Electrocardiogram-based monitoring of left ventricular dysfunction and ejection fraction

Abstract

The medical device system includes a medical device including one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes, a memory configured to store a machine learning model and multiple sets of training data, and a processing circuit in communication with the memory. The processing circuit is configured to apply the machine learning model to the electrical cardiac data to determine values ​​of metrics of left ventricular (LV) dysfunction. The machine learning model is trained based on the multiple sets of training data. Each set of training data in the multiple sets of training data includes a set of training electrical cardiac data and information indicative of values ​​of one or more metrics of LV dysfunction corresponding to the set of training electrical cardiac data.

Description

[Technical Field]

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 373,865, filed August 30, 2022, the entire contents of which are incorporated herein by reference.

[0002] FIELD OF THE DISCLOSURE The present disclosure relates to medical devices, and more particularly to systems including medical devices and methods of operating such systems to monitor a patient's condition. [Background technology]

[0003] Some types of medical devices can be used to monitor one or more physiological parameters of a patient, such as physiological parameters associated with cardiac function. Such medical devices may include, or be part of, a system that includes a sensor that detects signals associated with such physiological parameters, e.g., heart rate parameters. Values ​​determined based on such signals can be used to detect changes in a medical condition, assess the effectiveness of a therapy, or assist in generally assessing the patient's health.

[0004] A medical device that monitors a physiological parameter related to a patient's medical condition can evaluate a value associated with the physiological parameter, such as determining whether the value exceeds a threshold or has changed over time. A value that exceeds a threshold or has changed can indicate that the patient's medical condition has changed, e.g., worsened, or that a therapy being administered to the patient is not effectively managing the patient's medical condition. Summary of the Invention

[0005] Generally, the present disclosure is directed to techniques for determining metrics of left ventricular (LV) dysfunction, such as LV ejection fraction (LVEF), based on an electrocardiogram (ECG) or electrogram (EGM) sensed by a device such as an insertable cardiac monitor (ICM) or another implantable medical device (IMD). LV dysfunction can be associated with heart failure (HF). In some examples, processing circuitry of a system implementing the techniques of the present disclosure can further determine HF status, e.g., the likelihood of future worsening HF, based on the LV dysfunction metric.

[0006] In accordance with the techniques of the present disclosure, an insertable cardiac monitor (ICM) or other implantable medical device (IMD) can collect cardiac signals indicative of electrical cardiac data, such as ECG data or EGM data. The cardiac signals can include information relevant to monitoring various patient conditions. For example, the cardiac signals may include criteria indicative of events in the cardiac cycle. For example, P waves indicate atrial depolarization, R waves indicate ventricular depolarization, and T waves indicate ventricular repolarization. The system can analyze the cardiac signals to identify these criteria or monitor one or more patient conditions based on the identified criteria and / or other information. For example, the system can classify the cardiac signals as indicative of one or more patient conditions, such as atrial fibrillation (AF).

[0007] The techniques of the present disclosure may be implemented by a system including one or more IMDs that can autonomously and continuously collect biometric data while the one or more IMDs are implanted in a patient for months or years, store the biometric data, and perform numerous operations per second on the data to enable the system herein to analyze the biometric data. Using the techniques of the present disclosure with IMDs may be advantageous when a physician cannot be continuously present with the patient to assess physiological parameters and / or when performing the operations on the data described herein cannot be done in the physician's mind.

[0008] In some examples, the techniques and systems of the present disclosure can use one or more models, such as probability and / or machine learning models, to more accurately analyze biometric data to identify metric values, monitor patient conditions, or predict future status of a patient's condition. In some examples, the model is trained using a set of training instances, one or more of which include data indicating a relationship between biometric data and likelihood values. Because the model is constructed based on potentially thousands or millions of training instances, the model can reduce the amount of error in recurrence predictions performed by the system. Reducing error using the techniques of the present disclosure can provide one or more technical and clinical advantages, such as improving the likelihood that a system will accurately provide information based on analyzing biometric data.

[0009] A machine learning model may provide more accurate output when the model is trained using a larger set of training data compared to when the model is trained using a smaller set of training data. This is because a larger training dataset provides more training instances than a smaller training dataset, meaning that a model trained using a larger training dataset will generate output based on a greater amount of knowledge compared to a model trained using a smaller training dataset. When a machine learning model is trained using supervised learning, the training data is labeled so that the model learns associations between data features and labels. This means that the accuracy of a machine learning model trained using supervised learning may be limited if the amount of available labeled training data is limited.

[0010] In some examples, the health monitoring system can apply a machine learning model to electrical cardiac data generated by the medical device based on cardiac signals sensed from the patient by the medical device to determine a value of the LV dysfunction metric. When the health monitoring system applies the machine learning model to the electrical cardiac data, the machine learning model can process the electrical cardiac data to determine a value of the LV dysfunction metric. The machine learning model, in some examples, can output the determined value of the LV dysfunction metric. In some examples, the machine learning model can output a confidence that the LV dysfunction metric is less than a threshold LV dysfunction metric and / or a confidence that the LV dysfunction metric is not less than a threshold LV dysfunction metric.

[0011] One or more characteristics of the set of electrical cardiac data can indicate a value of a metric of LV dysfunction corresponding to the set of electrical cardiac data. For example, the electrical cardiac data can indicate characteristics corresponding to heart rate, heart rate variability, one or more arrhythmias, R waves, P waves, T waves, or other cardiac events, or any combination thereof. By applying a machine learning model to the electrical cardiac data including these characteristics, the health monitoring system can be configured to cause the machine learning model to output a value of the metric of LV dysfunction. It can be beneficial to use a machine learning model to generate a value of the metric of LV dysfunction based on the electrical cardiac data, such that the health monitoring system can determine the metric of LV dysfunction based on the electrical cardiac data collected by the medical device.

[0012] In some examples, the machine learning model applied by the health monitoring system may be trained using multiple sets of training data. Each set of training data among the multiple sets of training data may include a set of training electrical cardiac data and information indicative of one or more values ​​of LV dysfunction metrics corresponding to the set of training electrical cardiac data. That is, each value among the one or more values ​​of LV dysfunction metrics indicated by the set of training data may be associated with a time relative to the set of training electrical cardiac data. This means that the set of training data may be labeled with one or more values ​​of the LV dysfunction metrics at a time relative to a characteristic of the training electrical cardiac data. When the machine learning model is trained, the model may learn to recognize patterns corresponding to the characteristic of the electrical cardiac data and the LV dysfunction metrics.

[0013] The techniques of the present disclosure may provide one or more advantages. For example, by using a machine learning model trained on a set of training data including both electrical cardiac data and one or more values ​​of the LV dysfunction metric corresponding to the electrical cardiac data to determine the value of the LV dysfunction metric, the system can more accurately determine the value of the LV dysfunction metric compared to a system that does not use a machine learning model trained on electrical cardiac data labeled with the values ​​of the LV dysfunction metric. The machine learning model may also be evaluated on a patient-by-patient basis. For example, the system can determine whether the machine learning model performs adequately in determining the value of the LV dysfunction metric based on the electrical cardiac data collected from a particular patient and update the model if performance is inadequate. As a result, the machine learning model may perform better for individual patients compared to a system that does not update the machine learning model based on patient-by-patient performance.

[0014] In one example, a medical device system includes a medical device including one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes, a memory configured to store a machine learning model and multiple sets of training data, and a processing circuit in communication with the memory. The processing circuit is configured to apply the machine learning model to the electrical cardiac data to determine values ​​of a left ventricular (LV) dysfunction metric, the machine learning model being trained based on the multiple sets of training data, each set of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the LV dysfunction metric corresponding to the set of training electrical cardiac data. The processing circuit is also configured to output the determined values ​​of the LV dysfunction metric to a computing device.

[0015] In another example, a method of operating a medical device system including a medical device comprising one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes includes applying, by a processing circuit of the medical device system, a machine learning model to the electrical cardiac data to determine values ​​of a metric of left ventricular (LV) dysfunction, the machine learning model being trained based on a plurality of sets of training data, each set of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the metric of LV dysfunction corresponding to the set of training electrical cardiac data, the processing circuit being in communication with a memory of the medical device system, the memory being configured to store the machine learning model and the plurality of sets of training data. The method also includes outputting, by the processing circuit, the determined values ​​of the metric of LV dysfunction to a computing device.

[0016] In another example, a non-transitory computer-readable storage medium includes program instructions that, when executed by a processing circuit of a medical device system including one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes, cause the processing circuit to apply a machine learning model to the electrical cardiac data to determine values ​​of a metric of left ventricular (LV) dysfunction, the machine learning model being trained based on a plurality of sets of training data, each set of training data of the plurality of sets of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the metric of LV dysfunction corresponding to the set of training electrical cardiac data, the processing circuit being in communication with a memory of the medical device system, the memory being configured to store the machine learning model and the plurality of sets of training data. The program instructions also cause the processing circuit to output the determined values ​​of the metric of LV dysfunction to the computing device.

[0017] This summary is intended to provide an overview of the subject matter described in the present disclosure. It is not intended to provide an exclusive or comprehensive description of the systems, devices, and methods described in detail in the accompanying drawings and the following description. Further details of one or more examples of the present disclosure are set forth in the accompanying drawings and the following description. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a block diagram illustrating an example medical device system including an implantable medical device (IMD) in conjunction with a patient, in accordance with one or more techniques of the present disclosure.

[0019] [Figure 2A] 2 is a perspective view illustrating an IMD that may be an exemplary configuration of the IMD of FIG. 1 as an insertable cardiac monitor (ICM), in accordance with one or more techniques of the present disclosure. FIG.

[0020] [Figure 2B] 2 is a perspective view illustrating another IMD, which may be another example configuration of the IMD from FIG. 1 as an ICM, in accordance with one or more techniques of the present disclosure.

[0021] [Figure 3] 2 is a block diagram illustrating an example configuration of the IMD of FIG. 1 in accordance with one or more techniques of the present disclosure.

[0022] [Figure 4] 2 is a block diagram illustrating an example configuration of a computing device that may correspond to either of the computing devices of FIG. 1 (or both operating in cooperation), in accordance with one or more techniques of this disclosure.

[0023] [Figure 5] FIG. 2 is a block diagram illustrating an operational aspect of the health monitoring system (HMS) of FIG. 1 in accordance with one or more techniques of the present disclosure.

[0024] [Figure 6] FIG. 1 is a conceptual diagram illustrating an example machine learning model configured to output information corresponding to electrical cardiac data, in accordance with one or more techniques of the present disclosure.

[0025] [Figure 7] FIG. 1 is a block diagram illustrating an example of a machine learning model that has been trained using supervised learning and / or reinforcement learning, in accordance with one or more techniques of this disclosure.

[0026] [Figure 8] 1 is a flow diagram illustrating an example method for determining whether to use a machine learning model to process electrical cardiac data collected from a patient, in accordance with one or more techniques of this disclosure.

[0027] [Figure 9] 1 is a flow diagram illustrating an example method for applying a machine learning model to determine values ​​of metrics of left ventricular (LV) dysfunction, in accordance with one or more techniques of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0028] A device with one or more physiological sensors includes at least electrodes for sensing at least one channel of electrical cardiac data, such as electrocardiogram (ECG) data and / or electrogram (EGM) data. Processing circuitry of the device, or another device in communication with the sensing device, determines a metric of left ventricular (LV) dysfunction, such as LV ejection fraction (LVEF), based on the electrical cardiac data. In some examples, the device may be an implantable medical device (IMD). The IMD may include a housing configured for subcutaneous implantation on which one or more electrodes are disposed, such as an insertable cardiac monitor (ICM). In some examples, the IMD may be a leadless IMD. In other examples, the device may be one or more other implantable or external devices. Examples of the one or more other implantable or external devices may include an implantable multi-channel cardiac pacemaker, an implantable cardioverter defibrillator (ICD), an implantable pulse generator (IPG), a leadless (e.g., intracardiac) pacemaker, an extravascular pacemaker and / or ICD, or other IMD or combination of such IMDs, an external monitor, a drug pump, or a smartwatch or other IMD with physiological monitoring capabilities.

[0029] In any such example, the processing circuitry may transmit metric values ​​of the patient's LV dysfunction to a remote computer, receive instructions for medical intervention from the remote computer based on the patient's heart failure status, and transmit the instructions for medical intervention to the user interface. Such instructions for medical intervention may include at least one of a change in medication selection, a change in medication dosage, instructions to schedule a clinician visit, or instructions for the patient to seek treatment. In this manner, the patient's diagnosis and / or treatment for a condition may be modified as needed between clinic visits, which may help avoid adverse medical events, such as recurrent symptoms, acute heart failure (HF), or hospitalization.

[0030] In some examples, the techniques described herein may enable identification of changes in a patient's status before the changes lead to a progression of the patient's condition or the development of one or more additional medical conditions. Thus, the techniques described herein may help enable a determination of the likelihood that a patient will experience an adverse medical event, which may help clinicians prescribe personalized treatments that help avoid hospitalizations, improve clinical outcomes, and / or reduce the economic burden on the healthcare system.

[0031] This disclosure describes various systems, devices, and techniques for analyzing biometric data collected by medical devices to generate information related to a patient's health. In some examples, based on analysis of electrical cardiac data such as an ECG or EGM, it may be possible to identify one or more patient conditions, such as HF or cardiac arrhythmia, or predict the future occurrence of one or more patient conditions. The system may analyze the cardiac data using a machine learning model to generate information related to the patient condition, such as one or more values ​​of metrics of LV dysfunction. In some cases, the system may train the machine learning model using a large set of training data, including one or both of an unlabeled training dataset and a labeled training dataset.

[0032] In machine learning, "labeled" training data refers to training data accompanied by information identifying one or more aspects of the data. For example, electrical cardiac data, such as ECG data and EGM data, can include labels identifying P waves, R waves, and T waves. In another example, the electrical cardiac data can include labels indicating parameters such as characteristics such as heart rate, heart rate variability, or arrhythmia. In another example, the electrical cardiac data can be labeled with one or more values ​​of a metric of LV dysfunction, such as LVEF. A label including the value of the metric of LV dysfunction can indicate the time of the value of the LV metric relative to one or more times in the electrical cardiac data. Other exemplary labels may include patient information such as age, gender, pre-existing conditions, or any combination thereof. "Unlabeled" training data refers to training data that does not include information identifying aspects of the data. An unlabeled electrical cardiac training dataset can include electrical cardiac data including features such as P waves, R waves, and T waves without labels identifying such features or any other characteristics of the dataset.

[0033] The system may be configured to train a machine learning model based on multiple sets of training data. Each set of training data may include a set of training electrocardiac data and information indicative of one or more values ​​of LV dysfunction metrics corresponding to the set of training electrocardiac data. That is, the set of training electrocardiac data may include one or more labels indicative of values ​​of the LV dysfunction metrics. The set of training electrocardiac data may include one or more labels indicative of characteristics of the electrocardiac data, but this is not required. The system may train a machine learning model based on the multiple sets of training data such that the machine learning model learns patterns corresponding to the electrocardiac data and the LV dysfunction metrics.

[0034] LVEF is a clinically actionable cardiac function metric that can be measured in a clinical setting using echocardiograms, cardiac magnetic resonance imaging (MRI), or other modalities. In some instances, it may be beneficial to derive a metric of LV dysfunction, such as LVEF, based on electrical cardiac data, such as ECG data or EGM data. This is because an IMD can continuously record electrical cardiac data from a patient over long periods of time, and deriving a metric of LV dysfunction, such as LVEF, based on the electrical cardiac data allows the system to track the metric of LV dysfunction over long periods of time without requiring the patient to frequently visit a clinic for echocardiograms and / or cardiac MRI measurements.

[0035] Because electrical cardiac data, such as ECG data and EGM data, have diagnostic relevance for indicating LV dysfunction metrics, such as LVEF, it may be beneficial to monitor LV dysfunction metrics using electrical cardiac data. Using electrical cardiac data to monitor LV dysfunction metrics may enable a system to achieve continuous monitoring of LV dysfunction metrics over longer periods of time compared to systems using methods such as echocardiograms and / or MRI. Because electrical cardiac data may be easier to record over long periods of time by IMD compared to echocardiograms and / or MRI data, monitoring LV dysfunction metrics using electrical cardiac data may enable a system to more accurately monitor LV dysfunction metrics over longer periods of time in real time and / or via cloud-based artificial intelligence (AI) compared to systems that measure LV dysfunction metrics using echocardiograms and / or MRI.

[0036] FIG. 1 is a block diagram illustrating an exemplary medical device system 2 including an IMD 10 associated with a patient 4, in accordance with one or more techniques of the present disclosure. The IMD 10 is configured for continuous, longitudinal monitoring of the heart of the patient 4. For example, the IMD 10 is configured to sense cardiac signals indicative of electrical cardiac data, such as ECG data, EGM data, or other electrical cardiac data; identify criteria within the cardiac signals; identify one or more parameter values ​​(e.g., values ​​of metrics of LV dysfunction) based on the cardiac signals; detect a patient condition or arrhythmia (e.g., HF, ventricular fibrillation (VF), atrial fibrillation (AF), atrioventricular (AV) block) based on the cardiac signals; store the cardiac signals and information corresponding to the cardiac signals in a memory of the IMD 10; or any combination thereof. In some examples, the IMD 10 is configured to automatically sense cardiac signals continuously or intermittently according to a predetermined schedule. In some examples, the IMD 10 is configured to sense cardiac signals based on user input. For example, the IMD 10 can sense patient signals based on user input indicative of symptoms of a patient condition.

[0037] By way of example, the IMD 10 may be a pacemaker or implantable cardioverter-defibrillator, which may be coupled to intravascular or extravascular leads, or a pacemaker with a housing configured for implantation within the heart, which may be leadless. Some IMDs, such as implantable patient monitors, do not provide therapy. The IMD 10 may be such an IMD, for example, the Reveal LinQ™ or LinQ II™ insertable cardiac monitor (ICM), available from Medtronic, Inc. (Minneapolis, Minnesota), which may be inserted subcutaneously. Such IMDs may facilitate relatively long-term monitoring of a patient during normal daily activities and may periodically transmit collected data to a remote patient monitoring system, for example, the Medtronic CareLink™ network.

[0038] The IMD 10 can determine values ​​of patient parameters or metrics, for example, based on physiological signals sensed by the IMD or response therapies delivered by the IMD. Patient parameters can include, by way of example, fluid levels, heart rate, respiratory rate, patient activity, body temperature, heart sounds, oxygenation, and R-wave morphology. Other exemplary patient metrics include cough, speech, posture, tissue perfusion, hematocrit, thoracic impedance, subcutaneous impedance, intracardiac impedance, heart rate variability (HRV), weight, blood pressure, sleep apnea burden (which may be derived from respiratory rate), ischemic burden, sleep duration, sleep quality, pre-ventricular contraction (PVC) burden, incidence, metrics of LV dysfunction such as LVEF, and frequency or duration of cardiac arrhythmias or other events such as HF, VF, AF, or AV block, as well as sensed cardiac intervals (e.g., QT interval). Another exemplary patient metric is ventricular rate during AF. Concentrations or levels of various substances within the patient (eg, blood glucose, hematocrit, troponin and / or brain natriuretic peptide (BNP) levels) may also be used as one or more patient metrics.

[0039] The medical device system 2 may include one or more sensors (e.g., for sensing the activity state of the patient 4 and / or the cardiac function of the patient 4). The one or more sensors may collectively detect at least one first signal and at least one second signal that enable a processing circuit of the medical device system 2 to determine whether the activity state of the patient 4 meets at least one inactivity criterion and to determine the value of at least one LV dysfunction metric, such as LVEF, based on such signals. Although the processing circuit may be described herein as being a component of the IMD 10 for clarity, such processing circuit may be included within the IMD 10 and / or within another medical device of the medical device system 2, e.g., the external device 12.

[0040] In some examples, the one or more sensors may include one or more accelerometers or other sensors configured to detect at least one signal indicative of one or more aspects of the activity state of the patient 4, such as activity level, posture, and / or respiration rate. In some examples, the one or more sensors may include a plurality of electrodes that may be disposed on a housing of the IMD 10. The plurality of electrodes may be configured to detect electrical cardiac signals, such as an ECG and / or an EGM.

[0041] In addition to the IMD 10, the medical device system 2 includes one or more patient computing devices, such as patient computing devices 12A and 12B (collectively “patient computing devices 12”). The patient computing devices 12 are configured to wirelessly communicate with the IMD 10. The computing devices 12 can retrieve parameter data, ECG and / or EGM data, LV dysfunction metric data, and other data from the IMD 10. In some examples, the computing devices 12 take the form of the patient 4's personal computing devices. For example, the computing device 12A can take the form of the patient 4's smartphone, and the computing device 12B can take the form of the patient 4's smartwatch or other smart clothing. In some examples, the computing devices 12 can be any computing device configured to wirelessly communicate with the IMD 10, such as a desktop, laptop, or tablet computer. The computing devices 12 can communicate with the IMD 10 and each other according to, for example, Bluetooth® or Bluetooth® Low Energy (BLE) protocols. In some examples, only one of the computing devices 12, e.g., computing device 12A, is configured to communicate with the IMD 10, e.g., by executing software (e.g., part of a health monitoring application as described herein) that enables communication and interaction with the IMD.

[0042] 1, may include electrodes and other sensors for sensing physiological signals of patient 4 and may collect and store parameter data based on such signals. Computing device 12B may be incorporated into clothing of patient 4, such as clothing, shoes, glasses, a watch or wristband, a hat, etc. In some examples, computing device 12B is a smartwatch or other accessory or peripheral for smartphone computing device 12A.

[0043] One or more of computing devices 12 may be configured to communicate with various other devices or systems via network 16. For example, one or more of computing devices 12 may be configured to communicate with one or more computing systems, such as computing system 20, via network 16. Computing system 20 may be managed by the manufacturer of IMD 10 to provide, for example, cloud storage and analysis of collected data, maintenance and software services, or other networked functionality for their respective devices and their users. Computing system 20, in some examples, may include or be implemented by the Medtronic CareLink™ network.

[0044] As shown in FIG. 1 , computing system 20 may include processing circuitry 22 and memory 24. Processing circuitry 22 may include fixed-function circuitry and / or programmable processing circuitry. Processing circuitry 22 may include any one or more of a microprocessor, a controller, a digital signal processor (DSP), a graphics processing unit (DSP), a tensor processing unit (TPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or equivalent discrete or analog logic circuitry. In some examples, processing circuitry 22 may include multiple components, such as any combination of one or more microprocessors, one or more controllers, one or more DSPs, GPUs, TPUs, one or more ASICs, or one or more FPGAs, as well as other discrete or integrated logic circuits that may be physically located within one or more devices in one or more physical locations. Computing system 20 may be configured as a cloud computing system.

[0045] Processing circuitry 22 may be capable of processing instructions stored in memory 24. In some examples, memory 24 includes a computer-readable medium containing instructions that, when executed by processing circuitry 22, cause computing system 20 and processing circuitry 22 to perform various functions attributed thereto herein. In the example shown in FIG. 1 , computing system 20 implements a health monitoring system (HMS) 26. As described in more detail below, HMS 26 can generate information corresponding to one or more patient conditions of patient 4 based on physiological data collected by IMD 10. Memory 24 may include any volatile, nonvolatile, magnetic, optical, or electrical medium, such as random-access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically-erasable programmable ROM (EEPROM), ferroelectric RAM (FRAM), dynamic random-access memory (DRAM), flash memory, or any other digital medium.

[0046] Computing device 12 may transmit data, including data retrieved from IMD 10, to computing system 20 over network 16. The data may include sensed data, such as values ​​of physiological parameters measured by IMD 10 and possibly one or more of computing devices 12, as well as other physiological signals or data recorded by IMD 10 and / or computing device 12. Data transmitted from computing device 12 to computing system 20 may include electrical cardiac data, such as ECG data and / or EGM data.

[0047] Network 16 may include one or more computing devices, such as one or more non-edge switches, routers, hubs, gateways, security devices such as firewalls, intrusion detection and / or intrusion prevention devices, servers, cellular base stations and nodes, wireless access points, bridges, cable modems, application accelerators, or other network devices. Network 16 may include one or more networks managed by a service provider and thus form part of a larger public network infrastructure, such as the Internet. Network 16 may provide access to the Internet for computing devices and systems such as those shown in FIG. 1 and may provide a communications framework that allows computing devices and systems to communicate with each other. In some examples, network 16 may include a private network that provides a communications framework that allows computing devices and systems shown in FIG. 1 to communicate with each other but isolates some of the data flow from devices outside the private network for security purposes. In some examples, communications between computing devices and systems shown in FIG. 1 are encrypted.

[0048] As described herein, IMD 10 may be configured to generate diagnostic information for patient 4, such as information indicative of one or more values ​​of metrics of LV dysfunction. In some examples, IMD 10 may be configured to transmit such data to wireless access point 34 and / or computing device 12. Wireless access point 34 and / or computing device 12 may then communicate the retrieved data to computing system 20 via network 16.

[0049] In some cases, computing device 20 may be configured to provide a secure storage site for data collected from IMD 10 and / or computing device 12. In some cases, computing system 20 may include a database that stores medical- and health-related data. For example, computing system 20 may include a cloud server or other remote server that stores data collected from IMD 10 and / or computing device 12. In some cases, computing device 20 may assemble data in web pages or other documents for viewing by trained professionals, such as clinician 40, via clinician computing device 38. One or more aspects of the exemplary system described with reference to FIG. 1 may be implemented using common network technologies and capabilities, which may be similar to those provided by the Medtronic CareLink® network.

[0050] In some examples, one or more of the clinician computing devices 38 may be a tablet or other smart device at the clinician's location, enabling the clinician to program, receive alerts from, and / or interrogate the IMD 10. For example, the clinician may access data collected by the IMD 10 via the clinician computing device 38 to check the status of a medical condition, such as when the patient 4 is between clinical visits. For example, the computing system 20 may transmit, via the clinician computing device 38, data indicative of one or more sets of electrical cardiac data and / or one or more values ​​of metrics of LV dysfunction (e.g., determined by the computing system 20, the computing device 12, or other devices described herein) to the clinician 40.

[0051] As described herein, processing circuitry of one or more devices of medical device system 2 (e.g., computing device 12 and / or computing system 20 implementing HMS 26) may be configured to determine information corresponding to the electrical cardiac data (e.g., ECG data and / or EGM data) generated by IMD 10. In some examples, HMS 26 may include a machine learning model configured to generate an output based on receiving as input the electrical cardiac data generated by IMD 10. The machine learning model may, in some examples, generate an output including information corresponding to the health of patient 4. For example, the machine learning model may output one or more values ​​of a metric of LV dysfunction corresponding to the electrical cardiac data generated by IMD 10. In some cases, the machine learning model may output information indicating that patient 4 is likely to experience a certain level of the metric of LV dysfunction in the future. In some examples, the machine learning model may be stored by a memory of IMD 10, although this is not required. The machine learning model may also be stored by a device separate from IMD 10.

[0052] In some examples, computing system 20 may be configured to determine long-term trends of LV dysfunction metrics on a patient-specific basis. For example, computing system 20 may determine historical estimates of LV dysfunction metrics based on values ​​of LV dysfunction metrics determined based on electrical cardiac data collected by IMD 10 and actual values ​​of LV dysfunction metrics measured via echocardiography and / or cardiac MRI. For example, computing system 20 may be configured to determine long-term trends of LV dysfunction metrics based on values ​​of LV dysfunction determined based on electrical cardiac data collected by IMD 10 and actual values ​​of LV dysfunction metrics on a patient-specific basis using data corresponding to a given patient.

[0053] 1, the environment 28 includes one or more Internet of Things (IoT) devices, such as IoT devices 30. The IoT devices 30 may include, by way of example, so-called "smart" speakers, cameras, televisions, lights, locks, thermostats, appliances, actuators, controllers, or any other smart home (or building) devices. In the embodiment of FIG. 1, the IoT devices 30 are smart speakers and / or controllers, which may include displays. In some examples, the IoT devices 30 include cameras or other sensors and may activate the sensors to collect data about the patient 4, for example, to assess the condition of the patient 4.

[0054] The computing device 12 may be configured to communicate wirelessly with the IoT device 30 to cause the IoT device 30 to take actions described herein. In some examples, the HMS 26 communicates with the IoT device 30 over the network 16 to cause the IoT device 30 to take actions described herein. In some examples, the IMD 10 is configured to communicate wirelessly with the IoT device 30, for example, to communicate data to the computing system 20 over the network 16. In some examples, the IoT device 30 may be configured to provide some or all of the functionality attributed to the computing device 12 herein.

[0055] Environment 28 includes a computing facility, e.g., a local network 32, through which IMD 10, computing device 12, IoT device 30, and other devices within environment 28 may communicate with, e.g., HMS 26, via network 16. For example, environment 28 may be configured with wireless technologies such as Institute of Electrical and Electronics Engineers (IEEE) 802.11 wireless networks, IEEE 802.15 ZigBee networks, ultra-wideband protocols, near-field communication, etc. Environment 28 may include one or more wireless access points, e.g., wireless access point 34, that provide support for wireless communication throughout environment 28. Additionally or alternatively, for example, when a local network is unavailable, computing device 12, IoT device 30, and other devices within environment 28 may be configured to communicate with network 16, e.g., HMS 26, via cellular base station 36 and a cellular network.

[0056] In some examples, computing device 12 and / or computing system 20 may implement one or more algorithms to determine information corresponding to patient 4 based on data received from IMD 10. In some examples, computing device 12 and / or computing system 20 may have greater processing power than IMD 10, allowing for more complex analysis of the data. In some examples, computing device 12 and / or HMS 26 may apply probabilistic models, machine learning models, or other artificial intelligence-developed algorithms to the data to determine information corresponding to patient 4, for example, as described herein. Although described in the context of an example in which IMD 10 sensing a patient's cardiac activity may include an ICM, an exemplary system including any type of implantable, wearable, or external device configured to sense a patient's physiological parameters may be configured to implement the techniques of the present disclosure.

[0057] In some examples, the IMD 10 includes one or more electrodes. The IMD 10 can generate electrical cardiac data (e.g., ECG data and / or EGM data) based on cardiac signals sensed via the one or more electrodes. In some examples, the electrical cardiac data can be based on cardiac electrical signals indicative of one or more aspects of cardiac activity of the patient 4. For example, the electrical cardiac data can be indicative of one or more cardiac events, such as an atrial depolarization (indicated by a P wave), a ventricular depolarization (indicated by an R wave), and a ventricular repolarization (indicated by a T wave). The cardiac data can represent a sequence of cardiac data points, such that any given data point in the sequence of cardiac data points corresponds to the magnitude of the cardiac signal at a time. Based on the time at which a cardiac event occurs, the cardiac data can be indicative of one or more parameters, such as heart rate and heart rate variability. The IMD 10 can, in some examples, output the electrical cardiac signals to the computing system 20. The computing system 20 can, in some examples, store the electrical cardiac signals in memory 24.

[0058] In some examples, the memory 24 is configured to store a machine learning model. The HMS 26 may be configured to apply the machine learning model to electrical cardiac data (e.g., ECG data and / or EGM data) received as input from the IMD 10 to generate an output. In some examples, the machine learning model is trained based on multiple sets of training data. The memory 24 may be configured to store the multiple sets of training data. Each set of training data among the multiple sets of training data may include a set of training electrical cardiac data and information indicative of one or more values ​​of a metric of LV dysfunction corresponding to the set of training electrical cardiac data. In some examples, the one or more values ​​of the metric of LV dysfunction corresponding to the set of training electrical cardiac data among the sets of training data may each correspond to a time relative to one or more times of the set of training electrical cardiac data. In some examples, the set of training electrical cardiac data among the sets of training data may include one or more labels indicative of characteristics of the set of training electrical cardiac data, although this is not required.

[0059] In some examples, the electrical cardiac data received by computing system 20 from IMD 10 may include multiple sets of electrical cardiac data. In some examples, the multiple sets of electrical cardiac data may represent segments of long samples of electrical cardiac data. In some examples, each of the multiple sets of electrical cardiac data may represent individual data samples collected by IMD 10 at different times or time ranges. In some examples, each of the multiple sets of electrical cardiac data may correspond to the same duration. In some examples, one or more sets of electrical cardiac data may correspond to a duration that is different from the duration of one or more other sets of electrical cardiac data among the multiple sets of electrical cardiac data.

[0060] The HMS 26 can apply a machine learning model to each set of electrical cardiac data among the multiple sets of electrical cardiac data to determine a value of an LV dysfunction metric corresponding to each set of electrical cardiac data among the multiple sets of electrical cardiac data. This enables the HMS 26 to generate values ​​of the LV dysfunction metric corresponding to different points in time or time windows. By determining a value of the LV dysfunction metric corresponding to each set of electrical cardiac data among the multiple sets of electrical cardiac data, the HMS 26 can track the LV dysfunction metric over a period of time and determine one or more trends in the LV dysfunction metric over that period of time. In some examples, the HMS 26 can determine a trend in the LV dysfunction metric over a period of time based on the one or more determined values ​​of the LV dysfunction metric. For example, the HMS 26 can determine that the LV dysfunction metric is worsening over a period of time, improving over a period of time, or remaining stable over a period of time.

[0061] The IMD 10 may include a motion sensor, such as an accelerometer. The motion sensor of the IMD 10 may generate motion data based on movement of the IMD 10. In some examples, the IMD 10 may output the motion data to the computing system 20 for analysis. In some examples, the HMS 26 may determine a motion value indicative of the activity level of the patient 4 based on the motion data received from the IMD 10. The motion value corresponds to a value of an LV dysfunction metric determined based on the electrical cardiac data received from the IMD 10. The HMS 26 may determine whether to output an alert based on the motion value and the determined value of the LV dysfunction metric. For example, the HMS 26 may determine whether the value of the LV dysfunction metric is less than a threshold LV dysfunction metric. The HMS 26 may determine whether the motion value is greater than a threshold motion value. The HMS 26 may determine whether to output an alert based on whether the value of the LV dysfunction metric is less than a threshold LV dysfunction metric and whether the motion value is greater than the threshold motion value.

[0062] For example, if the LV dysfunction metric is LVEF, it may be beneficial to output an alert when the patient's LVEF is low and the patient's activity level is high. In this example, the HMS 26 may output an alert when the motion value is greater than a threshold motion value and when the value of the LV dysfunction metric is less than the LV dysfunction threshold metric. In some examples, the HMS 26 may not output an alert when the motion value is not greater than the threshold motion value or when the value of the LV dysfunction metric is not less than the LV dysfunction threshold metric.

[0063] The HMS 26 may train a machine learning model based on the multiple sets of training data. By training the machine learning model based on the multiple sets of training data, the processing circuitry is configured to cause the machine learning model to recognize one or more patterns corresponding to one or more characteristics of the LV dysfunction metric and the electrical cardiac data. Each set of training data among the multiple sets of training data may include a set of electrical cardiac data and one or more values ​​of the LV dysfunction metric corresponding to the set of training electrical cardiac data, such that each set of training data includes electrical cardiac data labeled with the values ​​of the LV dysfunction metric. Based on the characteristics of the electrical cardiac data and the labels indicating the values ​​of the LV dysfunction metric in each set of training data, the HMS 26 may train the machine learning model to recognize patterns corresponding to the electrical cardiac data and the LV dysfunction metric.

[0064] For example, once the machine learning model is trained, the HMS 26 can apply the machine learning model to a set of electrical cardiac data indicating one or more characteristics and / or parameters (e.g., P waves, T waves, R waves, heart rate, heart rate variability) to determine a value of the LV dysfunction metric. The HMS 26 can apply the machine learning model to process the set of electrical cardiac data to determine a value of the LV dysfunction metric based on one or more characteristics of the set of electrical cardiac data and one or more learned patterns between the characteristics of the electrical cardiac data and the LV dysfunction metric. In some examples, the machine learning model can output an estimated actual value of the LV dysfunction metric. If the LV dysfunction metric includes LVEF, the actual value of the LV dysfunction metric can include a ratio of the volume of blood pumped from the left ventricle to the total volume of blood in the left ventricle. In some examples, the machine learning model can output a confidence value indicating a confidence that the actual value of the LV dysfunction metric is accurate. In some examples, the machine learning model can output a confidence that the value of the LV dysfunction metric corresponding to the set of electrical cardiac data is lower than a threshold value for the LV dysfunction metric. In some examples, the machine learning model may output a confidence that the value of the LV dysfunction metric corresponding to the set of electrical cardiac data is less than the LV dysfunction metric threshold.

[0065] In some examples, HMS 26 may label each set of training data among the multiple sets of training data stored in memory 24. In some examples, to label each set of training data among the multiple sets of training data, HMS 26 may identify, for each set of training data among the multiple sets of training data, one or more characteristics of the set of training electrical cardiac data among the set of training data. For example, HMS 26 may identify one or more P waves, R waves, T waves, arrhythmia events, other electrical cardiac features, or any combination thereof. Additionally or alternatively, HMS 26 may identify one or more parameters corresponding to the set of training electrical cardiac data, such as heart rate, heart rate variability, pulse transit time (PTT), or any combination thereof. HMS 26 may label each set of training electrical cardiac data among the multiple sets of training data with one or more characteristics of the set of training electrical cardiac data.

[0066] In some examples, to label each set of training data among the multiple sets of training data, HMS 26 may identify, in the information indicating the values ​​of the one or more LV dysfunction metrics of the set of training data, a time corresponding to each of the one or more values ​​of the LV dysfunction metric. In some examples, the time corresponding to each of the one or more values ​​of the LV dysfunction metric corresponds to the time at which the value of the LV dysfunction metric was measured. HMS 26 may associate the time corresponding to each of the one or more values ​​of the LV dysfunction metric with the time of the set of training electrical cardiac data. That is, HMS 26 may label each set of training data among the multiple sets of training data with one or more times corresponding to the electrical cardiac data and a time corresponding to each of the one or more values ​​of the LV dysfunction metric. This means that when each set of training data among the multiple sets of training data is labeled, the labeled training data indicates the time at which the value of the LV dysfunction metric was measured relative to the time of the electrical cardiac data.

[0067] The technology described herein is not limited to examples in which the IMD 10 includes an ICM. In some examples, the IMD 10 of FIG. 1 can include any type of external or implantable medical device configured to collect electrical cardiac data, such as ECG data or EGM data. For example, the IMD 10 may include an implantable cardioverter-defibrillator (ICD), a pacemaker, a cardiac resynchronization therapy pacemaker (CRT-P), a cardiac resynchronization therapy defibrillator (CRT-D), or an external cardiac signal sensing device.

[0068] FIG. 2A is a perspective view illustrating an IMD 10A, which may be an exemplary configuration of the IMD 10 of FIG. 1 as an ICM, in accordance with one or more techniques of the present disclosure. In the example shown in FIG. 2A , the IMD 10A may be embodied as a monitoring device having a housing 112, a proximal electrode 116A, and a distal electrode 116B. The housing 112 may further include a first major surface 114, a second major surface 118, a proximal end 120, and a distal end 122. The housing 112 encloses electronic circuitry located within the IMD 10A and protects the circuitry contained therein from bodily fluids. The housing 112 may be hermetically sealed and configured for subcutaneous implantation. An electrical feedthrough provides electrical connection between the electrodes 116A and 116B.

[0069] In the example shown in FIG. 2A , the IMD 10A is defined by a length L, a width W, and a thickness or depth D, and is in the form of an elongated rectangular prism, with the length L being much greater than the width W, which is greater than the depth D. In one example, the geometry of the IMD 10A, particularly the width W being greater than the depth D, is selected to allow the IMD 10A to be inserted under a patient's skin using a minimally invasive procedure and to remain in a desired orientation during insertion. For example, the device shown in FIG. 2A may have a radial asymmetry (particularly a rectangular shape) along the longitudinal axis that maintains the device in a proper orientation after insertion. For example, the spacing between the proximal electrode 116A and the distal electrode 116B may be in the ranges of 5 millimeters (mm) to 55 mm, 30 mm to 55 mm, 35 mm to 55 mm, and 40 mm to 55 mm, or any range or individual spacing between 5 mm and 60 mm. Additionally, the IMD 10A may have a length L in the range of 30 mm to approximately 70 mm. In other examples, the length L may be in the range of 5 mm to 60 mm, 40 mm to 60 mm, or 45 mm to 60 mm, or any length or range of lengths between about 30 mm and about 70 mm. Additionally, the width W of the first major surface 114 may be in the range of 3 mm to 15 mm, 3 mm to 10 mm, or 5 mm to 15 mm, or any single width or range of widths between 3 mm and 15 mm. The depth D of the IMD 10A may be in the range of 2 mm to 15 mm, 2 mm to 9 mm, 2 mm to 5 mm, or 5 mm to 15 mm, or any single depth or range of depths between 2 mm and 15 mm. Additionally, an IMD 10A according to one example of the present disclosure has a geometry and size designed for ease of implantation and patient comfort. Examples of IMDs 10A described herein may have a volume of 3 cubic centimeters (cm) or less, 1.5 cubic cm or less, or any volume between 3 and 1.5 cubic centimeters.

[0070] 2A , when inserted into a patient, first major surface 114 faces outward, toward the patient's skin, and second major surface 118 is located opposite first major surface 114. Additionally, in the example shown in FIG. 2A , proximal end 120 and distal end 122 are rounded to reduce discomfort and irritation to surrounding tissue when inserted under a patient's skin. IMD 10A, including instruments and methods for inserting IMD 10A, are described, for example, in U.S. Patent No. 11,311,312, the entire contents of which are incorporated herein by reference.

[0071] The proximal electrode 116A is at or near the proximal end 120, and the distal electrode 116B is at or near the distal end 122. The proximal and distal electrodes 116A, 116B are used to sense electrical cardiac signals, e.g., ECG and / or EGM signals, and to measure interstitial impedance outside the thorax, which may be submuscular or subcutaneous. The electrical cardiac signals and impedance measurements may be stored in the memory of the IMD 10A, and the data may be transmitted via the integrated antenna 126A to another device, which may be another implantable device such as the computing device 12 or an external device. In some examples, the electrodes 116A and 116B may additionally or instead be used to sense any biopotential signal of interest, which may be, for example, an EGM, an electroencephalogram (EEG), an electromyogram (EMG), or a neural signal, from any implanted location. The housing 112 may house the circuitry of the IMD 10 shown in FIG. 3.

[0072] 2A , the proximal electrode 116A is at or very near the proximal end 120, and the distal electrode 116B is at or very near the distal end 122. In this example, the distal electrode 116B is not limited to a flattened outwardly facing surface, but may extend from the first major surface 114, around the rounded edge 124 and / or end face 126, and onto the second major surface 118, such that the electrode 116B has a three-dimensional curved configuration. In some examples, the electrode 116B is an uninsulated portion of the metal (e.g., titanium) portion of the housing 112.

[0073] 2A, the proximal electrode 116A is located on the first major surface 114 and is substantially flat and outward-facing. However, in other examples, the proximal electrode 116A may utilize a three-dimensional curved configuration of the distal electrode 116B to provide a three-dimensional proximal electrode (not shown in this example). Similarly, in other examples, the distal electrode 116B may utilize a substantially flat, outward-facing electrode located on the first major surface 114 similar to that shown with respect to the proximal electrode 116A.

[0074] These various electrode configurations allow for configurations in which the proximal electrode 116A and the distal electrode 116B are located on both the first major surface 114 and the second major surface 118. In other configurations, such as that shown in FIG. 2A , only one of the proximal electrode 116A and the distal electrode 116B is located on both the first major surface 114 and the second major surface 118, and in still other configurations, both the proximal electrode 116A and the distal electrode 116B are located on one of the first major surface 114 or the second major surface 118 (e.g., the proximal electrode 116A is located on the first major surface 114 and the distal electrode 116B is located on the second major surface 118). In another example, the IMD 10A may include electrodes on both the first major surface 114 and the second major surface 118 at or near the proximal and distal ends of the device, such that a total of four electrodes are included on the IMD 10A. Electrodes 116A and 116B may be formed from several different types of biocompatible conductive materials, such as stainless steel, titanium, platinum, iridium, or alloys thereof, and may utilize one or more coatings such as titanium nitride or fractal titanium nitride.

[0075] In the example shown in FIG. 2A , the proximal end 120 includes a header assembly 128 that includes one or more of the proximal electrode 116A, the integrated antenna 126A, the anti-migration protrusion 132, and / or the suture hole 134. The integrated antenna 126A is located on the same major surface as the proximal electrode 116A (i.e., the first major surface 114) and is included as part of the header assembly 128. The integrated antenna 126A enables the IMD 10A to transmit and / or receive data. In other examples, the integrated antenna 126A may be formed on the opposite major surface from the electrode 116A or may be incorporated into the housing 112 of the IMD 10A. In the example shown in FIG. 2A , the anti-migration protrusion 132 is located adjacent to the integrated antenna 126A and protrudes away from the first major surface 114 to prevent longitudinal movement of the device. In the example shown in FIG. 2A , anti-migration protrusion 132 includes a plurality (e.g., nine) small ridges or protrusions extending away from first major surface 114. As discussed above, in other examples, anti-migration protrusion 132 may be located on the opposite major surface from proximal electrode 116A and / or integrated antenna 126A. Additionally, in the example shown in FIG. 2A , header assembly 128 includes suture holes 134 that provide another means of securing IMD 10A to a patient to prevent migration after insertion. In the illustrated example, suture holes 134 are located adjacent proximal electrode 116A. In some examples, header assembly 128 is a molded header assembly made from a polymeric or plastic material that may be integral with or separable from the main portion of IMD 10A.

[0076] In some examples, IMD 10A may be configured to collect electrical cardiac signals, such as ECG signals or EGM signals, via electrodes 116A, 116B. In some examples, the electrical cardiac signals collected by IMD 10A may include one or more characteristics and / or be indicative of one or more patient parameters. IMD 10A may be configured to process the electrical cardiac signals, pre-process the electrical cardiac signals, output the electrical cardiac signals, store the electrical cardiac signals in memory, or any combination thereof.

[0077] 2B is a perspective view illustrating another IMD 10B, which may be another exemplary configuration of the IMD 10 of FIG. 1 as an ICM, in accordance with one or more techniques of the present disclosure. The IMD 10B of FIG. 2B may be configured substantially similarly to the IMD 10A of FIG. 2A, with the differences therebetween being described herein.

[0078] The IMD 10B may include a leadless subcutaneously implantable monitoring device, such as an ICM. The IMD 10B includes a housing having a base 140 and an insulating cover 142. The proximal electrode 116C and the distal electrode 116D may be formed or disposed on an outer surface of the insulating cover 142. For example, various circuits and components of the IMD 10B described with respect to FIG. 3 may be formed or disposed on an inner surface of the insulating cover 142 or within the base 140. In some examples, a battery or other power source for the IMD 10B may be included within the base 140. In the illustrated example, the antenna 126B may be formed or disposed on an outer surface of the insulating cover 142, but in some examples, the antenna 126B may be formed or disposed on an inner surface. In some examples, the insulating cover 142 may be positioned over the open base 140 such that the base 140 and the insulating cover 142 enclose the circuitry and other components and protect them from fluids, such as bodily fluids. The housing, including the base 140 and insulating cover 142, may be hermetically sealed and configured for subcutaneous implantation.

[0079] Circuitry and components may be formed on the inside of the insulating cover 142, such as by using flip-chip technology. The insulating cover 142 may be inverted onto the base 140. When inverted and placed on the container 140, the components of the IMD 10B formed on the inside of the insulating cover 142 may be positioned within the gap 144 defined by the base 140. The electrodes 116C and 116D and the antenna 126B may be electrically connected to the circuitry formed on the inside of the insulating cover 142 through one or more vias (not shown) formed through the insulating cover 142. The insulating cover 142 may be formed of sapphire (i.e., corundum), glass, parylene, and / or any other suitable insulating material. The base 140 may be formed of titanium or any other suitable material (e.g., a biocompatible material). The electrodes 116C and 116D may be formed of stainless steel, titanium, platinum, iridium, or any alloy thereof. Additionally, electrodes 116C and 116D may be coated with a material such as titanium nitride or fractal titanium nitride, although other suitable materials and coatings may be used for such electrodes.

[0080] In the example shown in FIG. 2B, the housing of the IMD 10B defines a length L, a width W, and a thickness or depth D. Similar to the IMD 10A of FIG. 2A, the housing is in the form of an elongated rectangular prism, with the length L being much greater than the width W, and the width W being greater than the depth D. For example, the spacing between the proximal electrode 116C and the distal electrode 116D may be in the range of 5 mm to 50 mm, 30 mm to 50 mm, 35 mm to 45 mm, or any single spacing or range of spacings between 5 mm and 50 mm, such as about 40 mm. Additionally, the IMD 10B may have a length L in the range of 5 mm to about 70 mm. In other examples, the length L may be in the range of 30 mm to 70 mm, 40 mm to 60 mm, 45 mm to 55 mm, or any single length or range of lengths between 5 mm and 50 mm, such as about 45 mm. Additionally, width W may range from 3 mm to 15 mm, 5 mm to 15 mm, 5 mm to 10 mm, or any single width or range of widths between 3 mm and 15 mm, such as about 8 mm. Thickness or depth D of IMD 10B may range from 2 mm to 15 mm, 5 mm to 15 mm, or 3 mm to 5 mm, or any single depth or range of depths between 2 mm and 15 mm, such as about 4 mm. IMD 10B may have a volume of 3 cubic centimeters (cm) or less, or 1.5 cubic cm or less, for example, about 1.4 cubic cm.

[0081] In the example shown in Figure 2B, when inserted under a patient's skin, the outer surface of insulating cover 142 faces outward, toward the patient's skin. Additionally, as shown in Figure 2B, proximal end 146 and distal end 148 are rounded to reduce discomfort and irritation to surrounding tissue when inserted under a patient's skin. Additionally, the edges of IMD 10B may be rounded.

[0082] In some examples, IMD 10B may be configured to collect electrical cardiac signals, such as ECG signals or EGM signals, via electrodes 116C, 116D. In some examples, the electrical cardiac signals collected by IMD 10B may include one or more characteristics and / or be indicative of one or more patient parameters. IMD 10B may be configured to process the electrical cardiac signals, pre-process the electrical cardiac signals, output the electrical cardiac signals, store the electrical cardiac signals in memory, or any combination thereof.

[0083]

[0023] Figure 3 is a block diagram illustrating an example configuration of IMD 10 of Figure 1 in accordance with one or more techniques of this disclosure. As shown in Figure 3, IMD 10 includes processing circuitry 150, memory 152, sensing circuitry 154 coupled to electrodes 116A and 116B (hereinafter "electrodes 116A, 116B") and one or more sensors 158, and communication circuitry 160.

[0084] Processing circuitry 150 may include fixed-function circuitry and / or programmable processing circuitry. Processing circuitry 150 may include any one or more of a microprocessor, controller, GPU, TPU, DSP, ASIC, FPGA, or equivalent discrete or analog logic circuitry. In some examples, processing circuitry 150 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 circuits. The functionality attributed to processing circuitry 150 herein may be embodied as software, firmware, hardware, or any combination thereof. In some examples, memory 152 includes computer-readable instructions that, when executed by processing circuitry 150, cause IMD 10 and processing circuitry 150 to perform various functions attributed to IMD 10 and processing circuitry 150 herein. The memory 152 may include any volatile, non-volatile, magnetic, optical, or electrical medium, such as RAM, ROM, NVRAM, EEPROM, flash memory, or any other digital medium.

[0085] Sensing circuit 154 can sense cardiac signals via electrodes 116A, 116B, for example, measuring the impedance of tissue proximate IMD 10. The measured impedance can vary based on respiration, cardiac pulse or flow, and the degree of perfusion or edema. Processing circuit 150 can determine patient metrics related to respiration, fluid retention, cardiac pulse or flow, perfusion, and / or edema based on the measured impedance. In some examples, processing circuit 150 can identify features of the sensed cardiac signals, such as heart rate, heart rate variability, T-wave alternans, beat intervals (e.g., QT interval), and / or morphological features, to detect the onset of cardiac arrhythmia in patient 4.

[0086] In some examples, IMD 10 includes one or more sensors 158, such as one or more accelerometers, gyroscopes, microphones, optical sensors, temperature sensors, pressure sensors, and / or chemical sensors. In some examples, sensing circuit 154 may include one or more filters and amplifiers for filtering and amplifying signals received from electrodes 116A, 116B and / or one or more of sensors 158. In some examples, sensing circuit 154 and / or processing circuit 150 may include a rectifier, a filter and / or amplifier, a sense amplifier, a comparator, and / or an analog-to-digital converter. Processing circuit 150 may determine parameter data 182, e.g., values ​​of physiological parameters of patient 4, based on signals from sensors 158, which may be stored as data 180 in memory 152. Patient parameters determined from signals from sensors 158 may include intravascular fluid levels, interstitial fluid levels, oxygen saturation, glucose levels, stress hormone levels, heart sounds, physical movement, activity intensity, sleep duration, sleep quality, body position, or blood pressure.

[0087] In addition to data 180, memory 152 may store an application 170 executable by processing circuit 150. Application 170 may include a data processing application 172. Processing circuit 150 may execute data processing application 172 to process parameter data 182 and electrical cardiac data 184. In some examples, data processing application 172 may identify information corresponding to parameter data 182 and / or electrical cardiac data 184. For example, data processing application 172 may identify the heart rate of patient 4 based on electrical cardiac data 184 by identifying the rate at which R waves occur in the electrical cardiac data 184.

[0088] The data 180 may include electrical cardiac data 184 sensed by the IMD 10 via the electrodes 116A, 116B. In some examples, the electrical cardiac data 184 may represent cardiac data such as ECG data or EGM data indicative of cardiac activity of the patient 4. In some examples, the electrical cardiac data 184 may be indicative of cardiac activity of the patient 4 over an extended period of time (e.g., weeks or months) collected continuously via the electrodes 116A, 116B while the IMD 10 is implanted under the skin of the patient 4. In some examples, the electrical cardiac data 184 may include multiple sets of cardiac data collected respectively via the electrodes 116A, 116B while the IMD 10 is implanted under the skin of the patient 4. The electrical cardiac data 184 may include data indicative of cardiac signals, heart rate information, RR interval information, morphological information, or other information.

[0089] Processing circuitry 150 may use communications circuitry 160 to communicate parameter data 182 and electrical cardiac data 184 to one or more other computing devices, such as computing device 12 and / or computing system 20. Communications circuitry 160 may include any suitable hardware, firmware, software, or any combination thereof for wirelessly communicating with another device. Communications circuitry 160 may be configured to transmit and / or receive signals via inductive coupling, electromagnetic coupling, near field communication (NFC), radio frequency (RF) communication, Bluetooth®, Wi-Fi, or other proprietary or non-proprietary wireless communication schemes.

[0090] In some examples, the memory 152 of the IMD 10 is configured to store a model 194 that includes a machine learning model 196. The IMD 10 may be configured to apply the machine learning model 196 to the electrical cardiac data 184 to determine a value for the metric of LV dysfunction that corresponds to the electrical cardiac data 184. In some examples, the machine learning model 196 may be trained by the computing system 20 of FIG. 1 and output to the IMD 10. The IMD 10 need not store the model 194 that includes the machine learning model 196. The IMD 10, in some examples, can output the data 180 to one or more other devices (e.g., the external device 12, the computing system 20, the clinician computing device 38, or any combination thereof) for processing.

[0091] 4 is a block diagram illustrating an example configuration of computing device 12, which may correspond to either computing device 12A or 12B (or both operating cooperatively) in accordance with one or more techniques of this disclosure. In some examples, computing device 12 takes the form of a smartphone, laptop, tablet computer, personal digital assistant (PDA), smart watch, or other wearable computing device. In some examples, IoT device 30 and / or clinician computing device 38 may be configured similarly to the configuration of computing device 12 shown in FIG. 4.

[0092] 4, computing device 12 may be logically divided into user space 202, kernel space 204, and hardware 206. Hardware 206 may include one or more hardware components that provide an operating environment for components executing in user space 202 and kernel space 204. User space 202 and kernel space 204 may represent different sections or segmentations of memory, with kernel space 204 providing higher privileges to processes and threads than user space 202. For example, kernel space 204 may include operating system 220, which operates with higher privileges than components executing in user space 202.

[0093] 4, hardware 206 includes processing circuitry 230, memory 232, one or more input devices 234, one or more output devices 236, one or more sensors 238, and communication circuitry 240. Although computing device 12 is shown in FIG. 4 as an example stand-alone device, it may be any component or system that includes processing circuitry or other suitable computing environment for executing software instructions, and may not necessarily include one or more elements shown in FIG.

[0094] Processing circuitry 230 is configured to execute functions and / or process instructions for execution within computing device 12. For example, processing circuitry 230 may be configured to receive and process instructions stored in memory 232 that provide the functionality of components included in kernel space 204 and user space 202 to perform one or more operations in accordance with the techniques of this disclosure. Examples of processing circuitry 230 may include any one or more microprocessors, controllers, GPUs, TPUs, DSPs, ASICs, FPGAs, or equivalent discrete or integrated logic circuits.

[0095] Memory 232 may be configured to store information within computing device 12 for processing during operation of computing device 12. Memory 232, in some examples, is described as a computer-readable storage medium. In some examples, memory 232 includes temporary or volatile memory. Examples of volatile memory include RAM, DRAM, static random access memory (SRAM), and other forms of volatile memory known in the art. Memory 232, in some examples, also includes one or more memories configured for long-term storage of information, including, for example, non-volatile storage elements. Examples of such non-volatile storage elements include magnetic hard disks, optical disks, floppy disks, flash memory, or forms of electrically programmable memories (EPROM) or EEPROM memory. In some examples, memory 232 includes cloud-related storage.

[0096] One or more input devices 234 of computing device 12 may receive input from, for example, patient 4, clinician 40, or another user. Examples of input are tactile input, audio input, kinetic input, and visual input. Input device 234 may include, by way of example, a mouse, keyboard, voice response system, camera, buttons, control pad, microphone, presence or touch sensing component (e.g., a screen), or any other device for detecting input from a user or machine.

[0097] One or more output devices 236 of computing device 12 may generate output, for example, to patient 4 or another user. Examples of output are haptic output, tactile output, audio output, and visual output. Output devices 236 of computing device 12 may include a presence sensing screen, a sound card, a video graphics adapter card, speakers, a cathode ray tube monitor, a liquid crystal display (LCD), light emitting diodes (LEDs), or any type of device for generating haptic, audio, and / or visual output.

[0098] One or more sensors 238 of computing device 12 may sense physiological parameters or signals of patient 4. Sensors 238 may include electrodes, accelerometers (e.g., triaxial accelerometers), optical sensors, impedance sensors, temperature sensors, pressure sensors, heart sound sensors (e.g., microphones or accelerometers), and other sensors and sensing circuitry (e.g., including analog-to-digital converters (ADCs)) similar to that described above with respect to IMD 10 and FIG.

[0099] The communications circuitry 240 of the computing device 12 may communicate with other devices by transmitting and receiving data. The communications circuitry 240 may receive data from the IMD 10, such as patient metrics and / or higher-resolution diagnostic information, from the communications circuitry within the IMD 10. The communications circuitry 240 may include a network interface card, such as an Ethernet card, an optical transceiver, a radio frequency transceiver, or any other type of device capable of transmitting and receiving information. For example, the communications circuitry 160 may include a wireless transceiver configured to communicate according to a standard or protocol, such as 3G, 4G, 5G, Wi-Fi (e.g., 802.11 or 802.15 ZigBee), Bluetooth, or BLE.

[0100] 4, the health monitoring application 250 executes in the user space 202 of the computing device 12. The health monitoring application 250 may be logically divided into a presentation layer 252, an application layer 254, and a data layer 256. The presentation layer 252 may include a user interface (UI) component 260 that generates and renders the user interface of the health monitoring application 250.

[0101] Data layer 256 may include parameter data 290 and electrical cardiac data 292, which may be received from IMD 10 via communications circuitry 240 and stored in memory 232 by processing circuitry 230. Application layer 254 may include, but is not limited to, a data analyzer 270 and a model configuration service 272. Data analyzer 270 may be configured to process parameter data 290 and / or electrical cardiac data 292 generated by IMD 10 to generate information corresponding to one or more patient conditions of patient 4. Data analyzer 270 may determine the information corresponding to patient 4 based on application of parameter data 290 and / or electrical cardiac data 292 as inputs to one or more models 294, which may include one or more probability models, machine learning models, algorithms, decision trees, and / or thresholds. In examples where models 294 include one or more machine learning models, data analyzer 270 may apply feature vectors derived from the data to models 294.

[0102] The model configuration service 272 may be configured to modify the model 294 based on feedback indicating whether the determination was accurate or updated parameters received, for example, from the HMS 26. In some examples, the model configuration service 272 may utilize a dataset from the patient 4 for supervised machine learning to further train a model included as part of the model 294. The model configuration service 272, or another component executed by the processing circuitry of the medical device system 2, may select a configuration for the model 294 based on the patient's etiological data. In some examples, different models 294 tuned for different cohorts of patients may be available for selection of the patient 4 based on such etiological data.

[0103] In some examples, model 294 includes a machine learning model configured to process electrical cardiac data 292 to determine a value of a metric of LV dysfunction. Computing device 12 may be configured to apply the machine learning model to electrical cardiac data 292 to determine a value of a metric of LV dysfunction corresponding to electrical cardiac data 292. In some examples, the machine learning model may be trained by computing system 20 and output computing device 12 of FIG. 1 . Computing device 12 need not store model 294, which includes a machine learning model. Computing device 12, in some examples, can output electrical cardiac data 292 to one or more other devices (e.g., computing system 20, clinician computing device 38, or any combination thereof) for processing and / or receive one or more values ​​determined based on processing of electrical cardiac data 292.

[0104] FIG. 5 is a block diagram illustrating an operational view of HMS 26 in accordance with one or more techniques of this disclosure. HMS 26 may be implemented in computing system 20, which may include hardware components such as processing circuitry 22, memory 24, and communication circuitry embodied in one or more physical devices. FIG. 5 provides an operational view of HMS 26 when hosted as a cloud-based platform. In the example of FIG. 5, the components of HMS 26 are arranged according to multiple logical layers that implement the techniques of this disclosure. Each layer may be implemented by one or more modules comprised of hardware, software, or a combination of hardware and software.

[0105] Computing devices, such as computing device 12, IoT device 30, and clinician computing device 38, act as clients that communicate with HMS 26 through interface layer 300. Computing devices typically run client software applications, such as desktop applications, mobile applications, and web applications. Interface layer 300 represents the set of application programming interfaces (APIs) or protocol interfaces exposed and supported by HMS 26 for client software applications. Interface layer 300 may be implemented using one or more web servers.

[0106] As shown in FIG. 5 , the HMS 26 also includes an application layer 302, which represents a collection of services 310 for performing the functions attributed to the HMS 26 herein. The application layer 302 receives information from client applications, e.g., data from the computing device 12 or the IoT device 30 (some or all of which may have been retrieved from the IMD 10), and processes the information according to one or more of the services 310 to respond to the information. The application layer 302 may be implemented as one or more discrete software services 310 running on one or more application servers (e.g., physical or virtual machines). That is, the application servers provide a runtime environment for the execution of the services 310. In some examples, the functionality of the functional interface layer 300 and the application layer 302 described above may be executed on the same server. The services 310 may communicate via a logical service bus 312. The service bus 312 generally represents a set of logical interconnects or interfaces that allow different services 310 to send messages to other services, such as via a publish / subscribe communication model.

[0107] The data layer 304 of the HMS 26 uses one or more data repositories to provide persistence of information within the HMS 26. A data repository can generally be any data structure or software that stores and / or manages data. Examples of data repositories include, but are not limited to, relational databases, multidimensional databases, maps, and hash tables, to name a few.

[0108] The services 310 may include a data analyzer 330, a model configuration service 332, and a record management service 334. As shown in FIG. 5 , each of the services 310 is implemented modularly within the HMS 26. While each service is shown as a separate module, in some embodiments, the functionality of two or more services may be combined into a single module or component. Each of the services 310 may be implemented in software, hardware, or a combination of hardware and software. Furthermore, the services 310 may generally be implemented as standalone devices, separate virtual machines or containers, processes, threads, or software instructions for execution on one or more physical processors. The record management service 334 may store the received patient data as parameter data 350 and electrical cardiac data 352.

[0109] The data analyzer 330 can determine information corresponding to cardiac activity of the patient 4 based on the electrical cardiac data 352 generated by the IMD 10, and possibly other parameter data 350. In some examples, the data analyzer 330 can identify one or more features (e.g., R waves, T waves, P waves) in the electrical cardiac data 352 received from the IMD 10. In some examples, the data analyzer 330 can determine whether the electrical cardiac data 352 is indicative of a patient condition or arrhythmia, such as HF, VF, AF, or AV block. The data analyzer 330 can determine information corresponding to the patient 4 based on application of the parameter data 350 and / or the electrical cardiac data 352 as input to a machine learning model 354. In some examples, the data analyzer 330 can apply feature vectors derived from the data to the machine learning model 354.

[0110] The machine learning model 354 may be developed by the model configuration service 332. Exemplary machine learning techniques that may be employed to generate the machine learning model 354 include various learning styles such as supervised learning, unsupervised learning, and semi-supervised learning. Representative types of algorithms include Bayesian algorithms, Markov models, Hawkes processes, clustering algorithms, decision tree algorithms, regularization algorithms, regression algorithms, example-based algorithms, artificial neural network algorithms, deep learning algorithms, dimensionality reduction algorithms, etc. Various examples of specific algorithms include Bayesian linear regression, boosted decision tree regression, and neural network regression, backpropagation neural networks, self-attention models, convolutional neural networks (CNN), long short-term networks (LSTM), a priori algorithms, k-means clustering, k-nearest neighbor (kNN), learning vector quantization (LVQ), self-organizing maps (SOM), locally weighted learning (LWL), ridge regression, least absolute shrinkage and selection operator (LASSO), elastic nets, and least-angle regression (LARS), principal component analysis (PCA), and principal component regression (PCR).

[0111] In some examples, the model configuration service 332 may be configured to train the machine learning model 354 using the training data 355. By training the machine learning model based on the training data 355, the model configuration service 332 is configured to cause the machine learning model 354 to recognize one or more patterns corresponding to one or more characteristics of the LV dysfunction (e.g., LVEF) metric and the electrical cardiac data. For example, the model configuration service 332 is configured to cause the machine learning model 354 to recognize one or more patterns in the electrical cardiac data that correspond to one or more values ​​of the LV dysfunction metric. This allows the HMS 26 to use the machine learning model 354 to process input samples of the electrical cardiac training data and determine values ​​of the LV dysfunction metric.

[0112] Training data 355 can include electrical cardiac training data 356 and LV dysfunction metric training data 358. In some examples, training data 355 can include multiple sets of training data, where each set of training data among the multiple sets of training data includes a set of electrical cardiac training data among electrical cardiac training data 356 and a set of LV dysfunction metric training data among LV dysfunction metric training data 358. In some examples, the set of electrical cardiac training data among the sets of training data and the set of LV dysfunction metric training data can correspond to the same patient. That is, the set of electrical cardiac training data can be collected from the patient, and the set of LV dysfunction metric training data can include one or more values ​​of LV dysfunction metrics measured from the patient.

[0113] To train the machine learning model 354, the model configuration service 332 may label and / or classify each set of training data among the multiple sets of training data among the training data 355. In some examples, to label each set of training data among the training data 355, the model configuration service 332 may label the set of electrical cardiac training data among the respective sets of training data with one or more characteristics, criteria, and / or features of the set of electrical cardiac training data. In some examples, to label each set of training data among the training data 355, the machine learning model 354 may label the set of electrical cardiac training data among the respective sets of training data with one or more parameters corresponding to the set of electrical cardiac training data, such as heart rate, heart rate variability, pulse wave transit time, or any combination thereof. In some examples, to label each set of training data among the training data 355, the machine learning model 354 may label the set of electrical cardiac training data among the respective sets of training data with one or more arrhythmias or patient conditions corresponding to the set of electrical cardiac training data, such as HF, AV, AF, AV block, or any combination thereof. In some examples, data analyzer 330 may identify one or more characteristics, criteria, features, parameters, patient conditions, arrhythmias, or any combination thereof, corresponding to each set of electrical cardiac training data in electrical cardiac training data 356.

[0114] In some examples, to label each set of training data in training data 355, model configuration service 332 may identify one or more values ​​of an LV dysfunction metric in a set of LV dysfunction metric training data in LV dysfunction metric training data 358 that corresponds to each set of training data in training data 355. In some examples, each value of the one or more values ​​of the LV dysfunction metric may correspond to a time or range of times at which the value of the LV dysfunction metric was measured from the patient. Model configuration service 332 may associate a time corresponding to each value of the one or more values ​​of the LV dysfunction metric in a set of LV dysfunction metric training data with a time or range of times of the corresponding set of electrical cardiac training data. For example, because each set of training data in training data 355 includes a set of electrical cardiac training data in electrical cardiac training data 356 and a set of LV dysfunction metric training data in LV dysfunction metric training data 358 collected from the same patient, model configuration service 332 may associate a time at which the electrical cardiac training data was collected from the patient with a time at which each value of the one or more values ​​of the LV dysfunction metric was measured from the patient. This may allow the model configuration service 332 to identify patterns in the electrical cardiac training data that are indicative of values ​​of metrics of LV dysfunction.

[0115] In some examples, the model configuration service 332 may create multiple sets of training data 355, each set including a set of electrical cardiac training data (e.g., ECG data and / or EGM data) and a set of LV dysfunction metric training data. In some examples, the set of electrical cardiac training data and the set of LV dysfunction metric training data may be collected from the same patient at different times. For example, the set of electrical cardiac training data and the set of LV dysfunction metric training data may be collected from the same patient less than one day apart, less than one week apart, less than two weeks apart, or another amount of time apart. In some examples, the set of electrical cardiac training data and the set of LV dysfunction metric training data may be collected from the same patient simultaneously and / or during overlapping time windows. In either case, the model configuration service 332 may associate, for each set of training data in the training data 355, a time or time range corresponding to the respective set of electrical cardiac training data in the electrical cardiac training data 356 with a time or time range corresponding to the respective set of LV dysfunction metric training data in the LV dysfunction metric training data 358. This may enable the model configuration service 332 to train the machine learning model 354 to recognize patterns in the electrical cardiac data that are indicative of values ​​of metrics of LV dysfunction.

[0116] In some examples, electrical cardiac training data 356 may include multiple sets of electrical cardiac training data. The multiple sets of electrical cardiac data may include one or more electrical cardiac measurements each collected from the human patient via a Holter monitor, one or more sets of electrical cardiac data collected via electrodes attached to the human patient's skin, one or more sets of electrical cardiac data collected from the patient by a wearable device (e.g., a smart watch), one or more sets of electrical cardiac data collected from the human patient via an IMD, or any combination thereof. In any case, each set of electrical cardiac data among the multiple sets of electrical cardiac training data in electrical cardiac training data 356 may include electrical data indicative of the patient's cardiac activity, such as one or more cardiac cycles of the patient's heart muscle.

[0117] The model configuration service 332, in some examples, can train the machine learning model 354 based on one or more sets of raw electrical cardiac training data from the electrical cardiac training data 356. The raw electrical cardiac training data can indicate one or more features, characteristics, and / or parameters of cardiac activity without labels identifying those features or characteristics. For example, the set of raw electrical cardiac training data can indicate one or more R waves, T waves, P waves, and other features or characteristics without labels, and the set of raw electrical cardiac training data can indicate identifying those characteristics. Additionally or alternatively, the set of raw electrical cardiac training data can indicate parameters such as heart rate and / or heart rate variability without labels identifying values ​​of those parameters.

[0118] Model configuration service 332, in some examples, can train machine learning model 354 based on one or more labeled sets of raw electrical cardiac training data from electrical cardiac training data 356. The labeled electrical cardiac data can include labels that identify one or more features, characteristics, and / or parameters of cardiac activity indicated by the electrical cardiac training data. For example, a set of labeled electrical cardiac training data can include one or more labels that identify one or more R waves, T waves, P waves, and other features or characteristics. Additionally or alternatively, a set of labeled electrical cardiac training data can include one or more labels that identify parameter values, such as heart rate and / or heart rate variability, that correspond to the labeled electrical cardiac training data.

[0119] In some examples, to train the machine learning model 354, it may be beneficial for the model configuration service to use information corresponding to the training data to associate patterns and aspects of the training data with known characteristics of the training data. For example, when the model configuration service 332 is training the machine learning model 354 to identify criteria in cardiac signals such as ECG and EGM signals, it may be beneficial for the training data to include a plurality of cardiac data samples with labels that identify the criteria (e.g., labels that identify P waves, R waves, T waves, and other criteria). Additionally or alternatively, when the model configuration service 332 is training the machine learning model 354 to identify arrhythmias such as AF or AV block in cardiac data, it may be beneficial for the training data to include one or more labels that indicate portions of the training data that are indicative of an arrhythmia.

[0120] In some examples, the model configuration service 332 can train a machine learning model 354 to process the electrical cardiac data and determine a value for LVEF. In some examples, the LVEF can represent the ratio of the volume of blood ejected from the left ventricle in response to a ventricular depolarization to the volume of blood present in the left ventricle immediately before the ventricular depolarization. The LVEF can be expressed as a percentage (e.g., 50% of the blood present in the left ventricle before the depolarization was ejected from the ventricle in response to the depolarization). In some examples, when the machine learning model 354 outputs a percentage and / or ratio value for LVEF, the machine learning model 354 can represent a regression model.

[0121] In some examples, the model configuration service 332 may represent a classification model when the machine learning model 354 outputs a confidence that the patient's LVEF is low (e.g., less than 35%). For example, the machine learning model 354 may process the electrical cardiac data 352 and output a confidence that the value of the LV dysfunction metric (e.g., the value of the LVEF) is less than a threshold value for the LV dysfunction metric. In some examples, the confidence output by the machine learning model 354 may represent a probability in a range between 0 and 1 that the electrical cardiac data 352 corresponding to the patient indicates that the patient is associated with a value of the LV dysfunction metric that is less than a threshold value for the LV dysfunction metric. In some examples, the machine learning model 354 may process the electrical cardiac data 352 and output a confidence that the value of the LV dysfunction metric (e.g., the value of the LVEF) is greater than or equal to a threshold value for the LV dysfunction metric.

[0122] The machine learning model 354 may be configured to process the electrical cardiac data of the electrical cardiac data 352 from any patient to determine the value of the LV dysfunction metric. For example, the machine learning model 354 may process a set of electrical cardiac data corresponding to a patient at rest to determine the value of the LV dysfunction metric. In another example, the machine learning model 354 may process a set of electrical cardiac data corresponding to a patient who is active to determine the value of the LV dysfunction metric. In another example, the machine learning model 354 may process a set of electrical cardiac data corresponding to a patient who exhibits symptoms of heart failure to determine the value of the LV dysfunction metric. The machine learning model 354 may process a set of electrical cardiac data corresponding to a patient in any condition to determine the value of the LV dysfunction metric.

[0123] In some examples, the model configuration service 332 may use the paired electrical cardiac training data 356 and the LV dysfunction metric training data 358 (e.g., point-in-time paired ECG-LVEF data) to develop the machine learning model 354 as a classification model for determining whether the value of the LV dysfunction metric is small or large (e.g., LVEF less than 35%). In some examples, the model configuration service 332 may use the paired electrical cardiac training data 356 and the LV dysfunction metric training data 358 (e.g., point-in-time paired ECG-LVEF data) to develop the machine learning model 354 as a regression model for estimating the exact value of the LV dysfunction metric (e.g., LVEF=64%). When the machine learning model 354 is a classification model, the machine learning model 354 can output a confidence that the value of the LV dysfunction metric is small and / or large than a threshold. When the machine learning model 354 is a regression model, the machine learning model 354 can output an accurate estimate of the value of the LV dysfunction metric.

[0124] The model configuration service 332 may collect longitudinal electrical cardiac training data (e.g., longitudinal ECG data), point-in-time pairings of electrical cardiac data with LV dysfunction metric data, and longitudinal LV dysfunction metric data (e.g., longitudinal EF data) to develop the machine learning model 354 as a classification model and / or a regression model, as the case may be. In one example, the machine learning model 354 may collect longitudinal electrical cardiac training data (e.g., longitudinal ECG data), point-in-time pairings of electrical cardiac data with LV dysfunction metric data, and longitudinal LV dysfunction metric data (e.g., longitudinal EF data) to develop the machine learning model 354 as a classification model and / or a regression model. n A set of ECG data (e.g., ECG(t1), ECG(t2), ...ECG(t n )) at time t m A set of LVEF data (e.g., LVEF(t)) from the LV dysfunction metric training data 358 in m )) to create sets of training data from training data 355. The paired sets of electrical cardiac training data and LV dysfunction metric training data may correspond to different time points, the same time point, overlapping time points and / or starting time windows, or any combination thereof.

[0125] In some examples, the model configuration service 332 can train the machine learning model 354 to be a classification model configured to output a confidence that the LVEF is less than a threshold LVEF. In some examples, the threshold LVEF is 35%, but this is not required. The threshold LVEF may be any value. In some examples, if the LVEF is less than the threshold LVEF, this may indicate that the patient is at risk of experiencing one or more conditions, such as heart failure. In some examples, a patient's LVEF may change depending on one or more factors, such as activity level, posture, whether the patient is awake or asleep, among other factors.

[0126] When model configuration service 332 trains machine learning model 354 to be a classification model, model configuration service 332 may use point-in-time paired sets of electrical cardiac training data and LV dysfunction metric training data. That is, model configuration service 332 can train machine learning model 354 using multiple sets of training data from training data 355, each set of training data including a set of electrical cardiac training data from electrical cardiac training data 356 and a set of LV dysfunction metric training data from LV dysfunction metric training data 358 collected from the patient at the same time, during the same time window, or during overlapping time windows. When model configuration service 332 trains machine learning model 354 using point-in-time paired training data to be a classification model configured to output a confidence level that an LVEF is less than an LVEF threshold, machine learning model 354 can process the electrical cardiac data from electrical cardiac data 352 collected from the patient and output a confidence level that the patient's LVEF is less than an LVEF threshold at the time or time window at which the electrical cardiac data was collected from the patient.

[0127] In some examples, the machine learning model 354 may accept as input a set of ECG data from the electrical cardiac data 352 corresponding to time k and output an estimate of the LVEF at time m. The machine learning model 354 may operate according to the following equation: EF(t m ):F(ECG(t k ))~=EF(t m ), (mk) < 1 week (Eq. 1)

[0128] As seen in Equation 1 above, the machine learning model 354 can accept as input a set of ECG data from the electrical cardiac data 352 corresponding to time k and output an estimate of LVEF at time m, where time k occurs less than one week before time m. That is, the model configuration service 332 can train the machine learning model 354 based on paired sets of electrical cardiac training data 356 and LV dysfunction metric training data 358. The set of electrical cardiac data corresponding to each paired set of electrical cardiac training data and LV dysfunction metric training data may be collected from the patient less than one week before the set of LV dysfunction metric training data is collected from the patient. This means that once the machine learning model 354 is trained, the machine learning model 354 can accept as input a set of ECG data from the electrical cardiac data 352 collected from the patient and output a confidence level that the value of the patient's LV dysfunction metric will be less than the LV dysfunction metric threshold less than one week after the set of ECG data is collected from the patient.

[0129] In some examples, the model configuration service 332 can train the machine learning model 354 based on one or more sets of training data, each including two or more sets of ECG data among the electrical cardiac training data 356 collected from the patient paired with a set of LV dysfunction metric training data among the LV dysfunction metric training data 358 collected from the patient. In some examples, each set of ECG data among the two or more sets of ECG data may correspond to a point in time or a time window, and the set of LV dysfunction metric training data may correspond to a point in time or a time window. In some examples, each of the two or more sets of ECG data may be collected from the patient before the set of LV dysfunction metric training data is collected from the patient. In some examples, each of the two or more sets of ECG data may be collected from the patient after the set of LV dysfunction metric training data is collected from the patient. In some examples, some of the two or more sets of ECG data may be collected from the patient before the set of LV dysfunction metric training data is collected from the patient, and some of the two or more sets of ECG data may be collected from the patient after the set of LV dysfunction metric training data is collected from the patient.

[0130] This means that once the machine learning model 354 is trained, it may output a value for a metric of LV dysfunction based on two or more sets of ECG data from the electrical cardiac data 352 collected from the patient. In some examples, when the machine learning model 354 is trained based on a set of training data that includes two or more sets of ECG data collected from the same patient, the machine learning model 354 may be more robust compared to when the machine learning model 354 is trained based on a set of training data that includes a single set of ECG data. The machine learning model 354, in some examples, may operate according to the following equation: F(ECG(t k ),ECG(tk -1),ECG(t k -2),ECG(t k +1))~EF(tm) where mk, (m-(k-1)), (m-(k-2)), (m-(k+1))<1 week (Formula 2)

[0131] As seen in Equation 2 above, the machine learning model 354(F) may accept as input a set of ECG data from the electrical cardiac data 352 collected from the patient at time k, a set of ECG data from the electrical cardiac data 352 collected from the patient at time k−1, a set of ECG data from the electrical cardiac data 352 collected from the patient at time k−2, and a set of ECG data from the electrical cardiac data 352 collected from the patient at time k+1. The machine learning model 354 may output a value of the LV dysfunction metric corresponding to the patient at time m. The value of the LV dysfunction metric may, in some examples, represent a classification of whether the LV dysfunction metric is less than an LV dysfunction threshold metric. The value of the LV dysfunction metric may, in some examples, represent an accurate estimate of the value of the LV dysfunction metric. In some examples, time k, time k−1, time k−2, and time k+1 may each be less than one week before time m.

[0132] The model configuration service 332, in some examples, can use the longitudinal data to train the machine learning model 354 and output an ECG-based classification of whether the LVEF is low or high (e.g., a confidence that the LVEF is lower than an LVEF threshold). The model configuration service 332, in some examples, can use the longitudinal data to train the machine learning model 354 and output an ECG-based determination of whether the LVEF has significantly increased or significantly decreased. For example, the machine learning model 354 can receive one or more sets of ECG data collected from a patient and output a determination of whether the patient's LVEF has increased beyond an LVEF increase threshold or decreased beyond an LVEF decrease threshold.

[0133] In some examples, the data analyzer 330 can identify one or more parameters and / or characteristics indicated by the electrical cardiac training data 356, such as heart rate, heart rate variability, arrhythmia, R waves, T waves, P waves, or any combination thereof. The model configuration service 332 can develop a machine learning model 354 to perform one or more LVEF classifications and / or one or more EF trend detections based on the parameter ranges. For example, the model configuration service 332 can train the machine learning model 354 to process the input electrical cardiac data differently based on heart rate, heart rate variability, or arrhythmia.

[0134] In some instances, a patient's LVEF may not change instantaneously over a short period of time, and LVEF changes may occur infrequently. This means that there may be hysteresis associated with increasing and decreasing trends in LVEF. The data analyzer 330 may quantify two parameters from the long-term EF data: hysteresis during EF decrease and hysteresis during EF increase. These two parameters may be used as parameters for ECG-based EF change detection to detect physiologically realistic EF changes while reducing the frequency of false positives.

[0135] In some examples, the ICM ECG-LVEF paired measurements in the training data 355 may be limited, while the Holter ECG-LVEF paired measurements in the training data 355 may be more readily available. In some examples, the training data 355 may include one or more paired Holter / wearable ECG-LVEF data sets. In some examples, the training data 355 may include one or more paired Holter / wearable ECG-ICM ECG data sets. That is, one or more sets of the training data 355 may include ECG data collected from a patient via a Holter electrocardiograph, a wearable device, an IMD (e.g., an ICM), or any combination thereof. For example, the HMD 26 may map M from the wearable / Holter ECG data to the ICM ECG. In some examples, the HMD 26 may map the ICM ECG to the Holter ECG according to the formula: ICM ECG = M(Holter ECG). In some examples, the HMD 26 may map M according to the formula: ICM ECG = 0.75 * Lead II Holter ECG +0.25 * According to the lead I Holter ECG, the ICM ECG can be mapped to the Holter ECG. This mapping can be used on the dataset D1 to train a machine learning model 354 to map M(D1 Holter ECG)~=D1 EF. After the machine learning model 354 is trained, it can be used for subsequent ICM ECG-based EF classification.

[0136] Patient mortality in heart failure may be correlated with contractile reserve. The term "contractile reserve" may refer to the difference between myocardial contractility and myocardial stress. In some examples, a higher contractile reserve may be associated with a lower mortality rate, and a lower contractile reserve may be associated with a higher mortality rate. The HMS 26 can estimate the contractile reserve of a heart failure patient based on ECG-derived LVEF, ECG-derived autonomic tone, medication information, activity information, or any combination thereof. In other words, the HMS 26 can monitor a metric of LV dysfunction (e.g., LVEF) of the patient 4 based on the electrical cardiac data 352 (e.g., ECG data) to determine the patient 4's contractile reserve and monitor the patient 4's risk of mortality.

[0137] The ECG-derived LVEF may represent a metric for monitoring HF worsening in patients with reduced LVEF. For example, if patient 4 has reduced LVEF, the HMS 26 may track the patient's LVEF based on the electrical cardiac data 352 collected from patient 4. The HMS 26 may identify one or more trends in LVEF based on the electrical cardiac data 352 collected from patient 4. If the one or more trends in LVEF indicate a decrease in LVEF, this may indicate worsening heart failure. If the one or more trends in LVEF indicate an increase in LVEF, this may indicate an improvement in heart failure. In some examples, the medical device system 2 may collect electrical cardiac data using a multi-lead ICM system. To preserve battery life, the ICM may be used primarily in a single-channel arrhythmia detection mode. For LVEF detection, a multi-channel ECG may be recorded periodically (e.g., once a day, once a week, or according to any other interval) for LVEF classification via post-processing.

[0138] A system for ECG-based EF monitoring can include a single-lead ICM device. During the implant procedure, real-time LVEF estimation metrics can be calculated to determine whether LVEF can be estimated based on data collected by the ICM in the implanted orientation. Because arrhythmia monitoring can function across multiple ICM orientations, the implant process can be optimized for ECG-based EF determination. In some examples, instead of analyzing routine ECG episodes from the device for LVEF estimation on the cloud (e.g., on the computing system 20), the system for ECG-based LVEF monitoring can use a highly sensitive onboard algorithm (e.g., an onboard IMD 10) to detect low LVEF or significant LVEF changes. The ECG collected by the IMD 10 can be post-processed by the HMD 26 to reduce false positives.

[0139] The HMS 26 can use ICM-ECG-based low LVEF detection to identify one or more patients who require further diagnosis and monitoring with echocardiography and / or cardiac MRI. A system for ECG-based LVEF can include a device that is configurable with respect to the number or frequency of ECG measurements (e.g., on-demand or once per period), the number of leads, and the ECG sampling rate. This can enable the system to switch EF monitoring modes depending on the patient monitoring need (e.g., infrequent monitoring before hospitalization for HF prediction, frequent monitoring during hospitalization and one week after discharge). The HMS 26 can identify whether a HF patient has a low or high LVEF for appropriate treatment follow-up.

[0140] 6 is a conceptual diagram illustrating an example machine learning model 400 configured to output information corresponding to electrical cardiac data, in accordance with one or more techniques of the present disclosure. The machine learning model 400 is an example of a deep learning model or algorithm. One or more of the IMD 10, the computing device 12, or the computing system 20 (e.g., the model configuration service 272 and / or the model configuration service 332) may train, store, and / or utilize the machine learning model 400, although in other examples, other devices may apply inputs associated with a particular patient to the machine learning model 400. Some non-limiting examples of machine learning techniques include Bayesian probability models, Hawkes processes, support vector machines, K-nearest neighbor algorithms, and multi-layer perceptrons.

[0141] 6, the machine learning model 400 can include an input layer 402, a hidden layer 404, and an output layer 406. The output layer 406 includes an output from a transfer function 405 of the output layer 406. The input layer 402 represents each of the input values ​​X1-X4 provided to the machine learning model 400. The number of inputs can be less than four or can be much more than four, for example, greater than four, including hundreds or thousands. In some examples, the input values ​​can be parameters determined based on electrical cardiac data 184, 292, 352, including those described herein, and possibly other parameter data 182, 290, 350.

[0142] Each input value for each node in the input layer 402 is provided to a respective node in the hidden layer 404. In the example of FIG. 6, the hidden layer 404 includes two layers, one with four nodes and the other with three nodes, although in other examples, fewer or more nodes may be used. Each input from the input layer 402 is multiplied by a weight and then summed at each node in the hidden layer 404. During training of the machine learning model 400, the weight for each input is adjusted to establish a relationship between the input physiological parameter value and one or more output values ​​indicative of the patient's health status. In some examples, one hidden layer may be incorporated into the machine learning model 400, or three or more hidden layers may be incorporated into the machine learning model 400, with each layer including the same or different number of nodes.

[0143] The result of each node in the hidden layer 404 is applied to a transfer function in the output layer 406. The transfer function may be linear or nonlinear, depending on the number of layers in the machine learning model 400. An exemplary nonlinear transfer function may be a sigmoid function or a rectifier function. The output 407 of the transfer function may be one or more values ​​indicating a classification of whether the LV dysfunction metric is below a threshold and / or a determination of the exact value of the LV dysfunction metric. By applying data to a machine learning model such as machine learning model 400, the processing circuitry of the medical device system 2 can determine information corresponding to the patient 4 with high specificity and sensitivity.

[0144] FIG. 7 is a block diagram illustrating an example of a machine learning model 400 being trained using supervised learning and / or reinforcement learning, in accordance with one or more techniques of this disclosure. The machine learning model 400 may be implemented using any number of models for supervised learning and / or reinforcement learning, such as, but not limited to, an artificial neural network, a decision tree, a naive Bayes network, a support vector machine, or a k-nearest neighbor model, to name a few examples. In some examples, one or more processing circuits of the IMD 10, the external device 12, and / or the computing system 20 first train the machine learning model 400 based on training data 500. The training data may, in some examples, include the training data 355 of FIG. 5. The output of the machine learning model 400 may be compared 504 to a target output 503, for example, as determined based on the labels. Based on an error signal representing the comparison, a processing circuit implementing a learning / training function 505 may send or apply a modification to the weights of the machine learning model 400 or otherwise modify / update the machine learning model 400. For example, one or more of IMD 10, external device 12, and / or computing system 20 may modify machine learning model 400 for each training instance in training set 500 to change the output generated by machine learning model 400 in response to data applied to machine learning model 400.

[0145] 8 is a flow diagram illustrating an example method for determining whether to use a machine learning model to process electrical cardiac data collected from a patient, in accordance with one or more techniques of the present disclosure. FIG. 8 is described with respect to medical device system 2 of FIG. 1. However, the techniques of FIG. 8 may be performed by different components of medical device system 2 or by additional or alternative medical devices.

[0146] The HMS 26 can train models and personalize and / or calibrate them to specific patients. For example, the HMS 26 can train a machine learning model to implement an ECG-based algorithm (e.g., for low LVEF classification) and validate the machine learning model using a large dataset. The HMS 26 can apply the machine learning model to electrical cardiac data collected from the patient for future use. In some examples, the HMS 26 can acquire the patient's ECG and paired EF. If the machine learning model performs adequately for the patient, it may be used as is. If the machine learning model does not perform adequately for the patient, it may be adjusted using an ECG-LVEF paired data measurement and up to N ECG-LVEF paired data measurements from that patient. A personalized algorithm may be used for the patient, where the algorithm parameters can be adjusted for that patient using patient-specific data to achieve adequate performance. If the machine learning model has been updated for N iterations and remains unsatisfactory, the machine learning model may be flagged as not applicable to the patient, and data may be collated for an overall model update. "Adequate performance" may correspond to the ability to detect low LVEF or the ability to detect significant changes in LVEF.

[0147] HMS 26 may receive 802 training data including electrical cardiac data and LV dysfunction metric data. In some examples, the electrical cardiac data and LV dysfunction metric data may correspond to a patient, such as patient 4. HMS 26 may train 804 a machine learning model based on the electrical cardiac data and LV dysfunction metric data corresponding to patient 4. In some examples, HMS 26 may train the machine learning model based on a set of training data including one or more paired sets of electrical cardiac data and LV dysfunction metric data corresponding to one or more human patients other than patient 4. HMS 26 may, in some examples, receive a set of incoming electrical cardiac data from patient 4.

[0148] The HMS 26 may apply 806 a machine learning model to the incoming electrical cardiac data corresponding to the patient 4 to determine a value for the LV dysfunction metric. In some examples, the machine learning model may represent a classification model configured to output a confidence that the value of the LV dysfunction metric is lower than a threshold value for the LV dysfunction metric. In some examples, the machine learning model may represent a regression model configured to output an estimate of the exact value of the LV dysfunction metric.

[0149] HMS26 may determine 808 whether the machine learning model performs adequately for patient 4 based on the output from the machine learning model. In some examples, to determine whether the machine learning model performs adequately for patient 4, HMS26 may compare the output from the machine learning model to LV dysfunction metric data corresponding to patient 4. When the machine learning model performs adequately for patient 4 (“Yes” at block 808), HMS26 may process the electrical cardiac data corresponding to patient 4 using the machine learning model.

[0150] When the machine learning model does not perform adequately for patient 4 (“no” at block 808), HMS26 may determine whether the machine learning model has been updated at least N times (810). When the machine learning model has been updated at least N times (“yes” at block 810), HMS26 may refuse to use the machine learning model to process data corresponding to patient 4 (812). When the machine learning model has not been updated at least N times (“no” at block 810), HMS26 may update the machine learning model (814). In some examples, to update the machine learning model, HMS26 may retrain the machine learning model with additional training data corresponding to patient 4. In some examples, to update the machine learning model, HMS26 may retrain the machine learning model with additional training data corresponding to one or more patients other than patient 4. Once HMS26 has updated the machine learning model, the process of FIG. 8 may return to block 808.

[0151] 9 is a flow diagram illustrating an example method for applying a machine learning model to determine values ​​of a metric of LV dysfunction, according to one or more techniques of the present disclosure. FIG. 9 is described with respect to medical device system 2 of FIG. 1. However, the techniques of FIG. 9 may be performed by different components of medical device system 2 or by additional or alternative medical devices.

[0152] The IMD 10 may generate electrical cardiac data based on cardiac signals sensed by the IMD 10 via one or more electrodes (902). The HMS 26 may apply a machine learning model to the electrical cardiac data to determine values ​​of LV dysfunction metrics (904). The techniques of this disclosure are not limited to the HMS 26 applying a machine learning model to the electrical cardiac data. The computing devices 12A, 12B and / or the IMD 10 may apply the machine learning model in some cases. In some examples, the machine learning model is trained based on multiple sets of training data. Each set of training data among the multiple sets of training data may include a set of training electrical cardiac data and information indicative of one or more values ​​of the LV dysfunction metrics corresponding to the set of training electrical cardiac data. The HMS 26 may output the determined values ​​of the LV dysfunction metrics to a computing device (e.g., computing device 12A, 12B) (906).

[0153] The following numbered sections may represent one or more aspects of the disclosure.

[0154] Paragraph 1: A medical device system includes a medical device including one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes; a memory configured to store a machine learning model and multiple sets of training data; and a processing circuit in communication with the memory. The processing circuit is configured to apply the machine learning model to the electrical cardiac data to determine values ​​of a metric of left ventricular (LV) dysfunction. The machine learning model is trained based on multiple sets of training data, each set of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the metric of LV dysfunction corresponding to the set of training electrical cardiac data. Additionally, the processing circuit is configured to output the determined values ​​of the metric of LV dysfunction to a computing device.

[0155] Item 2: The medical device system of item 1, wherein the electrical cardiac data includes a plurality of sets of electrical cardiac data, and wherein, to apply a machine learning model to the electrical cardiac data to determine a value of a metric of LV dysfunction, the processing circuitry is configured to apply the machine learning model to a set of electrical cardiac data among the plurality of sets of electrical cardiac data to determine a value of the metric of LV dysfunction corresponding to the set of electrical cardiac data among the plurality of sets of electrical cardiac data, and the processing circuitry is further configured to apply the machine learning model to each other set of electrical cardiac data among the plurality of sets of electrical cardiac data to determine a value of the metric of LV dysfunction corresponding to each other set of electrical cardiac data among the plurality of sets of electrical cardiac data, and output the determined value of the metric of LV dysfunction corresponding to each other set of electrical cardiac data among the plurality of sets of electrical cardiac data to the computing device.

[0156] Clause 3: The medical device system described in clause 1 or clause 2, wherein the medical device further comprises an accelerometer, the medical device further configured to generate motion data based on a motion signal sensed by the accelerometer, and the processing circuitry is further configured to determine a motion value indicative of the patient's activity level based on the motion data, the motion value corresponding to the determined value of the LV dysfunction metric, and to determine whether to output an alert based on the motion value and the determined value of the LV dysfunction metric.

[0157] Item 4: The medical device system described in Item 3, wherein to determine whether to output an alert, the processing circuit is configured to determine whether the value of the LV dysfunction metric is less than a threshold metric for LV dysfunction, determine whether the motion value is greater than a threshold motion value, and decide whether to output an alert based on whether the value of the LV dysfunction metric is less than the threshold metric for LV dysfunction and whether the motion value is greater than the threshold motion value.

[0158] Clause 5: A medical device system described in any of clauses 1 to 4, wherein the processing circuitry is further configured to train a machine learning model based on a plurality of sets of training data, and by training the machine learning model based on the plurality of sets of training data, the processing circuitry is configured to cause the machine learning model to recognize one or more patterns corresponding to one or more characteristics of a metric of LV dysfunction and the electrical cardiac data.

[0159] Clause 6: A medical device system described in any of clauses 1 to 5, wherein the processing circuitry is further configured to label each set of training data among the plurality of sets of training data.

[0160] Clause 7: The medical device system of clause 6, wherein to label each set of training data among the plurality of sets of training data, the processing circuitry is configured to identify, for each set of training data among the plurality of sets of training data, one or more characteristics of the set of training electrical cardiac data among the set of training data, and label the set of training electrical cardiac data among each set of training data among the plurality of sets of training data with the one or more characteristics of the set of training electrical cardiac data.

[0161] Clause 8: The medical device system described in clause 7, wherein the one or more characteristics of the set of training electrical cardiac data include any one or more of one or more R waves, one or more P waves, one or more T waves, a heart rate corresponding to the set of training electrical cardiac data, heart rate variability corresponding to the set of training electrical cardiac data, and arrhythmias indicated by the set of training electrical cardiac data.

[0162] Clause 9: A medical device system as described in clause 6 or clause 7, wherein, to label each set of training data among the plurality of sets of training data, the processing circuitry is configured to identify, in the information indicating the values ​​of one or more LV dysfunction metrics of the set of training data, a time corresponding to each value among the one or more values ​​of the LV dysfunction metrics, and associate the time corresponding to each value among the one or more values ​​of the LV dysfunction metrics with a time in the set of training electrical cardiac data.

[0163] Clause 10: A medical device system described in any of clauses 1 to 9, wherein the processing circuitry is configured to apply a machine learning model to the electrical cardiac data to determine a value of LV dysfunction, to determine a confidence level that the value of LV dysfunction is lower than an LV dysfunction threshold.

[0164] Item 11: A medical device system described in any of items 1 to 10, wherein the metric of LV dysfunction includes ejection fraction.

[0165] Clause 12: A method of operating a medical device system including a medical device having one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes, the method comprising: applying, by a processing circuit of the medical device system, a machine learning model to the electrical cardiac data to determine values ​​of a metric of left ventricular (LV) dysfunction, wherein the machine learning model is trained based on a plurality of sets of training data, each set of training data of the plurality of sets of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the metric of LV dysfunction corresponding to the set of training electrical cardiac data, the processing circuit being in communication with a memory of the medical device system, the memory being configured to store the machine learning model and the plurality of sets of training data; and outputting, by the processing circuit, the determined values ​​of the metric of LV dysfunction to a computing device.

[0166] Clause 13: The method of clause 12, wherein the medical device further comprises an accelerometer, and the medical device is further configured to generate motion data based on a motion signal sensed by the accelerometer, and the method further includes determining, by the processing circuitry, a motion value indicative of the patient's activity level based on the motion data, the motion value corresponding to the determined value of the LV dysfunction metric, and determining, by the processing circuitry, whether to output an alert based on the motion value and the determined value of the LV dysfunction metric.

[0167] Clause 14: The method of clause 12 or clause 13, wherein the medical device further comprises an accelerometer, and the medical device is further configured to generate motion data based on a motion signal sensed by the accelerometer, and the method further includes determining, by the processing circuitry, a motion value indicative of the patient's activity level based on the motion data, the motion value corresponding to the determined value of the LV dysfunction metric, and determining, by the processing circuitry, whether to output an alert based on the motion value and the determined value of the LV dysfunction metric.

[0168] Clause 15: The method of clause 14, wherein determining whether to output an alert includes: determining, by a processing circuit, whether the value of the LV dysfunction metric is less than a threshold LV dysfunction metric; determining, by a processing circuit, whether the motion value is greater than a threshold motion value; and determining, by a processing circuit, whether to output an alert based on whether the value of the LV dysfunction metric is less than the threshold LV dysfunction metric and whether the motion value is greater than the threshold motion value.

[0169] Clause 16: The method of any of clauses 12 to 15, further comprising training, by the processing circuitry, a machine learning model based on the plurality of sets of training data, wherein by training the machine learning model based on the plurality of sets of training data, the method includes causing the processing circuitry to recognize one or more patterns corresponding to one or more characteristics of a metric of LV dysfunction and the electrical cardiac data.

[0170] Clause 17: The method of any of clauses 12 to 16, further comprising labeling, by a processing circuit, each set of training data among the plurality of sets of training data.

[0171] Clause 18: The method of clause 17, wherein labeling each set of training data among the plurality of sets of training data includes: identifying, by the processing circuitry, for each set of training data among the plurality of sets of training data, one or more characteristics of the set of training electrical cardiac data among the set of training data; and labeling, by the processing circuitry, the set of training electrical cardiac data among each set of training data among the plurality of sets of training data with the one or more characteristics of the set of training electrical cardiac data.

[0172] Clause 19: The method of clause 17 or clause 18, wherein labeling each set of training data among the plurality of sets of training data includes: identifying, by the processing circuitry, in information indicating values ​​of one or more LV dysfunction metrics in the set of training data, a time corresponding to each value among the one or more values ​​of the LV dysfunction metrics; and associating, by the processing circuitry, the time corresponding to each value among the one or more values ​​of the LV dysfunction metrics with a time in the set of training electrical cardiac data.

[0173] Clause 20: A non-transitory computer-readable storage medium including program instructions that, when executed by a processing circuit of a medical device system comprising one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes, cause the processing circuit to apply a machine learning model to the electrical cardiac data to determine values ​​of metrics of left ventricular (LV) dysfunction, the machine learning model being trained based on a plurality of sets of training data, each set of training data of the plurality of sets of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the metrics of LV dysfunction corresponding to the set of training electrical cardiac data, the processing circuit being in communication with a memory of the medical device system, the memory being configured to store the machine learning model and the plurality of sets of training data, and causing the processing circuit to output the determined values ​​of the metrics of LV dysfunction to a computing device.

[0174] 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 present technology may be implemented within one or more microprocessors, DSPs, ASICs, FPGAs, or any other equivalent integrated or discrete logic QRS circuits, as well as any combination of such components, embodied in an external device such as a physician or patient programmer, stimulator, or other device. The terms "processor" or "processing circuitry" may generally refer to any of the foregoing logic circuits, alone or in combination with other logic circuits, or any other equivalent circuitry, alone or in combination with other digital or analog circuits.

[0175] For software-implemented aspects, at least a portion of the functionality attributed to the systems and devices described in this disclosure may be embodied as instructions on a computer-readable storage medium, such as in the form of RAM, DRAM, SRAM, magnetic disk, optical disk, flash memory, or EPROM or EEPROM, which may be executed to support one or more aspects of the functionality described in this disclosure.

[0176] Additionally, in some aspects, functionality described herein may be provided within dedicated hardware and / or software modules. The depiction of different features as modules or units is intended to emphasize different functional aspects and does not necessarily imply that such modules or units must be implemented by separate hardware or software components. Rather, functionality associated with one or more modules or units may be performed by separate hardware or software components, or may be integrated within common or separate hardware or software components. The technology may also be implemented entirely in one or more circuits or logic elements. The technology of this disclosure may be implemented in a wide variety of devices or apparatuses present in an IMD and / or external programmer, including an IMD, an external programmer, a combination of an IMD and external programmer, an integrated circuit (IC) or set of ICs, and / or discrete electrical circuits.

Claims

1. 1. A medical device system comprising: a medical device comprising one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from the patient via the one or more electrodes; a memory configured to store a machine learning model and a plurality of sets of training data; a processing circuit in communication with the memory, applying the machine learning model to the electrical cardiac data to determine values ​​of left ventricular (LV) dysfunction metrics, the machine learning model being trained based on a plurality of sets of training data, each set of training data of the plurality of sets of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the LV dysfunction metrics corresponding to the set of training electrical cardiac data; and a processing circuit configured to output the determined value of the LV dysfunction metric to a computing device.

2. The electrical cardiac data includes a plurality of sets of electrical cardiac data, and to apply the machine learning model to the electrical cardiac data to determine the value of the LV dysfunction metric, the processing circuitry: configured to apply the machine learning model to a set of electrocardial data among the plurality of sets of electrocardial data to determine the value of the LV dysfunction metric corresponding to the set of electrocardial data among the plurality of sets of electrocardial data; the processing circuitry applying the machine learning model to each other set of electrocardiac data among the plurality of sets of electrocardiac data to determine a value of the LV dysfunction metric corresponding to each other set of electrocardiac data among the plurality of sets of electrocardiac data; 10. The medical device system of claim 1, further configured to output to the computing device the determined value of the LV dysfunction metric corresponding to each other set of electrical cardiac data among the plurality of sets of electrical cardiac data.

3. The medical device further comprises an accelerometer, the medical device further configured to generate motion data based on motion signals sensed by the accelerometer, and the processing circuitry determining a motion value indicative of an activity level of the patient based on the motion data, the motion value corresponding to the determined value of the LV dysfunction metric; The medical device system of claim 1 , further configured to determine whether to output an alert based on the motion value and the determined value of the LV dysfunction metric.

4. To determine whether to output the alert, the processing circuitry: determining whether the value of the LV dysfunction metric is less than a threshold LV dysfunction metric; determining whether the motion value is greater than a threshold motion value; The medical device system of claim 3, configured to determine whether to output the alert based on whether the value of the LV dysfunction metric is less than a threshold metric of the LV dysfunction and whether the motion value is greater than the threshold motion value.

5. the processing circuitry is further configured to train the machine learning model based on the plurality of sets of training data; 2. The medical device system of claim 1, wherein the processing circuitry is configured to train the machine learning model based on a plurality of sets of the training data, so as to cause the machine learning model to recognize one or more patterns corresponding to one or more characteristics of the LV dysfunction metrics and electrical cardiac data.

6. The medical device system of claim 1 , wherein the processing circuitry is further configured to label each set of training data among the plurality of sets of training data.

7. To label each set of training data among the plurality of sets of training data, the processing circuitry: for each set of training data among the plurality of sets of training data, identifying one or more characteristics of the set of training electrical cardiac data among the sets of training data; 7. The medical device system of claim 6, configured to label the set of training electrical cardiac data in each set of training data in the plurality of sets of training data with the one or more characteristics of the set of training electrical cardiac data.

8. 8. The medical device system of claim 7, wherein the one or more characteristics of the set of training electrical cardiac data include any one or more of one or more R waves, one or more P waves, one or more T waves, a heart rate corresponding to the set of training electrical cardiac data, heart rate variability corresponding to the set of training electrical cardiac data, and arrhythmias exhibited by the set of training electrical cardiac data.

9. To label each set of training data among the plurality of sets of training data, the processing circuitry: identifying, in the information indicative of one or more values ​​of the LV dysfunction metric of the set of training data, a time corresponding to each value of the one or more values ​​of the LV dysfunction metric; 7. The medical device system of claim 6, configured to associate the time corresponding to each value of the one or more values ​​of the LV dysfunction metric with a time of the set of training electrical cardiac data.

10. 2. The medical device system of claim 1, wherein the processing circuitry is configured to apply the machine learning model to the electrical cardiac data to determine the value of LV dysfunction, and to apply the machine learning model to the electrical cardiac data to determine a confidence level that the value of LV dysfunction is below an LV dysfunction threshold.

11. The medical device system of claim 1 , wherein the LV dysfunction metric includes an ejection fraction.

12. 1. A method of operating a medical device system including a medical device comprising one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes, comprising: applying, by a processing circuit of the medical device system, a machine learning model to the electrical cardiac data to determine values ​​of left ventricular (LV) dysfunction metrics, the machine learning model being trained based on a plurality of sets of training data, each set of training data of the plurality of sets of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the LV dysfunction metrics corresponding to the set of training electrical cardiac data, the processing circuit being in communication with a memory of the medical device system, the memory being configured to store the machine learning model and the plurality of sets of training data; outputting, by the processing circuitry, the determined value of the LV dysfunction metric to a computing device.

13. The medical device further comprises an accelerometer, the medical device further configured to generate motion data based on motion signals sensed by the accelerometer, and the method further comprises: determining, by the processing circuitry, a motion value indicative of an activity level of the patient based on the motion data, the motion value corresponding to the determined value of the LV dysfunction metric; 13. The method of claim 12, further comprising determining, by the processing circuitry, whether to output an alert based on the motion value and the determined value of the LV dysfunction metric.

14. The medical device further comprises an accelerometer, the medical device further configured to generate motion data based on motion signals sensed by the accelerometer, and the method further comprises: determining, by the processing circuitry, a motion value indicative of an activity level of the patient based on the motion data, the motion value corresponding to the determined value of the LV dysfunction metric; 13. The method of claim 12, further comprising determining, by the processing circuitry, whether to output an alert based on the motion value and the determined value of the LV dysfunction metric.

15. determining whether to output the alert includes: determining, by the processing circuitry, whether the value of the LV dysfunction metric is less than a threshold LV dysfunction metric; determining, by the processing circuitry, whether the motion value is greater than a threshold motion value; and determining, by the processing circuitry, whether to output the alert based on whether the value of the LV dysfunction metric is less than the LV dysfunction threshold metric and whether the motion value is greater than the threshold motion value.

16. training, by the processing circuitry, the machine learning model based on the plurality of sets of training data; 13. The method of claim 12, wherein training the machine learning model based on the multiple sets of training data includes causing the processing circuitry to recognize one or more patterns corresponding to one or more characteristics of the LV dysfunction metric and electrical cardiac data.

17. The method of claim 12 , further comprising labeling, by the processing circuitry, each set of training data among the plurality of sets of training data.

18. Labeling each set of training data among the plurality of sets of training data includes: identifying, by the processing circuitry, for each set of training data among the plurality of sets of training data, one or more characteristics of the set of training electrical cardiac data among the sets of training data; and labeling, by the processing circuitry, the set of training electrical cardiac data in each set of training data in the plurality of sets of training data with the one or more characteristics of the set of training electrical cardiac data.

19. Labeling each set of training data among the plurality of sets of training data includes: identifying, by the processing circuitry, a time corresponding to each value of the one or more values ​​of the LV dysfunction metric in information indicative of the one or more values ​​of the LV dysfunction metric in the set of training data; and associating, by the processing circuitry, the time corresponding to each value of the one or more values ​​of the LV dysfunction metric with a time of the set of training electrocardiac data.

20. 1. A non-transitory computer-readable storage medium containing program instructions that, when executed by a processing circuit of a medical device system comprising one or more electrodes and configured to generate electrical cardiac data based on cardiac signals sensed from a patient via the one or more electrodes, cause the processing circuit to: applying a machine learning model to the electrical cardiac data to determine values ​​of left ventricular (LV) dysfunction metrics, the machine learning model being trained based on a plurality of sets of training data, each set of training data of the plurality of sets of training data including a set of training electrical cardiac data and information indicative of one or more values ​​of the LV dysfunction metrics corresponding to the set of training electrical cardiac data; the processing circuitry being in communication with a memory of the medical device system, the memory being configured to store the machine learning model and the plurality of sets of training data; A non-transitory computer-readable storage medium that causes a computing device to output the determined value of the LV dysfunction metric.

Images