Multi-layer Prediction of Cardiac Tachyarrhythmia

Abstract

Techniques for a multi-layer system for predicting arrhythmias in a patient are disclosed. In one example, a computing device processes parameterized patient data and provider data of the patient to generate a long-term probability that an arrhythmia will occur in the patient during a first time period. In response to determining that the arrhythmia is likely to occur during the first time period, the computing device causes a medical device to process the parameterized patient data to generate a short-term probability that the arrhythmia will occur in the patient during a second time period. In response to determining that the arrhythmia is likely to occur during the second time period, the medical device performs a remedial action to reduce the likelihood that the arrhythmia will occur.

Description

Technical Field

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

[0002] Malignant tachyarrhythmias, such as ventricular fibrillation, are uncoordinated contractions of the myocardium of the ventricles in the heart and are the most common arrhythmias in patients with cardiac arrest. If such an arrhythmia persists for more than a few seconds, it can lead to cardiogenic shock and the cessation of effective blood circulation. Thus, sudden cardiac death (SCD) can occur within minutes.

[0003] In patients at high risk of ventricular fibrillation, the use of implantable medical devices (IMDs), such as implantable cardioverter defibrillators (ICDs), has been shown to be beneficial in preventing SCD. An ICD is a battery-powered shock device that can include an electrical canister electrode (sometimes referred to as a can electrode) typically coupled to one or more electrical leads placed within the heart. If an arrhythmia is sensed, the ICD can send a pulse via the electrical lead to shock the heart and restore its normal rhythm. Some ICDs have been configured to attempt to terminate a detected tachyarrhythmia by delivering antitachycardia pacing (ATP) prior to delivering a shock. In addition, ICDs have been configured to deliver postshock pacing at a relatively high amplitude after the shock successfully terminates the tachyarrhythmia in order to support the heart's recovery from the shock. Some ICDs also deliver bradycardia pacing, cardiac resynchronization therapy (CRT), or other forms of pacing. Summary of the Invention

[0004] Generally, the present disclosure describes techniques for multi-layer prediction of arrhythmias in a patient. In some instances, a multi-layer system implements these techniques. In one instance, a computing device receives parameterized patient data collected by one or more electrodes and / or sensors of a patient's medical device. The computing device may additionally receive provider data of the patient from a database. In some instances, the computing device is a cloud computing system. The computing device applies a machine learning model trained using the parameterized patient data and provider data of multiple patients to the parameterized patient data and provider data of the patient to generate a long-term probability that an arrhythmia will occur in the patient within a first time period (e.g., typically from about 24 hours to about 48 hours). The computing device determines whether the long-term probability exceeds a long-term predetermined threshold, and in response to determining that the long-term probability exceeds the long-term predetermined threshold, sends an instruction to the medical device to cause the medical device to determine a short-term probability that an arrhythmia will occur in the patient within a second time period (e.g., shorter than the first time period and typically from about 30 minutes to about 60 minutes). In some instances, the long-term predetermined threshold is 50%.

[0005] In response to receiving an instruction from a computing device, a medical device processes subsequent parameterized patient data to generate a short-term probability that an arrhythmia will occur in a patient during a second time period. In response to determining that the short-term probability exceeds a short-term predetermined threshold, the medical device performs a remedial action to reduce the short-term probability that an arrhythmia will occur in the patient during the second time period. By way of example, the medical device may issue a notification of the short-term probability that an arrhythmia will occur in the patient during the second time period to the computing device, such that the patient or a clinician may become aware of the likelihood that an arrhythmia will occur in the patient during the second time period. In another example, the medical device (or another medical device local to the patient and communicating with the medical device that determines the short-term probability) initiates a drug delivery treatment or an electrical pacing treatment to reduce the likelihood that an arrhythmia will occur in the patient during the second time period. In some examples, the short-term predetermined threshold is 95%.

[0006] As a foregoing example, in response to determining that the likelihood that an arrhythmia will occur in a patient within the next 24 hours is greater than 50%, the computing device sends an instruction to the medical device to cause the medical device to determine the short-term probability. In response to determining that the likelihood that an arrhythmia will occur in the patient within the next 60 minutes is greater than 95%, the medical device performs a remedial action to reduce the likelihood that an arrhythmia will occur in the patient within the next 60 minutes.

[0007] Accordingly, the techniques disclosed herein may allow for enhanced patient care and an increased ability to prevent arrhythmias in patients. By way of example, knowledge that the likelihood of a tachyarrhythmia occurring within the next few days is relatively high may be used to help guide preventive care of a patient. Additionally, knowledge that the likelihood of a tachyarrhythmia occurring within the next few minutes or hours is relatively high may be used to guide preventive treatment or emergency care of a patient. Further, a system such as that disclosed herein may move computationally expensive and energy-consuming long-term arrhythmia prediction operations from a patient's personal medical device to a cloud computing system, and activate short-term cardiac prediction operations on the medical device only when the cloud computing system predicts that the likelihood of an arrhythmia occurring within the next few days is relatively high. Accordingly, the techniques of the present disclosure conserve power and extend the battery life of the medical device by using the medical device for cardiac prediction only when the likelihood of an arrhythmia occurring within a particular time span is relatively high.

[0008] In one example, the present disclosure describes a computing device that includes processing circuitry and a storage medium. The computing device is configured to: receive parameterized patient data of a patient; apply a machine learning model trained using parameterized patient data of multiple patients to the parameterized patient data to generate a first probability that an arrhythmia will occur in the patient within a first time period; determine that the first probability exceeds a predetermined threshold; and in response to determining that the first probability exceeds the predetermined threshold, send an instruction to a second device to cause the second device to determine a second probability that an arrhythmia will occur in the patient within a second time period.

[0009] In another example, the present disclosure describes a device that is configured to: collect parameterized patient data of a patient via one or more of a plurality of electrodes or sensors; receive an instruction from a computing device to generate a probability that an arrhythmia will occur in the patient within a certain time period; in response to the instruction, process the parameterized patient data to generate a probability that an arrhythmia will occur in the patient within a certain time period; determine that the probability exceeds a predetermined threshold; and in response to determining that the probability exceeds the predetermined threshold, perform a remedial action to reduce the probability that an arrhythmia will occur in the patient within a certain time period.

[0010] In another example, the present disclosure describes an external device that is configured to: receive an instruction to generate a first probability that an arrhythmia will occur in the patient within a first time period; in response to the instruction, process the parameterized patient data to generate a first probability that an arrhythmia will occur in the patient within a first time period; determine that the first probability exceeds a first predetermined threshold; and in response to determining that the first probability exceeds the first predetermined threshold, send an instruction to a medical device to cause the medical device to determine a second probability that an arrhythmia will occur in the patient within a second time period.

[0011] In another example, the present disclosure describes a method that includes: receiving, by a computing device including processing circuitry and a storage medium, parameterized patient data of a patient; applying, by the computing device, a machine learning model trained using parameterized patient data of multiple patients to the parameterized patient data to generate a first probability that an arrhythmia will occur in the patient within a first time period; determining, by the computing device, that the first probability exceeds a predetermined threshold; and in response to determining that the first probability exceeds the predetermined threshold, sending, by the computing device, an instruction to a second device to cause the second device to determine a second probability that an arrhythmia will occur in the patient within a second time period.

[0012] In another example, the present disclosure describes a method that includes: collecting parametric patient data of a patient by a device; receiving, by the device, from a computing device an instruction to generate a probability that an arrhythmia will occur in the patient over a period of time; in response to the instruction, processing, by the device, the parametric patient data to generate a probability that an arrhythmia will occur in the patient over a period of time; determining, by the device, that the probability exceeds a predetermined threshold; and in response to the determination that the probability exceeds the predetermined threshold, performing, by the device, a remedial action to reduce the probability that an arrhythmia will occur in the patient over a period of time.

[0013] 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 exhaustive interpretation of the devices and methods described in the following figures and description. Additional details of one or more examples are set forth in the following figures and description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A block diagram illustrating an example system for predicting arrhythmias according to the techniques of the present disclosure.

[0015] Figure 2 For a more detailed illustration Figure 1 A conceptual diagram of an IMD and leads of the system of

[0016] Figure 3 A block diagram of an example implantable medical device according to the techniques of the present disclosure.

[0017] Figure 4 A block diagram illustrating an example computing device operating in accordance with one or more techniques of the present disclosure.

[0018] Figure 5 A block diagram illustrating an example external device operating in accordance with one or more techniques of the present disclosure.

[0019] Figure 6 A flowchart illustrating an example operation according to the techniques of the present disclosure.

[0020] Figure 7 A flowchart illustrating an example operation according to the techniques of the present disclosure.

[0021] Figure 8A A flowchart illustrating an example operation according to the techniques of the present disclosure.

[0022] Figure 8B For a further detailed illustration Figure 8A A block diagram of an algorithm for long-term prediction of arrhythmias depicted in the example of

[0023] Figure 8C For a further detailed illustration Figure 8ABlock diagram of an algorithm for mid-term prediction of arrhythmia depicted in an example of

[0024] Figure 8D For further details Figure 8A Block diagram of an algorithm for short-term prediction of arrhythmia depicted in an example of

[0025] Figure 9 Block diagram of an example algorithm for short-term prediction of arrhythmia according to the technology of the present disclosure.

[0026] Throughout the drawings and the description, like reference numerals refer to like elements. Detailed description

[0027] Figure 1 Block diagram of an example system for predicting arrhythmia according to the technology of the present disclosure. System 10 includes a medical device. An example of such a medical device is Figure 1 the IMD 16 depicted in Figure 1 As illustrated by the example system 10 in

[0028] In some examples, the IMD 16 may be, for example, an implantable cardiac pacemaker, an implantable cardioverter / defibrillator (ICD), or a pacemaker / cardioverter / defibrillator. The IMD 16 is connected to leads 18, 20, and 22 and is communicatively coupled to an external device 27, which in turn is communicatively coupled to a computing device 24 on a communication network 25.

[0029] In some examples, the IMD 16 includes a communication circuitry 17 that includes any suitable circuitry, firmware, software, or any combination thereof for communicating with another device, such as Figure 1 the external device 27 of For example, the communication circuitry 17 may include one or more processors, memories, radios, antennas, transmitters, receivers, modulation and demodulation circuitry, filters, amplifiers, etc. for communicating with other devices, such as the computing device 24, via radio frequency. The IMD 16 may use the communication circuitry 17 to receive downlink data from the external device 27 to control one or more operations of the IMD 16 and / or to send uplink data to the external device 27.

[0030] Leads 18, 20, and 22 extend into the heart 12 of the patient 14 to sense the electrical activity of the heart 12 and / or deliver electrical stimulation to the heart 12. In Figure 1 the illustrated example, the right ventricle (RV) lead 18 extends through one or more veins (not shown), the superior vena cava (not shown), and the right atrium 26, and into the right ventricle 28. The left ventricle (LV) lead 20 extends through one or more veins, the vena cava, the right atrium 26, and into the coronary sinus 30, to a region adjacent to the free wall of the left ventricle 32 of the heart 12. The right atrium (RA) lead 22 extends through one or more veins and the vena cava, and into the right atrium 26 of the heart 12.

[0031] Although Figure 1 the example system 10 depicts the IMD 16, in other examples, the techniques of the present disclosure may be applied to other types of medical devices that are not necessarily implantable. For example, a medical device according to the techniques of the present disclosure may include a wearable medical device or "smart" clothing worn by the patient 14. For example, such a medical device may take the form of a watch worn by the patient 14 or a circuit system adhesively attached to the patient 14. In another example, a medical device as described herein may include an external medical device having implantable electrodes.

[0032] In some examples, the external device 27 takes the form of an external programmer or a mobile device, such as a mobile phone, "smart" phone, laptop computer, tablet computer, personal digital assistant (PDA), wearable electronic device, etc. In some examples, the external device 27 is a CareLink TM monitor available from Medtronic, Inc. A user, such as a doctor, technician, surgeon, electrophysiologist, or other clinician, may interact with the external device 27 to retrieve physiological or diagnostic information from the IMD 16. A user, such as the patient 14 or clinician as described above, may also interact with the external device 27 to program the IMD 16, e.g., select or adjust values for the operating parameters of the IMD 16. The external device 27 may include processing circuitry, memory, a user interface, and communication circuitry capable of sending information to and receiving information from each of the IMD 16 and the computing device 24.

[0033] In some instances, computing device 24 takes the form of a handheld computing device, a computer workstation, a server or other networked computing device, a smart phone, a tablet computer, or an external programmer, and includes a user interface for presenting information to a user and receiving input from the user. In some instances, computing device 24 may include one or more devices implementing a machine learning system, such as a neural network, a deep learning system, or other types of predictive analytics systems. A user, such as a doctor, technician, surgeon, electrophysiologist, or other clinician, may interact with computing device 24 to retrieve physiological or diagnostic information from IMD 16. The user may also interact with computing device 24 to program IMD 16, e.g., to select values for operating parameters of the IMD. Computing device 24 may include a processor configured to evaluate EGMs and / or other sensed signals sent from IMD 16 to computing device 24.

[0034] Network 25 may include one or more computing devices (not shown), such as one or more non-edge switches, routers, hubs, gateways, security devices (such as a firewall), intrusion detection and / or intrusion prevention devices, servers, computer terminals, laptop computers, printers, databases, wireless mobile devices (such as cellular phones or personal digital assistants), wireless access points, bridges, cable modems, application accelerators, or other network devices. Network 25 may include one or more networks managed by a service provider and may thus form part of a large-scale public network infrastructure (such as the Internet). Network 25 may provide access to the Internet to computing devices such as computing device 24 and IMD 16 and may provide a communication framework that allows the computing devices to communicate with each other. In some instances, network 25 may be a private network that provides a communication framework that allows computing device 24, IMD 16, provider database 66, and claim database 68 to communicate with each other, but isolates computing device 24, IMD 16, provider database 66, and claim database 68 from external devices for security purposes. In some instances, the communication between computing device 24, IMD 16, provider database 66, and claim database 68 is encrypted.

[0035] External device 27 and computing device 24 may communicate via wireless communication over network 25 using any technique known in the art. In some instances, computing device 24 is a remote device that communicates with external device 27 via an intermediate device (such as a local access point, wireless router, or gateway) located in network 25. Although in Figure 1 the instance where external device 27 and computing device 24 communicate over network 25, in some instances, external device 27 and computing device 24 communicate directly with each other. Examples of communication techniques may include, for example, according to or Communication using the Low Energy (BLE) protocol. Other communication technologies are also contemplated. Computing device 24 may also communicate with one or more other external devices using a variety of known wired and wireless communication technologies.

[0036] The provider database 66 stores provider data for patient 14. The claims database 68 may store claims or payer information for patient 14, such as health records stored by an insurance provider or other payer for patient 14. The provider database 66 and the claims database 68 may include processing circuitry and one or more storage media (e.g., random access memory (RAM), read only memory (ROM), programmable read only memory (PROM), erasable programmable read only memory (EPROM), electronically erasable programmable read only memory (EEPROM), or flash memory). In some instances, the provider database 66 is a cloud computing system. In some instances, the functionality of the provider database 66 is distributed across multiple computing systems.

[0037] In accordance with the techniques of the present disclosure, system 10 provides a multi-layer prediction of arrhythmias in a patient. In some instances, system 10 provides long-term and short-term predictions of ventricular arrhythmias. In some instances, the long-term time period is from about 24 hours to about 48 hours. In some instances, the short-term time period is from about 30 minutes to about 60 minutes. In some instances, the long-term time period is greater than about 1 day and less than about 1 week and the short-term time period is greater than about 1 minute and less than about 1 day. In some instances, the long-term time period is less than about 1 day and the short-term time period is less than about 1 hour.

[0038] In one example, computing device 24 receives parameterized patient data collected by a medical device of patient 14. In some examples, the parameterized patient data includes physiological data of patient 14, such as one or more of the following: activity level of patient 14, heart rate of patient 14, posture of patient 14, electrocardiogram of patient 14, blood pressure of patient 14, pulse transit time of patient 14, respiratory rate of patient 14, hypopnea index or apnea of patient 14, accelerometer data of patient 14, features derived from the accelerometer data of patient 14 (such as activity count, posture, statistically controlled process variables, etc.), raw electromyogram of patient 14, one or more features derived from the raw electromyogram of patient 14 (such as heart rate variability, t-wave alternans, QRS morphology, etc.), interval data and features derived from the interval data, heart sound, potassium level, blood glucose index, temperature of patient 14, or any data derivable from the above parameterized data, or any other type of parameterized patient data. In some examples, the parameterized patient data includes behavioral data or demographic data of patient 14, such as age of patient 14, gender of patient 14, activity pattern of patient 14, sleep pattern of patient 14, gait change of patient 14, or temperature trend of the medical device or patient 14. In some examples, the medical device can automatically generate parameterized patient data by processing information from one or more sensors. For example, the medical device can determine, via one or more sensors, an example where patient 14 has fallen, patient 14 is frail or has a disease, or patient 14 has sleep apnea.

[0039] In some examples, the parameterized patient data includes environmental data, such as air quality measurements, ozone level, particle count, or pollution level near patient 14, environmental temperature, or day time. In some examples, one of the medical device, sensor 80, or external device 27 can sense environmental data via one or more sensors. In another example, the environmental data is received by external device 27 via an application (such as a weather application) executed on external device 27 and uploaded to computing device 24 via network 25. In another example, computing device 24 directly collects environmental data from a cloud service having location-based data of patient 14.

[0040] In some examples, the parameterized patient data includes patient symptom data uploaded by patient 14 via an external device (such as external device 27). For example, patient 14 can upload patient symptom data via an application executed on a smart phone. In some examples, patient 14 can upload patient symptom data via a user interface ( Figure 1 not depicted), such as via a touch screen, keyboard, graphical user interface, voice command, etc. In other examples, an image of the patient can be obtained via a camera of external device 27 and processed to identify patient symptom data.

[0041] In some instances, the parameterized patient data includes device-related data, such as one or more of the following: the impedance of one or more electrodes of a medical device, the selection of electrodes, the drug delivery schedule of a medical device, the history of electro-pacing therapy delivered to the patient, the diagnostic data of a medical device, the detected activity level of patient 14, the detected posture of patient 14, the detected temperature of the medical device or the patient, or the detected sleep state of the patient, e.g., whether the patient is in a dormant or awake state. In some instances, the computing device 24 periodically (e.g., daily) receives the parameterized patient data. In some instances, the medical device that collects the parameterized patient data is an IMD, such as IMD 16. In other instances, the medical device that collects the parameterized patient data is another type of patient device. Examples of medical devices that can collect parameterized patient data include IMD 16, sensors 80A - 80B (collectively referred to as "sensors 80"), wearable devices, or external devices 27, such as the patient programmer, clinician programmer, or mobile device (e.g., smartphone) of patient 14. Wearable devices include wearable sensors 80A, wearable medical devices, or other wearable electronic devices.

[0042] In some instances, sensors 80 can be used to collect the parameterized patient data of patient 14. Each of sensors 80 can include one or more accelerometers, pressure sensors, optical sensors for O2 saturation, etc. In some instances, sensors 80 can sense parameterized patient data including one or more of the following: the activity level of the patient, the heart rate of the patient, the posture of the patient, the electrocardiogram of the patient, the blood pressure of the patient, the accelerometer data of the patient, or other types of parameterized patient data. In Figure 1 an instance, sensor 80A is a wearable sensor, and sensor 80B is a non-wearable sensor. As Figure 1 depicted, sensor 80A is positioned on the upper arm of patient 14. However, other types of wearable sensors 80 can be positioned on other body parts of patient 14, or incorporated into the clothing of patient 14, such as within clothes, shoes, glasses, watches or wristbands, hats, etc.

[0043] The computing device 24 additionally receives provider data of patient 14 from the provider database 66 and the claims database 68. For convenience, the term "provider data" is used throughout the text to refer to different types of medical information about patient 14, and includes data from the provider database 66, the claims database 68, or Figure 1Data stored by other health information sources not explicitly depicted. In some instances, the provider data may include many different types of historical medical information about patient 14. The historical medical information may include, for example, electronic medical record (EMR) data, electronic health record (EHR) data, data from different healthcare providers, laboratories, clinicians, insurance companies, etc. The historical medical information may be stored in multiple different databases managed by different unrelated entities. As a non-limiting example, the provider database 66 may store, as an example, the patient's medication history, the patient's surgical history, the patient's hospitalization history, the patient's potassium levels over time, one or more laboratory test results of patient 14, the cardiovascular history of patient 14, or the patient's comorbidities such as atrial fibrillation, heart failure, or diabetes. As another example, the provider database 66 may store medical images of patient 14, such as x-ray images, ultrasound images, echocardiograms, anatomical images, medical photos, radiographic images, etc. The claims database 68 may store claims or payer information for patient 14, such as health records stored by patient 14's insurance provider or other payers. Typically, the provider data is patient-specific, e.g., specifically referring to the medical history of patient 14. However, in some instances, the provider data may include broader demographic information or population type information for multiple patients. For example, the provider data may include medical records of multiple patients of one or more population types similar to patient 14 with patient-specific information removed.

[0044] The computing device 24 applies a machine learning model trained using parameterized patient data and provider data of multiple patients to the parameterized patient data and provider data of patient 14 to perform long-term monitoring of patient 14. In some instances, the computing device 24 performs long-term monitoring by generating a long-term probability that an arrhythmia will occur in patient 14 within a first time period (e.g., typically from about 24 hours to about 48 hours). The computing device 24 determines whether the long-term probability exceeds a long-term predetermined threshold, and in response to determining that the long-term probability exceeds the long-term predetermined threshold, sends an instruction for the IMD 16 to perform short-term monitoring of patient 14. In some instances, the computing device 24 sends the instruction to an external device 27, which in turn sends the instruction to the IMD 16. In some instances, a clinician selects the long-term predetermined threshold as a value to ensure high sensitivity in predicting arrhythmias. In some instances, the long-term predetermined threshold is 50%. In some instances, the long-term predetermined threshold is another threshold, such as 75%, 80%, 90%, or 95%. In some instances, the instruction causes the IMD 16 to perform short-term monitoring by determining a short-term probability that an arrhythmia will occur in patient 14 within a second time period (e.g., typically from about 30 minutes to about 60 minutes). In some instances, in response to determining that the long-term probability exceeds the long-term predetermined threshold, the computing device 24 may perform other actions, such as notifying a clinician that the long-term probability has exceeded the long-term predetermined threshold or that the computing system 24 has determined that patient 14 is likely to have an arrhythmia within the first time period.

[0045] In response to receiving an instruction from the computing device 24, the IMD 16 processes subsequent parameterized patient data to generate a short-term probability that an arrhythmia will occur in patient 14 within the second time period. In some instances, the IMD 16 generates the short-term probability by performing feature detection on the subsequent parameterized patient data. In some instances, the IMD 16 may analyze parameterized patient data similar to the parameterized patient data analyzed by the computing device 24 as described above. In other instances, the IMD 16 analyzes parameterized patient data different from the parameterized patient data analyzed by the computing device 24. For example, the computing device 24 may analyze parameterized patient data representing one or more values averaged over a long time period (e.g., from about 24 hours to about 48 hours), while the IMD 16 may analyze parameterized patient data representing one or more values averaged over a short time period (e.g., from about 30 minutes to about 60 minutes).

[0046] In response to determining that the short-term probability exceeds a short-term predetermined threshold, the IMD 16 performs a remedial action to reduce the short-term probability that an arrhythmia will occur in the patient 14 during a second time period. In some instances, the clinician selects the short-term predetermined threshold as a value that ensures high specificity in predicting arrhythmias. In some instances, the short-term predetermined threshold is 95%. In some instances, the short-term predetermined threshold is another threshold, such as 80%, 90%, 99%, 99.5%, or 99.9%.

[0047] For example, the IMD 16 may issue a notification of the short-term probability that an arrhythmia will occur in the patient 14 during a second time period to the computing device 24, such that the patient 14 or the clinician can become aware of the likelihood that an arrhythmia will occur in the patient 14 during the second time period. In another instance, the IMD 16 initiates a treatment for the patient 14, such as a drug delivery treatment or an electrical pacing treatment, to reduce the likelihood that an arrhythmia will occur in the patient 14 during the second time period.

[0048] In some instances, in response to an instruction from the computing device 24, the IMD 16 processes the parameterized patient data to generate a short-term probability that an arrhythmia will occur in the patient 14, and determines whether the short-term probability exceeds the short-term predetermined threshold at one time. In other instances, the IMD 16 processes the parameterized patient data to generate a short-term probability that an arrhythmia will occur in the patient 14, and determines whether the short-term probability continuously exceeds the short-term predetermined threshold during a first time period. In other instances, the IMD 16 processes the parameterized patient data to generate a short-term probability that an arrhythmia will occur in the patient 14, and periodically determines several times during a first time period (e.g., once every 10 minutes, once an hour, once a day, etc.) whether the short-term probability exceeds the short-term predetermined threshold.

[0049] In some instances, after determining that the likelihood that an arrhythmia will occur in the patient during a first time period has decreased, the IMD 16 stops performing short-term monitoring. For example, the IMD 16 may stop processing subsequent parameterized patient data to generate a short-term probability that an arrhythmia will occur in the patient 14 during the first time period. In some instances, the IMD 16 stops performing short-term monitoring after a certain amount of time (e.g., a long time period). In some instances, the IMD 16 stops performing short-term monitoring after a certain time period that is longer than the long time period. In some instances, the IMD 16 stops performing short-term monitoring after about 1 day. In some instances, the IMD 16 stops performing short-term monitoring after about 1 week. In some instances, the duration of the IMD 16 performing short-term monitoring of the patient 14 is approximately the same as the long time period. In other instances, the duration of the IMD 16 performing short-term monitoring of the patient 14 is greater than or less than the long time period.

[0050] In some instances, the IMD 16 stops performing short-term monitoring in response to receiving an instruction from the computing device 24. For example, in response to determining that the long-term probability of an arrhythmia occurring in the patient 14 within a first time period has decreased to less than a long-term predetermined threshold, the computing device 24 sends an instruction to the IMD 16 to stop performing short-term monitoring.

[0051] By processing the parameterized patient data to generate the short-term probability only when indicated by the computing device 24, the IMD 16 performs power-intensive operations, such as processing the parameterized patient data, only when such operations are beneficial to the patient 14 to prevent an impending arrhythmia. Thus, compared to conventional systems, the IMD 16 can conserve power and thus exhibit an increased battery life.

[0052] As Figure 1 depicted in the instance of, the IMD 16 is an implantable medical device that performs short-term monitoring of the patient 14. However, in other instances of the techniques of the present disclosure, another medical device, such as one or more of the external device 27, the sensor 80, a wearable medical device, or other types of devices external to the patient 14, may perform short-term monitoring of the patient 14 as described herein.

[0053] In the foregoing instance, the computing device 24 is a single device or a distributed system that receives the parameterized patient data, the provider data, and applies a machine learning model to the parameterized patient data and the provider data to generate the long-term probability of an arrhythmia occurring in the patient 14 within a first time period, and sends an instruction to the IMD 16. However, in other instances, the functions of the computing device 24 may be performed by a group of devices. For example, an external programmer may receive the parameterized patient data from the IMD 16 and upload the parameterized patient data to a local access point in the network 25. Additionally, the local access point may receive the provider data of the patient 14 from the provider database 66. The distributed computing system may receive the parameterized data and the provider data of the patient 14 from the local access point and apply a machine learning model to generate the long-term probability of an arrhythmia occurring in the patient 14 within a first time period. The distributed computing system may send the long-term probability to the external programmer via the local access point. The external programmer may determine that the long-term probability exceeds the long-term predetermined threshold, and in response, send an instruction to the IMD 16 to cause the IMD 16 to perform short-term monitoring of the patient 14.

[0054] The foregoing examples describe a two - layer prediction system. For example, computing device 24 makes a first generalized long - term prediction of arrhythmias in patient 14, and IMD 16 makes a second fine - grained short - term prediction of arrhythmias in patient 14. However, the techniques of the present disclosure can provide systems with different numbers of layers. For example, in a three - layer prediction system, computing device 24 makes a first generalized long - term prediction of arrhythmias in patient 14, external device 7 makes a second intermediate prediction, and IMD 16 makes a third fine - grained short - term prediction of arrhythmias in patient 14. Other examples of the techniques of the present disclosure can incorporate additional predictions at various levels to further improve the accuracy of the predicted arrhythmias in patient 14, resulting in four - layer, five - layer, or higher - layer systems.

[0055] Additionally, in the foregoing example of the two - layer prediction system, computing device 24 makes a first generalized long - term prediction of arrhythmias in patient 14, and IMD 16 makes a second fine - grained short - term prediction of arrhythmias in patient 14. Further, in the foregoing example of the three - layer prediction system, computing device 24 makes a first generalized long - term prediction of arrhythmias in patient 14, external device 27 makes a second intermediate prediction, and IMD 16 makes a third fine - grained short - term prediction of arrhythmias in patient 14. However, these predictions made by each of computing device 24, external device 27, and / or IMD 16 are provided only as examples. In other examples, each of computing device 24, external device 27, and / or IMD 16 can make predictions for similar or different time lengths. For example, each of computing device 24, external device 27, and / or IMD 16 may be capable of using the techniques described herein to perform long - term, intermediate, or mid - term, or short - term predictions of arrhythmias. For example, computing device 24 can use parameterized patient data and provider data to make a short - term prediction of arrhythmias in patient 14, and in response to determining that an arrhythmia is likely to occur, send instructions to IMD 16 to perform additional short - term predictions of the arrhythmia generated in patient 14.

[0056] Additionally, in the foregoing example of the two-tier prediction system, computing device 24 makes a first generalized long-term prediction of arrhythmias in patient 14, and IMD 16 makes a second finer short-term prediction of arrhythmias in patient 14. In response to determining that an arrhythmia is likely to occur in patient 14, IMD 16 may perform a remedial action to reduce the short-term probability that an arrhythmia will occur in patient 14. However, in other examples of the techniques of the present disclosure, the multi-tier systems as described herein may include different combinations of devices. For example, a two-tier prediction system may include computing device 24 that makes a first generalized long-term prediction of arrhythmias in patient 14; and an external device 27 that makes a second finer short-term prediction of arrhythmias in patient 14 and performs a remedial action to reduce the short-term probability that an arrhythmia will occur in patient 14. As another example, a two-tier prediction system may include external device 27 that makes a first generalized long-term prediction of arrhythmias in patient 14; and a wearable medical device (such as one of sensors 80) that makes a second finer short-term prediction of arrhythmias in patient 14 and performs a remedial action to reduce the short-term probability that an arrhythmia will occur in patient 14. In yet another example, a two-tier prediction system may include external device 27 that makes a first generalized long-term prediction of arrhythmias in patient 14; and IMD 16 that makes a second finer short-term prediction of arrhythmias in patient 14 and performs a remedial action to reduce the short-term probability that an arrhythmia will occur in patient 14.

[0057] Accordingly, the techniques disclosed herein may allow for enhanced patient care and increased ability to prevent arrhythmias in patients. For example, knowledge that the likelihood of a tachyarrhythmia occurring in the next few days is relatively high can be used to help guide preventive care of the patient. Additionally, knowledge that the likelihood of a tachyarrhythmia occurring in the next few minutes or hours is relatively high can be used to guide preventive treatment or emergency care of the patient. Further, a system such as that disclosed herein can move computationally expensive and energy-consuming long-term arrhythmia prediction operations from a patient's personal medical device to a cloud computing system, and activate short-term cardiac prediction operations on the medical device only when the cloud computing system predicts that the likelihood of an arrhythmia occurring in the next few days is relatively high. Thus, the techniques of the present disclosure can conserve power and extend the battery life of a medical device by using the medical device for cardiac prediction only when the likelihood of an arrhythmia occurring within a particular time span is relatively high.

[0058] Figure 2 For a more detailed illustration Figure 1Conceptual diagram of the IMD 16 and leads 18, 20, 22 of system 10. In the illustrated example, bipolar electrodes 40 and 42 are positioned adjacent the distal end of lead 18, and bipolar electrodes 48 and 50 are positioned adjacent the distal end of lead 22. Additionally, four electrodes 44, 45, 46, and 47 are positioned adjacent the distal end of lead 20. Lead 20 may be referred to as a quadripolar LV lead. In other examples, lead 20 may include more or fewer electrodes. In some examples, LV lead 20 includes segmented electrodes, e.g., where each of a plurality of longitudinal electrode positions of the lead, such as the positions of electrodes 44, 45, 46, and 47, includes a plurality of discrete electrodes disposed about the circumference of the lead at respective circumferential positions.

[0059] In the illustrated example, electrodes 40 and 44 - 48 take the form of annular electrodes, and electrodes 42 and 50 may take the form of extendable helical tip electrodes, telescopically mounted within insulating electrode heads 52 and 56, respectively. Leads 18 and 22 also respectively include elongate electrodes 62 and 64, which may take the form of coils. In some examples, each of electrodes 40, 42, 44 - 48, 50, 62, and 64 is electrically coupled to a respective conductor within the lead body of its associated lead 18, 20, 22, and thus to circuitry within the IMD 16.

[0060] In some examples, the IMD 16 includes one or more case electrodes, such as Figure 2 case electrode 4 as illustrated, which may be integrally formed with or otherwise coupled to the outer surface of the hermetic case 8 of the IMD 16. In some examples, case electrode 4 is defined by an uninsulated portion of the outward - facing portion of the case 8 of the IMD 16. Other divisions between insulated and non - insulated portions of the case 8 may be used to define two or more case electrodes. In some examples, the case electrode substantially comprises the entirety of the case 8.

[0061] Case 8 encloses signal - generating circuitry that generates therapeutic stimuli, such as cardiac pacing, cardioversion, and defibrillation pulses, as well as sensing circuitry for sensing electrical signals attendant to depolarization and repolarization of the heart 12. Case 8 may also enclose a memory for storing sensed electrical signals. Case 8 may also enclose communication circuitry 17 for communication between the IMD 16 and the computing device 24.

[0062] The IMD 16 senses electrical signals attendant to depolarization and repolarization of the heart 12 via electrodes 4, 40, 42, 44 - 48, 50, 62, and 64. The IMD 16 may sense such electrical signals via any bipolar combination of electrodes 40, 42, 44 - 48, 50, 62, and 64. Additionally, any one of electrodes 40, 42, 44 - 48, 50, 62, and 64 may be combined with the case electrode 4 for unipolar sensing.

[0063] The illustrated leads 18, 20, and 22 and the number and configuration of the electrodes are merely examples. Other configurations, namely the number and location of the leads and electrodes, are also possible. In some examples, system 10 may include additional leads or lead segments having one or more electrodes located at different positions within the cardiovascular system for sensing and / or delivering therapy to patient 14. For example, as an alternative to or in addition to the intracardiac leads 18, 20, and 22, system 10 may include one or more epicardial or extravascular (e.g., subcutaneous or subxiphoid) leads that are not positioned within heart 12.

[0064] In accordance with the techniques of the present disclosure, IMD 16 receives instructions from Figure 1 computing device 24 to perform short-term monitoring of patient 14. For example, in response to receiving instructions from computing device 24, IMD 16 processes subsequent parameterized patient data to generate a short-term probability that an arrhythmia will occur in patient 14 during a second time period. In some examples, IMD 16 generates the short-term probability by performing feature detection on the subsequent parameterized patient data. In some examples, IMD 16 may analyze parameterized patient data similar to the parameterized patient data analyzed by computing device 24 as described above. In other examples, IMD 16 analyzes parameterized patient data different from the parameterized patient data analyzed by computing device 24. For example, computing device 24 may analyze parameterized patient data representing one or more values averaged over a relatively long time period (e.g., about 24 hours to about 48 hours), while IMD 16 may analyze parameterized patient data representing one or more values averaged over a relatively short time period (e.g., about 30 minutes to about 60 minutes). As used throughout the disclosure, the terms "long time period" and "short time period" are used herein to distinguish two time periods of different lengths from one another, where the length of one of the two time periods (e.g., "long time period") is greater than the length of the other of the two time periods (e.g., "short time period").

[0065] In response to determining that the short-term probability exceeds a short-term predetermined threshold, IMD 16 performs a remedial action to reduce the short-term probability that an arrhythmia will occur in patient 14 during the second time period. For example, IMD 16 may publish a notification of the short-term probability that an arrhythmia will occur in patient 14 during the second time period to computing device 24 such that patient 14 or a clinician may become aware of the likelihood that an arrhythmia will occur in patient 14 during the second time period. In another example, IMD 16 initiates therapy for patient 14, such as drug delivery therapy or electrical pacing therapy, to reduce the likelihood that an arrhythmia will occur in patient 14 during the second time period.

[0066] Although this document is described in the context of providing an example IMD 16 for therapeutic electrical stimulation, the techniques disclosed herein for short-term prediction of arrhythmias can be used with other types of devices. For example, these techniques can be implemented with an additional cardiac defibrillator coupled to electrodes external to the cardiovascular system, a transcatheter pacemaker configured for implantation within the heart (such as the Micra TM transcatheter pacing system), an insertable cardiac monitor (such as the Reveal LINQ TM ICM, which is also commercially available from Medtronic PLC), a nerve stimulator, a drug delivery device, a wearable device (such as a wearable cardioverter defibrillator), a fitness tracker or other wearable device, a mobile device (such as a mobile phone), a "smart" phone, a laptop computer, a tablet computer, a personal digital assistant (PDA), or "smart" clothing, such as "smart" glasses or "smart" watches.

[0067] Thus, the techniques disclosed herein can allow for enhanced patient care and an increased ability to prevent arrhythmias in patients. For example, knowledge that the likelihood of a tachyarrhythmia occurring in the next few days is relatively high can be used to help guide a patient's preventive care. Additionally, knowledge that the likelihood of a tachyarrhythmia occurring in the next few minutes or hours is relatively high can be used to guide a patient's preventive treatment or emergency care. Further, a system such as that disclosed herein can activate short-term cardiac prediction operations on a medical device only in response to a cloud computing system predicting that the likelihood of an arrhythmia occurring in the next few days is relatively high. Thus, the techniques of the present disclosure can conserve power and extend the battery life of a medical device by using the medical device for cardiac prediction only when the likelihood of an arrhythmia occurring within a particular time span is relatively high.

[0068] Figure 3Block diagram of an example IMD 16 of the technology according to the present disclosure. In the illustrated example, IMD 16 includes processing circuitry 58, memory 59, communication circuitry 17, sensing circuitry 50A, therapy delivery circuitry 52A, sensor 57, and power source 54. Memory 59 includes computer-readable instructions that, when executed by processing circuitry 58, cause IMD 16 and processing circuitry 58 to perform various functions attributed to IMD 16 and processing circuitry 58 herein (e.g., perform short-term prediction of arrhythmias, deliver therapies such as antitachycardia pacing, bradycardia pacing, and post-shock pacing therapy, etc.). Memory 59 may include any volatile medium, non-volatile medium, magnetic medium, optical medium, or electrical medium, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), electrically erasable programmable ROM (EEPROM), flash memory, or any other digital or analog medium.

[0069] Processing circuitry 58 may include any one or more of a microprocessor, a controller, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or equivalent discrete or analog logic circuitry. In some examples, processing circuitry 58 may include multiple components (such as any combination of one or more microprocessors, one or more controllers, one or more DSPs, one or more ASICs, or one or more FPGAs) and other discrete or integrated logic circuitry. The functions attributed to processing circuitry 58 herein may be embodied as software, firmware, hardware, or any combination thereof.

[0070] Processing circuitry 58 controls therapy delivery circuitry 52A to deliver a stimulation therapy to heart 5 according to therapy parameters that may be stored in memory 59. For example, processing circuitry 58 may control therapy delivery circuitry 52A to deliver an electrical pulse having an amplitude, pulse width, frequency, or electrode polarity specified by the therapy parameters. In this manner, therapy delivery circuitry 52A may deliver pacing pulses (e.g., ATP pulses, bradycardia pacing pulses, or post-shock pacing therapy) to heart 5 via electrodes 34 and 40. In some examples, therapy delivery circuitry 52A may deliver pacing stimulation in the form of voltage or current electrical pulses, such as ATP therapy, bradycardia therapy, or post-shock pacing therapy. In other examples, therapy delivery circuitry 52A may deliver one or more of these types of stimulation in the form of other signals (such as sine waves, square waves, or other substantially continuous period signals).

[0071] The therapy delivery circuitry 52A is electrically coupled to electrodes 34 and 40 carried on the housing of the IMD 16. Although the IMD 16 may include only two electrodes, such as electrodes 34 and 40, in other instances, the IMD 16 may utilize three or more electrodes. The IMD 16 may use any combination of the electrodes to deliver therapy and / or detect electrical signals from the patient 12. In some instances, the therapy delivery circuitry 52A includes a charging circuit, one or more pulse generators, capacitors, transformers, switch modules, and / or other components capable of generating and / or storing energy for delivery as a pacing therapy, cardiac resynchronization therapy, other therapy, or a combination of therapies. In some instances, the therapy delivery circuitry 52A delivers the therapy as one or more electrical pulses in accordance with one or more sets of therapy parameters that define the amplitude, frequency, voltage, or current of the therapy or other parameters of the therapy.

[0072] The sensing circuitry 50A monitors signals from one or more combinations (also referred to as vectors) of two or more of the electrodes 4, 40, 42, 44 - 48, 50, 62, and 64 to monitor the electrical activity, impedance, or other electrical phenomena of the heart 12. In some instances, the sensing circuitry 50A includes one or more analog components, digital components, or a combination thereof. In some instances, the sensing circuitry 50A includes one or more sense amplifiers, comparators, filters, rectifiers, threshold detectors, analog-to-digital converters (ADCs), and the like. In some instances, the sensing circuitry 50A converts the sensed signals to digital form and provides the digital signals to the processing circuitry 58 for processing or analysis. In one instance, the sensing circuitry 50A amplifies the signals from the electrodes 4, 40, 42, 44 - 48, 50, 62, and 64 and converts the amplified signals to multi-bit digital signals via an ADC.

[0073] In some instances, the sensing circuitry 50A performs sensing of an electrocardiogram to determine heart rate or heart rate variability, or to detect arrhythmias (e.g., tachyarrhythmias or bradycardia) or to sense other parameters or events from the electrocardiogram. The sensing circuitry 50A may also include switching circuitry to select which available electrodes (and electrode polarities) to use for sensing cardiac activity based on the electrode combinations or electrode vectors used in the current sensing configuration. The processing circuitry 58 may control the switching circuitry to select the electrodes to be used as sensing electrodes and their polarities. The sensing circuitry 50 may include one or more detection channels, each of which may be coupled to a selected electrode configuration to detect cardiac signals via the electrode configuration. In some instances, the sensing circuitry 50A compares the processed signal to a threshold to detect the presence of atrial or ventricular depolarization and to indicate to the processing circuitry 58 the presence of atrial depolarization (e.g., the P wave) or ventricular depolarization (e.g., the R wave). The sensing circuitry 50A may include one or more amplifiers or other circuitry for comparing the electrocardiogram amplitude to a threshold, which may be adjustable.

[0074] The processing circuitry 58 may include a timing and control module, which may be implemented as hardware, firmware, software, or any combination thereof. The timing and control module may include dedicated hardware circuitry (such as an ASIC) separate from other components of the processing circuitry 58 (such as a microprocessor), or a software module executed by a component of the processing circuitry 58 that may be a microprocessor or an ASIC. The timing and control module may implement a programmable counter. If the IMD 16 is configured to generate bradycardia pacing pulses and deliver them to the heart 12, then such a counter may control the basic time intervals associated with DDD, VVI, DVI, VDD, AAI, DDI, DDDR, VVIR, DVIR, VDDR, AAIR, DDIR, and other pacing modes.

[0075] The memory 59 may be configured to store various operating parameters, therapy parameters, sensed and detected data, and any other information related to the treatment and care of the patient 12. In Figure 3 instances, the memory 59 may store, for example, sensed cardiac EGMs associated with detected or predicted arrhythmias, and therapy parameters that define the delivery of therapy provided by the therapy delivery circuitry 52A. In other instances, the memory 59 may act as a temporary buffer for storing data until it can be uploaded to the computing device 24.

[0076] The communication circuitry 17 includes means for communicating via Figure 1Any suitable circuitry, firmware, software, or any combination thereof for the network 25 to communicate with another device, such as computing device 24. By way of example, communication circuitry 17 may include one or more antennas, modulation and demodulation circuitry, filters, amplifiers, etc., for radio frequency communication with other devices, such as computing device 24, via network 25. Under the control of processing circuitry 58, communication circuitry 17 may receive downlink telemetry from computing device 24 and transmit uplink telemetry to computing device 24 by means of antennas that may be internal and / or external. Processing circuitry 58 may provide data and control signals to be uplinked to computing device 24 to the telemetry circuitry within communication circuitry 17, for example, via an address / data bus. In some instances, communication circuitry 17 may provide received data to processing circuitry 58 via a multiplexer.

[0077] Power source 54 may be any type of device configured to hold a charge to operate the circuitry of IMD 16. Power source 54 may be provided as a rechargeable or non-rechargeable battery. In other instances, power source 54 may incorporate an energy scavenging system that stores electrical energy from the movement of IMD 16 within patient 12.

[0078] According to the techniques of the present disclosure, the IMD 16 collects parameterized patient data of the patient 14 via the sensing circuitry 50A and / or the sensor 57. The sensor 57 may include one or more sensors, such as one or more accelerometers, pressure sensors, optical sensors for O2 saturation, etc. In some instances, the parameterized patient data includes one or more of the following: the patient's activity level, the patient's heart rate, the patient's posture, the patient's electrocardiogram, the patient's blood pressure, the patient's accelerometer data, or other types of parameterized patient data. The IMD 16 uploads the parameterized patient data via the communication circuitry 17 over the network 25 to the computing device 24 and / or one or more of the sensor 80 or the external device 27. In some instances, the activity level may be the sum of the activity over a certain time period (such as one second or several seconds or one minute or several minutes). In some instances, the IMD 16 uploads the parameterized patient data to the computing device 24, the external device 27, and / or the sensor 80 on a daily basis. In some instances, the parameterized patient data includes one or more values representing the average measurements of the patient 14 over a long time period (e.g., about 24 hours to about 48 hours). In this instance, the IMD 16 both uploads the parameterized patient data to the computing device 24 and performs short-term monitoring of the patient 14 (as described below). However, in other instances, the medical device that collects the parameterized patient data is different from the medical device that performs short-term monitoring of the patient 14. For example, one or more other devices, such as a wearable medical device or a mobile device (e.g., a smart phone) of the patient 14, may collect the parameterized patient data and upload the parameterized patient data to the computing device 24.

[0079] In some instances, the processing circuitry 58 receives instructions from Figure 1 the computing device 24 to perform short-term monitoring of the patient 14. For example, in response to receiving instructions from the computing device 24, the processing circuitry 58 executes the short-term prediction software 60A stored in the memory 59 as the short-term prediction software 60B. For example, the processing circuitry 58 may execute the short-term prediction software 60A to process subsequent parameterized patient data to generate a short-term probability of an arrhythmia occurring in the patient 14 within a short time period. In some instances, the processing circuitry 58 generates the short-term probability by performing feature detection on subsequent parameterized patient data sensed by the sensing circuitry 50A. In some instances, the processing circuitry 58 performs feature detection on one or more of electrocardiogram data, electrode impedance measurements, accelerometer data, temperature data of the patient 14, or audio data of the heart of the patient 14.

[0080] As an example, processing circuitry 58 performs feature detection on subsequent parameterized patient data, including electrocardiogram data. In this example, the parameterized patient data includes one or more of the average frequency or average amplitude of the T-wave of the electrocardiogram of patient 14. The processing circuitry 58 receives the raw electrocardiogram signal therefrom via the sensing circuitry 50A and / or the sensor 57, and extracts features from the raw electrocardiogram signal. In some examples, the processing circuitry 58 identifies one or more of T-wave alternans, QRS morphology measurements, etc. For example, the processing circuitry 58 identifies one or more features of the T-wave of the electrocardiogram of patient 14 and applies a model to the one or more identified features to generate a short-term probability that an arrhythmia will occur in patient 14 within a short time period. In some examples, the one or more identified features are one or more amplitudes of the T-wave. In some examples, the one or more identified features are the frequency of the T-wave. In some examples, the one or more identified features include at least the amplitude of the T-wave and the frequency of the T-wave.

[0081] In some examples, the processing circuitry 58 identifies one or more relative changes in one or more of the identified features of the parameterized patient data that predict a subsequent arrhythmia in patient 14. In some examples, the processing circuitry 58 identifies one or more interactions between the multiple identified features that predict a subsequent arrhythmia in patient 14. In some examples, the processing circuitry 58 analyzes parameterized patient data representing one or more values averaged over a short time period (e.g., about 30 minutes to about 60 minutes).

[0082] In some examples, the processing circuitry 58 may use the identified features as inputs to a second machine learning model that produces, as an output of the second machine learning model, a short-term probability that an arrhythmia will occur in patient 14 within a short time period. In another example, the processing circuitry 58 uses the raw signal obtained via the sensing circuitry 50A and / or the sensor 57 as a direct input to the second machine learning model that produces the short-term probability. In some examples, in a manner similar to the training process described above for the machine learning model of the computing device 24, the parameterized patient data and provider data of multiple patients are used to train the second machine learning model. In some examples, the second machine learning model performs error correction based on received feedback indicating whether an arrhythmia has occurred in patient 14, so as to gradually improve the predictions made by the second machine learning model by "learning" from the correct and incorrect predictions made by the second machine learning model in the past. Additionally, the training process can be used to further fine-tune the second machine learning model trained using population-based data, thereby providing more accurate predictions for a specific individual.

[0083] Processing circuitry 58 compares the short - term probability of an arrhythmia occurring in patient 14 over a short - term time period with a predetermined threshold. In response to determining that the short - term probability exceeds the predetermined threshold, processing circuitry 58 performs a remedial action to reduce the short - term probability of an arrhythmia occurring in patient 14 over the short - term time period. For example, processing circuitry 58 may publish a notification of the short - term probability of an arrhythmia occurring in patient 14 over the short - term time period via communication circuitry 17 to computing device 24, such that patient 14 or a clinician may become aware of the likelihood of an arrhythmia occurring in patient 14 over the short - term time period. In another example, processing circuitry 58 causes treatment delivery circuitry 52A to initiate delivery of treatment to patient 14. Although in Figure 3 the example of, treatment delivery circuitry 52A is configured to deliver an electrical stimulation treatment or an electrical pacing treatment, in other examples, processing circuitry 58 may cause, for example, a drug delivery system to deliver a drug treatment to patient 14. Thus, processing circuitry 58 may perform a remedial action to reduce the likelihood of an arrhythmia occurring in patient 14 over the short - term time period.

[0084] Although described herein in the context of an example IMD 16 that provides therapeutic electrical stimulation, the techniques for short - term prediction of arrhythmias disclosed herein may be used with other types of devices. For example, these techniques may be implemented with a transcatheter pacemaker configured for implantation within the heart (such as the Micra TM transcatheter pacing system commercially available from Medtronic PLC, Dublin, Ireland), an insertable cardiac monitor (such as the Reveal LINQ TM ICM, which is also commercially available from Medtronic PLC), a nerve stimulator, a drug delivery device, a wearable device (such as a wearable cardioverter defibrillator), a fitness tracker or other wearable device, a mobile device (such as a mobile phone), a “smart” phone, a laptop computer, a tablet computer, a personal digital assistant (PDA), or “smart” clothing, such as “smart” glasses or “smart” watches.

[0085] Thus, the techniques disclosed herein may allow for enhanced patient care and an increased ability to prevent arrhythmias in patients. For example, knowledge that the likelihood of a tachyarrhythmia occurring within the next few minutes or hours is relatively high can be used to guide preventive treatment or emergency care for the patient. Additionally, a system such as that disclosed herein may activate short - term cardiac prediction operations on a medical device only in response to the computing device 24 predicting a relatively high likelihood of an arrhythmia occurring within the next few days. Thus, the techniques of the present disclosure may conserve power and extend the battery life of the IMD 16 by using the IMD 16 for cardiac prediction only when the likelihood of an arrhythmia occurring within a particular time span is relatively high.

[0086] Figure 4 Block diagram of an example computing device 24 operating in accordance with one or more techniques of the present disclosure. In one example, computing device 24 includes processing circuitry 402 for executing an application 424 that includes a long-term prediction module 450 or any other application described herein. Although shown as a stand-alone computing device 24 for purposes of example, computing device 24 can be any component or system that includes processing circuitry or other suitable computing environment for executing software instructions and, for example, need not include Figure 4 one or more of the elements shown in Figure 4 (e.g., communication circuitry 406; and in some examples, components such as one or more storage devices 408 may not be co-located with other components or may be located in the same rack). In some examples, computing device 24 can be a cloud computing system distributed across multiple devices.

[0087] As Figure 4 shown in the example of

[0088]

[0089] shown, computing device 24 includes processing circuitry 402, one or more input devices 404, communication circuitry 406, one or more output devices 412, one or more storage devices 408, and one or more user interface (UI) devices 410. In one example, computing device 24 additionally includes one or more applications 424 (such as long-term prediction module 450) and an operating system 416 that can be executed by computing device 24. Each of components 402, 404, 406, 408, 410, and 412 is coupled (physically, communicatively, and / or operably) for inter-component communication. In some examples, communication channel 414 can include a system bus, network connection, interprocess communication data structure, or any other means for communicating data. As an example, components 402, 404, 406, 408, 410, and 412 can be coupled via one or more communication channels 414.

[0088] In one example, processing circuitry 402 is configured to implement functions and / or process instructions for execution within computing device 24. For example, processing circuitry 402 may be capable of processing instructions stored in storage device 408. Examples of processing circuitry 402 can include any one or more of the following: a microprocessor, a controller, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or equivalent discrete or integrated logic circuitry.

[0089] One or more storage devices 408 may be configured to store information within the computing device 24 during operation. In some instances, the storage device 408 is described as a computer-readable storage medium. In some instances, the storage device 408 is temporary memory, meaning that the primary purpose of the storage device 408 is not long-term storage. In some instances, the storage device 408 is described as volatile memory, meaning that when the computer is turned off, the storage device 408 does not maintain the stored content. Examples of volatile memory include random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), and other forms of volatile memory known in the art. In some instances, the storage device 408 is used to store program instructions executed by the processing circuitry 402. In one instance, the storage device 408 is used by software or an application 424 running on the computing device 24 to temporarily store information during program execution.

[0090] In some instances, the storage device 408 also includes one or more computer-readable storage media. The storage device 408 may be configured to store a larger amount of information than volatile memory. The storage device 408 may additionally be configured for long-term storage of information. In some instances, the storage device 408 includes non-volatile storage elements. Examples of such non-volatile storage elements include magnetic hard disks, optical disks, floppy disks, flash memory, or various forms of electrically programmable read-only memory (EPROM) or electrically erasable and programmable (EEPROM) memory.

[0091] In some instances, the computing device 24 also includes communication circuitry 406. In one instance, the computing device 24 utilizes the communication circuitry 406 to communicate with external devices such as Figure 1 the IMD 16 and the provider database 66. The communication circuitry 406 may include a network interface card, such as an Ethernet card, an optical transceiver, a radio frequency transceiver, or any other type of device that can send and receive information. Other examples of such network interfaces may include 3G, 4G, 5G, and WiFi radios.

[0092] In one instance, the computing device 24 also includes one or more user interface devices 410. In some instances, the user interface devices 410 are configured to receive input from a user through tactile, audio, or visual feedback. Examples of one or more user interface devices 410 include a presence-sensitive display, a mouse, a keyboard, a voice response system, a camera, a microphone, or any other type of device for detecting commands from a user. In some instances, the presence-sensitive display includes a touch-sensitive screen.

[0093] One or more output devices 412 may also be included in computing device 24. In some instances, output device 412 is configured to provide output to a user using tactile, audio, or visual stimuli. In one instance, output device 412 includes a presence-sensitive display, a sound card, a video graphics adapter card, or any other type of device for converting a signal into a suitable form understandable by a human or a machine. Additional instances of output device 412 include speakers, a cathode ray tube (CRT) monitor, a liquid crystal display (LCD), or any other type of device that can generate an understandable output to a user.

[0094] Computing device 24 may include an operating system 416. In some instances, operating system 416 controls the operation of the components of computing device 24. For example, in one instance, operating system 416 facilitates communication between one or more application programs 424 and long-term prediction module 450 and processing circuitry 402, communication circuitry 406, storage device 408, input device 404, user interface device 410, and output device 412.

[0095] Application program 424 may also include program instructions and / or data executable by computing device 24. One or more instances of application programs 424 executable by computing device 24 may include long-term prediction module 450. Alternatively or additionally, other additional applications (not shown) may be included to provide other functions described herein and are not depicted for simplicity.

[0096] According to the techniques of the present disclosure, application program 424 includes long-term prediction module 450. In one instance, processing circuitry 402 executes long-term prediction module 450 to provide Figure 1 a long-term prediction of ventricular arrhythmia for patient 14. In one instance, long-term prediction module 450 includes a machine learning model that generates a long-term probability of an arrhythmia occurring in patient 14 over a long time period (e.g., typically from about 24 hours to about 48 hours) based on parameterized patient data and provider data of patient 14. In some instances, the machine learning model is generated by a neural network system, a deep learning system, or other types of supervised or unsupervised machine learning systems. For example, the machine learning model may be generated by a feedforward neural network, such as a convolutional neural network, a radial basis function neural network, a recurrent neural network, a modular or associative neural network. In some instances, long-term prediction module 450 trains the machine learning model with parameterized patient data and provider data of multiple patients to generate the long-term probability. In some instances, after the machine learning model has been pre-trained with parameterized patient data and provider data of multiple patients, long-term prediction module 450 further trains the machine learning model with parameterized patient data and provider data specific to patient 14.

[0097] In some instances, the long-term prediction module 450 trains a machine learning model with parameterized patient data and provider data of multiple patients, determines an error rate of the machine learning model, and then feeds the error rate back to the machine learning model to enable the machine learning model to update its prediction based on the error rate. In some instances, the long-term prediction module 450 may receive feedback from patient 14 or a clinician indicating whether a predicted arrhythmia has occurred in patient 14 during a first time period. In some instances, the long-term prediction module 450 may receive a message from the IMD 16 indicating that the IMD 16 has detected (or has not detected) the occurrence of an arrhythmia in patient 14 and whether an arrhythmia in patient 14 has been predicted (or not predicted). In some instances, the long-term prediction module 450 may obtain feedback in other ways, such as by periodically checking provider data to determine whether an arrhythmia has occurred. The long-term prediction module 450 may update the machine learning model with feedback indicating whether a predicted arrhythmia has occurred in patient 14 during a first time period. Thus, the training process may occur iteratively to gradually improve the predictions made by the machine learning model by "learning" from the correct and incorrect predictions made by the machine learning model in the past. Additionally, the training process may be used to further fine-tune a machine learning model trained using population-based data to provide more accurate predictions for a specific individual.

[0098] Once the machine learning model has been trained to produce a long-term probability that an arrhythmia will occur in patient 14 over a long time period (which is a threshold accuracy selected by a clinician), the long-term prediction module 450 may use the machine learning model to provide a long-term prediction of ventricular arrhythmia for patient 14. For example, the long-term prediction module 450, executed by the processing circuitry 402, receives parameterized patient data collected by a medical device of patient 14 via the communication circuitry 406. In some instances, the parameterized patient data includes one or more of the following: the patient's activity level, the patient's heart rate, the patient's posture, the patient's electrocardiogram, the patient's blood pressure, the patient's accelerometer data, or other types of parameterized patient data. In some instances, the medical device that collects the parameterized patient data is an IMD, such as Figure 1 the IMD 16. In other instances, the medical device that collects the parameterized patient data is another type of patient device, such as a wearable medical device or a mobile device (e.g., a smart phone) of patient 14. In some instances, the long-term prediction module 450 receives the parameterized patient data from the IMD 16 on a daily basis.

[0099] The long-term prediction module 450 also receives provider data of patient 14 from the provider database 66 via the communication circuitry 406. In some instances, the provider data stored by the provider database 66 may include many different types of historical medical information about patient 14. For example, the provider database 66 may store the patient's medication history, the patient's surgical history, the patient's hospitalization history, the patient's potassium levels over time, or one or more laboratory test results of the patient, etc.

[0100] The long-term prediction module 450 applies the trained machine learning model to the parameterized patient data and provider data of patient 14 to perform long-term monitoring of patient 14. In some instances, the long-term prediction module 450 performs long-term monitoring by generating a long-term probability that an arrhythmia will occur in patient 14 over a long time period (e.g., typically from about 24 hours to about 48 hours). For example, the long-term prediction module 450 applies the trained machine learning model to the parameterized patient data and provider data of patient 14 to generate a long-term probability that an arrhythmia will occur in patient 14 over a long time period. For example, the machine learning model may transform the parameterized patient data and provider data into one or more vectors and tensors (e.g., multi-dimensional arrays) representing the parameterized patient data and provider data. The machine learning model may apply mathematical operations to the one or more vectors and tensors to generate a mathematical representation of the parameterized patient data and provider data. The machine learning model may determine different weights corresponding to the recognition relationship between the parameterized patient data and provider data and the occurrence of arrhythmia over a long time period. The machine learning model may apply the different weights to the parameterized patient data and provider data to generate a long-term probability that an arrhythmia will occur in patient 14 over a long time period.

[0101] In some instances, the long-term prediction module 450 determines whether the determined long-term probability that an arrhythmia will occur in patient 14 over a long time period exceeds a predetermined threshold set by a clinician. In response to determining that the long-term probability exceeds the predetermined threshold, the long-term prediction module 450 sends an instruction to the IMD 16 via the communication circuitry 406 to cause the IMD 16 to perform short-term monitoring of patient 14 as described above. Additionally or alternatively, in response to determining that the long-term probability exceeds the predetermined threshold, the long-term prediction module 450 sends an instruction to the external device 27 via the communication circuitry 406 to cause the external device 27 to perform medium-term monitoring of patient 14 as described below. In some instances, in response to determining that the long-term probability exceeds the predetermined threshold, the long-term prediction module 450 may perform other actions, such as notifying the clinician via the output device 412 that the long-term probability has exceeded the predetermined threshold or that the long-term prediction module 450 has determined that patient 14 is likely to have an arrhythmia over a long time period.

[0102] Accordingly, the techniques disclosed herein may allow for enhanced patient care and an increased ability to prevent arrhythmias in patients. For example, knowledge that the likelihood of a tachyarrhythmia occurring within the next few days is relatively high can be used to help guide a patient's preventive care. Additionally, a system such as that disclosed herein can move computationally expensive and power-consuming long-term arrhythmia prediction operations from a patient's personal medical device to a cloud computing system and activate short-term cardiac prediction operations on the medical device only when the cloud computing system predicts that the likelihood of an arrhythmia occurring within the next few days is relatively high. Thus, the techniques of the present disclosure can conserve power and extend the battery life of a medical device by using the medical device for cardiac prediction only when the likelihood of an arrhythmia occurring within a particular time span is relatively high.

[0103] Figure 5 Block diagram of an example external device 27 operating in accordance with one or more techniques of the present disclosure. In some examples, the external device 27 takes the form of an external programmer or a mobile device, such as a mobile phone, a "smart" phone, a laptop computer, a tablet computer, a personal digital assistant (PDA), a wearable electronic device, etc. In some examples, the external device 27 is a CareLink TM monitor available from Medtronic, Inc.

[0104] In one example, the external device 27 includes processing circuitry 502 for executing an application 524 that includes a mid-term prediction module 550 or any other application described herein. Although shown as a stand-alone external device 27 for purposes of example, the external device 27 can be any component or system that includes processing circuitry or other suitable computing environment for executing software instructions and, for example, need not include Figure 5 one or more of the elements shown in Figure 5 (e.g., communication circuitry 506; and in some examples, components such as one or more storage devices 508 may not be co-located with other components or may be located in the same rack).

[0105] As Figure 5As shown in the example of, the external device 27 includes a processing circuitry 502, one or more input devices 504, a communication circuitry 506, one or more output devices 512, one or more storage devices 508, and one or more user interface (UI) devices 510. In one example, the external device 27 further includes one or more application programs 524 (such as the medium-term prediction module 550) and an operating system 516 that can be executed by the external device 27. Each of the components 502, 504, 506, 508, 510, and 512 is coupled (physically, communicatively, and / or operably) for communication among the components. In some examples, the communication channel 514 may include a system bus, a network connection, an interprocess communication data structure, or any other means for communicating data. As an example, the components 502, 504, 506, 508, 510, and 512 may be coupled via one or more communication channels 514.

[0106] In one example, the processing circuitry 502 is configured to implement functions and / or process instructions for execution within the external device 27. For example, the processing circuitry 502 may be capable of processing instructions stored in the storage device 508. Examples of the processing circuitry 502 may include any one or more of the following: a microprocessor, a controller, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or equivalent discrete or integrated logic circuitry.

[0107] One or more storage devices 508 may be configured to store information within the external device 27 during operation. In some examples, the storage device 508 is described as a computer-readable storage medium. In some examples, the storage device 508 is a temporary memory, which means that the primary purpose of the storage device 508 is not long-term storage. In some examples, the storage device 508 is described as volatile memory, which means that when the computer is turned off, the storage device 508 does not maintain the stored content. Examples of volatile memory include random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), and other forms of volatile memory known in the art. In some examples, the storage device 508 is used to store program instructions executed by the processing circuitry 502. In one example, the storage device 508 is used by software or an application program 524 running on the external device 27 to temporarily store information during program execution.

[0108] In some instances, the storage device 508 also includes one or more computer-readable storage media. The storage device 508 can be configured to store a larger amount of information than volatile memory. The storage device 508 can additionally be configured for long-term storage of information. In some instances, the storage device 508 includes non-volatile storage elements. Examples of such non-volatile storage elements include magnetic hard disks, optical disks, floppy disks, flash memory, or various forms of electrically programmable read-only memory (EPROM) or electrically erasable and programmable (EEPROM) memory.

[0109] In some instances, the external device 27 also includes communication circuitry 506. In one instance, the external device 27 uses the communication circuitry 506 to communicate with external devices such as Figure 1 the IMD 16 and the computing device 24. The communication circuitry 506 can include a network interface card, such as an Ethernet card, an optical transceiver, a radio frequency transceiver, or any other type of device that can send and receive information. Other examples of such network interfaces can include 3G, 4G, 5G, and WiFi radios.

[0110] In one instance, the external device 27 also includes one or more user interface devices 510. In some instances, the user interface devices 510 are configured to receive input from a user via tactile, audio, or visual feedback. Examples of one or more user interface devices 510 include a presence-sensitive display, a mouse, a keyboard, a voice response system, a camera, a microphone, or any other type of device for detecting commands from a user. In some instances, the presence-sensitive display includes a touch-sensitive screen.

[0111] One or more output devices 512 can also be included in the external device 27. In some instances, the output devices 512 are configured to provide output to a user using tactile, audio, or visual stimuli. In one instance, the output devices 512 include a presence-sensitive display, a sound card, a video graphics adapter card, or any other type of device for converting a signal into an appropriate form understandable by a human or a machine. Additional examples of output devices 512 include speakers, a cathode ray tube (CRT) monitor, a liquid crystal display (LCD), or any other type of device that can generate an understandable output to a user.

[0112] The external device 27 can include an operating system 516. In some instances, the operating system 516 controls the operation of the components of the external device 27. For example, in one instance, the operating system 516 facilitates the communication of one or more application programs 524 and the mid-term prediction module 550 with the processing circuitry 502, the communication circuitry 506, the storage device 508, the input device 504, the user interface devices 510, and the output devices 512.

[0113] The application program 524 may also include program instructions and / or data executable by the external device 27. One or more instances of the application program 524 executable by the external device 27 may include a medium-term prediction module 550. Alternatively or additionally, other additional applications (not shown) may be included to provide other functions described herein and are not depicted for simplicity.

[0114] According to the techniques of the present disclosure, the application program 524 includes a medium-term prediction module 550. In one instance, the processing circuitry 502 executes the medium-term prediction module 550 to provide Figure 1 a medium-term prediction of the ventricular arrhythmia of the patient 14. In one instance, the medium-term prediction module 550 includes a machine learning model that generates a medium-term probability that an arrhythmia will occur in the patient 14 within a medium-term time period (e.g., typically about 24 hours to about 48 hours) based on the parameterized patient data and provider data of the patient 14. In some instances, the machine learning model is generated by a neural network system, a deep learning system, or other types of supervised or unsupervised machine learning systems. For example, the machine learning model may be generated by a feedforward neural network, such as a convolutional neural network, a radial basis function neural network, a recurrent neural network, a modular or associative neural network. In some instances, the medium-term prediction module 550 trains the machine learning model with the parameterized patient data and provider data of multiple patients to generate the medium-term probability. In some instances, after the machine learning model has been pre-trained with the parameterized patient data and provider data of multiple patients, the medium-term prediction module 550 further trains the machine learning model with the parameterized patient data and provider data specific to the patient 14. Although in the foregoing instances, the medium-term prediction module 550 trains the machine learning model with the parameterized patient data and provider data of multiple patients, in some instances, the medium-term prediction module 550 trains the machine learning model only with the parameterized patient data rather than the provider data or only with the provider data rather than the parameterized patient data.

[0115] In some instances, the mid-term prediction module 550 trains a machine learning model with parameterized patient data and provider data of multiple patients, determines an error rate of the machine learning model, and then feeds the error rate back to the machine learning model to enable the machine learning model to update its prediction based on the error rate. In some instances, the mid-term prediction module 550 may receive feedback from patient 14 or a clinician indicating whether a predicted arrhythmia has occurred in patient 14 during a first time period. In some instances, the mid-term prediction module 550 may receive a message from the IMD 16 indicating that the IMD 16 has detected (or not detected) the occurrence of an arrhythmia in patient 14 and whether an arrhythmia in patient 14 has been predicted (or not predicted). In some instances, the mid-term prediction module 550 may obtain feedback in other ways, such as by periodically checking provider data to determine whether an arrhythmia has occurred. The mid-term prediction module 550 may update the machine learning model with feedback indicating whether a predicted arrhythmia has occurred in patient 14 during the first time period. Thus, the training process may occur iteratively to progressively improve the prediction made by the machine learning model by "learning" from past correct and incorrect predictions made by the machine learning model. Additionally, the training process may be used to further fine-tune a machine learning model trained with population-based data to provide more accurate predictions for a specific individual.

[0116] Once the machine learning model has been trained to produce a mid-term probability that an arrhythmia will occur in patient 14 during a mid-term time period (which is a threshold accuracy selected by a clinician), the mid-term prediction module 550 may use the machine learning model to provide a mid-term prediction of ventricular arrhythmia for patient 14. For example, the mid-term prediction module 550, executed by the processing circuitry 502, receives parameterized patient data collected by a medical device of patient 14 via the communication circuitry 506. In some instances, the mid-term prediction module 550 may also receive other parameterized patient data collected by an external device 27, such as a geographical location, accelerometer data, or an input from patient 14. In some instances, the parameterized patient data includes one or more of the following: the patient's activity level, the patient's heart rate, the patient's posture, the patient's electrocardiogram, the patient's blood pressure, the patient's accelerometer data, or other types of parameterized patient data. In some instances, the medical device that collects the parameterized patient data is an IMD, such as Figure 1 the IMD 16. In other instances, the medical device that collects the parameterized patient data is another type of patient device, such as a wearable medical device of patient 14, a wearable sensor (such as Figure 1 the sensor 80) or a mobile device (e.g., a smart phone). In some instances, the mid-term prediction module 550 receives parameterized patient data from the IMD 16 and / or the sensor 80 on a daily basis.

[0117] In some instances, the mid-term prediction module 550 additionally receives provider data of patient 14 from the provider database 66 via the communication circuitry 506. In some instances, the provider data stored by the provider database 66 may include many different types of historical medical information about patient 14. For example, the provider database 66 may store the patient's medication history, the patient's surgical history, the patient's hospitalization history, the patient's potassium levels over time, or one or more laboratory test results of the patient, etc.

[0118] The mid-term prediction module 550 applies the trained machine learning model to the parameterized patient data and provider data of patient 14 to perform mid-term monitoring of patient 14. In some instances, the mid-term prediction module 550 performs mid-term monitoring by generating a mid-term probability that an arrhythmia will occur in patient 14 within a mid-term time period (e.g., typically from about 24 hours to about 48 hours). For example, the mid-term prediction module 550 applies the trained machine learning model to the parameterized patient data and provider data of patient 14 to generate a mid-term probability that an arrhythmia will occur in patient 14 within the mid-term time period. For example, the machine learning model may transform the parameterized patient data and provider data into one or more vectors and tensors (e.g., multi-dimensional arrays) representing the parameterized patient data and provider data. The machine learning model may apply mathematical operations to the one or more vectors and tensors to generate a mathematical representation of the parameterized patient data and provider data. The machine learning model may determine different weights corresponding to the recognition relationship between the parameterized patient data and provider data and the occurrence of arrhythmia within the mid-term time period. The machine learning model may apply the different weights to the parameterized patient data and provider data to generate a mid-term probability that an arrhythmia will occur in patient 14 within the mid-term time period. Although in the foregoing instance, the mid-term prediction module 550 applies the machine learning model to the parameterized patient data and provider data of multiple patients, in some instances, the mid-term prediction module 550 applies the machine learning model only to the parameterized patient data rather than the provider data or only to the provider data rather than the parameterized patient data.

[0119] In some instances, the mid-term prediction module 550 determines whether the determined mid-term probability that an arrhythmia will occur in patient 14 within the mid-term time period exceeds a predetermined threshold set by a clinician. In response to determining that the mid-term probability exceeds the predetermined threshold, the mid-term prediction module 550 sends an instruction to the IMD 16 via the communication circuitry 506 to cause the IMD 16 to perform short-term monitoring of patient 14 as described above. In some instances, in response to determining that the mid-term probability exceeds the predetermined threshold, the mid-term prediction module 550 may perform other actions, such as notifying the clinician via the output device 512 that the mid-term probability has exceeded the predetermined threshold or that the mid-term prediction module 550 has determined that patient 14 is likely to have an arrhythmia within the mid-term time period.

[0120] In the foregoing example, the mid-term prediction module 550 applies a machine learning model to determine a mid-term probability that an arrhythmia will occur in patient 14 during a mid-term time period. In other examples, the mid-term prediction module 550 may apply feature detection to the parameterized patient data to determine a mid-term probability that an arrhythmia will occur in patient 14 during a mid-term time period in a manner similar to how the IMD 16 applies feature detection to the parameterized patient data to determine a short-term probability that an arrhythmia will occur in patient 14 during a short-term time period, as described above.

[0121] As an example, the mid-term prediction module 550 performs feature detection on the parameterized patient data that includes electrocardiogram data. In this example, the parameterized patient data includes one or more of the average frequency or average amplitude of the T-wave of the electrocardiogram of patient 14. The mid-term prediction module 550 receives the raw electrocardiogram signal from Figure 1 the IMD 16 or the sensor 80 and extracts features from the raw electrocardiogram signal. In some examples, the mid-term prediction module 550 identifies one or more of T-wave alternans, QRS morphology measurements, etc. For example, the mid-term prediction module 550 identifies one or more features of the T-wave of the electrocardiogram of patient 14 and applies a model to the one or more identified features to generate a mid-term probability that an arrhythmia will occur in patient 14 during a mid-term time period. In some examples, one or more of the identified features are one or more amplitudes of the T-wave. In some examples, one or more of the identified features are the frequency of the T-wave. In some examples, one or more of the identified features at least include the amplitude of the T-wave and the frequency of the T-wave.

[0122] Although the IMD 16 may perform feature detection during a short-term time period due to battery and processing power limitations, the external device 27 may not be subject to such limitations. For example, the external device 27 may be easily rechargeable, have a larger battery, or have significantly more computational resources. Thus, compared to the IMD 16, the external device 27 may apply a feature detection algorithm that is computationally more expensive, algorithmically complex, consumes more power, or analyzes the parameterized patient data over a longer time period (e.g., a mid-term time period).

[0123] Accordingly, the techniques disclosed herein may allow for enhanced patient care and an increased ability to prevent arrhythmias in patients. For example, knowledge that the likelihood of a tachyarrhythmia occurring within the next few days is relatively high can be used to help guide preventive care for a patient. Additionally, a system such as that disclosed herein can offload computationally expensive and energy-consuming complex arrhythmia prediction operations from a medical device where battery life is a critical issue, and activate short-term cardiac prediction operations on the medical device only when the predicted likelihood of an arrhythmia occurring within the next few days is relatively high. Thus, the techniques of the present disclosure can conserve power and extend the battery life of a medical device by using the medical device for cardiac prediction only when the likelihood of an arrhythmia occurring within a particular time span is relatively high.

[0124] Figure 6 A flowchart illustrating an example operation of the techniques according to the present disclosure. For convenience, reference is made to Figure 1 described Figure 6 . In some examples, Figure 6 the operation of is an operation for providing a multi-layer prediction of arrhythmias in patient 14 (e.g., long-term and short-term predictions of ventricular arrhythmias in patient 14).

[0125] In one example, the IMD 16 collects parameterized patient data (600) of patient 114. In some examples, the parameterized patient data includes one or more of the following: the patient's activity level, the patient's heart rate, the patient's posture, the patient's electrocardiogram, the patient's blood pressure, the patient's accelerometer data, or other types of parameterized patient data. In some examples, the parameterized patient data includes one or more values representing the average measurements of patient 14 over a long time period (e.g., about 24 hours to about 48 hours). In some examples, the IMD 16 uploads the parameterized patient data to an external device 27. The external device 27 forwards the parameterized patient data to a computing device 24 (601). In some examples, the external device 27 collects additional parameterized patient data, such as geographical location, accelerometer data, or input from patient 14. In some examples, the external device 27 receives parameterized patient data from other sources, such as one or more of the sensors 80. The external device 27 forwards the parameterized patient data collected from each source to the computing device 24. The computing device 24 receives the patented parameterized data forwarded by the external device 27 via a network 25 (602). In some examples, the computing device 24 receives the parameterized patient data daily.

[0126] The computing device 24 also receives provider data (604) of the patient 14 from the provider database 66. In some instances, the provider data stored by the provider database 66 may include many different types of historical medical information about the patient 14. For example, the provider database 66 may store the patient's medication history, the patient's surgical history, the patient's hospitalization history, the patient's potassium levels over time, or one or more laboratory test results of the patient, etc.

[0127] The computing device 24 applies a machine learning model trained with the parameterized patient data and provider data of multiple patients to the parameterized patient data and provider data of the patient 14 to generate a long-term probability that an arrhythmia will occur in the patient 14 over a long time period (606). In some instances, the long time period is from about 24 hours to about 48 hours. The computing device 24 determines whether the long-term probability exceeds a long-term predetermined threshold (608). In response to determining that the long-term probability exceeds the long-term predetermined threshold, the computing device 24 sends an instruction for the IMD 16 to determine a short-term probability that an arrhythmia will occur in the patient 14 over a short time period (610). In some instances, the computing device 24 forwards the instruction to the external device 27, which in turn forwards the instruction to the IMD 16 (611). In some instances, the short time period is from about 30 minutes to about 60 minutes). In some instances, in response to determining that the long-term probability exceeds the long-term predetermined threshold, the computing device 24 may perform other actions, such as notifying a clinician that the long-term probability has exceeded the long-term predetermined threshold or that the computing system 24 has determined that the patient 14 is likely to have an arrhythmia over the long time period.

[0128] In response to receiving the instruction from the computing device 24, the IMD 16 processes the subsequent parameterized patient data to generate a short-term probability that an arrhythmia will occur in the patient 14 over a short time period (612). In some instances, the IMD 16 generates the short-term probability by performing feature detection on the subsequent parameterized patient data. In some instances, the IMD 16 may analyze parameterized patient data similar to the parameterized patient data analyzed by the computing device 24 as described above. In other instances, the IMD 16 analyzes parameterized patient data different from the parameterized patient data analyzed by the computing device 24. For example, the computing device 24 may analyze parameterized patient data representing one or more values averaged over a long time period (e.g., from about 24 hours to about 48 hours), while the IMD 16 may analyze parameterized patient data representing one or more values averaged over a short time period (e.g., from about 30 minutes to about 60 minutes).

[0129] The IMD 16 determines whether the short-term probability exceeds a short-term predetermined threshold (614). In response to determining that the short-term probability exceeds the short-term predetermined threshold, the IMD 16 performs a remedial action to reduce the short-term probability that an arrhythmia will occur in the patient 14 within a short-term period (616). For example, the IMD 16 may issue a notification of the short-term probability that an arrhythmia will occur in the patient 14 within a short-term period to the computing device 24, such that the patient 14 or a clinician can become aware of the likelihood that an arrhythmia will occur in the patient 14 within a short-term period. As an example, the notification may instruct the patient 14 to perform a breathing exercise or take a medication to reduce the likelihood that an arrhythmia will occur in the patient 14 within a short-term period. In another example, the IMD 16 initiates a treatment, such as a drug delivery treatment or an electrical pacing treatment, to the patient 14 to reduce the likelihood that an arrhythmia will occur in the patient 14 within a short-term period.

[0130] Figure 7 A flowchart illustrating an example operation of the techniques according to the present disclosure. For convenience, reference is made to Figure 1 Describe Figure 7 . In some examples, Figure 7 The operation is an operation for providing a multi-layer prediction of arrhythmia in the patient 14 (e.g., including long-term, medium-term, and short-term predictions of ventricular arrhythmia in the patient 14). In Figure 7 The examples, the long-term, medium-term, and short-term predictions of ventricular arrhythmia in the patient 14 may correspond to predictions that ventricular arrhythmia in the patient 14 is likely to occur in the patient 14 within a long-term period (e.g., about 1 week or longer), a medium-term period (e.g., about 24 hours to about 48 hours), or a short-term period (e.g., about 30 minutes to about 60 minutes).

[0131] In one example, the IMD 16 collects parameterized patient data (700) of patient 114. In some examples, the parameterized patient data includes one or more of the following: the patient's activity level, the patient's heart rate, the patient's posture, the patient's electrocardiogram, the patient's blood pressure, the patient's accelerometer data, or other types of parameterized patient data. In some examples, the parameterized patient data includes one or more values representing the average measurements of patient 14 over a long time period (e.g., about 24 hours to about 48 hours). In some examples, the IMD 16 uploads the parameterized patient data to an external device 501. The external device 27 forwards the parameterized patient data to a computing device 24 (701). In some examples, the external device 27 collects additional parameterized patient data, such as geographical location, accelerometer data, input from patient 14. In some examples, the external device 27 receives parameterized patient data from other sources, such as one or more of the sensors 80. The external device 27 forwards the parameterized patient data collected from each source to the computing device 24. The computing device 24 receives the patent parameterized data (702) forwarded by the external device 27 via a network 25. In some examples, the computing device 24 receives the parameterized patient data on a daily basis.

[0132] The computing device 24 also receives provider data of patient 14 from a provider database 66 (704). In some examples, the provider data stored in the provider database 66 may include many different types of historical medical information about patient 14. By way of example, the provider database 66 may store the patient's medication history, the patient's surgical history, the patient's hospitalization history, the patient's potassium level over time, or one or more laboratory test results of the patient, etc.

[0133] The computing device 24 applies a machine learning model trained using the parameterized patient data and provider data of multiple patients to the parameterized patient data and provider data of patient 14 to generate a long-term probability that arrhythmia will occur in patient 14 over a long time period (706). In some examples, the long time period is about one week or longer. The computing device 24 determines whether the long-term probability exceeds a long-term predetermined threshold (708).

[0134] In response to determining that the long-term probability exceeds the long-term predetermined threshold, the computing device 24 sends an instruction to cause the external device 27 to determine a medium-term probability that arrhythmia will occur in patient 14 over a medium time period (710). In some examples, in response to determining that the long-term probability exceeds the long-term predetermined threshold, the computing device 24 may perform other actions, such as notifying a clinician that the long-term probability has exceeded the long-term predetermined threshold or that the computing system 24 has determined that patient 14 is likely to have arrhythmia over a long time period.

[0135] In response to receiving instructions from computing device 24, external device 27 applies a machine learning model trained using parameterized patient data and provider data of multiple patients to the parameterized patient data and provider data of patient 14 to generate a medium-term probability (712) that arrhythmia will occur in patient 14 within a medium-term time period. In some instances, external device 27 applies a machine learning model trained using parameterized patient data rather than provider data to the parameterized patient data of patient 14 rather than provider data to generate a medium-term probability that arrhythmia will occur in patient 14 within a medium-term time period. In some instances, the medium-term time period is from about 24 hours to about 48 hours. External device 27 determines whether the medium-term probability exceeds a medium-term predetermined threshold (714). In some instances, the medium-term predetermined threshold is 50%. In some instances, the medium-term predetermined threshold is another threshold, such as 75%, 80%, 90%, 99%, or 99.5% or 99.9%.

[0136] In response to determining that the medium-term probability exceeds the medium-term predetermined threshold, external device 27 sends instructions for IMD 16 to determine a short-term probability (716) that arrhythmia will occur in patient 14 within a short-term time period. In some instances, the short-term time period is from about 30 minutes to about 60 minutes). In some instances, in response to determining that the medium-term probability exceeds the medium-term predetermined threshold, external device 27 may perform other actions, such as notifying a clinician that the medium-term probability has exceeded the medium-term predetermined threshold or that external device 27 has determined that patient 14 is likely to have arrhythmia within the medium-term time period.

[0137] In response to receiving instructions from external device 27, IMD 16 processes subsequent parameterized patient data to generate a short-term probability (718) that arrhythmia will occur in patient 14 within a short-term time period. In some instances, IMD 16 generates the short-term probability by performing feature detection on the subsequent parameterized patient data. In some instances, IMD 16 may analyze parameterized patient data similar to the parameterized patient data analyzed by computing device 24 and / or external device 27 as described above. In other instances, IMD 16 analyzes parameterized patient data different from the parameterized patient data analyzed by computing device 24 and / or external device 27. For example, computing device 24 may analyze parameterized patient data representing one or more values averaged over a long-term time period (e.g., about one week or longer), external device 27 may analyze parameterized patient data representing one or more values averaged over a medium-term time period (e.g., about 24 hours to about 48 hours), and IMD 16 may analyze parameterized patient data representing one or more values averaged over a short-term time period (e.g., about 30 minutes to about 60 minutes).

[0138] The IMD 16 determines whether the short-term probability exceeds a short-term predetermined threshold (720). In response to determining that the short-term probability exceeds the short-term predetermined threshold, the IMD 16 performs a remedial action to reduce the short-term probability (722) that an arrhythmia will occur in the patient 14 within a short-term period. For example, the IMD 16 may issue a notification of the short-term probability that an arrhythmia will occur in the patient 14 within a short-term period to the computing device 24, such that the patient 14 or the clinician can become aware of the likelihood that an arrhythmia will occur in the patient 14 within a short-term period. In another example, the IMD 16 initiates a treatment for the patient 14, such as a drug delivery treatment or an electrical pacing treatment, to reduce the likelihood that an arrhythmia will occur in the patient 14 within a short-term period.

[0139] As described herein, the use of the terms "long-term probability", "mid-term probability", and "short-term probability" and the use of the terms "long-term probability threshold", "mid-term probability threshold", and "short-term probability threshold" are used throughout this disclosure only to distinguish the terms from one another. These terms are not limited to any particular length of time. For example, the term "long-term probability" may be interchangeable with the term "first probability", the term "mid-term probability" may be interchangeable with the term "second probability", the term "short-term probability" may be interchangeable with the term "third probability", the term "long-term probability threshold" may be interchangeable with the term "first probability threshold", the term "mid-term probability threshold" may be interchangeable with the term "second probability threshold", and the term "short-term probability threshold" may be interchangeable with the term "third probability threshold".

[0140] Figure 8A A flowchart illustrating an example operation of the techniques according to the present disclosure. Specifically, Figure 8A the flowchart depicts operations for providing a multi-layer prediction of arrhythmia in the patient 14 (e.g., including long-term, mid-term, and short-term predictions of ventricular arrhythmia in the patient 14). For convenience, reference is made to Figure 1 describe FIG. 8.

[0141] As depicted in the example of FIG. 8, the cloud computing network 808 applies the long-term algorithm 802 to the provider data of the patient 14 to generate a long-term probability that an arrhythmia will occur in the patient 14 within a long-term period. Additional details of the long-term algorithm 802 are depicted in Figure 8B In some examples, the cloud computing network 808 includes Figure 1 one or more computing devices 24 of

[0142] In response to determining that the long-term probability that the patient 14 will experience an arrhythmia exceeds a long-term predetermined threshold, the cloud computing network 808 sends an instruction to cause the external device 27 to apply the mid-term algorithm 804 to the parameterized patient data of the patient 14 to generate a mid-term probability that an arrhythmia will occur in the patient 14 within a mid-term period. As depicted in Figure 8AAs described, the external device 27 may receive parameter data from the IMD 16, for example, daily or otherwise periodically. In Figure 8C Additional details of the medium-term algorithm 804 are depicted.

[0143] In response to determining that the medium-term probability that patient 14 will experience an arrhythmia exceeds a medium-term predetermined threshold, the external device 27 sends an instruction for the IMD 16 to apply the short-term algorithm 806 to the parameterized patient data of patient 14 to generate a short-term probability that an arrhythmia will occur in patient 14 within a short time period. In Figure 8D Additional details of the short-term algorithm 806 are depicted.

[0144] In some instances, each higher-level algorithm may supply risk factors to a lower-level algorithm, where the lower-level algorithm may include the risk factors when determining the probability that an arrhythmia will occur in patient 14. For example, the medium-term algorithm 804 may use the long-term probability that an arrhythmia will occur in patient 14 within a long time period generated by the long-term algorithm 802 when determining the medium-term probability that an arrhythmia will occur in patient 14 within a medium time period. Similarly, the short-term algorithm 806 may use the medium-term probability that an arrhythmia will occur in patient 14 within a medium time period generated by the medium-term algorithm 804 when determining the short-term probability that an arrhythmia will occur in patient 14 within a short time period.

[0145] Using Figure 8A The multi-level arrhythmia prediction depicted in may extend the battery life of the IMD 16, thus allowing the IMD 16 to be implanted in patient 14 for a longer time without replacement. In addition, using such multi-level arrhythmia prediction may provide a high degree of sensitivity, specificity, and accuracy to the needs and risks of patient 14.

[0146] Figure 8B To illustrate in further detail Figure 8A A block diagram of the algorithm 802 for long-term prediction of arrhythmia depicted in the example of. In Figure 8B In an example of, one or more computing devices 24 of the cloud computing network 808 apply various types of provider data and parameterized patient data w1, w2, w3,..., w m To a neural network layer including a plurality of neurons and a plurality of hidden layers 1-5. In some instances, each of w1, w2, w3,..., w m Corresponds to a different type of provider data or parameterized patient data, such as the patient's medication history, the surgical history of patient 14, the electrocardiogram of patient 14, etc. The neural network processes the provider data and parameterized patient data w1, w2, w3,..., w m By mapping the inputs on a plurality of neurons and a plurality of hidden layers 1-5 to generate an output P EMR, which is the long - term probability of arrhythmia occurring in patient 14 over a long time period.

[0147] Figure 8C To explain in further detail Figure 8A is a block diagram of algorithm 804 for mid - term prediction of arrhythmia depicted in the example of Figure 8B In the example of , external device 27 applies various types of provider data and parameterized patient data x1, x2, x3,..., x m to a neural network layer including multiple neurons and multiple hidden layers 1 - 5. In some examples, each of x1, x2, x3,..., x m corresponds to a different type of provider data or parameterized patient data, such as the patient's medication history, the surgical history of patient 14, the electrocardiogram of patient 14, etc. The neural network processes the provider data and parameterized patient data x1, x2, x3,..., x m by mapping the inputs over multiple neurons and multiple hidden layers 1 - 5 to produce an output P D24 , which is the mid - term probability of arrhythmia occurring in patient 14 over a mid - term time period.

[0148] Figure 8D To explain in further detail Figure 8A is a block diagram of algorithm 806 for short - term prediction of arrhythmia depicted in the example of Figure 8D As depicted in , IMD 16 generates a short - term probability of arrhythmia occurring in patient 14 by performing feature detection on each of multiple windows e1, e2, e3, e4,..., em of parameterized patient data. Each of the windows e1, e2, e3, e4,..., em represents a period of time during which IMD 16 applies criteria to determine whether one or more indicators of a subsequent arrhythmia are present. As depicted in Figure 8D In the example of , at decision window 820, IMD 16 determines the probability that arrhythmia is likely to occur in patient 14 during a short - term time period. Additional information regarding the use of multiple parameters to predict arrhythmia can be found, for example, in U.S. Patent Application Publication No. 2017 / 0354365, titled "MULTI - PARAMETER PREDICTION OF ACUTE CARDIAC EPISODES AND ATTACKS", filed on June 12, 2017 and published on December 14, 2017 by Zhou et al.

[0149] In some examples, IMD 16 may use the long - term probability P of arrhythmia occurring in patient 14 over a long time period generated by one or more computing devices 24EMR and a mid - term probability P of arrhythmia occurring in patient 14 during a mid - term time period generated by external device 27 D24 in combination with feature detection of the parameterized patient data to determine a probability that arrhythmia is likely to occur in patient 14 during a short - term time period. In some instances, IMD 16 applies different weights to the feature - detected long - term probability P EMR and the mid - term probability P D24 for each to determine a probability that arrhythmia is likely to occur in patient 14 during a short - term time period. In some instances, IMD 16 applies a decision tree to the feature - detected long - term probability P EMR and the mid - term probability P D24 for each to determine a probability that arrhythmia is likely to occur in patient 14 during a short - term time period.

[0150] Figure 9 A block diagram of an example algorithm 902 for short - term prediction of arrhythmia in accordance with the techniques of the present disclosure. For convenience, reference is made to Figure 1 describe Figure 9 . In some instances, example algorithm 902 operates in a manner substantially similar to the algorithm 806 for arrhythmia for short - term prediction Figure 8A and 8D . In some instances, IMD 16 can process multiple types of parameterized patient data to determine a short - term probability that arrhythmia is likely to occur in patient 14 during a short - term time period. In Figure 9 an instance, IMD 16 applies a machine - learning model to calculate a short - term probability that arrhythmia is likely to occur in patient 14 during a short - term time period. By way of example, as input, the machine - learning model of IMD 16 can receive feature detection performed on different types of parameterized patient data. By way of example, IMD can perform feature detection on one or more of electrocardiogram data, electrode impedance measurements, accelerometer data, temperature data of patient 14, audio data of the heart of patient 14, or a risk score calculated by other algorithms at different levels. Such risk scores can include, for example, a long - term probability P of arrhythmia occurring in patient 14 during a long - term time period generated by one or more computing devices 24 EMR and a mid - term probability P of arrhythmia occurring in patient 14 during a mid - term time period generated by external device 27 D24 .

[0151] Accordingly, the techniques disclosed herein can allow for enhanced patient care and an increased ability to prevent arrhythmias in patients. For example, knowledge that the likelihood of a tachyarrhythmia occurring in the next few days is relatively high can be used to help guide preventive care for a patient. Additionally, knowledge that the likelihood of a tachyarrhythmia occurring in the next few minutes or hours is relatively high can be used to guide preventive treatment or emergency care for the patient. Further, a system such as that disclosed herein can move computationally expensive and energy-consuming long-term arrhythmia prediction operations from a patient's personal medical device to a cloud computing system and activate short-term cardiac prediction operations on the medical device only when the cloud computing system predicts a relatively high likelihood of an arrhythmia occurring in the next few days. Thus, the techniques of the present disclosure can save power and extend the battery life of a medical device by using the medical device for cardiac prediction only when the likelihood of an arrhythmia occurring within a particular time span is relatively high.

[0152] The following examples can illustrate one or more aspects of the present disclosure.

[0153] Example 1. A computing device includes processing circuitry and a storage medium, the computing device being configured to: receive parameterized patient data of a patient; apply a machine learning model trained using parameterized patient data of a plurality of patients to the parameterized patient data to generate a first probability that an arrhythmia will occur in the patient within a first time period; determine that the first probability exceeds a predetermined threshold; and in response to the determination that the first probability exceeds the predetermined threshold, send an instruction to a second device to cause the second device to determine a second probability that an arrhythmia will occur in the patient within a second time period.

[0154] Example 2. The computing device according to Example 1, wherein the computing device is further configured to receive provider data of the patient, and wherein, to apply the machine learning model, the computing device is configured to apply a machine learning model trained using parameterized patient data and provider data of a plurality of patients to the parameterized patient data and the provider data of the patient to generate the first probability that the arrhythmia will occur in the patient within the first time period.

[0155] Example 3. The computing device according to any combination of Examples 1 to 2, wherein the parameterized patient data includes at least one of the following: physiological data of the patient; age of the patient; gender of the patient; activity pattern of the patient; sleep pattern of the patient; gait changes of the patient; device-related data of a medical device; temperature trend of the patient; or temperature trend of the medical device.

[0156] Example 4. The computing device according to Example 3, wherein the device-related data of the medical device includes at least some of the following: the impedance of one or more electrodes of the medical device; the selection of the electrodes; the drug delivery schedule of the medical device; the history of electro-pacing therapy delivered to the patient; the diagnostic data of the medical device; the detected activity level of the patient; the detected posture of the patient; the detected temperature of the patient; or the detected sleep state of the patient.

[0157] Example 5. The computing device according to any combination of Examples 1 to 4, wherein the provider data of the patient includes at least some of the following: the patient's medication history; the patient's surgical history; the patient's hospitalization history; the patient's potassium level; or the patient's one or more laboratory test results.

[0158] Example 6. The computing device according to any combination of Examples 1 to 5, wherein the first time period is greater than about 1 day, and the second time period is less than about 1 day.

[0159] Example 7. The computing device according to any combination of Examples 1 to 6, wherein the computing device is further configured to: after sending the instruction to the second device, receive feedback from at least one of the second devices indicating whether the arrhythmia has occurred in the patient during the first time period; and update the machine learning model using the feedback indicating whether the arrhythmia has occurred in the patient during the first time period.

[0160] Example 8. A device configured to: collect parameterized patient data of a patient via one or more of a plurality of electrodes or sensors; receive an instruction from a computing device to generate a probability that an arrhythmia will occur in the patient within a certain time period; in response to the instruction, process the parameterized patient data to generate the probability that the arrhythmia will occur in the patient within the certain time period; determine that the probability exceeds a predetermined threshold; and in response to the determination that the probability exceeds the predetermined threshold, perform a remedial action to reduce the probability that the arrhythmia will occur in the patient within the certain time period.

[0161] Example 9. The device according to Example 8, wherein in order to process the parameterized patient data to generate the probability that the arrhythmia will occur in the patient within the certain time period, the device is configured to apply a machine learning model trained using parameterized patient data of a plurality of patients to the parameterized patient data to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0162] Example 10. The apparatus according to Example 8, wherein to process the parameterized patient data to generate the probability that the arrhythmia will occur in the patient within the certain time period, the apparatus is configured to: identify one or more characteristics of the T wave of the electrocardiogram of the parameterized patient data; and apply a model to the one or more characteristics to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0163] Example 11. The apparatus according to Example 10, wherein the one or more characteristics of the T wave include the amplitude of the T wave, and wherein to apply the model to the one or more characteristics to generate the probability that the arrhythmia will occur in the patient within the certain time period, the apparatus is configured to apply the model to the amplitude of the T wave to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0164] Example 12. The apparatus according to any combination of Examples 10 to 11, wherein the one or more characteristics of the T wave include the frequency of the T wave, and wherein to apply the model to the one or more characteristics to generate the probability that the arrhythmia will occur in the patient within the certain time period, the apparatus is configured to apply the model to the frequency of the T wave to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0165] Example 13. The apparatus according to any combination of Examples 8 to 12, wherein to perform the remedial action, the apparatus is configured to issue a notification of the probability that the arrhythmia will occur in the patient within the certain time period to at least one of the computing device and the external device.

[0166] Example 14. The apparatus according to any combination of Examples 8 to 13, wherein to perform the remedial action, the apparatus is configured to deliver at least one of a drug delivery treatment and an electrical pacing treatment to the patient.

[0167] Example 15. An external device configured to: receive an instruction to generate a first probability that an arrhythmia will occur in a patient within a first time period; in response to the instruction, process parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period; determine that the first probability exceeds a first predetermined threshold; and in response to the determination that the first probability exceeds the first predetermined threshold, send an instruction to a medical device to cause the medical device to determine a second probability that an arrhythmia will occur in the patient within a second time period.

[0168] Example 16. The external device according to Example 15, wherein, in order to process the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period, the external device is configured to apply a machine learning model trained using the parameterized patient data of multiple patients to the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period.

[0169] Example 17. The external device according to Example 15, wherein, in order to process the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period, the external device is configured to perform feature detection of the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period.

[0170] Example 18. The external device according to any combination of Examples 15 to 17, wherein the external device is further configured to: receive the parameterized patient data from at least one of the medical device or one or more sensors; and send the received parameterized patient data to a computing device.

[0171] Example 19. The external device according to Example 18, wherein the external device is further configured to: determine the geographical location of the patient; and send the determined geographical location together with the parameterized patient data to the computing device.

[0172] Example 20. The external device according to any combination of Examples 15 to 19, wherein the external device includes at least one of a mobile device or a wearable electronic device.

[0173] Example 21. A method, comprising: receiving, by a computing device including processing circuitry and a storage medium, parameterized patient data of a patient; applying, by the computing device, a machine learning model trained using the parameterized patient data of multiple patients to the parameterized patient data to generate a first probability that an arrhythmia will occur in the patient within a first time period; determining, by the computing device, that the first probability exceeds a predetermined threshold; and in response to the determination that the first probability exceeds the predetermined threshold, sending, by the computing device, an instruction to a second device to cause the second device to determine a second probability that an arrhythmia will occur in the patient within a second time period.

[0174] Example 22. The method according to Example 21 further includes: receiving, by the computing device, provider data of the patient, wherein applying the machine learning model includes applying a machine learning model trained using parameterized patient data and provider data of a plurality of patients to the parameterized patient data and the provider data of the patient to generate the first probability that the arrhythmia will occur in the patient during the first time period.

[0175] Example 23. The method according to any combination of Examples 21 to 22 further includes: after sending the instruction to the second device, receiving, by the computing device and from the second device, feedback indicating whether the arrhythmia has occurred in the patient during the first time period; and updating, by the computing device, the machine learning model using the feedback indicating whether the arrhythmia has occurred in the patient during the first time period.

[0176] Example 24. A method includes: collecting, by a device, parameterized patient data of a patient; receiving, by the device from a computing device, an instruction to generate a probability that an arrhythmia will occur in the patient during a certain time period; in response to the instruction, processing, by the device, the parameterized patient data to generate the probability that the arrhythmia will occur in the patient during the certain time period; determining, by the device, that the probability exceeds a predetermined threshold; and in response to the determination that the probability exceeds the predetermined threshold, performing, by the device, a remedial action to reduce the probability that the arrhythmia will occur in the patient during the certain time period.

[0177] Example 25. The method according to Example 24, wherein processing the parameterized patient data to generate the probability that the arrhythmia will occur in the patient during the certain time period includes applying a machine learning model trained using parameterized patient data of a plurality of patients to the parameterized patient data to generate the probability that the arrhythmia will occur in the patient during the certain time period.

[0178] Example 26. The method according to Example 24, wherein processing the parameterized patient data to generate the probability includes: identifying one or more features of the T wave of the electrocardiogram of the parameterized patient data; and applying a model to the one or more features to generate the probability that the arrhythmia will occur in the patient during the certain time period.

[0179] Example 27. The method according to Example 26, wherein the one or more features of the T-wave include the amplitude of the T-wave, and wherein applying the model to the one or more features to generate the probability that the arrhythmia will occur in the patient within the certain time period includes applying the model to the amplitude of the T-wave to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0180] Example 28. The method according to any combination of Examples 26 to 27, wherein the one or more features of the T-wave include the frequency of the T-wave, and wherein applying the model to the one or more features to generate the probability that the arrhythmia will occur in the patient within the certain time period includes applying the model to the frequency of the T-wave to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0181] Example 1A. A method comprising: receiving, by a computing device including processing circuitry and a storage medium, parameterized patient data of a patient collected by a medical device; applying, by the computing device, a machine learning model trained using parameterized patient data of multiple patients to the parameterized patient data to generate a first probability that an arrhythmia will occur in the patient within a first time period; determining, by the computing device, that the first probability exceeds a predetermined threshold; and in response to the determination that the first probability exceeds the predetermined threshold, sending, by the computing device, an instruction to the medical device to cause the medical device to determine a second probability that the arrhythmia will occur in the patient within a second time period.

[0182] Example 2A. The method according to Example 1A, further comprising receiving, by the computing device, electronic medical record (EMR) data of the patient, wherein applying the machine learning model includes applying a machine learning model trained using parameterized patient data and EMR data of multiple patients to the parameterized patient data and the EMR data of the patient to generate the first probability that the arrhythmia will occur in the patient within the first time period.

[0183] Example 3A. The method according to any one of Examples 1A or 2A, wherein the parameterized patient data includes at least one of: physiological data of the patient; or device-related data of the medical device.

[0184] Example 4A. The method according to Example 3A, wherein the device-related data of the medical device includes at least some of: the impedance of one or more electrodes of the medical device; electrode selection; the drug delivery schedule of the medical device; the history of electro-pacing treatment delivered to the patient; or the diagnostic data of the medical device.

[0185] Example 5A. The method according to any one of Examples 1A to 4A, wherein receiving the parameterized patient data comprises receiving the parameterized patient data at least daily.

[0186] Example 6A. The method according to any one of Examples 1A to 5A, wherein the patient's EMR data comprises at least some of the following: the patient's medication history; the patient's surgical history; the patient's hospitalization history; the patient's potassium level; or one or more laboratory test results of the patient.

[0187] Example 7A. The method according to any one of Examples 1A to 6A, wherein the first time period is greater than 1 day and less than 1 week; and wherein the second time period is greater than 1 minute and less than 1 day.

[0188] Example 8A. The method according to any one of Examples 1A to 6A, wherein the first time period is less than 1 day; and wherein the second time period is less than 1 hour.

[0189] Example 9A. The method according to any one of Examples 1A to 8A, further comprising: after sending the instruction to the medical device, receiving, by the computing device and from the medical device, feedback indicating whether the arrhythmia has occurred in the patient during the first time period; and updating, by the computing device, the machine learning model using the feedback indicating whether the arrhythmia has occurred in the patient during the first time period.

[0190] Example 10A. The method according to any one of Examples 1A to 9A, wherein the machine learning model comprises a model generated by a convolutional neural network.

[0191] Example 11A. A method comprising: collecting, by a medical device, parameterized patient data of a patient; receiving, by the medical device from a computing device, an instruction to generate a probability that an arrhythmia will occur in the patient within a certain time period; in response to the instruction, processing, by the medical device, the parameterized patient data to generate the probability that the arrhythmia will occur in the patient within the certain time period; determining, by the medical device, that the probability exceeds a predetermined threshold; and in response to the determination that the probability exceeds the predetermined threshold, performing, by the medical device, a remedial action to reduce the probability that the arrhythmia will occur in the patient within the certain time period.

[0192] Example 12A. The method according to Example 11A, wherein processing the parameterized patient data to generate the probability that the arrhythmia will occur in the patient within the certain time period includes applying a machine learning model trained using the parameterized patient data of multiple patients to the parameterized patient data to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0193] Example 13A. The method according to any one of Examples 11A or 12A, wherein processing the parameterized patient data to generate the probability includes: identifying one or more features of the T wave of the electrocardiogram of the parameterized patient data; and applying a model to the one or more features to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0194] Example 14A. The method according to Example 13A, wherein the one or more features of the T wave include the amplitude of the T wave, and wherein applying the model to the one or more features to generate the probability that the arrhythmia will occur in the patient within the certain time period includes applying the model to the amplitude of the T wave to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0195] Example 15A. The method according to Example 13A, wherein the one or more features of the T wave include the frequency of the T wave, and wherein applying the model to the one or more features to generate the probability that the arrhythmia will occur in the patient within the certain time period includes applying the model to the frequency of the T wave to generate the probability that the arrhythmia will occur in the patient within the certain time period.

[0196] Example 16A. The method according to any one of Examples 11A to 15A, wherein the medical device includes a wearable medical device.

[0197] Example 17A. The method according to any one of Examples 11A to 16A, wherein the medical device includes an implantable medical device (IMD).

[0198] Example 18A. The method according to any one of Examples 11A to 17A, further comprising sending the parameterized patient data to the computing device at least daily by the medical device.

[0199] Example 19A. The method according to any one of Examples 11A to 18A, wherein performing the remedial action includes issuing a notification to the computing device of the probability that the arrhythmia will occur in the patient within the certain time period.

[0200] Example 20A. The method according to any one of Examples 11A to 19A, wherein performing the remedial action includes delivering at least one of a drug delivery treatment and an electrical pacing treatment to the patient.

[0201] Example 21A. A computing device comprising processing circuitry and a storage medium, the computing device being configured to: receive parameterized patient data of a patient collected by a medical device; apply a machine learning model trained using parameterized patient data of multiple patients to the parameterized patient data to generate a first probability that an arrhythmia will occur in the patient within a first time period; determine that the first probability exceeds a predetermined threshold; and in response to the determination that the first probability exceeds the predetermined threshold, send an instruction to the medical device to cause the medical device to determine a second probability that an arrhythmia will occur in the patient within a second time period.

[0202] Example 22A. The computing device according to Example 21A, wherein the computing device is further configured to receive electronic medical record (EMR) data of the patient, and wherein, in order to apply the machine learning model to generate the first probability, the computing device is further configured to apply a machine learning model trained using parameterized patient data and EMR data of multiple patients to the parameterized patient data and the EMR data of the patient to generate the first probability that the arrhythmia will occur in the patient within the first time period.

[0203] Example 23A. The computing device according to any one of Examples 21A or 22A, wherein the parameterized patient data includes at least one of the following: physiological data of the patient; or device-related data of the medical device.

[0204] Example 24A. The computing device according to any one of Examples 21A to 23A, wherein, in order to receive the parameterized patient data, the computing device is configured to receive the parameterized patient data at least once a day.

[0205] Example 25A. The computing device according to any one of Examples 21A to 24A, wherein the first time period is greater than 1 day and less than 1 week; and wherein the second time period is greater than 1 minute and less than 1 day.

[0206] Example 26A. The computing device according to any one of Examples 21A to 25A, wherein the first time period is less than 1 day; and wherein the second time period is less than 1 hour.

[0207] Example 27A. The computing device according to any one of Examples 21A to 26A, wherein the computing device is further configured to: after sending the instructions to the medical device, receive from the medical device feedback indicating whether the arrhythmia has occurred in the patient during the first time period; and update the machine learning model using the feedback indicating whether the arrhythmia has occurred in the patient during the first time period.

[0208] Example 28A. A medical device configured to: collect parameterized patient data of a patient via a plurality of electrodes or sensors; receive from a computing device an instruction to generate a probability that an arrhythmia will occur in the patient during a certain time period; in response to the instruction, process the parameterized patient data to generate the probability that the arrhythmia will occur in the patient during the time period; determine that the probability exceeds a predetermined threshold; and in response to the determination that the probability exceeds the predetermined threshold, perform a remedial action to reduce the probability that the arrhythmia will occur in the patient during the time period.

[0209] Example 29A. The medical device according to Example 28A, wherein in order to process the parameterized patient data to generate the probability, the medical device is configured to: identify one or more features of the T wave of an electrocardiogram of the parameterized patient data; and apply a model to the one or more features to generate the probability that the arrhythmia will occur in the patient during the certain time period.

[0210] Example 30A. The medical device according to Example 29A, wherein the one or more features of the T wave include the amplitude of the T wave, and wherein in order to apply the model to the one or more features to generate the probability that the arrhythmia will occur in the patient during the certain time period, the medical device is configured to apply the model to the amplitude of the T wave to generate the probability that the arrhythmia will occur in the patient during the certain time period.

[0211] Example 31A. The medical device according to Example 29A, wherein the one or more features of the T wave include the frequency of the T wave, and wherein in order to apply the model to the one or more features to generate the probability that the arrhythmia will occur in the patient during the certain time period, the medical device is configured to apply the model to the frequency of the T wave to generate the probability that the arrhythmia will occur in the patient during the certain time period.

[0212] Example 32A. The medical device according to any one of Examples 28A to 31A, wherein the medical device is further configured to send the parameterized patient data to the computing device at least daily.

[0213] Example 33A. The medical device according to any one of Examples 28A to 32A, wherein, in order to perform the remedial action, the medical device is configured to issue a notification to the computing device of the probability that the arrhythmia will occur in the patient within the certain time period.

[0214] Example 34A. The medical device according to any one of Examples 28A to 33A, wherein, in order to perform the remedial action, the medical device is configured to deliver to the patient at least one of a drug delivery treatment and an electrical pacing treatment.

[0215] Example 35A. A system comprising: an implantable medical device (IMD) configured to collect parameterized patient data of a patient via a plurality of electrodes or sensors; and a computing device comprising a storage medium and processing circuitry, wherein the computing device is configured to: receive the parameterized patient data collected by the IMD; receive electronic medical record (EMR) data of the patient; apply a machine learning model trained using parameterized patient data and EMR data of a plurality of patients to the parameterized patient data and the EMR data of the patient to generate a first probability that an arrhythmia will occur in the patient within a first time period; determine that the first probability exceeds a first predetermined threshold; and in response to the determination that the first probability exceeds the first predetermined threshold, send an instruction to the IMD via the computing device to cause the IMD to determine a second probability that the arrhythmia will occur in the patient within a second time period; wherein the IMD is further configured to: receive the instruction from the computing device; in response to the instruction, process the parameterized patient data to generate the second probability that the arrhythmia will occur in the patient within the second time period; determine that the second probability exceeds a second predetermined threshold; and in response to the determination that the second probability exceeds the second predetermined threshold, perform at least one of: issue a notification to the computing device of the second probability that the arrhythmia will occur in the patient within the second time period; and deliver to the patient at least one of a drug delivery treatment and an electrical pacing treatment.

[0216] In some examples, the techniques of the present disclosure include systems that include means for performing any of the methods described herein. In some examples, the techniques of the present disclosure include computer-readable media that include instructions that cause processing circuitry to perform any of the methods described herein.

[0217] It should be understood that the various aspects disclosed herein can be combined in different combinations rather than the combinations specifically presented in the specification and the drawings. It should also be understood that depending on the instance, certain actions or events of any of the processes or methods described herein can be performed in a different order, can be added, combined, or completely omitted (e.g., all of the described actions or events may not be necessary for performing the techniques). Additionally, although for clarity some aspects of the disclosure are described as being performed by a single module, unit, or circuit, it should be understood that the techniques of the disclosure can be performed by a combination of units, modules, or circuitry associated with, for example, a medical device.

[0218] In one or more instances, the described techniques can be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions can be stored as one or more instructions or code on a computer-readable medium and executed by a hardware-based processing unit. The computer-readable medium can include a non-transitory computer-readable medium corresponding to a tangible medium such as a data storage medium (e.g., RAM, ROM, EEPROM, flash memory, or any other medium that can be used to store the desired program code in the form of instructions or data structures and that can be accessed by a computer).

[0219] The instructions can be executed by one or more processors, which can be, for example, one or more digital signal processors (DSPs), general microprocessors, application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. Thus, as used herein, the term "processor" or "processing circuitry" can refer to any of the foregoing structures or any other physical structure suitable for implementing the described techniques. Moreover, the techniques can be fully implemented in one or more circuits or logic elements.

Claims

1. An implantable medical device configured to: Collect parameterized patient data of a patient via one or more of a plurality of electrodes or sensors; Receive instructions from a non-implantable computing device that generate a first probability of an arrhythmia occurring in the patient during a first time period, wherein, The non-implantable computing device does not have sensing circuitry for cardiac signals and is configured to receive the parameterized patient data of the patient from the implantable medical device, and the instructions are generated in response to a determination by the non-implantable computing device that a second probability that an arrhythmia will occur in the patient within a second time period exceeds a second predetermined threshold, the first time period being shorter than the second time period; In response to the instructions, process the parameterized patient data to generate a first probability that the arrhythmia will occur in the patient within the first time period that is shorter than the second time period; Determine that the first probability exceeds a first predetermined threshold; And In response to the determination that the first probability exceeds the first predetermined threshold, perform a remedial action to reduce the first probability that the arrhythmia will occur in the patient within the first time period.

2. The implantable medical device according to claim 1, wherein, in order to process the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period, the device is configured to apply a machine learning model trained using parameterized patient data of a plurality of patients to the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period.

3. The implantable medical device according to any one of claims 1 to 2, wherein, in order to process the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period, the device is configured to: Identify one or more features of the T wave of an electrocardiogram of the parameterized patient data; and Apply a model to the one or more features to generate the first probability that the arrhythmia will occur in the patient within the first time period.

4. The implantable medical device according to claim 3, Wherein the one or more features of the T wave include the amplitude of the T wave, and Wherein, in order to apply the model to the one or more features to generate the first probability that the arrhythmia will occur in the patient within the first time period, the device is configured to apply the model to the amplitude of the T wave to generate the first probability that the arrhythmia will occur in the patient within the first time period.

5. The implantable medical device according to claim 3, Wherein the one or more features of the T wave include the frequency of the T wave, and Wherein, in order to apply the model to the one or more features to generate the first probability that the arrhythmia will occur in the patient within the first time period, the device is configured to apply the model to the frequency of the T wave to generate the first probability that the arrhythmia will occur in the patient within the first time period.

6. The implantable medical device according to claim 5, wherein, in order to perform the remedial action, the device is configured to issue a notification of the probability that the arrhythmia will occur in the patient within a certain time period to at least one of the computing device and the external device.

7. The implantable medical device according to claim 1, wherein, in order to perform the remedial action, the device is configured to deliver to the patient at least one of a drug delivery treatment and an electrical pacing treatment.

8. A system including an implantable medical device and a non-implantable computing device, the implantable medical device being configured to: Collect parameterized patient data of a patient via one or more electrodes; Receive from the non-implantable computing device an instruction to generate a first probability that an arrhythmia will occur in the patient within a first time period; In response to the instruction, process the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient within the first time period; Determine that the first probability exceeds a first predetermined threshold; And In response to the determination that the first probability exceeds the first predetermined threshold, perform a remedial action to reduce the first probability that the arrhythmia will occur in the patient within the first time period, The non-implantable computing device includes processing circuitry and a storage medium, does not include sensing circuitry for cardiac signals, and the non-implantable computing device is configured to: Receive parameterized patient data of the patient from the implantable medical device; Apply a machine learning model trained using parameterized patient data of multiple patients to the received parameterized patient data to generate a second probability that an arrhythmia will occur in the patient within a second time period greater than the first time period; Determine that the second probability exceeds a second predetermined threshold; And In response to the determination that the second probability exceeds the second predetermined threshold, send the instruction to the implantable medical device to cause the implantable medical device to determine the first probability that the arrhythmia will occur in the patient within the first time period.

9. The system according to claim 8, wherein the computing device is further configured to receive provider data of the patient, wherein, in order to apply the machine learning model, the computing device is configured to apply a machine learning model trained using parameterized patient data and provider data of multiple patients to the parameterized patient data and the provider data of the patient to generate the second probability that the arrhythmia will occur in the patient within the second time period.

10. The system according to any one of claims 8 to 9, wherein the first time period is less than about 1 day and the second time period is greater than about 1 day.

11. The system according to any one of claims 8 to 9, wherein the computing device is further configured to: After sending the instruction to the implantable medical device, receive from the implantable medical device feedback indicating whether the arrhythmia has occurred in the patient within the second time period; and Update the machine learning model using the feedback indicating whether the arrhythmia has occurred in the patient during the second time period.

12. The system according to any one of claims 8 to 9, further comprising an external device configured to: Receive an instruction from the computing device to generate a third probability that an arrhythmia will occur in the patient during a third time period; In response to the instruction, process the parameterized patient data to generate the third probability that the arrhythmia will occur in the patient during the third time period; Determine that the third probability exceeds a third predetermined threshold; and In response to the determination that the third probability exceeds the third predetermined threshold, send the instruction to the implantable medical device to cause the implantable medical device to determine the first probability that an arrhythmia will occur in the patient during the first time period.

13. The system according to claim 12, wherein, in order to process the parameterized patient data to generate the third probability that the arrhythmia will occur in the patient during the first third time period, the external device is configured to apply a machine learning model trained using the parameterized patient data of multiple patients to the parameterized patient data to generate the third probability that the arrhythmia will occur in the patient during the third time period.

14. A computer-readable medium comprising instructions that, when executed by a processor of an implantable medical device, cause the implantable medical device to perform operations including: Collect parameterized patient data of a patient; Receive an instruction from a non-implantable computing device to generate a first probability that an arrhythmia will occur in the patient during a first time period, wherein the non-implantable computing device does not have a sensing circuit system for cardiac signals and is configured to receive the parameterized patient data of the patient from the implantable medical device, and the instruction is generated in response to the non-implantable computing device determining that a second probability that an arrhythmia will occur in the patient during a second time period exceeds a second predetermined threshold based on the parameterized patient data of the patient, and the first time period is shorter than the second time period; In response to the instruction, process the parameterized patient data to generate the first probability that the arrhythmia will occur in the patient during the first time period that is shorter than the second time period; Determine that the first probability exceeds a first predetermined threshold; and In response to the determination that the first probability exceeds the first predetermined threshold, perform a remedial action to reduce the first probability that the arrhythmia will occur in the patient during the first time period.

15. A computer-readable medium comprising instructions that, when executed by a processor of a non-implantable computing device that does not have a sensing circuit system for cardiac signals, cause the non-implantable computing device to perform operations including: Receive parameterized patient data of a patient from an implantable medical device; Apply a machine learning model trained with parameterized patient data of multiple patients to the received parameterized patient data to generate a first probability that arrhythmia will occur in the patient within a first time period; Determine that the first probability exceeds a first predetermined threshold; and In response to the determination that the first probability exceeds the first predetermined threshold, send the instruction to the implantable medical device to cause the implantable medical device to determine a second probability that arrhythmia will occur in the patient within a second time period shorter than the first time period.

Images