more segments of the patient physiological data corresponding to the at least one cardiac event and the stored set of cardiac morphology templates in templates data storage 63. In some examples, processing circuitry 50 may further generate a cardiac morphology template based on at least a portion of the patient physiological data stored in patient data storage 64 corresponding to the at least one cardiac event. As described herein, the generated cardiac morphology template may be considered a “snapshot” or representation (e.g., based on application of a transform) of at least a portion of patient physiological data (e.g., a QRS complex or other portion of ECG signal data) corresponding to the at least one cardiac event. As such, IMD 10 may generate one or more cardiac morphology templates that may include specific patterns and durations of cardiac events. The generated cardiac morphology template may additionally be stored, at least temporarily, in patient data storage 64.

[0070] The second model or second stage algorithm implemented by template matching 67 may determine similarity between the one or more segments of the patient physiological data corresponding to the at least one cardiac event and the stored set of cardiac morphology templates in templates data storage 63. The second model or algorithm applied by template matching 67 maybe more computationally intensive than the first stage algorithm applied by detection logic 66, and may be an edge detection algorithm that uses template matching and morphology features for identifying true cases of a cardiac event (such as ventricular tachycardia (VT)), e.g., by comparing morphology of individual beats in the patient data to one or more of the morphology templates stored in templates data storage 63. Template matching 67 may determine a number of beats in the suspected event that sufficiently match the template(s). As described herein, multiple templates may be stored in templates storage 63, such that multiple morphologies generated by different foci within the heart can be matched. In some examples, the second stage algorithm applied by template matching unit 67 may filter out a portion of the false positives detected by the first stage algorithm applied by detection logic 66.

[0071] Specifically, template matching 67 may apply a second model or algorithm to the set of cardiac morphology templates stored in templates data storage 63 and the one or more segments of the patient physiological data to determine a first instance of a cardiac morphology template match. Template matching 67 may be configured to determine a similarity score between the one or more segments of the patient physiological data and one or more cardiac morphology templates from the set of stored cardiac morphology templates. The similarity score may be based off a comparison of wavelet decompositions. Specifically, template matching 67 may be configured to assess a partial waveform displayed in the patient data by comparing it to one or more waveforms displayed in one or more cardiac morphology templates from the set of stored cardiac morphology templates. If the similarity score meets a threshold score for a match, IMD 10 may record the number of beats between the match. In some examples, template matching 67 may assess subsequent beats from the match for similarity. In some examples, template matching 67 may repeat an assessment for similarity three times for confirmation. In some examples, processing circuitry 50 is further configured to count each instance of a cardiac morphology template match for each cardiac morphology template from the set of stored cardiac morphology templates. In some examples, a count may be stored with each cardiac morphology template from the set of stored cardiac morphology templates in templates data storage 63.

[0072] In some examples, template matching 67 may apply one or more feature matching algorithms (e.g., a scale-invariant feature transform, speeded up robust features, binary robust invariant scalable keypoints, etc.) to determine similarity. In some examples, template matching 67 may apply one or more template matching algorithms (e.g., normalized cross-correlation, sum of squared differences., etc.) to determine similarity. In some examples, template matching 67 maycompare a transformation of the sensed signal, e.g., a wavelet transformation, to template wavelet transformation values.

[0073] In some examples, template matching 67 may apply a rules-based algorithm or logic to determine similarity. In some examples, template matching 67 may apply histogram-based matching (e.g., color-based image matching). In some examples, template matching 67 may apply geometric transformation matching (e.g., random sample consensus). In some examples, template matching 67 may apply local binary pattern for image matching.

[0074] In some examples, the second model applied by template matching 67 may apply an edge detection algorithm (e.g., Sobel operator, Prewitt operator, canny edge detector, Laplacian of Gaussian (LoG), Marr-Hildreth or LoG with zero-crossing, Kirsch operator, Roberts Cross operator, Frei-Chen operator, edge detection with Hough Transform, Craigs Edge Dectector, etc.) to determine similarity.

[0075] In some examples, template matching 67 may apply one or more machine learning models to determine similarity. Some non-limiting examples of machine learning techniques include Support Vector Machines, K-Nearest Neighbor algorithm, and Multi-layer Perceptron. In some examples, template matching 67 may apply a deep learning algorithm (e.g., convolutional neural networks, neural feature embeddings, a CNN model of ResNet-18, Al exNet, VGGNet, GoogleNet, ResNet50, DenseNet, Transformers, Attention models, ViT, BERT, GPT, etc.).

[0076] In some examples, IMD 10 may be configured to implement all stages of a multi-stage algorithm to detect events. For example, while detection logic 66 and template matching 67 may implement a first and second model, respectively, IMD 10 may be configured to also implement a third stage model. In these examples, IMD 10 may be configured to collect continuous ECG signal data and implement a deep learning algorithm (e.g., a deep learning classification algorithm) that further filters out false positives by segmenting the ECG signal data corresponding to a template match and determining if the data constitutes a true cardiac event (e.g., the third stage model may determine an extent to which patient parameter data is indicative of an acute health event, such as a ventricular tachyarrhythmia or SC A). For example, IMD 10 implement a deep learning algorithm (e.g., a deep learning classification algorithm or segmentation algorithm) that further filters out false positives by segmenting the ECG signal data corresponding to a template match and determining if the data constitutes a true cardiac event. In some examples, IMD 10 may switch between a high computation mode and a low computation mode based on the detection of data that triggers the implementation of the second algorithm or the third algorithm.

[0077] IMD 10 may store data indicative of the first instance of the cardiac morphology template match in detected event data storage 69. Detected event data storage 69 may include any volatile, non-volatile, magnetic, optical, or electrical media, such as a random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically-erasable programmable ROM (EEPROM), flash memory, or any other digital media. In general, detected event data storage 69 may be configured as a database, flat file, table, or other data structure stored within storage device 60. In some examples, detected event data storage 69 may operate, at least in part, as a cache or temporary storage for data indicative of a first instance of a cardiac morphology template match. In some examples, the data indicative of the first instance of the cardiac morphology template match may include recorded data and meta data that is associated with the at least one cardiac event. For example, the data indicative of the first instance of the cardiac morphology template match may include one or more of the one or more segments of the generated patient physiological data corresponding to the at least one cardiac event (e.g., onset and offset triggers or physiological data indicative of the start and termination of the at least one cardiac event) a frequency of the at least one cardiac event, a number of heartbeats associated with the at least one cardiac event, a total number of heartbeats, a heartrate, a time duration, and any other data or meta data corresponding to the at least one cardiac event. For example, data indicative of the first instance of the cardiac morphology template match may include the episode of PVC pattern (e.g., bigeminy, trigeminy, every beat, etc.) and PVC morphology.

[0078] In some examples, IMD 10 may transmit, to a computing system (e.g., computing system 6) or device (e.g., external device 12) in communication with IMD 10, the data indicative of the first instance of the cardiac morphology template match that is stored in detected event data storage 69. As described above, the transmitted data may include the one or more segments of the patient physiological data and one or more stored cardiac morphology templates corresponding to the particular morphology template match, one or more short segments of ECG signal corresponding to the particular morphology template match and / or at least one cardiac event, and / or other patient physiological data stored in patient data storage 64 that corresponds to the particular morphology template match and / or at least one cardiac event. IMD 10 may transmit the data to the computing system based on a time schedule, immediately after a first instance of a cardiac morphology template match has been determined, or, in some examples, based on whether a certain amount of memory has been consumed by IMD 10. Although many examples described herein discuss a transmission of data indicative of a first instance of a cardiac morphology template match, in some examples, IMD 10 may be configured to determine one or more instances of acardiac morphology template match, and may transmit, to computing system 6, data indicative of the one or more instances of the cardiac morphology template match. Furthermore, in some examples, IMD 10 may not perform template matching, any may transmit data indicative of a cardiac event as determined by detection logic 66.

[0079] FIG. 3 is a conceptual side-view diagram illustrating an example configuration of the IMD of FIGS. 1 and 2, in accordance with one or more examples of the present disclosure. The descriptions of FIGS. 1 and 2 are equally applicable to FIG. 3. While different examples of IMD 10 may include leads, in the example shown in FIG. 3, IMD 10 may include a leadless, subcutaneously-implantable monitoring device having a housing 15 and an insulative cover 76. Electrode 16A and electrode 16B may be formed or placed on an outer surface of cover 76. Components 50-70, described above with respect to FIG. 2, may be formed or placed on an inner surface of cover 76, or within housing 15. In some examples, IMD 10 may include an antenna (not shown in FIG. 3) that may be formed or placed on the inner surface of cover 76, but may be formed or placed on the outer surface in some examples. In some examples, insulative cover 76 may be positioned over an open housing 15 such that housing 15 and cover 76 enclose an antenna and components 50-70, and protect the antenna and circuitries from fluids such as body fluids.

[0080] Components 50-70 may be formed on the inner side of insulative cover 76, such as by using flip-chip technology. Insulative cover 76 may be flipped onto a housing 15. When flipped and placed onto housing 15, the components of IMD 10 formed on the inner side of insulative cover 76 may be positioned in a gap 78 defined by housing 15. Electrodes 16 may be electrically connected to switching circuitry 58 through one or more vias (not shown) formed through insulative cover 76. Insulative cover 76 may be formed of sapphire (i.e., corundum), glass, parylene, and / or any other suitable insulating material. Housing 15 may be formed from titanium or any other suitable material (e.g., a biocompatible material). Electrodes 16 may be formed from any of stainless steel, titanium, platinum, iridium, or alloys thereof. In addition, electrodes 16 may be coated with a material such as titanium nitride or fractal titanium nitride, although other suitable materials and coatings for such electrodes may be used.

[0081] FIG. 4A is a perspective drawing illustrating an IMD 10A, which may be an example configuration of IMD 10 of FIGS. 1-3 as an ICM. In the example shown in FIG. 4 A, IMD 10A may be embodied as a monitoring device having housing 1412, proximal electrode 1416A and distal electrode 1416B. Housing 1412 may further comprise first major surface 1414, second major surface 1418, proximal end 1420, and distal end 1422. Housing 1412 encloses electronic circuitry located inside the IMD 10A and protects the circuitry contained therein from body fluids.Housing 1412 may be hermetically sealed and configured for subcutaneous implantation. Electrical feedthroughs provide electrical connection of electrodes 1416A and 1416B.

[0082] In the example shown in FIG. 4A, IMD 10A is defined by a length / ., a width W and thickness or depth D and is in the form of an elongated rectangular prism wherein the length L is much larger than the width W, which in turn is larger than the depth D. In one example, the geometry of the IMD 10A - in particular a width W greater than the depth D - is selected to allow IMD 10A to be inserted under the skin of the patient using a minimally invasive procedure and to remain in the desired orientation during insertion. For example, the device shown in FIG. 4A includes radial asymmetries (notably, the rectangular shape) along the longitudinal axis that maintains the device in the proper orientation following insertion. For example, the spacing between proximal electrode 1416A and distal electrode 1416B may range from 5 millimeters (mm) to 55 mm, 30 mm to 55 mm, 35 mm to 55 mm, and from 40 mm to 55 mm and may be any range or individual spacing from 5 mm to 60 mm. In addition, IMD 10A may have a length L that ranges from 30 mm to about 70 mm. In other examples, the length L may range from 5 mm to 60 mm, 40 mm to 60 mm, 45 mm to 60 mm and may be any length or range of lengths between about 30 mm and about 70 mm. In addition, the width W of major surface 1414 may range from 3 mm to 15, mm, from 3 mm to 10 mm, or from 5 mm to 15 mm, and may be any single or range of widths between 3 mm and 15 mm. The thickness of depth D of IMD 10A may range from 2 mm to 15 mm, from 2 mm to 9 mm, from 2 mm to 5 mm, from 5 mm to 15 mm, and may be any single or range of depths between 2 mm and 15 mm. In addition, IMD 10A according to an example of the present disclosure is has a geometry and size designed for ease of implant and patient comfort. Examples of IMD 10A described in this disclosure may have a volume of three cubic centimeters (cm) or less, 1.5 cubic cm or less or any volume between three and 1.5 cubic centimeters.

[0083] In the example shown in FIG. 4 A, once inserted within the patient, the first major surface 1414 faces outward, toward the skin of the patient while the second major surface 1418 is located opposite the first major surface 1414. In addition, in the example shown in FIG. 4A, proximal end 1420 and distal end 1422 are rounded to reduce discomfort and irritation to surrounding tissue once inserted under the skin of the patient. IMD 10A, including instrument and method for inserting IMD 10A is described, for example, in U.S. Patent Publication No. 2014 / 0276928, incorporated herein by reference in its entirety.

[0084] Proximal electrode 1416A is at or proximate to proximal end 1420, and distal electrode 1416B is at or proximate to distal end 1422. Proximal electrode 1416A and distal electrode 1416B are used to sense cardiac EGM signals, e.g., ECG signals, thoracically outside the ribcage, whichmay be sub-muscularly or subcutaneously. Cardiac signals may be stored in a memory of IMD 10 A, and data may be transmitted via integrated antenna 1430 A to another device, which may be another implantable device or an external device, such as external device 14. In some example, electrodes 1416A and 1416B may additionally or alternatively be used for sensing any biopotential signal of interest, which may be, for example, an ECG, EEG, EMG, or a nerve signal, or for measuring impedance, from any implanted location.

[0085] In the example shown in FIG. 4A, proximal electrode 1416A is at or in close proximity to the proximal end 1420 and distal electrode 1416B is at or in close proximity to distal end 1422. In this example, distal electrode 1416B is not limited to a flattened, outward facing surface, but may extend from first major surface 1414 around rounded edges 1424 and / or end surface 1426 and onto the second major surface 1418 so that the electrode 1416B has a three-dimensional curved configuration. In some examples, electrode 1416B is an uninsulated portion of a metallic, e.g., titanium, part of housing 1412.

[0086] In the example shown in FIG. 4A, proximal electrode 1416A is located on first major surface 1414 and is substantially flat, and outward facing. However, in other examples proximal electrode 1416A may utilize the three-dimensional curved configuration of distal electrode 1416B, providing a three dimensional proximal electrode (not shown in this example). Similarly, in other examples distal electrode 1416B may utilize a substantially flat, outward facing electrode located on first major surface 1414 similar to that shown with respect to proximal electrode 1416A. The various electrode configurations allow for configurations in which proximal electrode 1416A and distal electrode 1416B are located on both first major surface 1414 and second major surface 1418. In other configurations, such as that shown in FIG. 4 A, only one of proximal electrode 1416A and distal electrode 1416B is located on both major surfaces 1414 and 1418, and in still other configurations both proximal electrode 1416A and distal electrode 1416B are located on one of the first major surface 1414 or the second major surface 1418 (e.g., proximal electrode 1416A located on first major surface 1414 while distal electrode 1416B is located on second major surface 1418). In another example, IMD 10A may include electrodes on both major surface 1414 and 1418 at or near the proximal and distal ends of the device, such that a total of four electrodes are included on IMD 10 A. Electrodes 1416A and 1416B may be formed of a plurality of different types of biocompatible conductive material, e.g., stainless steel, titanium, platinum, iridium, or alloys thereof, and may utilize one or more coatings such as titanium nitride or fractal titanium nitride.

[0087] In the example shown in FIG. 4 A, proximal end 1420 includes a header assembly 1428 that includes one or more of proximal electrode 1416A, integrated antenna 1430 A, anti-migrationprojections 1482, and / or suture hole 1434. Integrated antenna 1430A is located on the same major surface (i.e., first major surface 1414) as proximal electrode 1416A and is also included as part of header assembly 1428. Integrated antenna 1430 A allows IMD 10A to transmit and / or receive data. In other examples, integrated antenna 1430 A may be formed on the opposite major surface as proximal electrode 1416A, or may be incorporated within the housing 1412 of IMD 10A. In the example shown in FIG. 4A, anti-migration projections 1432 are located adjacent to integrated antenna 1430 A and protrude away from first major surface 1414 to prevent longitudinal movement of the device. In the example shown in FIG. 4A, anti-migration projections 1432 include a plurality (e.g., nine) small bumps or protrusions extending away from first major surface 1414. As discussed above, in other examples anti-migration projections 1432 may be located on the opposite major surface as proximal electrode 1416A and / or integrated antenna 1430A. In addition, in the example shown in FIG. 4A, header assembly 1428 includes suture hole 1434, which provides another means of securing IMD 10A to the patient to prevent movement following insertion. In the example shown, suture hole 1434 is located adjacent to proximal electrode 1416A. In one example, header assembly 1428 is a molded header assembly made from a polymeric or plastic material, which may be integrated or separable from the main portion of IMD 10 A.

[0088] FIG. 4B is a perspective drawing illustrating another IMD 10B, which may be another example configuration of IMD 10 from FIGS. 1, 3, and 4 as an ICM. IMD 10B of FIG. 4B may be configured substantially similarly to IMD 10A of FIG. 4A, with differences between them discussed herein.

[0089] IMD 10B may include a leadless, subcutaneously-implantable monitoring device, e.g. an ICM. IMD 10B includes housing having a base 1440 and an insulative cover 1442. Proximal electrode 1416C and distal electrode 1416D may be formed or placed on an outer surface of cover 1442. Various circuitries and components of IMD 10B, e.g., described above with respect to FIG. 2, may be formed or placed on an inner surface of cover 1442, or within base 1440. In some examples, a battery or other power source of IMD 10B may be included within base 1440. In the illustrated example, antenna 1430B is formed or placed on the outer surface of cover 1442, but may be formed or placed on the inner surface in some examples. In some examples, insulative cover 1442 may be positioned over an open base 1440 such that base 1440 and cover 1442 enclose the circuitries and other components and protect them from fluids such as body fluids. The housing including base 1440 and insulative cover 1442 may be hermetically sealed and configured for subcutaneous implantation.

[0090] Circuitries and components may be formed on the inner side of insulative cover 1442, such as by using flip-chip technology. Insulative cover 1442 may be flipped onto a base 1440. When flipped and placed onto base 1440, the components of IMD 10B formed on the inner side of insulative cover 1442 may be positioned in a gap 1444 defined by base 1440. Electrodes 1216C and 1216D and antenna 1230B may be electrically connected to circuitry formed on the inner side of insulative cover 1442 through one or more vias (not shown) formed through insulative cover 1442. Insulative cover 1442 may be formed of sapphire (i.e., corundum), glass, parylene, and / or any other suitable insulating material. Base 1440 may be formed from titanium or any other suitable material (e.g., a biocompatible material). Electrodes 1416C and 1246D may be formed from any of stainless steel, titanium, platinum, iridium, or alloys thereof. In addition, electrodes 1246C and 1246D may be coated with a material such as titanium nitride or fractal titanium nitride, although other suitable materials and coatings for such electrodes may be used.

[0091] In the example shown in FIG. 4B, the housing of IMD 10B defines a length / ., a width W and thickness or depth D and is in the form of an elongated rectangular prism wherein the length L is much larger than the width W, which in turn is larger than the depth D, similar to IMD 10A of FIG. 4A. For example, the spacing between proximal electrode 1416C and distal electrode 1416D may range from 5 mm to 50 mm, from 30 mm to 50 mm, from 35 mm to 45 mm, and may be any single spacing or range of spacings from 5 mm to 50 mm, such as approximately 40 mm. In addition, IMD 10B may have a length L that ranges from 5 mm to about 70 mm. In other examples, the length L may range from 30 mm to 70 mm, 40 mm to 60 mm, 45 mm to 55 mm, and may be any single length or range of lengths from 5 mm to 50 mm, such as approximately 45 mm. In addition, the width may range from 3 mm to 15 mm, 5 mm to 15 mm, 5 mm to 10 mm, and may be any single width or range of widths from 3 mm to 15 mm, such as approximately 8 mm. The thickness or depth D of IMD 10B may range from 2 mm to 15 mm, from 5 mm to 15 mm, or from 3 mm to 5 mm, and may be any single depth or range of depths between 2 mm and 15 mm, such as approximately 4 mm. IMD 10B may have a volume of three cubic centimeters (cm) or less, or 1.5 cubic cm or less, such as approximately 1.4 cubic cm.

[0092] In the example shown in FIG. 4B, once inserted subcutaneously within the patient, outer surface of cover 1442 faces outward, toward the skin of the patient. In addition, as shown in FIG. 4B, proximal end 1446 and distal end 1448 are rounded to reduce discomfort and irritation to surrounding tissue once inserted under the skin of the patient. In addition, edges of IMD 10B may be rounded.

[0093] FIG. 5 is a block diagram illustrating an example configuration of components of computing system 6. In the example of FIG. 4, computing system 6 includes processing circuitry 80, communication system 82, storage device 85, and user interface module 86.

[0094] Processing circuitry 80 may include fixed function circuitry and / or programmable processing circuitry. Processing circuitry 80 may include any one or more of a microprocessor, a controller, a graphics processing unit (GPU), a tensor processing unit (TPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or equivalent discrete or analog logic circuitry. In some examples, processing circuitry 80 may include multiple components, such as any combination of one or more microprocessors, one or more controllers, one or more GPUs, one or more TPUs, one or more DSPs, one or more ASICs, or one or more FPGAs, as well as other discrete or integrated logic circuitry. The functions attributed to processing circuitry 80 herein may be embodied as software, firmware, hardware, or any combination thereof. In some examples, storage device 85 includes computer-readable instructions that, when executed by processing circuitry 80, cause computing system 6 and processing circuitry 80 to perform various functions attributed herein to computing system 6 and processing circuitry 80. Storage device 85 may include any volatile, non-volatile, magnetic, optical, or electrical media, such as a random-access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically-erasable programmable ROM (EEPROM), flash memory, or any other digital media.

[0095] Communication system 82 may include any suitable hardware, firmware, software, or any combination thereof for communicating with another device, such as IMD 10 and / or external device 12 using any of the secure or non-secure communication protocols described above in relation to FIGS 1 - 4B. Under the control of processing circuitry 80, communication system 82 may receive downlink telemetry from, as well as send uplink telemetry to, IMD 10, external device 12, or another device. Communication system 82 may be configured to transmit or receive signals via inductive coupling, electromagnetic coupling, Near Field Communication (NFC), RF communication, Bluetooth®, WI-FI™, or other proprietary or non-proprietary wireless communication schemes. Communication system 82 may also be configured to communicate with devices other than IMD 10 or external device 12 via any of a variety of forms of wired and / or wireless communication and / or network protocols.

[0096] Storage device 85 may be configured to store information within computing system 6 during operation. Storage device 85 may include a computer-readable storage medium or computer-readable storage device. In some examples, storage device 85 includes one or more of ashort-term memory or a long-term memory. Storage device 85 may include, for example, RAM, DRAM, SRAM, magnetic discs, optical discs, flash memories, or forms of EPROM or EEPROM. In some examples, storage device 85 is used to store data indicative of instructions for execution by processing circuitry 80. Storage device 85 may be used by software or applications running on computing system 6 to temporarily store information during program execution.

[0097] Data exchanged between computing system 6, external device 12, and IMD 10 may include operational parameters (e.g., such as a communication rate). Computing system 6 may transmit data including computer readable instructions which, when implemented by IMD 10, may control IMD 10 to change one or more operational parameters and / or export collected data. For example, processing circuitry 80 may transmit an instruction to IMD 10 which requests IMD 10 to export collected data (e.g., diagnostic data or patient physiological data) to computing system 6 and / or external device 12. Computing system 6 may receive the collected data from IMD 10 and store the collected data in storage device 85. As described herein, the data computing system 6 and / or external device 12 receives from IMD 10 may include data indicative of a first instance of a cardiac morphology template match, such as patient physiological data and one or more stored cardiac morphology templates corresponding to the particular morphology template match, one or more short segments of ECG signal corresponding to the particular morphology template match and / or at least one cardiac event, and / or other patient physiological data stored in patient data storage 64 that corresponds to the particular morphology template match and / or at least one cardiac event. In one example, IMD 10 may generate sensor data from captured signals and forward that sensor data to computing system 6 for storage in patient data storage 84 with other sensor data for a same patient. Processing circuitry 80 may implement any of the techniques described herein to analyze data from IMD 10 to determine whether the data constitutes as evidence of a cardiac event that occurred in the patient at some point in time, e.g., to determine whether the patient is experiencing a change in health. In some examples, processing circuitry 80 may implement any of the techniques described herein to generate predictions regarding the patient’s health.

[0098] As described herein, computing system 6 may be employed by system 2 to further determine or classify cardiac events detected by IMD 10. In examples in which computing system 6 is a cloud computing system, computing system 6 may be equipped with a greater amount of memory space than IMD 10. As such, computing system 6 may apply machine learning techniques or algorithms that may be too computationally expensive to execute on IMD 10. Furthermore, computing system 6 may apply machine learning techniques or algorithms that can more accurately determine (e.g., reduce false positives detected by IMD 10) and / or predict cardiac events.

[0099] Processing circuitry 80 executes machine learning module 88 configured to perform analysis on patient data storage 84. Specifically, machine learning module 88 may apply one or more machine learning models or algorithms to data received from IMD 10 via communication circuitry 82 that indicates a positive cardiac event detection (e.g., data indicative of a first instance of a cardiac morphology template match, one or more short segments of ECG signal corresponding to at least one cardiac event, and / or other patient physiological data that corresponds to at least one cardiac event). Machine learning module 88 may confirm or reject the data received from IMD 10 as indicative of a cardiac event. As such, computing system 6 may be configured to further filter out false positives initially detected by IMD 10 by applying one or more deep learning models (e.g., convolutional neural networks, neural feature embeddings, a CNN model of ResNet-18, AlexNet, VGGNet, GoogleNet, ResNet50, DenseNet, etc.) to the patient physiological data. In examples in which a first stage and second stage algorithm is applied by IMD 10, computing system 6 may apply a third stage model or algorithm that may significantly improve the specificity of the overall event detection method, as the third stage model or algorithm may comprise multiple layers of neurons and achieve greater accuracy with increased computational power. Additionally, in some examples, if computing system 6 determines that IMD 10 incorrectly identified a template match, computing system 6 may send an updated a cardiac morphology template, e.g., segmented ECG signal data or values representing a transform thereof, back to IMD 10 in order to improve the accuracy of template matching 67.

[0100] As described herein, in some examples, additionally or alternatively to confirming or rejecting the presence of cardiac events, computing system 6 may be configured to determine or predict risk scores for cardiac events. For example, computing system 6 may be configured to determine a risk score for patient 4 (who may be an HCM patient) for sudden cardiac death based on NSVT events detected by IMD 10. In this example, IMD 10 may determine a first instance of a cardiac morphology template match that corresponds to an NSVT event. IMD 10 may transmit, to computing system 6, data indicative of the first instance of the cardiac morphology template match, such as NSVT episode duration, NSVT burden (frequency of NSVT episodes, rate (bpm), number of beats in each NVST episode), ECG signals recorded by IMD 10 during each episode, and / or patient physiological characteristics. Computing system 6 may receive data from IMD 10 and store the data in patient data storage 84. In some examples, computing system 6 may transform the transmitted data and store the data in patient data storage 84. For example, machine learning module 88 may normalize time-series ECG signal data prior to applying one or more machine learning algorithms to the data.

[0101] As described above, machine learning module 88 of computing system 6 may apply one or more machine learning models to the data, in which the one or more machine learning models may be trained on relevant features to various cardiac events (such as accurate R-R intervals relevant to NSVT). In some examples, the training regimen of the model may include multi-task learning of fiducial point segmentation and NSVT segmentation.

[0102] In some examples, the one or more machine learning models applied by machine learning module 88 may include a deep learning model, which may be a supervised learning model. The model may be a segmentation neural network that segments time-series ECG signals and classifies each sample as a cardiac event or not. The model may be a convolutional neural network with a UNet architecture or a transformer attention-based model. The segmentation neural network may segment continuous streams of cardiac data into discernible segments, in which each segment may represent a specific sample or frame (e.g., set of consecutive samples) of the ECG signal. In some examples, the signal data may be manipulated or pre-processed, e.g., noise reduction, normalization, and / or feature extraction (in which key characteristics of the ECG signal, such as QRS complex, T-wave, etc. are isolated) may be performed. The segmentation neural network may process each segment, in which layers of the neural network may be leveraged to detect patterns indicative of various cardiac conditions. The segmentation neural network may output a classification for each segment, e.g., a label indicative of a normal event or a cardiac anomaly.

[0103] A Unet architecture may provide an ability to capture both local and broader contextual information in the ECG signal, as the Unet architecture may utilize a series of convolutional layers for feature extraction, and then up-sample layers to reconstruct the classification of the signal at a high resolution. As such, the Unet architecture may provide more precise segmentation.

[0104] A transformer attention-based model may identify dependencies in the data, regardless of position within the signal. The transformer attention-based model may utilize attention mechanisms to weigh the importance of different segments of the ECG signal, thus better capturing long-range dependencies and subtle nuances that are crucial for accurate classification of cardiac events.

[0105] The model may transform the signal data into a multi-channel representation such that the model predicts with multiple different segmentation masks. For example, in some examples, the model may be a convolutional neural network with a UNet architecture with two output channels: a binary mask for NSVT, and a binary mask for PVCs. Specifically, in some examples, theECG waveforem may be normalized and passed into a neural network that includes aconvolutional UNet architecture, in which the neural network outputs a probability score for each sample or each event type (e.g., a probability score for premature ventricular contraction and a probability score for non-sustained ventricular tachycardia). A threshold may be applied to the probability scores to convert them to a binary mask for each event type (e.g., the probability score for premature ventricular contraction may be converted to a first binary mask for premature ventricular contraction, and the probability score for non-sustained ventricular tachycardia may be converted to a second binary mask for non-sustained ventricular tachycardia). The first binary mask and the second binary mask may be used to output a severity score. Specifically, the binary masks may be converted to a duration or burden of each event type, in which the burden may then be converted to a severity score.

[0106] In some examples, the model may perform tasks such as locating QRS complexes. In some examples, the model may predict multiple probabilities for each sample in time-series data, such as one probability per channel representing each characteristic of the data (e.g., different parts of the wave, such as complex T and P waves). The model may further predict the duration and presence of a cardiac event based on the probabilities exceeding a detection threshold. Machine learning module 88 may further apply smoothing, rolling median, etc. to the probability masks to produce a continuous output. A minimum number of consecutive probabilities exceeding the threshold may need to be met in order for the data to be classified as a true cardiac event. The model may determine the start time and stop time of a cardiac event.

[0107] Machine learning module 88 may determine, based on the output of the one or more machine learning models, a risk score (e.g., of sudden cardiac death or lethal arrhythmia) that weighs the frequency of true cardiac event occurrence and cardiac event duration. For example, longer and more frequent cardiac events may result in a higher risk score.

[0108] In some examples, machine learning module 88 may apply one or more machine learning models or algorithms to one or more waveforms in ECG data, and based on prediction values generated by the one or more machine learning models or algorithms, determine whether the one or more waveforms indicate a cardiac event. In some examples, ML module 88 may detect a cardiac event type based on the data received from IMD 10. For example, in some examples, ML module 88 may determine a classification of one or more waveforms by applying a modular machine learning architecture that includes modules that correspond to classifying respective cardiac arrhythmia types. In these examples, each module may include a neural network ensemble that includes a component neural network for predicting a likelihood of each cardiac arrhythmia type. Each component neural network of that ensemble may be configured to classify the patientphysiological data (e.g., sample(s) of ECG data) as a respective one of, for example, atrial fibrillation, atrial flutter, atrial flutter and atrial fibrillation, atrioventricular block, intraventricular conduction delay, premature contraction, premature ventricular contraction, premature atrial or atrioventricular junctional contraction, asystole / pause sinus bradycardia, sinus rhythm, sinus tachycardia, supraventricular tachycardia, non-sustained ventricular tachycardia, and ventricular fibrillation. It should be noted that the samples of data may or may not correspond to a suspected cardiac event.

[0109] Computing system 6 may also receive supplementary patient metadata (e.g., patient’s age, heart failure, ejection fraction) from one or more devices (e.g., IMD 10 or external device 12). Machine learning module 88 may additionally post-process the determined risk score using the supplementary patient metadata to further adjust the risk score. In some examples, the R-R intervals used in the model training may also be used in a post-processing logic or algorithm, as cardio-acceleration and cardio blunting may be valuable diagnostic features. In some examples, the post-processing logic or algorithm applied by machine learning module 88 may further analyze data outside of the final predicted cardiac event duration to identify cardio-acceleration and cardio blunting. In some examples, secondary data features may be used to further improve the diagnostic accuracy of the output. The post-processing logic or algorithm applied by machine learning module 88 may determine whether the predicted cardiac event duration meets the clinical definition for an episode (e.g., whether the predicted NSVT episode duration meets the clinical definition for an NSVT episode). The post-processing logic or algorithm may also determine if predicted events or samples that exceed the threshold represent noise rather than true events.

[0110] In examples in which machine learning module 88 applies a model with a convolutional UNet architecture that produces a binary mask for NSVT and a binary mask for PVCs, machine learning module 88 may implement post-processing logic that utilizes the binary masks and outputs a severity score. In these examples, the severity score may be a function of both NSVT presence and PVC presence and may be weighted based on duration of beat patterns, frequency of occurrence, and / or types of beat patterns observed. The score may represent severity of a single event or multiple events. In some examples, machine learning module 88 may further evaluate events for patterns (e.g., patterns of every beat, bigeminy, or trigeminy) and store this meta data with the corresponding ECG signal or patient physiological data in patient data storage 84.[OHl] In some examples, processing circuitry 80 generates, for output, data indicative of the cardiac event having a prediction value that exceeds a threshold value. Additionally, in some examples, if processing circuitry 80 determines that the cardiac event initially detected by IMD 10is a true cardiac event, processing circuitry 80 may further generate, for output, data indicative of a confirmation of IMD 10’s initial detection. For example, in response to a determination of a true positive cardiac event, processing circuitry 80 generates, for display, output data such as the one or more waveforms spanning an occurrence of the cardiac event. In contrast, if processing circuitry 80 determines that the cardiac event initially detected by IMD 10 is not a true cardiac event, processing circuitry 80 may generate, for output, data indicative of a rejection of IMD 10’s initial detection.

[0112] Additionally, any output from machine learning module 88 may be used to determine whether an ICD or other therapy should be recommended as treatment for patient 4. In some examples, machine learning module 88 may further include a recommender system or model configured to generate intelligent recommendations based on meta data or any other data stored in patient data storage 84 and any risk scores or predictions generated by other models included in machine learning module 88. In some examples, computing system 6 may send information pertaining to the patient physiological data and / or treatment recommendations to external device 12 and / or other devices in communication with computing system 6 via a network, such that the information may be displayed to patient 4, a clinician, or other user for review. For example, in some examples, a user may query a report of the data stored in patient data storage 84 and receive, from computing system 6, a display of each unique pattern or beat occurrence that was stored. For example, if a patient has an episode of PVCs in a bigeminy pattern and a run of NSVT, stacked plots of the corresponding ECG segments may be displayed in the report.

[0113] User interface module 86 may generate data indicative of a user interface for display on external device, such as external device 12. Computing system 6 may send, to external device 12, the data indicative of the user interface, and a user, such as a clinician or patient 4, may interact with external device 12 through the user interface. External device 12 may include a display (not shown), such as a liquid crystal display (LCD) or a light emitting diode (LED) display or other type of screen, which may present information related to IMD 10, e.g., patient physiological data, data indicative of a cardiac morphology template match, an indication of a cardiac event, prediction values of potential classes (e.g., types) of cardiac events, a classification of the patient physiological data as indicative of a cardiac event type based on the prediction values, indications of changes in patient health that are correlated to the classification, and / or treatment or therapy recommendations. In addition, user interface module 86 may be configured to receive input from a user of external device 12, such as, for example, feedback related to the accuracy of detections and / or predictions, updates to patient information, changes to IMD 10 settings / parameters, etc. Insome examples, external device 12 may comprise components that are similar if not substantially similar to the components described with IMD 10 and computing system 6. For example, external device 12 may include processing circuitry similar if not substantially similar to processing circuitry 50 and / or processing circuitry 50 of FIG. 2 and FIG. 5, respectively, communication system(s) similar if not substantially similar to communication system 56 and / or communication system 82 of FIG. 2 and FIG. 5, respectively, and storage device(s) similar if not substantially similar to storage device 60 and / or storage device 85 of FIG. 2 and FIG. 5, respectively. As such, in these examples, processing circuity of external device 12 may be configured to execute one or more components described with respect to IMD 10 and / or computing device 6, e.g., external device 12 may implement one or more of the models or algorithms discussed above with respect to IMD 10 and computing system 6. For example, responsive to determining at least one cardiac event, IMD 10 may transmit, to external device 12, one or more segments of the patient physiological data corresponding to the at least one cardiac event. External device 12 may be configured to apply the second model (e.g., a template matching algorithm) to a set of cardiac morphology templates and the one or more segments of the patient physiological data to determine a first instance of a cardiac morphology template match. External device 12 may then transmit, to computing system 6, data indicative of a first confirmed instance of a particular morphology template match (e.g., a segment of ECG signal corresponding to a particular morphology template match) for further analysis (e.g., implementation of the third model or algorithm).

[0114] FIG. 6 is a conceptual diagram illustrating an example machine learning model 1200 configured to determine an extent to which patient parameter data is indicative of an acute health event, such as a ventricular tachyarrhythmia or SCA. Machine learning model 1200 is an example of a set of rules implemented by any rules engine, algorithm, or machine learning model described herein with respect to IMD 10, such as detection logic 66 of FIG. 2 and ML module 88 of FIG. 5, any of which may be implemented by processing circuitry of any computing systems or devices in wireless communication with IMD 10, as discussed above. Machine learning model 1200 is an example of a deep learning model, or deep learning algorithm, trained to determine whether a particular set of patient parameter data indicates the presence of an acute health event, e.g., whether a particular segment of ECG signal data indicates SCA or a certain classification related to ventricular tachyarrhythmia or other cardiac event, as described herein.

[0115] One or more of IMD 10, computing system 6, or an external device 12 may train, store, and / or utilize machine learning model 1200, but other devices may apply inputs associated with a particular patient to machine learning model 1200 in other examples. As discussed above, othertypes of machine learning and deep learning models or algorithms may be utilized in other examples. For example, a CNN model of ResNet-18 may be used. Some non-limiting examples of models that may be used for transfer learning include AlexNet, VGGNet, GoogleNet, ResNet50, or DenseNet, etc. Some non-limiting examples of machine learning techniques include Support Vector Machines, K-Nearest Neighbor algorithm, and Multi-layer Perceptron.

[0116] As shown in the example of FIG. 6, machine learning model 1200 may include three layers. These three layers include input layer 1202, hidden layer 1204, and output layer 1206. Output layer 1206 comprises the output from the transfer function 1205 of output layer 1206. Input layer 1202 represents each of the input values XI through X4 provided to machine learning model 1200. The number of inputs may be equal to, less than, or greater than 4, including much greater than 4, e.g., hundreds or thousands. In some examples, the input values may any of the of values input into a machine learning model, as described above. In some examples, input values may include samples of an ECG signal. In addition, in some examples input values of machine learning model 1200 may include additional data, such as R-wave data, R-R interval data, or other data relating to one or more additional parameters of patient 4, as described herein.

[0117] Each of the input values for each node in the input layer 1202 is provided to each node of hidden layer 1204. In the example of FIG. 6, hidden layers 1204 include two layers, one layer having four nodes and the other layer having three nodes, but fewer or greater number of nodes may be used in other examples. Each input from input layer 1202 is multiplied by a weight and then summed at each node of hidden layers 1204. During training of machine learning model 1200, the weights for each input are adjusted to establish the relationship between the inputs, e.g., input ECG segment, to determining whether a particular set of inputs represents an acute health event and / or determining a score indicative of whether a set of inputs may be representative of SCA, MVT, PVT, VR, or another acute health event. In some examples, one hidden layer may be incorporated into machine learning model 1200, or three or more hidden layers may be incorporated into machine learning model 1200, where each layer includes the same or different number of nodes.

[0118] The result of each node within hidden layers 1204 is applied to the transfer function of output layer 1206. The transfer function may be liner or non-linear, depending on the number of layers within machine learning model 1200. Example non-linear transfer functions may be a sigmoid function or a rectifier function. The output 1207 of the transfer function may be a classification that indicates whether the particular ECG segment or other input set represents an acute health event, e.g., ventricular tachyarrhythmia, and / or a score indicative of an extent to whichthe input data set represents an acute health event. In some examples, output 1207 may include respective probabilities for a likelihood of a true cardiac event, e.g., as discussed herein with respect to FIGS. 2 and 6.

[0119] By applying the ECG signal data and / or other patient parameter data to a machine learning model, such as machine learning model 1200, processing circuitry, such as processing circuitry 80 of computing system 6, is able to determine a patient is experiencing or will soon experience an acute health event with great accuracy, specificity, and sensitivity. This may facilitate determinations of risk of sudden cardiac death and may lead to alerts and other interventions as described herein, such as determining an indication for ICD 18 of FIG. 7 for patient 4, and determining one or more operating parameters for ICD 18. Machine learning model 1200 may correspond to any one or more of rules, logic, algorithms, and machine learning models described herein.

[0120] FIG. 7 is an example of a machine learning model 1200 being trained using supervised and / or reinforcement learning techniques. Machine learning model 1200 may be implemented using any number of models for supervised and / or reinforcement learning, such as but not limited to, an artificial neural network, a decision tree, naive Bayes network, support vector machine, or k- nearest neighbor model, to name only a few of the examples discussed above. In some examples, processing circuitry of one or more of IMD 10, computing system 6, and / or an external device 12 initially trains the machine learning model 1200 based on training set data 1300 including numerous instances of input data corresponding to acute health events and non-acute health events, e.g., as labeled by an expert. A prediction or classification by the machine learning model 1200 may be compared 1304 to the target output 1303, e.g., as determined based on the label. Based on an error signal representing the comparison, the processing circuitry implementing a learning / training function 1305 may send or apply a modification to weights of machine learning model 1200 or otherwise modify / update the machine learning model 1200. For example, one or more of IMD 10, computing system 6, and / or an external device 12 may, for each training instance in the training set 1300, modify machine learning model 1200 to change a score generated by the machine learning model 1200 in response to data applied to the machine learning model 1200.

[0121] FIG. 8 is a block diagram illustrating an example system that includes an access point 90, a network 92, and one or more other computing devices 100A-100N (collectively, “computing devices 100”), which may be coupled to IMD 10, computing system 6, and external device 12 via network 92, in accordance with one or more techniques described herein. In this example, IMD 10 may use communication system 56 to communicate with external device 12 via afirst wireless connection, and to communicate with an access point 90 via a second wireless connection. In the example of FIG. 8, access point 90, external device 12, computing system 6, and computing devices 100 are interconnected and may communicate with each other through network 92. Network 92 may comprise a local area network, wide area network, or global network, such as the Internet. One or more aspects of the illustrated system of FIG. 8 may be implemented with general network technology and functionality, which may be similar to that provided by the Medtronic CareLink® Network.

[0122] Access point 90 may include a device that connects to network 92 via any of a variety of connections, such as telephone dial-up, digital subscriber line (DSL), or cable modem connections. In other examples, access point 90 may be coupled to network 92 through different forms of connections, including wired or wireless connections. In some examples, access point 90 may be a user device, such as a tablet or smartphone that may be co-located with the patient. IMD 10 may be configured to transmit data, such as patient physiological data, to access point 90. Access point 90 may then communicate the retrieved data to computing system 6 via network 92.

[0123] In some examples, computing system 6 may be configured to provide a secure storage site for data that has been collected from IMD 10 and / or external device 12. In some examples, computing system 6 may assemble data in web pages or other documents for viewing by trained professionals, such as clinicians, via computing devices 100 and / or external device 12.

[0124] In some examples, external device 12 may be operated by a patient (such as patient 4 of FIG. 1) or a clinician. External device 12 may be a computing device with a display viewable by a user (e.g., patient 4, a physician, technician, surgeon, electrophysiologist, clinician) and an interface for providing input to external device 12 (i.e., a user input mechanism). In some examples, external device 12 may be a notebook computer, tablet computer, computer workstation, one or more servers, cellular phone, personal digital assistant, handheld computing device, networked computing device, or another computing device that may run an application that enables the computing device to interact with IMD 10 and / or computing system 6. External device 12 may communicate with one or more other devices or systems shown in FIG. 8 via near-field communication technologies (e.g., inductive coupling, NFC or other communication technologies operable at ranges less than 10-20 cm) and far-field communication technologies (e.g., radiofrequency (RF) telemetry according to the 802.11 or Bluetooth® specification sets, or other communication technologies operable at ranges greater than near-field communication technologies).

[0125] In some examples, external device 12 may be a personal computing device (e.g., mobile phone) located with a patient, and one or more of computing devices 100 may be a tablet or other smart device located with a clinician, by which the clinician may program, receive data from, and / or interrogate IMD 10. Although not illustrated in FIG. 8, each of computing devices 100 may include a storage device and processing circuitry similar if not substantially similar to the storage devices and processing circuitries described herein. Furthermore, one or more of computing devices 100 may operate similarly to external device 12 as described with respect to FIGS. 1 and 5. For example, a user of one or more of computing devices 100 may query a report of data evaluated by computing system 6. User interface module 86 may generate data indicative of a user interface for display on one or more of computing devices 100, and computing system 6 may send, to one or more of computing devices 100, the data indicative of the user interface. A user, such as a clinician, may then interact with one or more of computing devices 100 through the user interface to view or analyze the report of data. Specifically, the report or information presented on computing devices 100 may include patient physiological data, data indicative of a cardiac morphology template match, an indication of a cardiac event, prediction values of potential classes (e.g., types) of cardiac events, a classification of the patient physiological data as indicative of a cardiac event type based on the prediction values, indications of changes in patient health that correlated to the classification, and / or treatment or therapy recommendations. In addition, user interface module 86 may be configured to receive input from a user of one or more of computing devices 100, such as, for example, feedback related to the accuracy of detections and / or predictions, updates to patient information, changes to IMD 10 settings / parameters, etc.

[0126] External device 12 and one or more of computing devices 100 can additionally or alternatively include a peripheral pointing device, such as a mouse, via which a user or clinician may interact with a user interface (e.g., a user interface generated based on data received from user interface module 86 of computing system 6). In some examples, a display of external device 12 and one or more of computing devices 100 may include a touch screen display, and a user or clinician may interact with external device 12 and one or more of computing devices 100 via the display. It should be noted that the user or clinician may also interact with external device 12 and one or more of computing devices 100 remotely via a networked computing device. The input mechanisms may include, for example, any one or more of buttons, a keypad (e.g., an alphanumeric keypad), a peripheral pointing device, a touch screen, or another input mechanism that allows the user or clinician to navigate through user interfaces generated based on data received from computing system 6 and provide input. In other examples, user interface module 86of computing system 6 may also include audio circuitry for providing audible notifications, instructions or other sounds to external device 12 and one or more of computing devices 100, receiving voice commands from the user or clinician, or both.

[0127] As described herein, a clinician may access data collected by IMD 10 through a computing device 100, such as when patient 4 is in between clinician visits, to check on a status of a medical condition. In some examples, the clinician may enter instructions for a medical intervention for patient 4 into an application executed by computing device 100, such as based on patient data known to the clinician. Device 100 then may transmit the instructions for medical intervention to external device 12 location with patient 4, another of computing devices 100 located with another clinician, or another of computing devices 100 located with a caregiver of patient 4. For example, such instructions for medical intervention may include an instruction to change a drug dosage, timing, or selection, to schedule a visit with the clinician, or to seek medical attention. In further examples, a computing device 100 may generate an alert to external device 12 operated by patient 4 based on a status of a medical condition of patient 4, which may enable patient 4 proactively to seek medical attention prior to receiving instructions for a medical intervention. In this manner, patient 4 may be empowered to act, as needed, to address his or her medical status, which may help improve clinical outcomes for patient 4.

[0128] FIG. 9 is a flow diagram illustrating an example operation for determining a cardiac event, in accordance with one or more examples of the present disclosure. As described above in relation to FIG. 2, the processing circuitry in the description of FIG. 9 may refer to processing circuitry in one or more of the components of a system of this disclosure in some examples. In other examples, one or more steps in the blocks of FIG. 9, or portions of a block of FIG. 9, may be distributed among several components. The description of FIG. 9 may focus on the example of FIG. 2 to simplify the explanation, however, any of the medical systems configured to sense cardiac activity of a patient, and described above in relation to FIGS. 1 -8 may perform the steps of FIG. 9.

[0129] In the example of FIG. 9, processing circuitry 50 of IMD 10 receives, from computing system 6, a set of cardiac morphology templates based on historical patient physiological data (200). In some examples, IMD includes sensors 62 configured to detect at least one physiological parameter corresponding to cardiac physiology of patient 4 and sensing circuitry 52 configured to generate patient physiological data based on the at least one physiological parameter. Processing circuitry 50 applies detection logic 66 to the generated patient physiological data to determine at least one cardiac event (202). In some examples, the patient physiological data compriseselectrocardiogram data for patient 4. In some examples, sensors 62 comprise one or more electrodes 16A-16N for sensing electrical activity of a heart of patient 4, in which the generated patient physiological data comprises recorded digitized signals representing the electrical activity. In some examples, processing circuitry 50 is further configured to identify one or more waveforms from the generated patient physiological data, apply detection logic 66 to the one or more waveforms, and based on one or more prediction values generated by detection logic 66, determine whether the one or more waveforms indicate the at least one cardiac event. In some examples, the at least one cardiac event includes one or more of premature ventricular contraction or nonsustained ventricular tachycardia. In some examples, detection logic 66 includes at least one of a cardiac rate or cardiac interval threshold. Responsive to determining the at least one cardiac event, processing circuitry 50 applies template matching 67 to the set of cardiac morphology templates and one or more segments of the generated patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match (204). In some examples, template matching 67 includes application of edge detection to compare the one or more segments of the patient physiological data to the set of cardiac morphology templates to determine the first instance of the cardiac morphology template match. In some examples, processing circuitry 50 is further configured to count each instance of a cardiac morphology template match. Processing circuitry transmits, to computing system 6, data indicative of the first instance of the cardiac morphology template match (206).

[0130] In some examples, computing system 6 is configured to apply machine learning module 88 to the data indicative of the first instance of the cardiac morphology template match. In some examples, machine learning module 88 includes a neural network configured to determine whether the at least one cardiac event is a true cardiac event. In some examples, the neural network is trained on data corresponding to one or more of at least one previous cardiac event, one or more cardiac morphology templates, the patient physiological data, and historical patient physiological data.

[0131] FIG. 10 is a flow diagram illustrating another example operation for determining a cardiac event, in accordance with one or more examples of the present disclosure. As described above in relation to FIG. 5, the processing circuitry in the description of FIG. 10 may refer to processing circuitry in one or more of the components of a system of this disclosure in some examples. In other examples, one or more steps in the blocks of FIG. 10, or portions of a block of FIG. 10, may be distributed among several components. The description of FIG. 10 may focus on the example of FIG. 5 to simplify the explanation, however, any of the medical systems configuredto sense cardiac activity of a patient, and described above in relation to FIGS. 1 -8 may perform the steps of FIG. 10.

[0132] In the example of FIG. 10, processing circuitry 80 receives, from IMD 10, data indicative of a first instance of a cardiac morphology template match, in which the data indicative of the first instance of the cardiac morphology template match includes patient physiological data corresponding to at least one cardiac event (300). In some examples, the data indicative of the first instance of the cardiac morphology template match further includes one or more segments of the generated patient physiological data corresponding to the at least one cardiac event, one or more of a frequency of the at least one cardiac event, a number of heartbeats associated with the at least one cardiac event, a total number of heartbeats, a heartrate, or a time duration. Processing circuitry 80 applies machine learning module 88 to the data indicative of the first instance of the cardiac morphology template match (302).

[0133] In some examples, machine learning module 88 includes a neural network configured to determine whether the at least one cardiac event is a true cardiac event. In some examples, the neural network is trained on data corresponding to one or more of at least one previous cardiac event, one or more cardiac morphology templates, the patient physiological data, and historical patient physiological data. In some examples, the at least one cardiac event includes premature ventricular contraction and non-sustained ventricular tachycardia. In some examples, the neural network includes a convolutional UNet architecture, in which the neural network outputs a probability score for premature ventricular contraction and a probability score for non-sustained ventricular tachycardia. The probability score for premature ventricular contraction may be converted to a first binary mask for premature ventricular contraction, and the probability score for non-sustained ventricular tachycardia may be converted to a second binary mask for non-sustained ventricular tachycardia. The first binary mask and the second binary mask may be used to output a severity score. In some examples, the severity score is a function of non-sustained ventricular tachycardia presence and premature ventricular contraction presence (e.g., number and count of PVC morphologies). In some examples, the severity score is weighted based on at least one of a type of beat pattern, a duration of a beat pattern, frequency of non-sustained ventricular tachycardia occurrence, and frequency of premature ventricular contraction occurrence. Processing circuitry 80 determines, based on at least one output of machine learning module 88 (e.g., the severity score), whether the at least one cardiac event is a true cardiac event (304). In some examples, the output of machine learning module 88, e.g., the severity score, may represent a patient’s risk, e.g., of more lethal arrhythmias.

[0134] In some examples, machine learning module 88 may further include a recommender system or model configured to generate intelligent recommendations (e.g., a recommendation for an ICD or other therapy) based on patient data and / or any output from machine learning module 88 (e.g., a confirmation of the at least one cardiac event and / or the severity score). In some examples, computing system 6 may send output from machine learning module 88 and / or treatment recommendations to external device 12 and / or other devices in communication with computing system 6 via network 92, such that the information may be displayed to patient 4, a clinician, or other user for review. In some examples, external device 12 and / or the other devices may include a display which may present output from IMD 10 and / or computing device 6, such as information related to IMD 10 (e.g., patient physiological data, data indicative of a cardiac morphology template match, an indication of a cardiac event, prediction values of potential classes (e.g., types) of cardiac events, a classification of the patient physiological data as indicative of a cardiac event type based on the prediction values, indications of changes in patient health that are correlated to the classification) and / or output from machine learning module 88 (a confirmation of the at least one cardiac event, the severity score, and / or treatment or therapy recommendations). In some examples, IMD 10 may receive feedback related to the accuracy of detections and / or predictions, updates to patient information, an updated set of cardiac morphology templates, and / or changes to IMD 10 settings / parameters from computing device 6 and / or external device 12 that may be based on the output from IMD 10 and / or machine learning module 88.

[0135] FIG. 11 is a flow diagram illustrating another example operation for determining a cardiac event, in accordance with one or more examples of the present disclosure. As described above in relation to FIG. 8, the processing circuitry in the description of FIG. 11 may refer to processing circuitry in one or more of the components of a system of this disclosure in some examples. In other examples, one or more steps in the blocks of FIG. 11, or portions of a block of FIG. 11, may be distributed among several components. The description of FIG. 11 may focus on the example of FIG. 8 to simplify the explanation, however, any of the medical systems configured to sense cardiac activity of a patient, and described above in relation to FIGS. 1 -8 may perform the steps of FIG. 11.

[0136] In the example of FIG. 11, processing circuitry 50 of IMD 10 receives, from computing system 6 via access point 90 and / or external device 12 and network 92, a set of cardiac morphology templates based on historical patient physiological data (400). In some examples, IMD includes sensors 62 configured to detect at least one physiological parameter corresponding to cardiac physiology of patient 4 and sensing circuitry 52 configured to generate patientphysiological data based on the at least one physiological parameter. Processing circuitry 50 applies a first model (e.g., detection logic 66) to the generated patient physiological data to determine at least one cardiac event (402). Responsive to determining the at least one cardiac event, processing circuitry 50 sends, to external device 12, the generated patient physiological data corresponding to the at least one cardiac event (404). External device 12 applies a second model (e.g., template matching) to a set of cardiac morphology templates and one or more segments of the generated patient physiological data to determine a first instance of a cardiac morphology template match (406). In some examples, the second model includes application of edge detection to compare the one or more segments of the patient physiological data to the set of cardiac morphology templates to determine the first instance of the cardiac morphology template match. External device 12 transmits, to computing system 6 via network 92, data indicative of the first instance of the cardiac morphology template match (408). Computing system 6 applies a third model (e.g., machine learning module 88) to the data indicative of the first instance of the cardiac morphology template match (410). In some examples, machine learning module 88 includes a neural network configured to determine whether the at least one cardiac event is a true cardiac event. For example, in some examples, machine learning module 88 may be configured to perform the steps or functions attributed to machine learning module 88 with respect to FIG. 10.

[0137] The techniques described in this disclosure may be implemented, at least in part, in hardware, software, firmware, or any combination thereof. For example, various aspects of the techniques may be implemented within one or more microprocessors, DSPs, ASICs, FPGAs, or any other equivalent integrated or discrete logic QRS circuitry, as well as any combinations of such components, embodied in external devices, such as physician or patient programmers, stimulators, or other devices. The terms “processor” and “processing circuitry” may generally refer to any of the foregoing logic circuitry, alone or in combination with other logic circuitry, or any other equivalent circuitry, and alone or in combination with other digital or analog circuitry.

[0138] For aspects implemented in software, at least some of the functionality ascribed to the systems and devices described in this disclosure may be embodied as instructions on a computer- readable storage medium such as RAM, ROM, NVRAM, DRAM, SRAM, Flash memory, magnetic discs, optical discs, flash memories, or forms of EPROM or EEPROM. The instructions may be executed to support one or more aspects of the functionality described in this disclosure.

[0139] In addition, in some aspects, the functionality described herein may be provided within dedicated hardware and / or software components. Depiction of different features as components or units is intended to highlight different functional aspects and does not necessarily imply that suchcomponents or units must be realized by separate hardware or software components. Rather, functionality associated with one or more components or units may be performed by separate hardware or software components, or integrated within common or separate hardware or software components. Also, the techniques could be fully implemented in one or more circuits or logic elements. The techniques of this disclosure may be implemented in a wide variety of devices or apparatuses, including an IMD, an external programmer, a combination of an IMD and external programmer, an integrated circuit (IC) or a set of ICs, and / or discrete electrical circuitry, residing in an IMD and / or external programmer.

[0140] The techniques of this disclosure may also be understood based on the following examples.

[0141] Example 1. A medical system comprising: an implantable medical device comprising: a communication system configured for wireless communication; one or more sensors configured to detect at least one physiological parameter corresponding to cardiac physiology of a patient; sensing circuitry configured to generate patient physiological data based on the at least one physiological parameter; and processing circuitry operatively coupled to the communication system, the one or more sensors, and the sensing circuitry, wherein the processing circuitry is configured to: receive, from a computing system in communication with the implantable medical device via the communication system, a set of cardiac morphology templates based on historical patient physiological data; apply a first model to the generated patient physiological data to determine at least one cardiac event; responsive to determining the at least one cardiac event, apply a second model to the set of cardiac morphology templates and one or more segments of the generated patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match; and transmit, to the computing system, data indicative of the first instance of the cardiac morphology template match.

[0142] Example 2. The medical system of example 1, wherein the patient physiological data comprises electrocardiogram data for the patient.

[0143] Example 3. The medical system of any of examples 1-2, wherein the one or more sensors comprise one or more electrodes for sensing electrical activity of a heart of the patient, wherein the generated patient physiological data comprises recorded digitized signals representing the electrical activity, and wherein the processing circuitry is further configured to identify one or more waveforms from the generated patient physiological data, apply the first model to the one or more waveforms, and based on one or more prediction values generated by the first model, determine whether the one or more waveforms indicate the at least one cardiac event.

[0144] Example 4. The medical system of any of examples 1-3, wherein the at least one cardiac event includes one or more of premature ventricular contraction or non-sustained ventricular tachycardia.

[0145] Example 5. The medical system of any of examples 1-4, wherein the first model includes at least one of a cardiac rate or cardiac interval threshold.

[0146] Example 6. The medical system of any of examples 1-5, wherein the second model includes application of edge detection to compare the one or more segments of the patient physiological data to the set of cardiac morphology templates to determine the first instance of the cardiac morphology template match.

[0147] Example 7. The medical system of any of examples 1-6, wherein the data indicative of the first instance of the cardiac morphology template match includes one or more of the patient physiological data corresponding to the at least one cardiac event, a frequency of the at least one cardiac event, a number of heartbeats associated with the at least one cardiac event, a total number of heartbeats, a heartrate, or a time duration.

[0148] Example 8. The medical system of any of examples 1-7, wherein the processing circuitry is further configured to count each instance of a cardiac morphology template match.

[0149] Example 9. The medical system of any of examples 1-8, further comprising the computing system, wherein the computing system is configured to apply third model to the data indicative of the first instance of the cardiac morphology template match, and wherein the third model is a neural network configured to determine whether the at least one cardiac event is a true cardiac event, and wherein the third model is trained on data corresponding to one or more of at least one previous cardiac event, one or more cardiac morphology templates, the patient physiological data, and historical patient physiological data.

[0150] Example 10. The medical system of example 9, wherein the at least one cardiac event includes premature ventricular contraction and non-sustained ventricular tachycardia, wherein the neural network includes a convolutional UNet architecture, wherein the neural network outputs a probability score for premature ventricular contraction and a probability score for nonsustained ventricular tachycardia, wherein the probability score for premature ventricular contraction is converted to a first binary mask for premature ventricular contraction, wherein the probability score for non-sustained ventricular tachycardia is converted to a second binary mask for non-sustained ventricular tachycardia, and wherein the first binary mask and the second binary mask are used to output a severity score.

[0151] Example 11. The medical system of example 10, wherein the severity score is a function of non-sustained ventricular tachycardia presence and premature ventricular contraction presence, and wherein the severity score is weighted based on at least one of a type of beat pattern, a duration of a beat pattern, frequency of non-sustained ventricular tachycardia occurrence, and frequency of premature ventricular contraction occurrence.

[0152] Example 12. The medical system of any of examples 1-11, wherein implantable medical device is an insertable cardiac monitor comprising: a housing configured for subcutaneous implantation in the patient, the housing having a length between 40 millimeters (mm) and 60 mm between a first end and a second end, a width less than the length, and a depth less than the width; a first electrode at or proximate to the first end; a second electrode at or proximate to the second end; and circuitry within the housing and configured to sense an electrocardiogram corresponding to patient physiological data via the first electrode and the second electrode and detect the at least one cardiac event based on the electrocardiogram.

[0153] Example 13. A method comprising: receiving, by an implantable medical device and from a computing system, a set of cardiac morphology templates based on historical patient physiological data; applying, by the implantable medical device, a first model to generated patient physiological data to determine at least one cardiac event; responsive to determining the at least one cardiac event, applying, by the implantable medical device, a second model to the set of cardiac morphology templates and one or more segments of the generated patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match; and transmitting, by the implantable medical device and to the computing system, data indicative of the first instance of the cardiac morphology template match.

[0154] Example 14. The method of example 13, wherein the generated patient physiological data is based on at least one physiological parameter corresponding to cardiac physiology of a patient.

[0155] Example 15. The method of any of examples 13-14, wherein the implantable medical device comprises one or more sensors configured to detect the at least one physiological parameter, wherein the one or more sensors comprise one or more electrodes for sensing electrical activity of a heart of the patient, wherein the generated patient physiological data comprises recorded digitized signals representing the electrical activity, and wherein the method further comprises: identifying, by the implantable medical device, one or more waveforms from the generated patient physiological data; applying, by the implantable medical device, the first model to the one or more waveforms; and determining, by the implantable medical device and based onone or more prediction values generated by the first model, whether the one or more waveforms indicate the at least one cardiac event.

[0156] Example 16. The method of any of examples 13-15, wherein the patient physiological data comprises electrocardiogram data for a patient.

[0157] Example 17. The method of any of examples 13-16, wherein the at least one cardiac event includes one or more of premature ventricular contraction or non-sustained ventricular tachycardia.

[0158] Example 18. The method of any of examples 13-17, wherein the first model includes at least one of a cardiac rate or cardiac interval threshold.

[0159] Example 19. The method of any of examples 13-18, wherein the second model includes application of edge detection to compare the one or more segments of the patient physiological data to the set of cardiac morphology templates to determine the first instance of the cardiac morphology template match.

[0160] Example 20. The method of any of examples 13-19, wherein the data indicative of the first instance of the cardiac morphology template match includes one or more of the patient physiological data corresponding to the at least one cardiac event, a frequency of the at least one cardiac event, a number of heartbeats associated with the at least one cardiac event, a total number of heartbeats, a heartrate, or a time duration.

[0161] Example 21. The method of any of examples 13-20, wherein the processing circuitry is further configured to count each instance of a cardiac morphology template match.

[0162] Example 22. A method comprising: receiving, by a computing system and from an implantable medical device, data indicative of a first instance of a cardiac morphology template match, wherein the data indicative of the first instance of the cardiac morphology template match includes patient physiological data corresponding to at least one cardiac event; applying, by the computing system, a model to the data indicative of the first instance of the cardiac morphology template match; and determining, by the computing system and based on at least one output of the model, whether the at least one cardiac event is a true cardiac event.

[0163] Example 23. The method of example 22, wherein the data indicative of the first instance of the cardiac morphology template match further includes one or more of a frequency of the at least one cardiac event, a number of heartbeats associated with the at least one cardiac event, a total number of heartbeats, a heartrate, or a time duration.

[0164] Example 24. The method of any of examples 22-23, wherein the model is a neural network configured to determine whether the at least one cardiac event is a true cardiac event, andwherein the model is trained on data corresponding to one or more of at least one previous cardiac event, one or more cardiac morphology templates, the patient physiological data, and historical patient physiological data.

[0165] Example 25. The method of any of examples 22-24, wherein the at least one cardiac event includes premature ventricular contraction and non-sustained ventricular tachycardia, wherein the neural network includes a convolutional UNet architecture, wherein the neural network outputs a probability score for premature ventricular contraction and a probability score for nonsustained ventricular tachycardia, wherein the probability score for premature ventricular contraction is converted to a first binary mask for premature ventricular contraction, wherein the probability score for non-sustained ventricular tachycardia is converted to a second binary mask for non-sustained ventricular tachycardia, and wherein the first binary mask and the second binary mask are used to output a severity score.

[0166] Example 26. The method of any of examples 22-25, wherein the severity score is a function of non-sustained ventricular tachycardia presence and premature ventricular contraction presence, and wherein the severity score is weighted based on at least one of a type of beat pattern, a duration of a beat pattern, frequency of non-sustained ventricular tachycardia occurrence, and frequency of premature ventricular contraction occurrence.

[0167] Example 27. A non-transitory computer-readable storage medium comprising instructions that, when executed, cause processing circuitry of an implantable medical device to: receive, from a computing system, a set of cardiac morphology templates based on historical patient physiological data; apply a first model to generated patient physiological data to determine at least one cardiac event; responsive to determining the at least one cardiac event, apply a second model to the set of cardiac morphology templates and one or more segments of the generated patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match; and transmit, to the computing system, data indicative of the first instance of the cardiac morphology template match.

[0168] Example 28. A non-transitory computer-readable storage medium comprising instructions that, when executed, cause processing circuitry of a computing system to: receive, from an implantable medical device, data indicative of a first instance of a cardiac morphology template match, wherein the data indicative of the first instance of the cardiac morphology template match includes patient physiological data corresponding to at least one cardiac event; apply a model to the data indicative of the first instance of the cardiac morphology template match;and determine, based on at least one output of the model, whether the at least one cardiac event is a true cardiac event.

[0169] Furthermore, although described primarily with reference to examples that provide an infection status to indicate a device pocket infection in response to detecting temperature changes in the device pocket, other examples may additionally or alternatively automatically modify a therapy in response to detecting the infection status in the patient. The therapy may be, as examples, a substance delivered by an implantable pump, a delivery of antibiotics, etc. These and other examples are within the scope of the following claims.

Claims

WHAT IS CLAIMED IS:

1. A medical system comprising: an implantable medical device comprising: a communication system configured for wireless communication; one or more sensors configured to detect at least one physiological parameter corresponding to cardiac physiology of a patient; sensing circuitry configured to generate patient physiological data based on the at least one physiological parameter; and processing circuitry operatively coupled to the communication system, the one or more sensors, and the sensing circuitry, wherein the processing circuitry is configured to: receive, from a computing system in communication with the implantable medical device via the communication system, a set of cardiac morphology templates based on historical patient physiological data; apply a first model to the generated patient physiological data to determine at least one cardiac event; responsive to determining the at least one cardiac event, apply a second model to the set of cardiac morphology templates and one or more segments of the generated patient physiological data corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match; and transmit, to the computing system, data indicative of the first instance of the cardiac morphology template match.

2. The medical system of claim 1, wherein the patient physiological data comprises electrocardiogram data for the patient.

3. The medical system of claim 1 or claim 2, wherein the one or more sensors comprise one or more electrodes for sensing electrical activity of a heart of the patient, wherein the generated patient physiological data comprises recorded digitized signals representing the electrical activity, and wherein the processing circuitry is further configured to identify one or more waveforms from the generated patient physiological data, apply the first model to the one or more waveforms,and based on one or more prediction values generated by the first model, determine whether the one or more waveforms indicate the at least one cardiac event.

4. The medical system of any of claims 1 to 3, wherein the at least one cardiac event includes one or more of premature ventricular contraction or non-sustained ventricular tachycardia.

5. The medical system of any of claims 1 to 4, wherein the first model includes at least one of a cardiac rate or cardiac interval threshold.

6. The medical system of any of claims 1 to 5, wherein the second model includes application of edge detection to compare the one or more segments of the generated patient physiological data to the set of cardiac morphology templates to determine the first instance of the cardiac morphology template match.

7. The medical system of any of claims 1 to 6, wherein the data indicative of the first instance of the cardiac morphology template match includes one or more of the one or more segments of the generated physiological data corresponding to the at least one cardiac event, a frequency of the at least one cardiac event, a number of heartbeats associated with the at least one cardiac event, a total number of heartbeats, a heartrate, or a time duration.

8. The medical system of any of claims 1 to 7, wherein the processing circuitry is further configured to count each instance of a cardiac morphology template match.

9. The medical system of any of claims 1 to 8, further comprising the computing system, wherein the computing system is configured to apply third model to the data indicative of the first instance of the cardiac morphology template match, and wherein the third model is a neural network configured to determine whether the at least one cardiac event is a true cardiac event, and wherein the third model is trained on data corresponding to one or more of at least one previous cardiac event, one or more cardiac morphology templates, the patient physiological data, and historical patient physiological data.

10. The medical system of claim 9, wherein the at least one cardiac event includes premature ventricular contraction and non-sustained ventricular tachycardia, wherein the neural networkincludes a convolutional UNet architecture, wherein the neural network outputs a probability score for premature ventricular contraction and a probability score for non-sustained ventricular tachycardia, wherein the probability score for premature ventricular contraction is converted to a first binary mask for premature ventricular contraction, wherein the probability score for nonsustained ventricular tachycardia is converted to a second binary mask for non-sustained ventricular tachycardia, and wherein the first binary mask and the second binary mask are used to output a severity score.

11. The medical system of claim 10, wherein the severity score is a function of non-sustained ventricular tachycardia presence and premature ventricular contraction presence, and wherein the severity score is weighted based on at least one of a type of beat pattern, a duration of a beat pattern, frequency of non-sustained ventricular tachycardia occurrence, and frequency of premature ventricular contraction occurrence.

12. The medical system of any of claims 1 to 11, wherein implantable medical device is an insertable cardiac monitor comprising: a housing configured for subcutaneous implantation in the patient, the housing having a length between 40 millimeters (mm) and 60 mm between a first end and a second end, a width less than the length, and a depth less than the width; a first electrode at or proximate to the first end; a second electrode at or proximate to the second end; and circuitry within the housing and configured to sense an electrocardiogram corresponding to patient physiological data via the first electrode and the second electrode and detect the at least one cardiac event based on the electrocardiogram.

13. A non-transitory computer-readable storage medium comprising instructions that, when executed, cause processing circuitry of an implantable medical device to: receive, from a computing system, a set of cardiac morphology templates based on historical patient physiological data; apply a first model to generated patient physiological data to determine at least one cardiac event; responsive to determining the at least one cardiac event, apply a second model to the set of cardiac morphology templates and one or more segments of the generated patient physiologicaldata corresponding to the at least one cardiac event to determine a first instance of a cardiac morphology template match; and transmit, to the computing system, data indicative of the first instance of the cardiac morphology template match.

14. A non-transitory computer-readable storage medium comprising instructions that, when executed, cause processing circuitry of a computing system to: receive, from an implantable medical device, data indicative of a first instance of a cardiac morphology template match, wherein the data indicative of the first instance of the cardiac morphology template match includes patient physiological data corresponding to at least one cardiac event; apply a model to the data indicative of the first instance of the cardiac morphology template match; and determine, based on at least one output of the model, whether the at least one cardiac event is a true cardiac event.

Images