Method and apparatus for conductive system pacing capture classification

Abstract

A medical device system includes a memory configured to store cardiac signals sensed after delivery of ventricular conduction system pacing pulses. The medical device system includes processing circuitry configured to determine a capture type classification of the ventricular conduction system pacing pulses, and generate an output based on the capture type classification. The medical device system may include a user interface configured to present a representation of the capture type classification associated with the delivered ventricular conduction system pacing pulses.

Description

Technical Field

[0001] The present disclosure relates to methods and apparatus for classifying the capture type of cardiac conduction system pacing pulses. Background Art

[0002] During normal sinus rhythm (NSR), the heartbeat is regulated by the electrical signal generated by the sinoatrial (SA) node located in the wall of the right atrium. Each intrinsic atrial depolarization signal generated by the SA node spreads in the atrium, causing depolarization and contraction of the atrium, and reaches the atrioventricular (AV) node. The AV node responds by propagating the ventricular depolarization signal through the bundle of His (Bundle ofHis or "His bundle") of the ventricular septum and thereafter reaching the Purkinje branches and Purkinje muscle fibers of the right and left ventricles. This native conduction system includes the bundle of His, the left and right branches (sometimes referred to as the left and right bundle branches), and the Purkinje fibers can be referred to as the "His-Purkinje conduction system" or "His-Purkinje system".

[0003] Patients with conduction system abnormalities, such as AV node malconduction, SA node dysfunction, or other conduction abnormalities, may receive a pacemaker to restore a more normal heart rhythm and AV synchrony. Ventricular pacing may be performed to maintain the ventricular rate in patients with conduction abnormalities. A single-chamber ventricular pacemaker may be coupled to a transvenous ventricular lead carrying an electrode placed in the right ventricle (e.g., in the right ventricular apex). The pacemaker itself is typically implanted in a subcutaneous pocket, with the transvenous ventricular lead tunneled into the subcutaneous pocket. Intracardiac pacemakers have been introduced or proposed for complete implantation within the patient's heart, thereby eliminating the need for a transvenous lead. An intracardiac pacemaker can provide sensing and pacing from within a patient's heart chamber (e.g., from within the right ventricle of a patient with AV block or other conduction abnormalities) to provide ventricular rate support.

[0004] A dual chamber pacemaker may be provided that includes a transvenous atrial lead carrying an electrode that is placed in the right atrium and a transvenous ventricular lead carrying an electrode that is advanced through the right atrium into the right ventricle. The dual chamber pacemaker senses atrial and ventricular electrical signals and can provide both atrial pacing and ventricular pacing as needed to promote normal atrial and ventricular rhythms and to promote AV synchronization when SA and / or AV node or other conduction abnormalities are present.

[0005] Cardiac pacing using the His-Purkinje system has been proposed to provide ventricular pacing along the native conduction system of the heart. Long-term ventricular pacing via electrodes at or near the right ventricular apex may be associated with an increased risk of atrial fibrillation or heart failure. Alternative pacing sites along the His-Purkinje system have been studied or proposed, such as pacing at or near the His bundle or in an area near the left and / or right bundle branches. Ventricular pacing via the His-Purkinje system of the heart allows recruitment along the native conduction system of the heart and is hypothesized to promote more physiologically normal cardiac activation than other pacing sites (such as the ventricular apex). Summary of the invention

[0006] The technology disclosed in the present invention generally relates to a medical device system configured to classify the capture type of cardiac pacing pulses of the His-Purkinje conduction system delivered to the heart of a patient. The processing circuit system of the medical device system classifies the capture type of the delivered His-Purkinje conduction system pacing pulse (also referred to herein as a "ventricular conduction system" (VCS) pacing pulse) into one of a plurality of capture types. Among the capture type classifications are, for example, selective His-Purkinje system capture without ventricular myocardial capture, non-selective His-Purkinje system capture with ventricular myocardial capture, ventricular myocardial capture without His-Purkinje system capture, or loss of capture. The processing circuit system of the medical device system can receive a cardiac signal obtained during His-Purkinje system pacing. The processing circuit system can extract data from the received cardiac signal for applying an artificial intelligence or machine learning model to the extracted data. In some examples, the model can be trained using a cardiac signal data set to classify a cardiac signal (e.g., a cardiac signal waveform after pacing) and an associated VCS pacing pulse into one of a plurality of capture types. The processing circuit system may generate pacing capture information based on the capture classification for visual representation to a user in a user interface, the capture classification being determined based on the machine learning model. Additionally or alternatively, the processing circuit system may generate pacing capture information based on the capture classification for use in programming operating parameters into a medical device and controlling VCS pacing by the medical device. Such pacing capture information may include, but is not limited to, capture thresholds and / or recommended VCS pacing output settings associated with one or more capture type classifications.

[0007] In one example, the present disclosure provides a medical device system comprising: a memory configured to store a first cardiac signal sensed after delivering a VCS pacing pulse; and a processing circuit system configured to receive at least the first cardiac signal and apply a pacing capture classification machine learning model to at least the first cardiac signal. The processing circuit system can be configured to determine a capture type classification of a VCS pacing pulse from a plurality of capture types based on the applied pacing capture classification machine learning model. The processing circuit system can be configured to generate an output based on the capture type classification. The medical device system can include a user interface configured to present a representation of a capture type classification associated with a delivered ventricular conduction system pacing pulse in response to the output generated by the processing circuit system.

[0008] In another example, the present disclosure provides a method, the method comprising: storing in a memory a cardiac signal sensed after delivering a VCS pacing pulse; inputting at least the cardiac signal to a pacing capture classification machine learning model; and outputting, by the pacing capture classification machine learning model, a capture type classification of the VCS pacing pulse from a plurality of capture types. The method may further comprise generating, by a processing circuit system, an output based on the capture type classification, and presenting, by a user interface, a representation of the capture type classification associated with the delivered ventricular conduction system pacing pulse.

[0009] In yet another example, the present disclosure provides a non-transitory computer-readable storage medium storing a set of instructions that, when executed by a processing circuit system of a medical device system, causes the medical device system to: store in a memory of the medical device system a cardiac signal sensed after delivery of a VCS pacing pulse; apply a pacing capture classification machine learning model to at least the cardiac signal; and determine a capture type classification of the VCS pacing pulse from a plurality of capture types based on the applied pacing capture classification machine learning model. The instructions may further cause the medical device system to: generate an output based on the capture type classification; and present, by a user interface of the medical device system, a representation of the determined capture type associated with the delivered ventricular conduction system pacing pulse.

[0010] In another example, the present disclosure provides a medical device system, the medical device system comprising: a sensing circuit configured to sense a cardiac electrical signal; a therapy delivery circuit configured to deliver a VCS pacing pulse; and a processing circuit system configured to receive the cardiac electrical signal. The processing circuit system may be configured to apply a machine learning model to at least the sensed cardiac electrical signal, and the machine learning model is trained to determine a pacing capture type using multiple cardiac signal data sets. The processing circuit system may be further configured to determine a pacing capture type of the VCS pacing pulse based on the applied machine learning model. The medical device system may further include a user interface configured to present a representation of the determined pacing capture type of the VCS pacing pulse.

[0011] In another example, the present disclosure provides a medical device system, the medical device system comprising: a sensing circuit configured to sense at least one cardiac electrical signal; a therapy delivery circuit configured to deliver a VCS pacing pulse; and a memory configured to store a cardiac signal segment sensed by the sensing circuit during at least one post-pacing time interval after delivering a first VCS pacing pulse delivered by the therapy delivery circuit. The medical device system further comprises a processing circuit system, the processing circuit system being configured to apply a machine learning model to at least the first cardiac signal segment, and based on the machine learning model, determine a pacing capture type of the first VCS pacing pulse from a plurality of capture types. The processing circuit system may select an operational pacing pulse output based on at least the determined pacing capture type. The therapy delivery circuit may be further configured to deliver a second VCS pacing pulse according to the selected operational pacing pulse output.

[0012] The present invention also discloses the following embodiments:

[0013] Embodiment 1. A medical device system, the medical device system comprising: a memory configured to store a first cardiac signal sensed after delivering a VCS pacing pulse; and a processing circuit system configured to receive the first cardiac signal sensed after delivering the VCS pacing pulse. The processing circuit system is further configured to apply a pacing capture classification machine learning model to at least the first cardiac signal, and determine a capture type classification of the VCS pacing pulse from a plurality of capture types based on the applied pacing capture classification machine learning model. The processing circuit system may generate an output based on the capture type classification. The medical device system may include a user interface configured to present a representation of a capture type classification associated with a delivered ventricular conduction system pacing pulse in response to the generated output.

[0014] Embodiment 2. The medical device system of embodiment 1, wherein the memory is further configured to store a first template beat signal corresponding to a first VCS pacing pulse output. The processing circuit system may be further configured to input at least the first template beat signal and the first cardiac signal to the pacing capture classification machine learning model; and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the first template beat signal and the first cardiac signal.

[0015] Embodiment 3. A medical device system according to any one of embodiments 1 to 2, wherein the memory is further configured to store a first template beat signal corresponding to a first VCS pacing pulse output. The processing circuit system is further configured to determine a first template difference signal from the first cardiac signal and the first template beat signal, input at least the first template difference signal and the first cardiac signal to the pacing capture classification machine learning model; and determine the capture type classification based on the pacing capture classification machine applied to at least the first template difference signal and the first cardiac signal model.

[0016] Embodiment 4. The medical device system of any one of embodiments 1 to 3, wherein the memory is further configured to store a plurality of template beat signals, wherein each template beat signal corresponds to one of a plurality of VCS pacing pulse outputs. The processing circuit system may be further configured to input each of the plurality of template beat signals and the first cardiac signal into the pacing capture classification machine learning model, and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the plurality of template beat signals and the first cardiac signal.

[0017] Embodiment 5. A medical device system according to any one of embodiments 1 to 4, wherein the memory is further configured to store a plurality of template beat signals, wherein each template beat signal corresponds to one VCS pacing pulse output of a plurality of VCS pacing pulse outputs. The processing circuit system may be further configured to determine a plurality of template difference signals by determining a template difference signal from each template beat signal of the plurality of template beat signals and the first cardiac signal. The processing circuit system may input the plurality of template difference signals and the first cardiac signal into the pacing capture classification machine learning model, and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the plurality of template difference signals and the first cardiac signal.

[0018] Example 6. A medical device according to any one of Examples 4 to 5, wherein the memory is further configured to store the multiple template beat signals by storing a first template beat signal corresponding to a first VCS pacing pulse output, a second template beat signal corresponding to a second VCS pacing pulse output that is smaller than the first VCS pacing pulse output, and a third template beat signal corresponding to a third VCS pacing pulse output between the first VCS pacing pulse output and the second VCS pacing pulse output.

[0019] Example 7. A medical device system according to any one of Examples 1 to 6, wherein the processing circuit system can be further configured to determine a derivative signal from the first cardiac signal; input the first cardiac signal into the pacing capture classification machine learning model by inputting at least the derivative signal; and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the derivative signal.

[0020] Embodiment 8. The medical device system of any one of Embodiments 1 to 7, wherein the memory is further configured to store a second cardiac signal sensed after delivery of the VCS pacing pulse, wherein the first cardiac signal is sensed by a first sensing electrode vector and the second cardiac signal is sensed by a second sensing electrode vector different from the first sensing electrode vector. The processing circuit system may be further configured to input each of the first cardiac signal and the second cardiac signal into a pacing capture classification machine learning model, and determine a capture type classification based on the pacing capture classification machine learning model applied to at least the first cardiac signal and the second cardiac signal.

[0021] Embodiment 9. The medical device system of any one of embodiments 1 to 8, wherein the memory is further configured to store a pacing pulse output of the delivered VCS pacing pulse. The processing circuit system may be further configured to input at least the pacing pulse output and the first cardiac signal to the pacing capture classification machine learning model, and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the pacing pulse output and the first cardiac signal.

[0022] Example 10. A medical device system according to any one of Examples 1 to 9, wherein the processing circuit system is further configured to determine a characteristic of the first cardiac signal, input the characteristic of the first cardiac signal and the first cardiac signal into the pacing capture classification machine learning model, and determine the capture type classification based on the characteristic of the first cardiac signal and the pacing capture classification machine learning model applied to at least the first cardiac signal.

[0023] Embodiment 11. A medical device system according to any one of embodiments 1 to 10, wherein the processing circuit system is further configured to receive a training cardiac signal data set obtained from a plurality of patients, the training cardiac signal data set comprising a plurality of training cardiac signals, each training cardiac signal being sensed after delivery of a VCS pacing pulse, wherein the VCS pacing pulses associated with the plurality of training cardiac signals comprise VCS pacing pulses delivered at a plurality of different pacing pulse outputs. The processing circuit system may be configured to train the pacing capture classification machine learning model with the training cardiac signal data set according to a machine learning algorithm, and apply the pacing capture classification machine learning model trained with the training cardiac signal data set to at least the first cardiac signal.

[0024] Embodiment 12. A medical device system according to any one of Embodiments 1 to 11, wherein the processing circuit system is further configured to receive a plurality of cardiac signals including the first cardiac signal, each of the plurality of cardiac signals being associated with a VCS pacing pulse, wherein the VCS pacing pulse associated with the plurality of cardiac signals comprises VCS pacing pulses delivered with a plurality of different pacing pulse outputs. The processing circuit system may be further configured to input each of the plurality of cardiac signals into the pacing capture classification machine learning model, determine a capture type classification of each of the VCS pacing pulses delivered with the plurality of different pacing pulse outputs associated with the plurality of cardiac signals based on the pacing capture classification machine learning model, and determine a capture threshold of at least one of the plurality of capture types based on the capture type classification.

[0025] Embodiment 13 The medical device system of Embodiment 12, wherein the user interface is further configured to present the capture threshold of the at least one capture type.

[0026] Example 14. The medical device system of any one of Examples 12 to 13, wherein the processing circuit system is further configured to determine an operational pacing pulse output based on the capture threshold determined for the at least one capture type.

[0027] Example 15. The medical device system of Example 14, further comprising a therapy delivery circuit configured to generate a VCS pacing pulse based on the operational pacing pulse output.

[0028] Example 16. A medical device system according to any one of Examples 1 to 15, wherein the processing circuit system is further configured to classify the capture type as one of: selective His-Purkinje capture without ventricular myocardial capture, non-selective His-Purkinje capture with ventricular myocardial capture, ventricular myocardial capture without His-Purkinje capture, or loss of capture.

[0029] Embodiment 17. A method, the method comprising storing in a memory a first cardiac signal sensed after delivery of a ventricular conduction system (VCS) pacing pulse, and receiving, by a processing circuit system, the first cardiac signal sensed after delivery of the VCS pacing pulse. The method may further comprise applying a pacing capture classification machine learning model to at least the first cardiac signal, and determining a capture type classification of the VCS pacing pulse from a plurality of capture types based on the applied pacing capture classification machine learning model. The method may comprise generating, by the processing circuit system, an output based on the capture type classification, and presenting, by a user interface, a representation of the capture type classification associated with the delivered VCS pacing pulse in response to the generated output.

[0030] Example 18. According to the method described in Example 17, the method further includes storing a first template beat signal corresponding to the first VCS pacing pulse output in the memory, inputting at least the first template beat signal and the first cardiac signal into the pacing capture classification machine learning model, and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the first template beat signal and the first cardiac signal.

[0031] Example 19. A method according to any one of Examples 17 to 18, further comprising storing a first template beat signal corresponding to a first VCS pacing pulse output in the memory; determining a first template difference signal from the first cardiac signal and the first template beat signal by a processing circuit system; inputting at least the first template difference signal and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the first template difference signal and the first cardiac signal.

[0032] Embodiment 20. The method of any one of Embodiments 17 to 19, further comprising storing a plurality of template beat signals in the memory, wherein each template beat signal corresponds to a VCS pacing pulse output of a plurality of VCS pacing pulse outputs. The method may further comprise inputting each of the plurality of template beat signals and the first cardiac signal into a pacing capture classification machine learning model, and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the plurality of template beat signals and the first cardiac signal.

[0033] Embodiment 21. The method of any one of Embodiments 17 to 20, further comprising storing a plurality of template beat signals in the memory, wherein each template beat signal corresponds to one of a plurality of VCS pacing pulse outputs. The method may further comprise determining a plurality of template difference signals by determining a template difference signal from each of the plurality of template beat signals and the first cardiac signal; inputting the plurality of template difference signals and the first cardiac signal into a pacing capture classification machine learning model; and determining a capture type classification based on the pacing capture classification machine learning model applied to at least the plurality of template difference signals and the first cardiac signal.

[0034] Example 22. According to the method described in any one of Examples 20 to 21, the method further includes storing multiple template beat signals by storing a first template beat signal corresponding to a first VCS pacing pulse output, a second template beat signal corresponding to a second VCS pacing pulse output that is smaller than the first VCS pacing pulse output, and a third template beat signal corresponding to a third VCS pacing pulse output between the first VCS pacing pulse output and the second VCS pacing pulse output.

[0035] Example 23. A method according to any one of Examples 17 to 22, further comprising determining a derivative signal from the first cardiac signal; inputting the first cardiac signal into the pacing capture classification machine learning model by inputting at least the derivative signal; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the derivative signal.

[0036] Embodiment 24. The method of any one of embodiments 17 to 23, further comprising storing a second cardiac signal sensed after delivery of the ventricular conduction system (VCS) pacing pulse, wherein the first cardiac signal is sensed by a first sensing electrode vector and the second cardiac signal is sensed by a second sensing electrode vector different from the first sensing electrode vector. The method may further comprise inputting each of the first cardiac signal and the second cardiac signal into the pacing capture classification machine learning model, and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the first cardiac signal and the second cardiac signal.

[0037] Example 25. A method according to any one of Examples 17 to 24, further comprising storing the pacing pulse output of the delivered VCS pacing pulse in the memory; inputting at least the pacing pulse output and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the pacing pulse output and the first cardiac signal.

[0038] Example 26. According to the method described in any one of Examples 17 to 25, the method further includes determining a characteristic of the first cardiac signal; inputting the characteristic of the first cardiac signal and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the characteristic of the first cardiac signal and the pacing capture classification machine learning model applied to at least the first cardiac signal.

[0039] Embodiment 27. A method according to any one of embodiments 17 to 26, further comprising receiving, by the processing circuit system, a training cardiac signal data set obtained from a plurality of patients. The training cardiac signal data set may include a plurality of training cardiac signals, each training cardiac signal being sensed after delivery of a VCS pacing pulse, wherein the VCS pacing pulse associated with the plurality of training cardiac signals includes VCS pacing pulses delivered with a plurality of different pacing pulse outputs. The method may further comprise training the pacing capture classification machine learning model with the training cardiac signal data set according to a machine learning algorithm, and applying the pacing capture classification machine learning model trained with the training cardiac signal data set to at least the first cardiac signal.

[0040] Embodiment 28. The method of any one of Embodiments 17 to 27, further comprising receiving, by the processing circuit system, a plurality of cardiac signals including the first cardiac signal, each of the plurality of cardiac signals being associated with a VCS pacing pulse, wherein the VCS pacing pulses associated with the plurality of cardiac signals include VCS pacing pulses delivered with a plurality of different pacing pulse outputs. The method may further comprise inputting each of the plurality of cardiac signals into the pacing capture classification machine learning model; determining, based on the pacing capture classification machine learning model, a capture type classification for each of the VCS pacing pulses delivered with the plurality of different pacing pulse outputs associated with the plurality of cardiac signals; and determining, based on the capture type classification, a capture threshold for at least one of the plurality of capture types.

[0041] Embodiment 29. The method according to Embodiment 28 further comprises presenting, by the user interface, the capture threshold of the at least one capture type.

[0042] Example 30. A method according to any one of Examples 28 to 29, further comprising determining an operational pacing pulse output based on the capture threshold determined for the at least one capture type.

[0043] Example 31. A method according to Example 30, further comprising generating a VCS pacing pulse based on the operating pacing pulse output.

[0044] Example 32. A method according to any one of Examples 17 to 31, further comprising classifying the capture type as one of: selective His-Purkinje capture without ventricular myocardial capture, non-selective His-Purkinje capture with ventricular myocardial capture, ventricular myocardial capture without His-Purkinje capture, or loss of capture.

[0045] The details of one or more aspects of the present disclosure are set forth in the following drawings and description. Other features, objectives, and advantages of the techniques described in the present disclosure will be apparent from the description and drawings, and from the claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1A is a conceptual diagram of a medical device system including an implantable medical device (IMD) capable of pacing and sensing in a patient's heart.

[0047] Figure 1B is a conceptual diagram of an IMD connected to a pacing lead that is advanced to an alternate location within the heart to deliver VCS pacing pulses and sense cardiac electrical signals.

[0048] Figure 1C is a conceptual diagram of an IMD connected to a pacing lead that is advanced to another implantation location within the heart for delivery of VCS pacing pulses according to another example.

[0049] Figure 2A is a conceptual illustration of a leadless pacemaker positioned within the RA for providing ventricular pacing via the His-Purkinje system.

[0050] Figure 2B yes Figure 2A Conceptual diagram of a leadless pacemaker shown implanted in an alternative location for pacing the VCS.

[0051] Figure 3 is a schematic diagram of circuitry that may be packaged within an IMD configured to sense cardiac electrical signals and to perform VCS pacing.

[0052] Figure 4is a conceptual diagram of a processing circuit configured to analyze a cardiac signal to determine a pacing capture classification of the cardiac signal and an associated VCS pacing pulse, according to some examples.

[0053] Figure 5 is a flow chart of a method performed by processing circuitry of a medical device system in some examples for classifying VCS pacing pulses and corresponding post-paced QRS waveforms according to capture type.

[0054] Fig. 6A and Figure 6B is a diagram of a PCC configured to receive multiple unknown signal inputs for each VCS pacing pulse to be classified according to some examples.

[0055] Figure 7 is a flow chart of a method that may be performed by processing circuitry of a medical device system to classify VCS pacing pulses according to another example.

[0056] Figure 8 is a conceptual diagram of a processing circuit configured to analyze a cardiac signal to determine a pace capture classification according to another example.

[0057] Fig. 9 is a conceptual diagram of a processing circuit configured to analyze a cardiac signal to determine a pace capture classification according to another example.

[0058] Fig.10 is a flow chart of a method for determining VCS pacing capture type according to another example.

[0059] Fig.11 is a conceptual diagram of a processing circuit configured to analyze a cardiac signal to determine a pace capture classification according to yet another example.

[0060] Fig.12 is a conceptual diagram of a processing circuit configured to analyze a cardiac signal to determine a pace capture classification according to another example.

[0061] Fig.13 is a flow chart of a method for generating a VCS pacing pulse capture type classification during a pacing capture threshold test according to some examples.

[0062] Fig.14 is a conceptual diagram of a graphical user interface displaying cardiac electrical signals and labeled channels annotated by capture type classification output by a PCC, according to one example.

[0063] Fig.15 is a flow chart of a VCS capture monitoring method that may be performed by processing circuitry of a medical device system, according to some examples. DETAILED DESCRIPTION

[0064] In general, the present disclosure relates to devices, systems and methods for classifying the type of cardiac capture achieved by pacing pulses delivered to or near the His-Purkinje system. An example of a medical device system for receiving a cardiac electrical signal by a cardiac signal analyzer is disclosed herein, and the cardiac signal analyzer processes the cardiac electrical signal to classify the signal according to the capture type. The cardiac signal analyzer can be implemented as a processing circuit system that is configured to apply an artificial intelligence or machine learning model to the cardiac electrical signal to classify the capture type associated with the cardiac pacing pulse delivered to the His-Purkinje conduction system of the patient's heart. The cardiac signal analyzer can be a machine learning model that is implemented in the processing circuit system of the medical device system and trained using cardiac electrical signals obtained from a patient population during His-Purkinje system pacing. The established artificial intelligence or machine learning model (e.g., a neural network model) of the trained cardiac signal analyzer receives at least one cardiac electrical signal sensed after a pacing pulse delivered to or near a portion of the His-Purkinje system of the patient's heart, and outputs a capture type classification of the delivered pacing pulse. The cardiac signal analyzer is also referred to herein as a "pacing capture classifier" (PCC) because the cardiac signal analyzer can be implemented as a processing circuit system that is programmed or trained to analyze and classify cardiac signals based on the type of capture achieved by the delivered VCS pacing pulses.

[0065] For example, according to any of the examples described herein, the medical device may be configured to generate pacing pulses for delivery to the His-Purkinje conduction system of the patient's heart. As used herein, the term "ventricular conduction system" (VCS), which may be used to refer to "VCS pacing," "VCS pacing pulses," "VCS capture," and the like, may collectively refer to any portion of the His-Purkinje conduction system, including the His bundle, the right and left Purkinje branches, and the Purkinje fibers, such that "VCS pacing" may generally refer to pacing anywhere along the His-Purkinje conduction system, and "His-Purkinje pacing pulses" may refer to pacing pulses delivered anywhere along the His-Purkinje conduction system. "VCS capture" refers to capture of any portion of the His-Purkinje conduction system, which may be capture at or below the His bundle, such as along the left bundle branch and / or the right bundle branch, and is also referred to herein as "His-Purkinje system capture."

[0066] Cardiac tissue is "captured" by a pacing pulse having sufficient electrical energy to cause depolarization of cardiac tissue at the pacing site, thereby causing an electrical "induced response". The pacing-induced depolarization at the pacing site is then carried out by the native conduction system of the heart and / or the myocardial tissue, resulting in mechanical contraction of the cardiac chamber. In order to effectively capture and pace the heart to achieve the desired therapeutic effect, the cardiac pacing pulse needs to have a pulse energy equal to or greater than the capture threshold of the cardiac tissue at the pacing site. A pacing capture threshold test can be performed to determine the minimum pacing pulse amplitude for a given pacing pulse width (or vice versa) that captures the cardiac chamber. Determination of the capture threshold enables appropriate programming of the pacing pulse amplitude and pulse width (collectively referred to herein as "pacing pulse output") to promote effective pacing and avoid loss of capture. When a loss of capture or a change in capture type is detected, capture monitoring performed by a medical device configured to deliver VCS pacing can be performed at least periodically to allow the pacing pulse amplitude and / or pulse width to be adjusted to suprathreshold (greater than the capture threshold).

[0067] When pacing pulses are delivered by electrodes positioned in the heart to pace the VCS, it may be possible to capture only a portion of the His-Purkinje system, such as complete or partial capture of the His bundle or complete or partial capture of one or both of the right and left bundle branches, without capturing ventricular myocardial tissue. In other cases, the VCS pacing pulses may capture both a portion of the His-Purkinje system and the surrounding ventricular myocardium, and at other times, the VCS pacing pulses may capture the surrounding ventricular myocardium without capturing any portion of the His-Purkinje system.

[0068] Capture of only the His-Purkinje system without ventricular myocardial capture is referred to herein as "selective" His-Purkinje system (SHP) capture. Capture of the His-Purkinje system and at least a portion of the surrounding ventricular myocardial tissue is referred to herein as "non-selective" His-Purkinje system (NSHP) capture. Capture of the surrounding ventricular myocardium without capture of the His-Purkinje system may be referred to as ventricular myocardial only (VMO) capture. Loss of capture (LOC) occurs when the pacing pulse energy is below both the His-Purkinje system capture threshold and the VM capture threshold. The VCS pacing pulse does not cause depolarization of cardiac cells and no induced response occurs.

[0069] As disclosed herein, a PCC that may include a machine learning model may receive multiple channels of cardiac signal data, including at least one cardiac electrical signal sensed after delivery of a VCS pacing pulse of unknown capture type. The PCC is trained to output a capture type classification for the delivered VCS pacing pulse. The output capture type classification may be used for a variety of monitoring and / or pacing control functions. For example, the output capture type may be used to annotate a display of VCS pacing pulse markers and sensed cardiac signals. The output capture type may be used to determine a pacing capture threshold for a given capture type and / or select a pacing pulse amplitude and / or pulse width for delivering VCS pacing to a patient to facilitate a high likelihood of achieving a desired capture type (e.g., SHP capture or NSHP capture).

[0070] Figure 1A 1 is a conceptual diagram of a medical device system 10 capable of pacing and sensing in a patient's heart 8. System 10 includes an implantable medical device (IMD) 14 coupled to the patient's heart 8 via transvenous electrical leads 16, 17, and 18. IMD 14 is shown as a dual-chamber device capable of delivering cardiac pacing pulses and sensing cardiac electrical signals from the right atrium (RA) and from the ventricular chambers. Housing 15 encloses a housing 15 that is combined with, for example, the following Figure 3 The various circuits and components described herein correspond to internal circuit systems for sensing cardiac signals from heart 8 and delivering cardiac pacing therapy. In some examples, IMD 14 is capable of detecting tachyarrhythmias from the sensed cardiac signals and delivering a high-voltage cardioversion / defibrillation (CV / DF) shock to the patient's heart 8, e.g., to terminate a detected ventricular tachycardia or ventricular fibrillation.

[0071] IMD 14 includes a connector block 12 that can be configured to receive the proximal ends of RA lead 16 and VCS pacing lead 18, which are both advanced transvenously for positioning electrodes for sensing cardiac electrical signals and delivering pacing therapy. RA lead 16 is positioned with its distal end near the right atrium. RA lead 16 carries pacing electrode 20 and sensing electrode 22, both of which are shown as tip electrode 20 and ring electrode 22 spaced proximally from tip electrode 20. Electrodes 20 and 22 provide sensing and pacing in the RA and are each connected to a corresponding insulated conductor extending within the elongated body of RA lead 16. Each insulated conductor is coupled at its proximal end to a connector carried by proximal lead connector 40. Connector block 12 is configured to receive lead connector 40, which is used to electrically couple conductors extending from distal electrodes 20 and 22 to circuitry within housing 15 via electrical feedthroughs passing through housing 15.

[0072] As shown, from a right atrial approach, the VCS lead 18 can be advanced within the right atrium to position electrodes 32 and 34 for pacing and sensing near the VCS (e.g., at or near the His bundle). The VCS tip electrode 32 can be a spiral electrode that is advanced to the lower end of the atrial septum, below the AV node, and near the tricuspid annulus to position the tip electrode 32 in, along, or near the His bundle. A ring electrode 34 spaced proximally from the tip electrode 32 can be used as a return electrode, together with the cathode tip electrode 32, for pacing the right and left ventricles via the native His-Purkinje conduction system.

[0073] An intracardiac electrogram (EGM) signal may be generated by cardiac electrical signal sensing circuitry included in IMD 14 from cardiac electrical signals received via a sensing electrode vector that may include tip electrode 32 and ring electrode 34 of VCS lead 18. Electrodes 32 and 34 are coupled to respective insulated conductors extending within the elongated body of VCS lead 18, which provide electrical connection to a proximal lead connector 44 that is coupled to the internal IMD circuitry via connector block 12.

[0074] Housing 15 may act as a return electrode for use with any of the electrodes carried by leads 16 and 18 in a unipolar sensing or pacing configuration. Electrodes 32 and 34 may be used in a bipolar pacing electrode pair for delivering VCS pacing pulses and for receiving cardiac electrical signals to sense intrinsic and pacing-induced QRS waveforms. In some examples, IMD 14 may be configured to sense far-field (FF) cardiac signals, for example using electrodes 32 and housing 15 or using electrodes 34 and housing 15, and / or to sense near-field (NF) cardiac signals, for example using electrodes 32 and 34, for processing and analysis to determine capture type. Electrodes 32 and 34 may be used in a sensing electrode vector for sensing intrinsic R waves for determining cardiac rhythm and the need for electrical stimulation therapy.

[0075] In addition to pacing and sensing capabilities, in some examples, the IMD 14 may also be capable of delivering high voltage cardioversion or defibrillation (CV / DF) shocks. In this case, the IMD 14 may be coupled to at least one lead carrying one or more defibrillation electrodes, which may be elongated coil electrodes for delivering high voltage CV / DF shocks. In the example shown, a third lead 17 is shown coupled to the IMD 14, with its distal end advanced into the right ventricle (RV). The RV lead 17 may include a coil electrode 24 for delivering CV / DF shocks, for example in combination with the IMD housing 15. The IMD housing 15 may be used as an active electrode together with the coil electrode 24 during CV / DF shock delivery. The RV lead 17 (when present) may carry a distal tip electrode 28 and a ring electrode 30 to deliver ventricular pacing pulses in the RV and / or sense cardiac electrical signals from the RV. Electrodes 28, 30, and 24 are each connected to a corresponding insulated conductor extending within the elongated body of the RV lead 17. Each insulated conductor is coupled at its proximal end to a connector carried by proximal lead connector 42. Connector block 12 is configured to receive lead connector 42 for electrically coupling conductors extending from distal electrodes 24, 28, and 30 to circuitry within housing 15 via electrical feedthroughs passing through housing 15. In other examples, RC lead 17 is optional. When IMD 14 is implemented as an implantable cardioverter-defibrillator (ICD) having CV / DF shock capability in addition to cardiac pacing and sensing functions, one or more CV / DF electrodes may be carried by RA lead 16 and / or VCS lead 18.

[0076] It should be understood that although the IMD 14 is described as a dual-chamber pacemaker with or without CV / DF shock capability (which is capable of sensing and pacing the RA via VCS and sensing ventricular signals and pacing the ventricular chambers), the IMD 14 can be a single-chamber pacing device with single-chamber or dual-chamber sensing. For example, the IMD 14 can be coupled only to the VCS lead 18 to sense cardiac electrical signals and deliver VCS pacing pulses to maintain at least a minimum ventricular rate. When the lead 18 is positioned for delivering VCS pacing pulses, the VCS lead 18 can carry an additional sensing electrode positioned within the RA to sense the RA EGM signal, enabling the IMD 14 to perform dual-chamber (atrial P wave and ventricular R wave) sensing and VCS pacing, which can be atrial synchronized ventricular pacing delivered via the His-Purkinje system.

[0077] External device 50 is shown in telemetric communication with IMD 14 via communication link 60. External device 50 may be embodied as a programmer used in a hospital, clinic, or physician's office to retrieve data from IMD 14 and program operating parameters and algorithms in IMD 14 to control IMD functions. External device 50 may alternatively be embodied as a home monitor or handheld device for retrieving data from IMD 14. External device 50 may be used to program cardiac signal sensing parameters, cardiac rhythm detection parameters, cardiac pacing, and CV / DF therapy control parameters used by IMD 14.

[0078] The external device 50 may include a processor 52, a memory 53, a display unit 54, a user interface 56, and a telemetry unit 58. The processor 52 is configured to control the operation of the external device and receive patient data and IMD data for analysis and / or for generating a visual representation of the data to a clinician or other user. The processor 52 includes a processing circuit, shown as PCC 51, for classifying VCS pacing pulses according to capture types based on analysis of received cardiac electrical signals, as further described below. The processor 52 may process data and signals received from the IMD 14 via the telemetry unit 58 during a telemetry session. The processor 52 may be configured to control the telemetry unit 58 to transmit user-entered programming commands to the IMD 14 and process data received from the IMD 14 for display by the display unit 54. In some examples, the processor 52 may be configured to control the telemetry unit 58 to transmit automatically generated programming commands or other information to the IMD 14, which are generated by the processor 52 based on the pacing capture classification determined by the PCC 51.

[0079] Display unit 54, which may include a graphical user interface (GUI), displays data and other information to the user for viewing IMD operation and programmed parameters, and may display programmable parameters to the user for selection and programming of IMD 14. Display unit 54 may generate a display of EGM signals received from IMD 14 and / or data derived therefrom. Display unit 54 may be configured to display a GUI including various windows, icons, user-selectable menus, etc. to facilitate user interaction with external device 50 and IMD 14. Display unit 54 may be used as an input and / or output device using technology including a liquid crystal display (LCD), a quantum dot display, a dot matrix display, a light emitting diode (LED) display, an organic light emitting diode (OLED) display, a cathode ray tube display, electronic ink, or a monochrome, color, or any other type of display capable of generating tactile, audio, and / or visual output. In some examples, display unit 54 is a presence-sensitive display that may be used as a user interface device, the presence-sensitive display operating as one or more input devices and one or more output devices.

[0080] As described below, display unit 54 can be configured to present a representation of the capture type classification made by PCC 51 associated with a VCS pacing pulse. Such a representation reduces the burden on a user or clinician to interpret cardiac electrical signals during VCS pacing and simplifies programming of IMD 14 in a patient-specific manner, thereby facilitating effective VCS pacing. Determining the capture type of a VCS pacing pulse based on observation of cardiac signals can be challenging and require significant user expertise.

[0081] Therefore, the technology for classifying VCS pacing pulses according to capture type described herein provides specific improvements in the computer-related field of medical device programming and reporting of medical device related information and data with practical applications. For example, the use of the technology herein can enable an external device 50 to generate a visualization of cardiac electrical signals and / or VCS pacing pulse data annotated according to the capture type classification of the delivered VCS pacing pulse. Such visualization can more accurately inform the clinician or user about how the expected IMD 14 performs in delivering VCS pacing pulses to effectively capture the VCS of the patient's heart. The capture type annotation or other related reports of the VCS pacing pulse capture type classification can reduce the possibility of human error in programming IMD operating parameters (such as VCS pacing pulse output). In addition, the technology disclosed herein can reduce the complexity of programming IMD 14. Therefore, the technology disclosed herein can enable a medical device (such as IMD14) to be programmed to deliver VCS pacing, which can achieve the desired type of cardiac capture in a simplified, flexible and patient-specific manner.

[0082] User interface 56 may include a mouse, touch screen, keyboard, etc. to enable a user to interact with external programmer 50 to initiate a telemetry session with IMD 14 to retrieve data from and / or transmit data to IMD 14. Telemetry unit 58 includes a transceiver and antenna configured for bidirectional communication with telemetry circuitry included in IMD 14, and is configured to operate in conjunction with processor 52 for transmitting and receiving data related to IMD functionality via communication link 60, which may include data related to VCS pacing and capture type determinations.

[0083] You can use A wireless radio frequency (RF) link of Wi-Fi or Medical Implant Communications Service (MICS) or other RF or communications frequency bandwidth or communications protocol establishes a communications link 60 between the IMD 14 and the external device 50. Data stored or acquired by the IMD 14 (including EGM signals or associated data derived therefrom, results of device diagnostics, and a history of detected cardiac rhythm episodes), delivered therapy, or other data may be retrieved from the IMD 14 by the external device 50 following an interrogation command.

[0084] External device 50 may include external interface 55 including ports for electrical connection to surface ECG leads and electrodes 57a and 57b that may be coupled to a patient implanted with IMD 14. Although not shown for clarity, Figure 1A Only two ECG electrodes 57a and 57b are illustrated in the figure, but it should be understood that multiple surface ECG electrodes can be positioned on the patient as needed to record one, two, three or other desired number of ECG signals, and these surface ECG electrodes can include any of the limb leads I, II, III, aVR, aVL, aVF and / or any of the precordial leads V1, V2, V3, V4, V5 and V6. The processor 52 can receive the ECG signal for display by the display unit 54 for the user to observe during VCS pacing. The ECG signal can be processed and analyzed by the processor 52 (which can be performed in combination with the analysis of one or more EGM signals received from the IMD 14) to predict and classify the VCS pacing pulse capture type. The visual representation of the ECG signal by the display unit 54 may enable the user to confirm the patient's intrinsic conduction abnormalities (when VCS pacing is not being delivered) and to observe improvements in the ECG signal during VCS pacing, which may be annotated based on the capture classification output by the PCC 51. The capture type classification annotation may enable the user to more easily recognize and identify improvements in the ECG signal with VCS pacing, such as narrowed QRS signals, reduced activation time between QRS waveforms induced in the RV and left ventricle (LV) indicating more synchronous depolarization of the left and right ventricles, or the disappearance of QRS abnormalities such as QRS features indicative of left bundle branch block, right bundle branch block, or other ventricular conduction abnormalities.

[0085] The PCC 51 may be implemented in a processing circuit system of the medical device system 10, such as in a processor 52 of an external device 50. The PCC 51 may be trained using a machine learning algorithm to receive at least one cardiac signal input to be classified according to a capture type. The cardiac signal input may be an EGM signal input transmitted from the IMD 14. The EGM signal input may be sensed by the IMD 14 during a post-pacing time interval after a VCS pacing pulse is delivered by the IMD 14. The PCC 51 may receive multiple inputs (e.g., combined into multi-channel data points that may be aligned in time relative to the delivered VCS pacing pulse) for classifying post-pacing cardiac signal segments and associated VCS pacing pulses according to a pacing capture type (e.g., SHP capture, NSHP capture, VMO capture, or LOC). In some examples, additional capture types that may be classified by PCC 51 may include fusion beats, partial His bundle capture, complete His bundle capture, selective left bundle branch (LBB) capture without myocardial capture, non-selective LBB capture with myocardial capture, right bundle branch (RBB) capture without LBB capture, LBB capture with RBB block, RBB capture without LBB capture, RBB capture with LBB block, bilateral bundle branch capture (selective and / or non-selective), selective RBB capture without myocardial capture, non-selective RBB capture with myocardial capture, retrograde atrial capture, cathodal capture without anodal capture, cathodal and anodal capture, or an unknown capture type, but is not intended to be limiting.

[0086] For ease of illustration, the examples presented herein refer to the classification of SHP capture, NSHP capture, VMO capture, and LOC as four possible capture type classifications that may be output by the PCC 51. It should be understood that the PCC 51 may be trained to classify fewer than or more than four capture types, or to classify combinations of possible capture types that are different than the four capture types listed above. The PCC 51 may be trained to classify capture types using the techniques disclosed herein, which may include any of the examples of capture type classifications listed above or any other type of capture or beat that may occur following a VCS pacing pulse. The PCC 51 may apply artificial intelligence (AI) techniques to analyze the inputs in order to determine the capture type. The AI ​​technology may include deep learning technology, such as convolutional neural networks (CNN), residual CNN, feed-forward neural networks (FFNN), recurrent neural networks (RNN), transformers, or other machine learning technologies, such as decision trees, random forest models, or other machine learning methods for establishing a machine learning model for classifying the delivered VCS pacing pulses according to capture type based on post-pacing unknown cardiac signal inputs obtained during a post-pacing time interval after the delivered VCS pacing pulses.

[0087] The capture classification model implemented in the PCC 51 can learn from cardiac electrical signal data obtained from a patient population using machine learning. In some examples, the PCC 51 is trained to classify a post-pacing EGM waveform received from the IMD 14 as an unknown EGM signal input by executing a capture classification algorithm using at least one other template beat input representing a QRS waveform following a VCS pacing pulse with a preselected pacing pulse output. The template beat input can be a patient-specific template generated from one or more QRS waveforms acquired after a VCS pacing pulse delivered at a specified pacing pulse amplitude. Various examples of input signals for which the PCC 51 can be trained and which can subsequently be received for use in generating pacing capture classification for unknown post-pacing signal inputs are described below.

[0088] External device 50 may be implemented as a plurality of computing systems configured to receive cardiac electrical signals, which may include at least one EGM signal input received directly or indirectly from IMD 14. External device 50 may be a personal computer, a medical device programmer, a home monitor, a wearable patient monitor, a mobile device such as a smart phone, a laptop computer, a tablet computer, a personal digital assistant, etc. In some examples, external device 50 is a computing device in a remote patient monitoring system, such as the CARELINK® available from Medtronic, Inc., Dublin, Ireland. TM A computing device is included in the monitoring system.

[0089] Figure 1B 1 is a conceptual diagram of an IMD 14 coupled to a VCS lead 18 that is advanced to an alternative location within heart 8 for delivery of VCS pacing pulses and sensing of cardiac electrical signals. In this example, a distal portion of VCS lead 18 is advanced within the right ventricle (RV) from a right ventricular approach for sensing cardiac electrical signals and delivering VCS pacing pulses at or near the His bundle. IMD 14 may be a single chamber device coupled only to VCS lead 18. In other examples, IMD 14 may be a dual chamber device and coupled to RA lead 16 ( Figure 1A ) and VCS lead 18 to enable sensing of atrial P waves and delivery of atrial pacing pulses in AV synchronous pacing mode and delivery of VCS pacing pulses at an AV delay after an atrial event (sensed or paced).

[0090] In this example, the tip electrode 32 of the VCS lead 18 is placed in the ventricular septum 19, for example, high along the ventricular septum near the lower portion of the His bundle. The tip electrode 32 can be paired with a return anode ring electrode 34 for delivering VCS pacing pulses and for sensing raw cardiac electrical signals, which can be processed to obtain NF EGM signals. In some examples, the post-pacing NF EGM signal can be provided as an input to the PCC 51 for classification of capture type using the techniques disclosed herein. The tip electrode 32 or the ring electrode 34 can be paired with the IMD housing 15 for receiving raw cardiac electrical signals, which are processed to obtain FF EGM signals, which can be received as input by the PCC 51 for classification of capture type using the techniques disclosed herein.

[0091] In some examples, VCS pacing can be delivered in combination with LV myocardial pacing, which can be delivered via LV lead 46 for further improving mechanical synchronization of the RV and LV, for example, during cardiac resynchronization therapy (CRT). LV lead 46 can be advanced into the RA, through the coronary sinus ostium and into the cardiac veins of the LV to position electrodes 48a, 48b, 48c, and 48d (collectively referred to as "LV electrodes 48") along the LV myocardium to sense ventricular electrical signals and pace the LV myocardium. LV lead 46 is shown as a quadripolar lead carrying four electrodes 48a to 48d, which can be selected from a variety of bipolar pacing electrodes for pacing the LV myocardial tissue and sensing LV signals. One of the LV electrodes 48 can be selected in combination with the pacemaker housing 15 for delivering unipolar LV myocardial pacing in some circumstances and / or for sensing raw cardiac electrical signals which can be processed and analyzed as FF EGM signals for input to the PCC 51 for classification of VCS pacing capture type in accordance with the techniques disclosed herein.

[0092] When the VCS lead 18 is positioned for delivering His bundle or bundle branch pacing of one or both bundle branches, VCS pacing can be combined with ventricular myocardial pacing of the LV (using the LV lead 46) to correct LV conduction delays and achieve electrical and mechanical synchronization of the ventricles. Thus, in some examples, the IMD 14 can control the combination of VCS pacing pulse delivery with LV myocardial pacing pulse delivery at a specific time interval, which can include an AV delay and / or a VV delay. The AV delay can control the timing of the VCS pulse and / or the LV myocardial pacing pulse relative to an atrial event (e.g., a sensed P wave or a delivered atrial pacing pulse). In some examples, the VV delay can control the timing between the VCS pacing pulse delivered via the VCS lead 18 and the LV myocardial pacing pulse delivered via the LV lead 46.

[0093] It should be understood that LV lead 46 is optional. In some examples, IMD 14 is coupled only to VCS lead 18 advanced into the RA or RV for delivering VCS pacing and sensing ventricular EGM signals. In other examples, Figure 1A RA lead 16 shown in the figure is implanted in combination with a VCS lead 18 advanced into the RA or RV for delivery of VCS pacing in a dual chamber sensing and pacing system. External device 50 can receive one or more EGM signals sensed using any available ventricular EGM sensing electrode vectors from IMD 14 for processing and analysis according to the techniques disclosed herein to classify the VCS capture type and display one or more cardiac electrical signals and / or VCS pacing pulse markers annotated according to the determined VCS capture type on external device display unit 54.

[0094] Figure 1C 1 is a conceptual diagram of a medical device system 10 including a VCS lead 18 coupled to an IMD 14 according to another example. VCS lead 18 is shown positioned in an alternative implantation site for pacing the VCS. Figure 1C , IMD 14 is a dual chamber device configured to receive RA lead 16, which can be positioned in the right atrial chamber to deliver atrial pacing pulses and sense atrial electrical signals via electrodes 20 and 22. IMD 14 can be configured to sense intrinsic atrial P waves, and deliver atrial pacing pulses in the absence of sensed P waves. IMD 14 can be configured to deliver atrial synchronized ventricular pacing by setting an AV delay in response to each sensed P wave or delivered atrial pacing pulse and deliver a ventricular pacing pulse to the VCS via lead 18 at the end of the AV delay.

[0095] The VCS lead 18 can be advanced transvenously into the RV via the RA for use in situ in a position relatively inferior to (lower than) Figure 1B The tip electrode 32 is positioned in the ventricular septum 19. When the tip electrode 32 is relatively advanced upward in the ventricular septum 19, as shown in FIG. Figure 1B As shown, the tip electrode 32 can be positioned along the lower portion of the His bundle for delivering pacing pulses for complete or partial capture of the His bundle. Figure 1C In the embodiment of the present invention, the tip electrode 32 can be advanced within the ventricular septum 19 near the bundle branches of the His-Purkinje system (e.g., at the LBB pacing site in the LBB region or at the RBB pacing site in the right bundle branch (RBB) region) for delivering pacing pulses to capture one or both bundle branches.

[0096] In various examples, the tip electrode 32 may be selected as a pacing cathode electrode, combined with the ring electrode 34 as a return anode electrode, for pacing and capturing the LBB and / or RBB. In some cases, the pacing pulse amplitude and pulse width may be selected to achieve cathode capture at the cathode electrode for capturing at least one bundle branch. In other cases, the pacing pulse amplitude and pulse width may be selected to achieve cathode and anode capture, which may simultaneously capture both the LBB and the RBB to provide dual or bilateral bundle branch (BB) pacing using a single bipolar electrode pair. In other examples, the tip electrode 32 or the ring electrode 34 may be selected as a cathode electrode paired with the housing 15 in a unipolar pacing electrode vector. Unipolar pacing may capture a single BB. However, in some cases, when a unipolar pacing pulse directly captures one bundle branch, unipolar pacing may capture both the RBB and the LBB, while a virtual current or disconnected excitation generated by the pacing electrode may excite another bundle branch, thereby potentially causing unipolar bilateral BB pacing while capturing both the LBB and the RBB.

[0097] Although the VCS lead 18 is shown as carrying a pair of pacing and sensing electrodes, a tip electrode 32 and a ring electrode 34, it should be understood that in other examples, the VCS lead 18 may include multiple electrodes along its distal portion to provide one or more selectable bipolar pacing electrode vectors and / or one or more unipolar pacing electrode vectors (e.g., in conjunction with the housing 15) for delivering pacing pulses to one or both of the RBB and the LBB.

[0098] In addition, when the IMD 14 is configured as an ICD capable of delivering high voltage shock therapy, the VCS lead 18 may include one or more coil electrodes, such as the coil electrode 35. The coil electrode 35 may be used, for example, in a sensing electrode vector with either the tip electrode 32 or the ring electrode 34 for sensing a ventricular EGM signal that may be analyzed by the PCC 51 for classification of the VCS pacing pulse capture type. Other examples of pacing lead configurations that may be positioned for pacing in the region of the LBB and / or RBB are generally disclosed in U.S. Patent Application No. 17 / 370,303 (Zhou et al.) and U.S. Publication No. 2019 / 0111270 (Zhou).

[0099] As described above, IMD 14 can communicate with external device 50 via wireless telemetry. External device 50 can receive information from IMD 14 using Figure 1C EGM signals sensed by any available electrodes shown in or in other examples described or shown in the accompanying drawings may be processed and analyzed to classify VCS pacing pulses according to capture type and / or displayed by display unit 54 according to the techniques disclosed herein.

[0100] Figure 2A is a conceptual diagram of a leadless pacemaker 100 positioned within the RA for providing ventricular pacing via the His-Purkinje system according to another example. The pacemaker 100 may include a distal tip electrode 102 extending away from a distal end 112 of a pacemaker housing 105. The pacemaker 100 is shown implanted in the RA of a patient's heart 8 to position the distal tip electrode 102 for delivering pacing pulses to the His bundle. For example, the distal tip electrode 102 may be inserted at the lower end of the atrial septum, below the AV node, and near the tricuspid annulus to position the tip electrode 102 in, along, or near the His bundle. The distal tip electrode 102 may be a spiral electrode that provides fixation to anchor the pacemaker 100 at the implantation site. In other examples, the pacemaker 100 may include a fixation member including one or more tines, hooks, barbs, spirals, or other fixation members that anchor the distal end of the pacemaker 100 at the implantation site.

[0101] A portion of the distal tip electrode 102 may be electrically insulated so that only the distal-most end of the tip electrode 102 (farthest from the housing distal end 112) is exposed to provide targeted pacing at a tissue site that includes a portion of the His bundle. One or more housing-based electrodes 104 and 106 may be carried on a surface of the housing of the pacemaker 100. The electrodes 104 and 106 are shown as ring electrodes that surround the longitudinal sidewall 107 of the pacemaker housing 105. The longitudinal sidewall 107 extends from the distal end 112 of the housing 105 to the proximal end 110. In other examples, a return anode electrode for sensing and pacing may be positioned on the housing proximal end 110. Pacing of VCS may be achieved using the distal tip electrode 102 as a cathode electrode and using either of the housing-based electrodes 104 and 106 as a return anode.

[0102] The cardiac electrical signal may be sensed by the pacemaker 100 using a sensing electrode pair selected from electrodes 102, 104, and 106. For example, the cardiac electrical signal may be sensed using the distal tip electrode 102 and the distal housing-based electrode 104. The second cardiac electrical signal may be sensed using electrodes 104 and 106 as a relatively more FF signal. The raw cardiac electrical signal may be processed to generate a NF EGM signal and a relatively more FF EGM signal. The EGM signal may be used to provide an input signal to the PCC 51 for classifying the capture type according to the techniques disclosed herein.

[0103] In some examples, an atrial P wave may be sensed from signals received via electrodes 104 and 106, and / or an atrial pacing pulse may be delivered via electrodes 104 and 106. An atrial synchronized VCS pacing pulse may be delivered via electrodes 102 and 104 to capture at least a portion of the His-Purkinje system at an AV delay after the sensed atrial P wave and / or the delivered atrial pacing pulse.

[0104] The pacemaker 100 is shown as communicating with an external device 50 via a communication link 60. The external device 50 can receive EGM signals sensed by the pacemaker 100, which can be provided as input to the PCC 51 for classifying the capture type of VCS pacing pulses, as further described below. It should be understood, therefore, that the EGM signals obtained for input to the PCC 51 can be sensed by electrodes carried by leads coupled to an IMD or by electrodes carried on a housing of the IMD, which can be a leadless pacemaker configured to deliver VCS pacing or an IMD including housing-based electrodes for sensing cardiac electrical signals, which housing-based electrodes can also be coupled to one or more leads for delivering VCS pacing pulses.

[0105] Figure 2B yes Figure 2A 1 is a conceptual diagram of a leadless pacemaker 100 shown implanted in an alternative location for pacing the VCS. In some examples, the pacemaker 100 can be implanted within the RV along the interventricular septum 19 for providing VCS pacing. The techniques disclosed herein can be used in conjunction with a leadless pacemaker, such as the pacemaker 100, having pacing electrodes coupled to and extending from a pacemaker housing 105, without the need for invasive medical leads coupled to the pacemaker for carrying pacing and sensing electrodes.

[0106] In this example, the pacemaker 100 can be positioned within the RV for advancing a pacing tip electrode 102 extending from a distal end 112 of a pacemaker housing 105 into the ventricular septum 19 for pacing the VCS, for example, in the region of the lower portion of the His bundle or along one or both of the RBB and the LBB (depending on the relative positioning of the tip electrode 102). The tip electrode 102 is shown as a "screw-in" spiral electrode, but can be configured as other types of electrodes capable of being advanced within septal tissue. The proximal portion of the pacing tip electrode 102 can be electrically insulated, such as having a coating in some examples, such that only the distal portion of the tip electrode 102 farthest from the pacemaker housing distal end 112 is exposed to provide targeted pacing at a tissue site including the His bundle, the LBB, and the RBB.

[0107] In other examples, the tip electrode 102 can be formed to have a straight shaft with a distal active electrode portion or other types of electrodes that can be tissue-piercing electrodes that can be advanced through the ventricular septum 19 to deliver pacing in the left portion of the septum 19 in the area of ​​the LBB, for example. In some examples, the pacemaker 100 can include a fixation member that includes one or more tines, hooks, barbs, spirals, or other fixation members that anchor the distal end 112 of the pacemaker 100 at the implantation site and that can not serve as an electrode. Examples of leadless intracardiac pacemakers that can be configured to deliver cardiac pacing pulses to a His-Purkinje system that can be used in conjunction with the techniques described herein are generally disclosed in U.S. Publication No. 2019 / 0111270 (Zhou) and U.S. Publication No. 2019 / 0083800 (Yang et al.).

[0108] The pacemaker 100 may include a distal housing-based ring electrode 104 along or near a distal end 102 of a pacemaker housing 105. The distal housing-based ring electrode 104 may be selected as a return anode electrode for use with the tip electrode 102 for bipolar pacing of the LBB or RBB near the tip electrode 102. Simultaneous bipolar bilateral BB pacing of both the RBB and the LBB may be achieved by cathode capture of the LBB at the tip electrode 102 and anodal capture of the RBB by the distal ring electrode 104. The polarity of the tip electrode 102 and the distal ring electrode 104 may be reversed to achieve cathode capture of the RBB and anodal capture of the LBB in some examples. The distal ring electrode 104 is shown as a ring electrode surrounding the distal portion of the housing 105, but alternatively may be a distal housing-based electrode in the form of a button electrode, a hemispherical electrode, a segmented electrode, etc. and may be along a surface of the distal end 112 of the housing 105 and / or along the longitudinal sidewall 107.

[0109] In the example shown, a housing-based proximal ring electrode 106 that may surround all or part of the longitudinal sidewall 107 of the housing 105 may be provided as a return anode electrode. In other examples, a return anode electrode for sensing and pacing may be positioned on the housing proximal end 110 and may be a button, ring, or other type of electrode. VCS pacing of the LBB may be achieved using the tip electrode 102 as the cathode electrode and the proximal ring electrode 106 as the return anode. Pacing of the myocardial tissue of the RBB and / or the ventricular septum 19 may be achieved using the distal ring electrode 104 as the cathode electrode and the proximal ring electrode 106 as the return anode. In this way, bilateral or dual-bundle branch pacing may be achieved using two different bipolar pacing electrode vectors carried by the housing 105.

[0110] The cardiac electrical signals generated by the heart 8 can be sensed by the pacemaker 100 using electrodes 102, electrodes 104, and / or electrodes 106. The cardiac electrical signals received via electrodes 102 and 104, electrodes 104 and 106, and / or electrodes 102 and 106 can be sensed by the pacemaker 100 and processed by the processing circuit system of the IMD 14, for example, and / or wirelessly transmitted to the external device 50 via the communication link 60, for example, as EGM signals. These signals can then be displayed and / or further processed and analyzed by the processor 52 of the external device 50 to provide a user with a visual representation of the sensed EGM signals. As described herein, the EGM signals sensed by the pacemaker 100 can be provided as input to the PCC 51 for classification of VCS pacing pulses according to pacing capture type, and can be displayed by the external device display unit 54 together with the pacing capture type annotation.

[0111] Figures 1A to 2B The examples of the present invention present various lead and / or electrode configurations that may be implemented for delivering VCS pacing in a system configured to perform the techniques disclosed herein for analyzing sensed cardiac electrical signals to classify VCS pacing pulses according to capture type. The specific lead and electrode configurations described and shown in the accompanying drawings are intended to be illustrative in nature. It should be understood that Figures 1A to 2B The leads and electrodes illustrated in the drawings may be implanted in different combinations than the example combinations shown, and some leads and / or electrodes may be omitted, or more electrodes may be carried by a given lead or IMD housing than in the examples shown. In some examples, a leadless IMD (e.g., IMD 100) may be implanted in a patient in combination with another implanted IMD (e.g., an IMD connected to an RA lead for pacing and sensing in the right atrium) for VCS pacing. Various lead-based and leadless IMD configurations are contemplated for sensing cardiac electrical signals and delivering VCS pacing pulses, which may be used in conjunction with the techniques disclosed herein to classify VCS pacing pulses and associated post-pacing cardiac electrical signals according to one of a plurality of possible capture types.

[0112] Figure 3 is a schematic diagram of circuitry that may be packaged within an IMD configured to sense cardiac electrical signals and to perform VCS pacing. Figure 3 The block diagram of FIG. 1 is described with reference to an IMD 14 coupled to electrodes carried by RA lead 16 and VCS lead 18, such as, for example, as an illustrative example. Figure 1C However, it should be understood that Figure 3The functionality of the various circuits and components shown in for sensing cardiac signals and delivering VCS pacing may be implemented in conjunction with other lead and electrode configurations, including Figure 2A and Figure 2B The leadless pacemaker 100 or other medical device configured to deliver VCS pacing pulses and sense cardiac electrical signals.

[0113] Shell 15 Figure 3 , which are represented as electrodes for cardiac electrical signal sensing and, in some examples, for delivering cardiac electrical stimulation pulses, such as unipolar pacing pulses. The electronic circuit system encapsulated in the housing 15 includes software, firmware, and hardware that work together to monitor cardiac electrical signals, determine when pacing therapy is needed, and deliver pacing electrical pulses to the patient's heart as needed according to the programmed pacing mode and pacing pulse control parameters. The electronic circuit system includes a control circuit 80, a memory 82, a therapy delivery circuit 84, a cardiac electrical signal sensing circuit 86, a telemetry circuit 88, and a power supply 98.

[0114] Power supply 98 provides power as needed to the circuitry of IMD 14, including each of components 80, 82, 84, 86, and 88. Power supply 98 may include one or more energy storage devices, such as one or more rechargeable or non-rechargeable batteries. Connections between power supply 98 and each of the other components 80, 82, 84, 86, and 88 will be Figure 3 84, but are not shown for clarity. For example, power supply 98 may be coupled to one or more charging circuits included in therapy delivery circuit 84 to provide the power required to charge a holding capacitor included in therapy delivery circuit 84, which is discharged at the appropriate time to deliver a pacing pulse under the control of control circuit 80. Power supply 98 is also optionally coupled to components of sensing circuit 86 for sensing cardiac electrical signals, such as sense amplifiers, analog-to-digital converters, switching circuit systems, etc. Power supply 98 can provide power to various components and circuits of telemetry circuit 88 and memory 82 as needed, which can be under the control of control circuit 80.

[0115] Figure 3The circuits shown are representative of the functionality included in the IMD 14, and may include any discrete and / or integrated electronic circuit components that implement analog and / or digital circuits capable of producing the functions attributed herein to the IMD 14 (or pacemaker 100). The various components may include application specific integrated circuits (ASICs), electronic circuits, processors (shared, dedicated, or groups) and memories that execute one or more software or firmware programs, combinational logic circuits, state machines, or other suitable components or combinations of components that provide the described functionality. In view of the disclosure herein, it is within the capabilities of those skilled in the art to provide software, hardware, and / or firmware to implement the described functionality in the context of any modern cardiac medical device system.

[0116] The control circuit 80 may communicate with the therapy delivery circuit 84 and the sensing circuit 86, for example, via a data bus, to cooperatively sense cardiac electrical signals and control the delivery of cardiac electrical stimulation therapy in response to sensed cardiac event signals (e.g., a P wave accompanying atrial depolarization and an R wave accompanying ventricular depolarization, or the absence thereof). Available electrodes are electrically coupled to the therapy delivery circuit 84 for delivering electrical stimulation pulses and / or the sensing circuit 86 for sensing cardiac electrical signals generated by the heart. The sensed cardiac electrical signals may include intrinsic signals generated by the heart in the absence of a pacing pulse that captures the heart (such as an intrinsic R wave), as well as induced response signals following a delivered pacing pulse of sufficient energy to cause capture of cardiac tissue.

[0117] The sensing circuit 86 may include two or more sensing channels for receiving raw cardiac electrical signals from two or more sensing electrode vectors. For example, RA signals may be sensed using RA lead electrodes 20 and 22 coupled to atrial sensing (A sensing) channel 87. Ventricular signals may be sensed through ventricular sensing (V sensing) channel 89 using VCS lead electrodes 32, 34, and / or 35. In some examples, the V sensing channel 89 may include multiple ventricular sensing channels for receiving raw signals from multiple sensing electrode vectors, which may include at least one electrode in or near the ventricular chamber. For example, the V sensing channel 89 may include a NF sensing channel for receiving raw NF signals (e.g., using electrodes 32 and 34 of the VCS lead 18 in a bipolar sensing electrode pair). The V sensing channel 89 may include a FF sensing channel for receiving raw FF signals. For example, a second electrode vector may be used to receive raw FF signals, the second electrode vector having electrodes spaced farther apart than the electrodes of the NF sensing electrode vector. For example, the FF signal may be sensed using the VCS lead tip electrode 32 or the ring electrode 34 paired with the IMD housing 15. In some examples, the V sensing channel 89 may receive a raw FF signal sensed using the tip electrode 32 or the ring electrode 34 paired with the coil electrode 35. In other examples, the FF signal may be sensed using the coil electrode 35 paired with the IMD housing 15. Any other available unipolar sensing electrode pair that senses a relatively more global or FF signal than the bipolar electrode pair of the tip electrode 32 and the ring electrode 34 may be used to sense a ventricular EGM signal that can be used to classify the VCS capture type, either alone or in combination with the ventricular NFEGM signal. The FF sensing electrode vector may have a larger inter-electrode distance than the NF sensing electrode vector, so that the FFEGM signal generated by the sensing circuit 86 from the raw FF signal may be more representative of the global activation of the ventricle or a relatively larger myocardial mass than the NF signal. The NF signal may be more representative of local tissue activation (depolarization) at or near the VCS pacing site.

[0118] The sensing circuit 86 may include switching circuitry for selectively coupling sensing electrode pairs among the available electrodes to corresponding sensing channels among the A sensing channels 87 or the V sensing channels 89. The switching circuitry may include a switch array, a switch matrix, a multiplexer, or any other type of switching device suitable for selectively coupling components of the sensing circuit 86 to selected electrodes.

[0119] Each of the sensing channels 87 and 89 of the sensing circuit 86 may include an input filter, a preamplifier, an analog-to-digital converter, and a bandpass filter for receiving raw cardiac electrical signals from a corresponding pair of sensing electrodes to generate a multi-bit digital cardiac electrical signal, which may be referred to as an "EGM" signal when the raw signal is sensed from within the cardiac chamber. The multi-bit EGM signal may be passed from the sensing circuit 86 to the control circuit 80 for processing and analysis, and / or for transmission to an external device 50 (e.g., Figure 1A ) for processing and analysis and / or display on the display unit 54. As described below, the control circuit 80 can extract segments of the ventricular FFEGM signal and / or NF EGM signal during the post-pacing time interval after the delivery of the VCS pacing pulse for classification of the capture type of the VCS pacing pulse. The post-pacing EGM signal sensed after the VCS pacing pulse can be obtained from

[0120] The FF ventricular sensing channel of the V sensing channel 89 receives and / or the NF ventricular sensing channel of the V sensing channel 89 receives for providing input to the PCC 51. It should be understood that when the V sensing channel 89 includes a NF sensing channel and a FF sensing channel, or other sensing channels for sensing multiple ventricular EGM signals, each sensing channel may include the same or different filter cutoff frequencies for passing the EGM signal to the control circuit 80, where the desired QRS signal is enhanced and other undesired cardiac signals or noise are attenuated.

[0121] Each sensing channel 87 and 89 may include cardiac event detection circuitry that may include one or more sense amplifiers, filters, rectifiers, threshold detectors, comparators, analog-to-digital converters (ADCs), timers, or other analog or digital components for detecting cardiac electrical events. For example, an atrial event detector may be included in the A sensing channel 87 for sensing an intrinsic P wave associated with intrinsic atrial depolarization using one or both of electrodes 20 and 22 carried by the RA lead 16. A ventricular event detector may be included in the V sensing channel 89 for sensing an intrinsic R wave associated with intrinsic ventricular depolarization using electrodes 32 and 34 carried by the VCS lead 18.

[0122] The cardiac event sensing threshold (such as a P wave sensing threshold or an R wave sensing threshold) can be automatically adjusted by the sensing circuit 86 under the control of the control circuit 80, for example, based on a timing period and a sensing threshold determined by the control circuit 80, stored in the memory 82, and / or controlled by the hardware, firmware and / or software of the control circuit 80 and / or the sensing circuit 86. For example, the R wave sensing threshold can be controlled to start at a starting threshold voltage after the post-ventricular blanking period and then decrease according to the decay curve until a minimum sensing threshold is reached. The minimum R wave sensing threshold can be set to the programmed sensitivity of the R wave detection circuit system. The sensitivity, which is programmed to a voltage level, typically in millivolts, is the minimum voltage level above which the cardiac event detection circuit system of the corresponding A sensing channel 87 or V sensing channel 89 can sense a P wave or an R wave.

[0123] Upon detecting a cardiac electrical event based on a sensing threshold crossing, the sensing circuit 86 may generate a sensed event signal that is communicated to the control circuit 80. For example, an atrial event detector may generate a P-wave sensed event signal in response to a P-wave sensing threshold crossing. A ventricular event detector may generate an R-wave sensed event signal in response to an R-wave sensing threshold crossing. The sensed event signal may be used by the control circuit 80 to set a pacing escape interval timer that controls a base time interval for scheduling cardiac pacing pulses (e.g., atrial pacing pulses) and VCS pacing pulses.

[0124] The control circuit 80 may include various timers or counters for counting down AV delays, VV delays, atrial lower rate intervals, ventricular lower rate intervals, etc. The sensed event signal may trigger or inhibit pacing pulses depending on the specific programmed pacing mode. For example, a P wave sensed event signal received from the sensing circuit 86 may cause the control circuit 80 to inhibit the scheduled atrial pacing pulse and schedule a VCS pacing pulse at the selected AV delay. If the AV delay ends before the control circuit 80 receives the R wave sensed event signal from the sensing circuit 86, the therapy delivery circuit 84 may generate and deliver a VCS pacing pulse at the AV delay after the sensed P wave, and in this manner deliver atrial synchronized ventricular pacing. If an R wave sensed event signal is received from the sensing circuit 86 before the AV delay ends, the scheduled VCS pacing pulse may be inhibited. The AV delay controls the amount of time between an atrial event (paced or sensed) and a VCS pacing pulse to promote electrical and mechanical synchronization of the heart chambers.

[0125] In other cases, a ventricular lower rate interval (LRI) may be set by control circuitry 80 to schedule a VCS pacing pulse after a delivered VCS pacing pulse or a sensed R-wave. The LRI may correspond to a programmed ventricular lower rate, or may be adjusted by control circuitry 80 to occur after a programmed ventricular lower rate, such as by an accelerometer signal or other patient activity sensor included in IMD 14 ( Figure 3 The control circuit 80 may be configured to control the therapy delivery circuit 84 to deliver rate-responsive pacing when an increase in patient activity level is detected by the IMD 14 (not shown). When the IMD 14 is operating in a dual-chamber pacing mode (e.g., DDD mode), when a P wave is sensed during an LRI, a VCS pacing pulse may be triggered to occur at the AV delay, and the LRI may be restarted when the VCS pacing pulse is delivered. If the LRI ends without a sensed P wave or a sensed R wave, a VCS pacing pulse is delivered at the end of the LRI, and the LRI may be restarted. The control circuit 80 may be configured to control the therapy delivery circuit 84 to deliver VCS pacing pulses according to various pacing modes and pacing therapies, which may include bradycardia pacing, post-shock pacing, anti-tachycardia pacing (ATP), cardiac resynchronization therapy (CRT), rate-responsive pacing, and the like.

[0126] The therapy delivery circuit 84 may include charging circuitry, one or more charge storage devices (such as one or more holding capacitors), an output capacitor, and switching circuitry that controls when the holding capacitor is charged and discharged across the output capacitor to deliver pacing pulses to a selected pacing electrode vector coupled to the therapy delivery circuit 84. The therapy delivery circuit 84 may include one or more pacing channels. In the example of the IMD 14, the therapy delivery circuit 84 may include an atrial pacing channel and a ventricular pacing channel, each channel including one or more holding capacitors, one or more switches, and an output capacitor for generating pacing pulses delivered by a corresponding RA lead 16 (e.g., via electrodes 20 and 22) or VCS lead 18 (e.g., via electrodes 32 and 34).

[0127] Charging the holding capacitor to the programmed pacing voltage amplitude and discharging the capacitor for the programmed pacing pulse width may be performed by therapy delivery circuitry 84 in accordance with control signals received from control circuitry 80. For example, a pacing timing circuit included in control circuitry 80 may include a programmable digital counter set by a microprocessor of control circuitry 80 to control, for example, various single-chamber and / or dual-chamber pacing modes, when LV lead 46 is connected to IMD 14 (e.g., Figure 1BThe microprocessor of the control circuit 80 may also set the amplitude, pulse width, polarity or other characteristics of the cardiac pacing pulses which may be based on programmed values ​​stored in the memory 82.

[0128] As noted above, the IMD 14 can be configured to detect non-sinus tachycardia and deliver ATP. When the IMD 14 is configured as an ICD for detecting tachycardia and delivering CV / DF shocks, in addition to the low-voltage therapy circuit system for generating low-voltage pacing pulses, the therapy delivery circuit 84 may also include a high-voltage therapy delivery circuit system for generating high-voltage shock pulses. In response to detecting ventricular tachycardia or fibrillation, the control circuit 80 can control the therapy delivery circuit 84 to deliver CV / DF shocks. The high-voltage therapy circuit system may include a high-voltage capacitor and a high-voltage charging circuit system for generating and delivering CV / DF shock pulses using a coil electrode (e.g., coil electrode 35) and / or housing 15 carried by one or more leads coupled to the IMD 14.

[0129] Control parameters used by control circuit 80 to sense cardiac event signals (e.g., P and R waves) and control pacing therapy delivery may be programmed into memory 82 via telemetry circuit 88. Telemetry circuit 88 includes a communication device for communicating with external device 50 (e.g., Figure 1A ) for communicating with the external device 50. Under the control of the control circuit 80, the telemetry circuit 88 may receive downlink telemetry from the external device 50 and send uplink telemetry to the external device. In some cases, the telemetry circuit 88 may be used to transmit communication signals to and receive communication signals from another medical device implanted in the patient. The telemetry circuit 88 may transmit the EGM signal, for example, as a continuously sampled signal or as any of an extracted post-pacing EGM segment sensed during a post-pacing time interval following a delivered VCS pacing pulse, for receipt by the external device 50 for processing and analysis by the PCC 51 implemented in the external device processor 52. In some examples, the extracted post-pacing time EGM segment includes one or more sample points beginning before the delivery of the VCS pacing pulse to include an isoelectric baseline of the EGM signal just before the delivery of the VCS pacing pulse.

[0130] In some examples, PCC 51 may be implemented in control circuit 80. PCC 51 may be implemented in control circuit 80 to classify EGM signals received from sensing circuit 86 following a VCS pacing pulse delivered by therapy delivery circuit 84. PCC 51 may be implemented in control circuit 80 to classify EGM signals received from sensing circuit 86 following a VCS pacing pulse delivered by therapy delivery circuit 84. Figure 1A After a machine learning model (e.g., a neural network or other machine learning model) is trained in an external device 50 or another computing system, VCS pacing pulses are classified according to capture type based on the machine learning model applied to the cardiac signals. In other examples, PCC 51 can be trained during a training period using signals from a patient implanted with IMD 14. The machine learning model can be established during supervised training and / or verified by a user during a validation process after the training period.

[0131] Figure 4 is a conceptual diagram of a processing circuit according to some examples that is configured to analyze a cardiac signal to determine a pacing capture classification of the cardiac signal and an associated VCS pacing pulse. The processing circuit, shown as PCC 51, may include a machine learning model trained using a machine learning algorithm using multiple channels of inputs 62, 64, 66, and 68. PCC 51 may be implemented in an external device processor 52 or in another computing system, which may be a computing device included in a remote patient monitoring system or other external device. However, it should be understood that the description of PCC 51 and the various features and functions attributed to PCC 51 may be implemented in whole or in part in the control circuit 80 of IMD 14.

[0132] The PCC 51 is configured to receive an unknown signal input 62, which includes a post-pacing cardiac electrical signal waveform to be classified as one of a plurality of pacing capture types. The unknown signal input 62 may be a time segment of an ECG signal sensed using a surface or subcutaneous electrode, a submuscular or other extracardiac electrode received by the external device processor 52. For example, the ECG signal may be received by the processor 52 from ECG electrodes coupled to leads connected to the external device 50 via the interface 55. When an ECG signal is used as the input signal 62, the ECG sensing electrode vector may correspond to any 12-lead ECG electrode vector, which may be, for example, one of the V1 to V6 ECG leads or one of the V1, V2, V5 or V6 ECG leads. The processor 52 may receive the ECG signal and obtain a time segment of the ECG signal including a post-pacing QRS waveform following a preceding VCS pacing pulse. Processor 52 may provide the ECG signal segment to PCC 51 as unknown signal input 62 .

[0133] The unknown signal input 62 may alternatively be an EGM signal sensed using intracardiac electrodes. The EGM signal may be sensed by the IMD 14, and transmitted by the IMD telemetry circuit 88, and received by the processor 52 via the external device telemetry unit 58. In some cases, one or more implanted leads (such as the VCS lead 18) may be coupled to the external device 50 via the interface 55 prior to being connected to the IMD 14 so that the external device processor 52 may receive the EGM signal in order to obtain the unknown signal input 62.

[0134] When the input signal 62 is an EGM signal, the EGM signal may be a FF signal that may be sensed using any available unipolar sensing electrode vector. Figures 1A to 1C 15 and IMD housing 15. In an example of an LV lead, unknown signal input 62 may be obtained from a FF signal sensed using VCS lead tip electrode 32 and IMD housing 15 or VCS lead ring electrode 34 and IMD housing 15. When other leads and electrodes are available, another FF or monopolar sensing electrode pair may be used to sense the EGM signal from which unknown signal input 62 is obtained. In an IMD configuration that includes an LV lead, for example, Figure 1B As shown, unknown signal input 62 may be obtained from EGM signals sensed using one or more LV lead electrodes 48 in a sensing electrode vector. For example, a monopolar sensing electrode vector may be selected using one of LV lead electrode 48 and IMD housing 15, or a combination of one LV lead electrode in LV lead electrode 48 and any other electrode available from RA lead 16, VCS lead 18, or RV lead 17 for sensing an EGM signal from which unknown signal input 62 is obtained. In a system including CV / DF coil electrodes (e.g., such as Figure 1A The RV lead coil electrode 24 or Figure 1C In the example of the VCS lead coil electrode 35 shown in FIG. 1 , the unknown signal input can be obtained from the EGM signal sensed using one or more CV / DF coil electrodes in the sensing electrode vector. For example, the EGM signal can be sensed from the coil electrode 24 and the housing 15 or the coil electrode 35 and the housing 15.

[0135] In other examples, relatively higher NF EGM signals that can be vector sensed using bipolar sensing electrodes can be received by the external device processor 52 for use in obtaining the unknown signal input 62. The NF EGM signals can be sensed using the VCS lead tip electrode 32 and the ring electrode 34. In the case of a leadless pacemaker 100, NV EGM signals can be sensed by the housing-based electrodes 102, 104, and / or 106. In other examples, bipolar EGM signals can be sensed using electrodes that can be positioned away from the VCS pacing site. For example, when the LV lead 46 is available, such as Figure 1BAs shown, the bipolar sensing electrode vector can be selected from any pair of electrodes 48. In yet other examples, when RV lead 17 is connected to IMD 14, EGM signals can be sensed between any pair of RV lead tip electrode 28 and RV lead ring electrode 30, RV lead tip electrode 28 and RV lead coil electrode 24, RV ring electrode 30 and RV lead coil electrode 24, or between any pair of RV lead electrodes 24, 28, or 30 and IMD housing 15.

[0136] For each VCS pacing pulse delivered for which capture classification is desired, an unknown signal input 62 may be obtained from a cardiac electrical signal (e.g., an ECG or EGM signal) sensed within at least a predetermined post-pacing time interval (e.g., including at least a predetermined number of post-pacing samples). In various examples, the unknown signal input is 35 to 100 samples or between 50 and 70 samples of an ECG or EGM signal sampled at 256 Hz. In an illustrative example, when the sampling rate is 256 Hz and 62 samples of the ECG or EGM signal are obtained (which may include at least one sample point before the time point of the delivered VCS pacing pulse), the signal input 62 may be 242 milliseconds (ms) long. In other examples, the input signal 62 may be sampled at 128 Hz or 512 Hz, and the number of samples may be adjusted accordingly to obtain, for example, an ECG or EGM signal input having a length between 150 ms and 600 ms, a length between 200 ms and 500 ms, or a length of approximately 200 ms to 350 ms.

[0137] The input signal 62 may be sampled starting from one or more sample points prior to the time point of VCS pacing pulse delivery to capture information about the isoelectric baseline of the ECG or EGM signal at or just prior to the time point of VCS pacing pulse delivery. By zeroing the first input signal sample point obtained prior to VCS pacing pulse delivery, the intrinsic cardiac electrical signal baseline difference between different input signals received by the PCC 51 may be corrected. In other examples, sampling of the input signal 62 may begin after a delay of one or more sample points, for example to reduce pacing artifact content in an unknown signal input. When the cardiac electrical signal is received directly by the external device 50 (e.g., via the interface 55), a higher sampling rate of the input signal 62 may be utilized because the processing power and burden may be less limited than when the EGM signal is received from the IMD 14 or when the PCC 51 is implemented in the IMD 14.

[0138] The post-pacing time interval classified by PCC 51 may extend a predetermined time interval from a sample point of VCS pacing pulse delivery, or from a predetermined delay of one or more sample points after the VCS pacing pulse. Sample points collected before the VCS pacing pulse may be used only to zero the baseline of the input signal. Therefore, as used herein, the term "post-pacing time interval" may not only be "post-pacing" after the delivered pacing pulse, but may include one or more sample points before the delivered VCS pacing pulse to enable baseline zeroing and time alignment of multiple signal inputs received by PCC 51. However, present in the post-pacing time interval after the delivered VCS pacing pulse is a QRS waveform (or lack of a QRS waveform), which is analyzed by PCC 51 to classify the delivered VCS pacing pulse and the associated post-pacing waveform according to the capture type.

[0139] Although in Figure 4 5 shows one unknown input signal 62 sensed using one sensing electrode vector, but it is contemplated that multiple unknown input signals may be received by the PCC 51 and sampled during at least the post-pacing time interval following a VCS pacing pulse for use in classifying the VCS pacing pulse according to one of a plurality of capture types. For example, the PCC 51 may receive one input signal sensed from a FF EGM signal sensing electrode vector, and a second input signal sensed from a NF EGM signal sensing electrode vector or an ECG signal sensing electrode vector. Figure 1A , the PCC 51 can receive a first unknown signal input sensed from a FF (e.g., unipolar) sensing electrode vector (which can be referred to as a “tip-to-can” or “ring-to-can” sensing vector, respectively) between the VCS lead tip electrode 32 or the VCS lead ring electrode 34 and the IMD housing 15. The PCC 51 can receive a second unknown signal input sensed from a NF (e.g., bipolar) sensing electrode vector between the VCS lead tip electrode 32 and the VCS lead ring electrode 34. In other examples, any combination of any one of the example sensing electrode vectors described herein or that can be obtained from electrodes coupled to the IMD 14 and / or external device 50 or two or more sensing electrode vectors can be used to sense cardiac electrical signals sampled within at least one post-pacing time interval in order to obtain multiple unknown signal inputs received by the PCC 51 for capture type classification.

[0140] In some examples, the PCC 51 receives two additional unknown signal inputs sampled from an ECG signal, such as any one of the V1, V2, V3, V4, V5, and / or V6 ECG signals or any one of the 12-lead ECG signals. Multiple ECG signals may be sampled within at least one post-pacing time interval to provide multiple unknown signal inputs that, along with the preceding VCS pacing pulse, are to be collectively classified by the PCC 51 as one of a plurality of possible capture types. In an illustrative example, one of the V1, V2, or V3 ECG signals and one of the V4, V5, or V6 ECG signals are each sampled within a time interval that includes one or more sample points before a VCS pacing pulse and a predetermined number of sample points after a VCS pacing pulse, and are provided as two unknown signal inputs for capture type classification by the PCC 51.

[0141] In still other examples, a combination of one or more EGM signals and one or more ECG signals may be provided as unknown input signals to be collectively classified by the PCC 51. To illustrate, a FF signal and a V5 or V6 ECG signal (or any other selected ECG signal) sensed from the tip-to-shell of the VCS lead tip electrode 32 to the housing 15 may be sampled within a post-pacing time interval (which may include one or more sample points prior to a delivered VCS pacing pulse) and received by the PCC 51 as unknown signal inputs.

[0142] In addition to receiving at least one unknown signal input 62 to be classified, the PCC 51 may also be configured to receive at least one template beat input. The template beat input may be sampled from a selected cardiac electrical signal during a post-pacing time interval following a VCS pacing pulse delivered with a selected pacing pulse output. For example, Figure 4 Three template beat inputs 64, 66 and 68 are shown in FIG. However, as described below, one, two, three or more template beat inputs may be received as inputs by the PCC 51. Each of the input signals received by the PCC 51 may be combined into a series of multi-channel data points. Figure 4 In the example of , the unknown signal to be classified and each of the three template beat inputs can be combined into a multi-channel input as represented by inputs 62 , 64 , 66 , and 68 .

[0143] Each template beat input 64, 66, and 68 can be generated from a cardiac electrical signal sensed by the same sensing electrode vector used to obtain the unknown signal input 62 (or one of the unknown signal inputs when the PCC 51 receives multiple unknown signal inputs). Each template beat input 64, 66, and 68 can be obtained during a post-pacing time interval using the same sampling rate as the unknown signal input 62. The template beat inputs 64, 66, and 68 can each correspond to a known pacing pulse output, but can be of an unknown capture type for a given patient. As such, the template beat inputs 64, 66, and 68 can be of an unknown capture type, but can be established during VCS pacing with a preselected pacing pulse output, such as a maximum pacing pulse output setting, a minimum pacing pulse output setting, and / or an intermediate pacing pulse output setting that can be obtained from a programmable pacing pulse output setting or used during a capture threshold search. As described in the examples given below, a template beat input can be established during VCS pacing with a relatively high pacing pulse amplitude (e.g., 5 volts or greater) that can be expected to capture at least a portion of the His-Purkinje conduction system and / or ventricular myocardium.

[0144] Each template beat input 64, 66, and 68 can be obtained by the IMD control circuit 80 or by the external device processor 52 from a post-pacing time interval following a single VCS pacing pulse delivered with a known pacing pulse output. In other examples, each template beat input can be obtained by averaging a plurality of individual cardiac signal segments, each of which is sensed within a predetermined post-pacing time interval following a VCS pacing pulse delivered with a selected pacing pulse output. As described above, the sampled signal can include one or more sample points prior to the delivered VCS pacing pulse, as well as sample points acquired within a predetermined time interval after the VCS pacing pulse, such that the "post-pacing" time interval includes at least the predetermined time interval after the VCS pacing pulse, and can include one or more sample points prior to the VCS pacing pulse. Thus, each template beat input 64 , 66 , and 68 may include an equal number of one or more sample points preceding the VCS pacing pulse and post-pacing signal template to enable processor 52 to perform baseline zeroing and time alignment of the template beat inputs 64 , 66 , and 68 .

[0145] Once established, the template beat inputs 64, 66, and 68 can be stored in the IMD memory 82 or the external device memory 53 and retrieved by the PCC 51 when the unknown signal input 62 is received for capture type classification. In other examples, the template beat inputs 64, 66, and 68 can be obtained by creating each template beat signal from a post-pacing sensed cardiac electrical signal after each of a predetermined number of VCS pacing pulses at a corresponding selected pacing pulse output whenever a capture threshold test is performed by the medical device system.

[0146] One or more of the unknown signal input 62 and / or the template beat inputs 64, 66, and 68 may include a leading (or trailing) data value. The data value may be associated with a signal characteristic of the corresponding unknown signal input or template beat, or with a corresponding VCS pacing pulse output associated with the unknown input signal or template beat input. For example, the unknown signal input may include a sampled post-pacing cardiac electrical signal that is pre-added or supplemented with a pacing pulse amplitude or pacing pulse width of a delivered VCS pacing pulse to be classified. The template beat input 64, 66, or 68 may include a template beat signal obtained within a predetermined post-pacing time interval (during one or more VCS pacing cycles) that is pre-added or supplemented with a data value corresponding to a pacing pulse output associated with the template beat signal. For example, the template beat input may be acquired during VCS pacing at a predetermined pacing pulse voltage amplitude (e.g., between 1 volt and 10 volts or between 2 volts and 8 volts). In one example, during VCS pacing, at least one template beat input 64, 66, or 68 is established at 5 volts or the highest pacing voltage amplitude used during capture threshold testing. The value of the pacing pulse amplitude (e.g., 5 volts) can be pre-added or appended to the associated template beat signal.

[0147] The pre-added or supplemented pacing pulse output data value (e.g., pacing pulse voltage amplitude or pacing pulse width) can be separated from the corresponding unknown signal sample point value or template sample point value by zero. Zero padding can be included before and / or after the pacing pulse output data value to separate the pacing pulse output data value from the cardiac electrical signal sample point value, thereby avoiding the pacing pulse output value and the cardiac signal sample point value from affecting each other during training and during VCS pacing pulse classification after the machine learning model is established. Zero padding before and after the pacing pulse output data value can avoid or reduce undesirable effects on the edges of the cardiac electrical signal of the unknown signal input or template beat input during machine learning.

[0148] In other examples, features of the unknown signal after pacing and / or the template beat signal to be classified can be pre-added or appended to the corresponding input signal provided to the PCC 51. For example, the processor 52 (or the control circuit 80) can determine the unknown signal segment to be classified and / or the features of the template beat signal to be provided as pre-added or appended data values. The signal feature can be the maximum peak amplitude, the activation time determined as the time interval from the VCS pacing pulse to the maximum peak amplitude, the signal width, the signal area, the maximum positive slope, the maximum negative slope, the time between the maximum positive slope and the maximum negative slope, or any other selected cardiac electrical signal feature. For example, the LV activation time can be pre-added to the input signal. The LV activation time can be determined from the time of the delivered VCS pacing pulse to the maximum peak amplitude of the post-pacing QRS signal (or another reference feature of the QRS signal), which can be sensed from the LV sensing electrode vector, for example, using the VCS pacing tip electrode positioned for pacing the LBB or using the electrode carried by the LV lead 46.

[0149] In some examples, a first data value can be pre-added to an unknown signal input and / or one or more template beat signals, and a second data value can be appended to the unknown signal input and / or one or more template beat signals to provide a PCC input. These data values ​​can include values ​​related to an associated VCS pacing pulse output, and / or values ​​related to cardiac electrical signal characteristics of a corresponding unknown input signal or template beat input, and / or patient demographic information, such as age, gender, or other patient-related information. The external device processor 52 or the IMD control circuit 80 can determine the data values ​​to be pre-added or appended to the corresponding input signal.

[0150] The PCC 51 may be trained for VCS pacing pulse capture type classification using a cardiac signal dataset from a patient population, using multiple unknown signal inputs and template beat inputs to be received by the PCC 51, using supervised deep learning techniques, such as convolutional neural networks (CNNs), residual CNNs, feed-forward neural networks (FFNNs), recurrent neural networks (RNNs), transformers, or other machine learning techniques, such as decision trees, random forest models, or other machine learning methods for building machine learning models (e.g., neural network models). In a similar manner, for example, the inputs received by the PCC 51 during the training session are acquired using the same sensing electrode vectors, sampling rates, and post-pacing time intervals as the signals to be acquired for input to the PCC 51, so as to classify VCS pacing pulse capture types after the training is completed and the machine learning model is fixed.

[0151] A PCC training data set may be obtained from each of a plurality of patients, each data set comprising a desired input signal. The plurality of data sets obtained from the plurality of patients define "epochs" that may be input to the PCC 51 multiple times during a training session. Training may be completed after a fixed number of epochs, e.g., after all training data sets from a patient population have been passed to the PCC 51 a fixed number of times. In other examples, PCC training may be stopped once the output and associated confidence level have not changed by more than a specified amount in one or more recent epochs.

[0152] The trained machine learning model of PCC 51 can be verified by inputting a validation data set obtained from one or more patients. The validation data set may be different from the training data set, but acquired in a similar manner. The validation data set may come from a different or smaller group of patients than the training data set. The output 70 of PCC 51 obtained from the input validation data set may be verified by expert truthing of the output capture type classification. The training of PCC 51 is completed when the expert's verification of the PCC output during training and / or validation has reached a certain threshold, which may be the minimum percentage of accurate classification and / or the maximum percentage of inaccurate classification of the capture classification obtained according to the expert truthing.

[0153] The output 70 of the PCC 51 includes at least a capture type classification of the VCS pacing pulse associated with the unknown input signal 62. The output 70 may include a confidence level for the classified capture type, which may be expressed as a percentage. In some examples, the output 70 may include a confidence level corresponding to each of the possible capture type classifications. The processor 52 may determine the capture type classification of the delivered VCS pacing pulse by applying the machine learning model of the PCC 51 to the input signal. The processor 52 may determine the capture type classification associated with the highest confidence level output of the applied PCC machine learning model. In some examples, if the highest confidence level among all capture type classifications is less than a threshold percentage, the capture type classification may be determined by the processor 52 as "unknown".

[0154] The output 70 of the PCC 51 can be used by the external device processor 52 to generate one or more outputs or information related to the VCS pacing capture classification. In some examples, the processor 52 and the display unit 54 can collaboratively generate a visual representation of the capture type classification, for example in a GUI of the display unit 52. Each classified VCS pacing pulse and / or the corresponding post-pacing cardiac electrical signal waveform can be displayed and annotated according to the classified capture type. In some examples, a confidence level corresponding to the classified capture type can be displayed for each classified VCS pacing pulse. As described below, the PCC output 70 can be obtained by the processor 52 during the pacing capture threshold test to display the annotated VCS pacing markers and / or cardiac electrical signals according to the capture type. The output 70 obtained during the capture threshold test can be used by the processor 52 to select a VCS pacing pulse output setting to achieve capture of at least a portion of the VCS or to achieve a desired type of VCS pacing capture. The selected VCS pacing pulse output setting may be displayed as a recommended setting in a GUI of display unit 54 or automatically programmed into IMD 14 via telemetry unit 58. Processor 52 may generate an output for display by display unit 54 in a GUI that includes a table or graphical representation of the pacing pulse outputs delivered during the capture threshold test and a corresponding capture type classification associated with each pacing pulse output.

[0155] Additionally or alternatively, PCC output 70 can be used by external device processor 52 to determine the percentage of delivered VCS pacing pulses corresponding to each of the possible capture types for reporting in a tabular display or a graphical display. Reporting of the percentage of each type of capture type classification for VCS pacing pulses can inform a clinician for use in patient management and pacing therapy control parameter optimization. When PCC 51 is implemented in IMD control circuitry 80, memory 82 can store a count of each type of capture classification so that control circuitry 80 can determine the percentage of each determined type of capture associated with VCS pacing pulse delivery, and the results can be transmitted to external device 50 for use in generating a report for display by display unit 54.

[0156] Figure 5 is a flow chart 200 of a method performed by processing circuitry of a medical device system in some examples for classifying VCS pacing pulses and corresponding post-paced QRS waveforms according to capture type. Figure 5The techniques are primarily described as being performed by processor 52 of external device 50, where IMD 14 delivers VCS pacing pulses and senses EGM signals, which may be received by processor 52 via telemetry transmissions between IMD 14 and external device 50. In other examples, another processor may be executed alone or in combination with processor 52. Figure 5 For example, control circuitry 80 of IMD 14 or a processor of another computing device, either alone or in combination with processor 52, may execute Figure 5 The input data used to train the PCC 51 at block 201 may be obtained from a plurality of different patients, each of whom is implanted with an IMD and associated lead and electrode configurations, such as the IMD 14 and above in combination. Figures 1A to 1C Thus, the control circuit 80 of the IMD 14 implanted in each patient can participate in obtaining the execution Figure 5 Each IMD delivers VCS pacing pulses with a varying pacing pulse output and senses EGM signals from which cardiac electrical signal segments can be obtained for generating training and validation data sets that are used as input to the PCC 51 at box 201 during training sessions and validation sessions.

[0157] In other examples, an external pacing device, which may be implemented in external device 50 or a pacing system analyzer (PSA), may be configured to generate VCS pacing pulses that may be delivered using implanted leads, such as VCS pacing lead 18, that are positioned for VCS pacing prior to connection to IMD 14. VCS pacing lead 18 may be electrically connected to interface 55 so that electrical pulses generated by external device 50 may be delivered as VCS pacing pulses, enabling acquisition of post-pacing cardiac electrical signals that are used to obtain input for training PCC 51.

[0158] At block 201, the PCC 51 is trained using training data including post-pacing cardiac electrical signal inputs obtained from a patient population. For example, the post-pacing cardiac electrical signal inputs may be obtained from ECG and / or EGM signals following a VCS pacing pulse at a predetermined post-pacing time interval and a selected sampling rate, according to any of the examples described above. As indicated above, the "post-pacing" cardiac electrical signal inputs may be more than just post-pacing sample points, as each signal input may include one or more pre-pacing sample points for time alignment and baseline zeroing.

[0159] Input training data acquired from multiple patients may then be loaded into the memory 53 of the external device 50 for retrieval by the processor 52 and provided as input to the PCC 51 during a machine learning training session. In some examples, the PCC 51 is trained to classify capture types using two or more unknown signal inputs obtained from two or more different cardiac signal sensing electrode vectors, which may be ECG and / or EGM signal sensing electrode vectors, during the same time interval following a VCS pacing pulse.

[0160] In other examples, the PCC 51 is trained to classify the capture type using at least one unknown signal input for each VCS pacing pulse to be classified and using one or more template beat inputs, as described above in conjunction with Figure 4 As described and further described below in conjunction with the accompanying flow charts and illustrations. For example, as described below, the processor 50 may determine a template beat difference signal as a difference between an unknown input signal to be classified and a template beat signal established from at least one cardiac electrical signal sensed during a post-pacing time interval following a VCS pacing pulse delivered with a preselected pacing pulse output. One or more template beat difference signals (each template beat difference signal corresponding to a different preselected pacing pulse output) may be combined with one or more unknown input signals during a machine learning training session and provided as an input to the PCC 51. In some examples, the PCC inputs additionally include one or more template beat inputs, which may or may not correspond to the template beat signal used to determine the template beat difference signal.

[0161] Typically, training data may be obtained from groups or populations of different patients. PCC 51 may be operable to analyze and process the received data to determine VCS pacing pulse classifications according to different capture types with high accuracy. PCC 51 may be trained under supervision, where an expert confirms the classified capture type based on visual inspection or analysis of input signal features. Typically, the greater the amount of input signal data on which PCC 51 is trained, the more accurate the machine learning model of PCC 51 is in classifying the capture type of a new, unknown signal input.

[0162] After the machine learning model (e.g., neural network or other model) of PCC 51 is trained using the training data set, a validation data set obtained from one or more patients different from the training data set can be input to PCC 51. An expert can confirm that the trained PCC output captures the type classification based on visual inspection.

[0163] Once the machine learning model is fixed, VCS pacing pulses can be delivered by the IMD 14 implanted in the individual patient at box 202 for capturing type classification. At box 204, the external device processor 52 obtains the post-pacing cardiac electrical signals required to provide input to the PCC 51. The IMD 14 can transmit one or more EGM signals to the external device 50. The processor 52 can receive the EGM signals for extracting post-pacing cardiac signal data to provide as input to the PCC 51. When VCS pacing pulses are being delivered by the IMD 14, the external device processor 52 can receive the EGM signals transmitted in real time by telemetry and extract post-pacing EGM signal segments to provide input to the PCC 51. In other examples, post-pacing EGM signal segments can be stored in the IMD memory 82, obtained during VCS pacing, and transmitted to the external device 50 for post-processing.

[0164] Processor 52 obtains at least one unknown input signal to be classified based on the machine learning model established during the training session and any other data inputs required by PCC 51. Other signal inputs are described herein in conjunction with the figures and flow charts, and may include template beat inputs, template difference inputs, derivative signal inputs, VCS pacing pulse related data inputs, and / or cardiac signal feature related inputs.

[0165] At box 206, the PCC 51 applies the trained model to the individual patient data to determine a capture type classification for the delivered VCS pacing pulses. At box 202, a series of VCS pacing pulses can be delivered. The processor 52 receives the desired cardiac electrical signals to extract the post-pacing unknown signal input for each delivered VCS pacing pulse in the series of pulses. The processor 52 can receive or determine any other desired data from the individual patient cardiac signals for input to the PCC 51 and classification by the PCC based on the applied machine learning model. At box 206, each delivered VCS pacing pulse in a series of (continuous or non-continuous) VCS pacing pulses can be classified according to the capture type.

[0166] The PCC 51 outputs each VCS pacing pulse capture classification for use by the external device processor 52 in generating information for display by the display unit 54 and / or transmission to the IMD 14. The PCC output may include a confidence level, e.g., as a percentage confidence, associated with the capture type classification or associated with each possible capture type that includes the capture type classification determined for the delivered VCS pacing pulse. The processor 52 may use the confidence levels to further post-process the PCC output, e.g., changing the capture type classification to "unknown" if the confidence level of the capture type classification with the highest confidence level is below a predetermined threshold. The confidence level threshold may be programmable by the user for use by the external device processor 52 in determining when a determined capture type classification is unknown.

[0167] At box 208, the external device processor 52 can generate an output useful in managing or controlling VCS pacing therapy, such as for automatically selecting or user selecting a VCS pacing pulse output setting. The external processor 52 can generate an output based on the capture type classification determined by the PCC 51. For example, each classification and associated confidence level can be stored in the external device memory 53. Each classification can be stored in association with the pacing pulse output of the classified VCS pacing pulse. However, it should be understood that the pacing pulse output of each VCS pacing pulse to be classified may be provided as an input to the PCC 51 or may not be provided as an input to the PCC. The PCC output can be compiled by the processor 52 along with other VCS pacing pulse data and / or cardiac signal information for generating information and data displayed in the GUI by the display unit 54. For example, the external device processor 52 may combine the PCC output with VCS pacing pulse timing and pacing output data and with received cardiac electrical signals to generate a visual representation in a GUI of the display unit 54 that includes cardiac electrical signals sensed during VCS pacing and VCS pacing pulse markers annotated with capture type classifications output by the PCC 51, which may include annotations of associated confidence levels.

[0168] In some examples, at block 208, the external device processor 52 may generate an output based on the PCC capture type classification, the output including a capture threshold for achieving one or more types of capture with a VCS pacing pulse. The external device processor 52 may be configured to perform an analysis of the capture type classification in combination with associated pacing pulse output data received from the IMD 14 for determining a minimum pacing pulse output, e.g., a minimum pacing pulse amplitude for a given pulse width, at which a particular capture type classification is made. The capture threshold may be determined by the processor 52 based on a threshold confidence level for the capture type classification. For example, the capture threshold for a given capture type may be determined to be the minimum VCS pacing pulse amplitude that produces a given capture type classification with at least 70%, 80%, 90%, or other threshold confidence level. One or more capture thresholds may be reported for each of the one or more capture type classifications, e.g., in a GUI displayed by the display unit 54. By way of example, capture thresholds that may be determined by processor 52 and reported in the GUI of display unit 54 may include a SHP capture threshold, a NSHP capture threshold, and / or a VMO capture threshold.

[0169] In yet other examples, external device processor 52 can generate an output indicating a recommended pacing pulse output setting based on the PCC capture type classification at block 208. Based on one or more determined capture thresholds, external device processor 52 can generate an indication of a recommended pacing pulse output (e.g., a recommended pacing pulse amplitude for a given pulse width) that is a safety margin greater than the determined capture threshold to facilitate capturing at least a portion of the VCS for each VCS pacing pulse delivered by IMD 14.

[0170] At block 208, the output generated by the external device processor 52 may include a communication signal transmitted by the external device telemetry unit 58 to the IMD 14. The external device processor 52 may transmit a signal confirming VCS capture (e.g., SHP or NSHP capture) or indicating another capture type (e.g., VMO capture or LOC). The IMD 14 may receive the transmitted signal and respond by maintaining the VCS pacing control parameter at a current value when a desired capture type is confirmed (e.g., when SHP or NSHP capture is confirmed), or by adjusting the VCS pacing control parameter when an undesirable capture type is confirmed (e.g., when VMO capture or LOC is indicated). Additionally or alternatively, the external device processor 52 may control the telemetry circuit 58 to transmit a programming command to adjust the VCS pacing pulse output setting to a value determined based on an analysis of the output of the PCC 51.

[0171] In some cases, the output generated at box 208 may include a displayed notification of a recommended change in the VCS pacing electrode vector (e.g., when more than one pacing electrode vector is available for delivering VCS pacing pulses) or a programmed command transmitted to the IMD 14 to change the pacing electrode vector. For example, when the pacing pulse output is at the maximum available (or maximum acceptable) pacing output, if it is not a SHP or NSHP capture based on the capture classification output by the PCC 51, a different VCS pacing electrode vector may be selected to attempt to deliver a VCS pacing pulse that effectively captures at least a portion of the His-Purkinje system as an acceptable pacing pulse output. In other cases, when the IMD 14 receives a VMO or LOC notification, the IMD 14 may perform a capture threshold test, which may include classifying VCS pacing pulses delivered during the capture threshold test with different pacing pulse outputs for one or more different VCS pacing electrode vectors. For example, the following in conjunction with Figures 13 to 15 Describes a technique for performing capture threshold testing.

[0172] Fig. 6A and Figure 6B Schematics 250 and 251 of a PCC 51 configured to receive multiple unknown signal inputs for each VCS pacing pulse to be classified according to some examples. In these examples, the PCC 51 is configured to receive a first unknown signal input 252a obtained from a FF EGM signal and a second unknown signal input 252b obtained from a NF EGM signal. The two unknown signal inputs 252a and 252b are sensed during the same post-pacing time interval after the same VCS pacing pulse of unknown capture type, but using different sensing electrode vectors. One sensing electrode vector can be selected for obtaining a relatively more global or FF signal, and a second sensing electrode vector can be selected to sense a relatively more local or NF signal. The PCC 51 is previously trained to output a determined capture type classification 270 by applying a machine learning model to the FF EGM and NF EGM unknown signal inputs 252a and 252b. Each of the signal inputs 252a and 252b can be time aligned starting from a first sample point of the signal input, which can be a sample point before the VCS pacing pulse. For each unknown signal input 252a and 252b, the baseline may be zeroed by zeroing the first pre-pacing sample point of each signal segment.

[0173] exist Fig. 6A In FIG. 5 , two EGM signal segments 254a and 254b are received as first unknown signal input 252a and second unknown signal input 252b and are classified by PCC 51 as NSHP capture 272a with a confidence level 272b of 98%, as shown by output 270 in this illustrative example. Figure 6B In FIG. 2 , two EGM signal segments 256a and 256b received as first unknown signal input 252a and second unknown signal input 252b, respectively, are classified as VMO capture 274a with a confidence level 274b of 92%, as indicated by output 270. Fig. 6A and Figure 6B In the illustrative example of, the PCC classifier 51 has been trained using two channels of unknown cardiac electrical signal inputs sensed using two different sensing vectors following the same VCS pacing pulse for classifying a VCS pacing pulse as, for example, one of SHP capture, NSHP capture, VMO capture, or LOC. The PCC output 270 may be used by the external device processor 52 to generate data, information, or other output according to any of the examples described herein, such as for generating a display of annotated cardiac electrical signals with VCS pacing markers and capture type labels, determining and reporting capture thresholds, and / or determining and recommending VCS pacing pulse output settings.

[0174] Fig. 6A and Figure 6B 5 shows an example of a PCC 51 trained to classify VCS pacing pulses using two unknown cardiac electrical signal inputs. The two inputs may or may not be pre-pended or appended with other data, such as a value or values ​​indicating the pacing pulse voltage amplitude and / or pacing pulse width of the associated VCS pacing pulse. Figure 4 As described, PCC 51 may be trained to classify VCS pacing pulses using at least one unknown cardiac electrical signal input and one or more template beat inputs.

[0175] Figure 7 Flowchart 300 is a method that may be performed by a medical device system for classifying VCS pacing pulses according to another example of using template beat input. In this example, one or more template beats are established by the external device processor 52 or the IMD control circuit 80 for providing input to the PCC 51. At block 302, VCS pacing is delivered, for example, by the therapy delivery circuit 84 of the IMD 14 or by an external pulse generator. The one or more VCS pacing pulses may be delivered with a selected pacing pulse output. The external device processor 52 or the IMD control circuit 80 may receive a cardiac electrical signal sensed within at least a predetermined post-pacing time interval after at least one VCS pacing pulse delivered with the selected pacing pulse output.

[0176] At block 304, a template beat signal may be established by the medical device system processing circuitry (e.g., the external device processor 52 or the IMD control circuitry 80). In some examples, the template beat signal may be sampled starting from one or more sample points prior to a VCS pacing pulse and during a post-pacing time interval after a single paced beat. For example, a series of 2 to 10 VCS pacing pulses may be delivered with a selected pacing pulse output, and the template beat signal may be established using the nth beat (e.g., the 3rd, 5th, or other preselected nth paced beat). In other examples, a cardiac signal is obtained during at least a post-pacing time interval after each of two or more (e.g., two to twelve or three to six) VCS paced beats, wherein each VCS pacing pulse is delivered with the same pacing pulse output. Multiple cardiac signal segments may be ensemble averaged by the processor 52 or the control circuitry 80 to obtain a template beat signal for the selected pacing pulse output. In some examples, when multiple cardiac signal segments are used to create a template beat signal, each signal segment can be compared with at least one other signal segment to determine cross-correlation between the signal segments. Any signal segment with a low cross-correlation with other signal segments can be discarded.

[0177] In some examples, at block 304, a single template beat signal is established for one selected pacing pulse output (e.g., a pacing pulse amplitude that may be the highest pulse amplitude used during a pacing capture threshold search performed by the IMD 14). In other examples, multiple template beat signals are established for each of a plurality of selected pacing pulse outputs (e.g., two or more pacing pulse amplitudes having the same pulse width, two or more pulse widths having the same pulse amplitude, or a plurality of different combinations of pulse amplitudes and pulse widths). Thus, the control circuit 80 of the IMD 14 may be configured to control the therapy delivery circuit 84 to deliver a predetermined number (e.g., 1 to 10 or 3 to 8) of VCS pacing pulses with each selected pacing pulse output at block 302. The template beat signal may be established using cardiac electrical signals sensed during at least a post-pacing time interval following at least one VCS pacing pulse delivered with each pacing pulse output.

[0178] The template pulsation signal may be sensed from the same sensing electrode vector as the sensing electrode vector used to obtain the unknown signal input to be classified. For example, the template pulsation signal may be established using the FF EGM signal sensed between the VCS lead tip electrode 32 or the VCS ring electrode 34 and the IMD housing. The unknown signal input (or at least one unknown signal input when the PCC 51 receives multiple unknown signal inputs) may be sensed using the same FF EGM signal sensing electrode vector. In other examples, the template pulsation signal may be sensed using an EGM or ECG sensing electrode vector that is different from the sensing electrode vector used to obtain the unknown signal input.

[0179] It should be understood that the template beat signal established at block 304 may be of an unknown capture type. The template beat signal provided to PCC 51 with different selected pacing pulse outputs provides PCC 51 with information about how much the unknown signal input to be classified differs from the post-pacing signal segments acquired during different VCS pacing pulse amplitudes (or pulse widths). In some examples, the pacing pulse amplitude (or pulse width) may be known to PCC 51 as pre-added or appended data.

[0180] At box 306, a VCS pacing pulse is delivered that will be classified by the PCC 51. The VCS pacing pulse may be delivered for use with the currently programmed pacing pulse output for delivering the VCS pacing therapy. At other times, the pacing pulse output may be tested to deliver the VCS pacing pulse as part of a capture threshold search. At box 308, the VCS pacing pulse may be delivered by the therapy delivery circuit 84 of the IMD 14, and the post-pacing signal may be sensed by the IMD sensing circuit 86. The control circuit 80 may obtain the unknown input signal sensed during the post-pacing time interval as an EGM signal segment, and store the EGM signal segment in the memory 82 for transmission to the external device 50. In other examples, the control circuit 80 may transmit the EGM signal received from the sensing circuit 86 in real time or after being stored in the memory 82, and the external device processor 52 may extract the desired EGM signal segment from the transmitted EGM signal, the EGM signal segment including at least the post-pacing time interval. As described above, an EGM signal segment may include one or more pre-pacing sample points for providing a reference time point for aligning multiple input signals and for baseline zeroing.

[0181] At box 312, the unknown signal input obtained after the VCS pacing pulse to be classified and the previously established template beat input are passed to the PCC 51. The PCC 51 is pre-trained to classify the unknown signal input and the associated VCS pacing pulse according to the capture type using the number and type of inputs received. At box 314, the PCC 51 generates a capture classification output, which can include a confidence level for the capture classification, which is determined by applying the machine learning model to the unknown signal input and the template beat input. As described above, the capture classification can be one of an SHP capture, an NSHP capture, a VMO capture, or a LOC, or any other capture or beat type that the PCC 51 is trained to classify. In some examples, the PCC output includes a confidence level associated with each possible capture type classification, where the highest confidence level is the determined capture type classification of the delivered VCS pacing pulse. For example, the PCC output may be stored in external device memory 53 for use by external device processor 52 (or transmitted and stored in IMD memory 82 for use by control circuitry 80) in accordance with any of the example monitoring, reporting and / or pacing therapy control functions described herein.

[0182] Figure 8 is a conceptual diagram 350 of a processing circuit PCC 51 configured to analyze cardiac signals to determine a pacing capture classification according to another example. In this example, the PCC 51 is trained to classify a delivered VCS pacing pulse based on a FF EGM signal segment received as a first unknown signal input 352a, a NF EGM signal segment received as a second unknown signal input 352b, and a template beat input 360. Each of the first unknown signal input 352a and the second unknown signal input 352b is obtained from a respective FF EGM signal and a NF EGM signal sensed during at least a post-pacing time interval after delivery of the VCS pacing pulse being classified. The PCC 51 is trained to process three inputs by applying a machine learning model to the three inputs to determine a capture type classification output 370 for the VCS pacing pulse, for example, as described above in conjunction with Figure 7 Generally described. Each of the signal inputs 352a, 352b, and 360 may be time aligned starting from a first sample point of the corresponding signal input, which may be a pre-pacing sample point. The first sample point may be zeroed for each signal input to zero the baseline of each signal 352a, 352b, and 360.

[0183] Template pulse input 360 can be combined as above Figure 7The template beat input 360 is established from a cardiac electrical signal (e.g., one of the FF EGM signal or the NFEGM signal) sensed during one or more post-pacing time intervals following a VCS pacing pulse delivered with a predetermined selected pacing pulse output. The template beat input 360 may be established for a VCS pacing pulse having a relatively high pacing pulse amplitude (e.g., 4 volts or higher or 5 volts or higher, with a pulse width of at least 0.25 ms to 0.5 ms, as an example). The template beat may be established for a cardiac signal sensed during VCS pacing with a relatively high pacing pulse output, such that the template beat corresponds to a QRS waveform when at least a portion of the VCS is expected to be captured. However, the actual capture type of the template beat may be unknown to the PCC 51.

[0184] In other examples, the PCC 51 can be trained to determine a capture type classification output 370 based on applying a machine learning model to a template beat input 360 and only one unknown signal input (e.g., a FF EGM signal input 352a or a NF EGM signal input 352b or an ECG signal input). Classification of a VCS pacing pulse and subsequent cardiac electrical signal waveform does not necessarily require multiple unknown signal inputs. In one example, the PCC 51 is trained to classify an unknown signal input 352a and a template beat input 360, the unknown signal input being received as a FF EGM signal input obtained after a VCS pacing pulse of unknown capture type, the template beat input being established from a sensed FF EGM signal after at least one VCS pacing pulse delivered using a 5.0 volt pulse amplitude and a 0.2 ms to 0.5 ms pulse width (or any other selected pacing pulse output).

[0185] PCC 51 may additionally receive data inputs indicative of the pacing pulse voltage amplitude of the VCS pacing pulse being classified and / or the pacing pulse voltage amplitude of the template beat input. In various examples, the pacing voltage amplitude may be pre-added or appended to at least one of the unknown signal inputs 352a or 352b and / or the template beat input 360, or provided as a separate input signal. Additionally or alternatively, PCC 51 may receive data inputs indicative of signal characteristics of the unknown input signals 352a and / or 352b and / or characteristics of the template beat input 360, as described above in conjunction with Figure 4 In general.

[0186] Fig. 9380 is a diagram of PCC 51 according to one example. In this example, PCC 51 can be trained to provide a capture type classification output 370 based on an unknown signal input 382a, a derivative signal input 382b, and a template beat input 390. In various examples, the unknown signal input 382a can be a FF EGM signal, a NF EGM signal, or an ECG. In one example, the unknown signal input 382a can be derived from, for example, a VCS lead tip electrode 32 to an IMD housing 15 (e.g., Figures 1A to 1C ) is obtained by sensing the FF EGM signal from the tip to the shell sensing electrode vector. The derivative signal input 382 can be determined by the control circuit 80 or the external device processor 52 as the derivative of the unknown signal input 382a. In some examples, the derivative signal input 382 is a first-order derivative signal, but in other examples, the derivative signal input 382 can be determined as a higher-order derivative, such as a second-order derivative, a third-order derivative, or a higher-order derivative. The processing circuit system included in the control circuit 80 may include a differentiator 384, which is used, for example, to obtain a derivative signal from the unknown signal input 382a. The differentiator 384 can be implemented in hardware, firmware, or software in the control circuit 80 or the external device processor 52 to determine the first-order derivative of the unknown signal input 382a to be provided as the derivative signal input 382b. It is contemplated that in some examples, PCC 51 is trained to classify VCS pacing pulse capture types in response to receiving only derivative signal input 382 or any combination of derivative signal input 382 and one or more unknown signal inputs, other derivative signal inputs, and / or template beat inputs.

[0187] So, in Figure 7 At block 308 of FIG. 1 , the control circuit 80 or external device processor 52 may be operable to obtain a post-pacing signal for providing input to the PCC classifier 51 by obtaining an unknown input signal 382a sensed within at least a post-pacing time interval following the VCS pacing pulse being classified, and determining a derivative of the unknown input signal 382a. The template beat input 390 may be combined with the VCS pacing pulse input 390 as described above. Figure 7 The foregoing is pre-established such that the control circuit 80 or external device processor 52 is configured to provide each of the three inputs 382a, 382b and 390 to the PCC 51 for use in generating the capture type classification output 370.

[0188] Although in Fig. 9390, but it is contemplated that in addition to or as an alternative to the derivative signal 382b determined from the unknown signal input 382a, the control circuit 80 or external device processor 52 may be configured to determine a derivative of the template beat input 390. In some examples, the derivative of the established template beat (rather than the template beat signal itself) may be provided to the PCC 51 as input 390. In other examples, both the template beat signal and the derivative of the template beat signal may be provided as inputs to the PCC 51.

[0189] In addition, although Fig. 9 382a, but it should be understood that the first-order derivative signal and the second-order derivative signal or other higher-order derivative signal determined from the unknown input signal 382a may be provided as input to the PCC 51. Additionally or alternatively, the first-order derivative signal and the second-order derivative signal or other higher-order derivative signal determined from the template beat signal 390 may be provided as input to the PCC 51. Thus, the PCC 51 may be trained to receive one or more derivative signals determined from the unknown input signal 382a and / or one or more derivative signals determined from the template beat input 390 as additional input signals. The derivative signal may be determined as an n-order difference signal from the corresponding unknown input signal or template beat signal.

[0190] It should be appreciated that additional data inputs such as pacing voltage amplitude, pacing pulse width, or other VCS pacing pulse information, and / or one or more signal characteristics of the unknown signal input 382a, derivative signal input 382b, template beat input 390, and / or template beat derivative input (if included) may be determined by the control circuit 80 or external device processor 52 and provided as inputs to the PCC 51. The additional data may be provided as a separate input, and / or pre-added or appended to one of the signal inputs 382a, 382b, or 390. The signal characteristics determined may be maximum peak amplitude, peak-to-peak amplitude, maximum (positive) slope, minimum (negative) slope, signal width, signal area, time between maximum and minimum slopes, activation time from the delivered VCS pacing pulse to the maximum peak of a signal segment or another reference point, or any other signal characteristic listed herein.

[0191] Fig.10 is a flow chart 301 of a method for determining VCS pacing capture type according to another example. Fig.10 The same numbered boxes shown in Figure 7 The same numbered boxes as shown and described above. However, in Fig.10In the example of FIG. 5 , after acquiring a post-pacing signal for delivering at least one unknown signal input to the PCC 51 at block 308, the control circuit 80 or the external device processor 52 may determine a template difference input as the difference between the unknown signal input and the established template beat signal at block 310. At block 312, the PCC 51 may receive at least one unknown input signal to be classified (e.g., at least one post-pacing FF EGM signal segment, NF EGM signal segment, and / or ECG signal segment, or any combination thereof) (along with the delivered VCS pacing pulse), and the at least one template difference input determined at block 310.

[0192] In various examples, a template difference input between each of the plurality of template beat signals established at block 304 and an unknown input signal sensed from the same sensing electrode vector may be determined. At block 312, the plurality of template difference inputs may be received by the PCC 51 for use in classifying the capture type of the VCS pacing pulse delivered at block 306. When the PCC 51 is trained using a plurality of unknown signal inputs, each of which is obtained from a different sensing electrode vector, the PCC 51 may additionally receive a plurality of template difference inputs corresponding to the difference between the template beat signal established with a predetermined pacing pulse output and the post-pacing unknown signal input obtained at block 308 and sensed from the same sensing electrode vector as the established template beat signal. The PCC 51 may be trained to classify the VCS pacing pulse capture type using at least one unknown input signal and at least one template difference signal. In some examples, the template beat signal used to determine the difference signal may additionally be provided to the PCC 51 as an input signal.

[0193] Fig.114 is a conceptual diagram 450 of a PCC 51 according to yet another example. In this example, the PCC 51 receives at least one unknown input signal 452 derived from a cardiac electrical signal sensed during at least a post-pacing time interval after delivery of a VCS pacing pulse to be classified (shown as an unknown signal segment 451). A template beat signal 458 was previously established by the control circuit 80 or the external device processor 52 (as described above) and may be provided as a template beat input 460. The template beat signal 458 may be determined from a post-pacing signal sensed from the same sensing electrode vector as the unknown signal segment 451 after a VCS pacing pulse delivered at a predetermined pacing pulse output (e.g., a pulse amplitude of 5.0 volts or higher or another relatively high pacing pulse output expected to typically capture cardiac tissue), which predetermined pacing pulse output may include at least a portion of the VCS. In other examples, template beat signal 458 may correspond to a post-pacing signal obtained after a VCS pacing pulse delivered with any predetermined or preselected pacing pulse output. Unknown signal segment 451 and template beat signal 458 may be time aligned starting from the first sample point of the corresponding signal segment, which may be a pre-pacing sample point. The baselines of unknown signal segment 451 and template beat signal 458 may be zeroed by zeroing the first sample point of each signal. The time aligned and zeroed signals 451 and 460 may be provided to PCC 51 as multi-channel data point inputs 452 and 460.

[0194] In addition, Fig.11 In an example of , the control circuit 80 or the external device processor 52 may be configured to determine the template difference signal 461 by subtracting the template beat signal 458 from the unknown signal segment 451 (or vice versa) to obtain the template difference signal 461 .

[0195] The control circuit 80 or external device processor 52 may include a summation (or difference) circuit that may be implemented as hardware, firmware, and / or software configured to determine the difference between each sample point of the time-aligned and zeroed unknown signal segment 451 and the template beat signal 458. Fig.11 In the example of , the control circuit 80 or external device processor 52 includes a summing circuit 464 that is configured to perform a summation of a positive input of the unknown signal input 452 and a negative input of the template beat signal 458 (or vice versa) for providing a template difference input 462 to the PCC 51.

[0196] The PCC 51 is trained to output a capture type classification 470 based on applying a machine learning model to at least an unknown signal input 452 and a template difference input 462. In some examples, such as Fig.11As shown in , the PCC 51 may additionally receive a template beat input 460 and be trained to output a capture type classification 470 based on the three inputs of the unknown signal input 452, the template beat input 460, and the template difference input 462. However, when the template difference input 462 is provided to the PCC 51, the template beat input 460 may be optional and may be omitted in some examples.

[0197] As generally described above, the control circuit 80 or external device processor 52 can be configured to provide one or more additional data inputs (such as signal characteristics determined from one or more of the unknown signal segment 451, the template beat signal 458, and / or the template difference signal 461), and / or the pacing pulse output of the delivered VCS pacing pulse to be classified or the pacing pulse output associated with the template beat signal to the PCC 51. The PCC 51 can be trained to provide a capture type classification output 470 based on the input signals 452, 460, and 462 and one or more data inputs that provide information about specific signal characteristics of the delivered VCS pacing pulse and / or the unknown signal segment 451, the template beat signal 458, and / or the template difference signal 461 to be classified. The additional data inputs can be received by the PCC 51 as pre-added or supplemental data to one of the inputs 452, 460, or 462, or as one or more separate input signals.

[0198] In the illustrative example, the PCC 51 may receive the characteristics of the template beat signal 458, and when the template beat characteristics are received, the template beat input 460 may or may not be received. For example, the control circuit 80 or the external device processor 52 may determine the signal characteristics of the template beat signal 458, such as the maximum peak amplitude, the signal width, the maximum slope, and / or the time from the VCS pacing pulse to the maximum peak, the maximum slope, or any other signal characteristic or combination of characteristics of the template beat signal 458. In some examples, the PCC 51 may receive the determined template beat signal characteristics as inputs along with at least the unknown signal input 452 and the template difference input 462 for determining the capture type classification output 470.

[0199] Fig.12480 is a conceptual diagram of a PCC 51 according to another example. In this example, the PCC 51 is trained to classify an unknown signal input 452 (and a corresponding delivered VCS pacing pulse) based on the unknown signal input 452 and a plurality of template difference inputs. In the illustrated example, the PCC 51 receives three template difference inputs 462, 466, and 468. Each template difference input 462, 466, and 468 can be determined by the control circuit 80 or the external device processor 52 as a difference between the unknown signal input 452 and a template beat signal established with a corresponding predetermined pacing pulse output during VCS pacing. The four signal inputs (the unknown signal input to be classified and the three template difference inputs) can be combined into a 4-channel data point represented by the inputs 452, 462, 466, and 468 received by the PCC 51.

[0200] In the example shown, the processing circuit system of the control circuit 80 or the external device processor 52 can determine three different template beat signals corresponding to three predetermined different VCS pacing pulse outputs, such as a relatively high output, a relatively low output, and an intermediate output that can be in an intermediate range between the high output and the low output. In other examples, two template beat signals can be established for the relatively high output and the relatively low output. In still other examples, more than three template beat signals can be established, including a high output template beat signal, a low output template beat signal, and two or more intermediate output template beat signals. For each template beat signal, the processing circuit system can determine a template difference signal as the difference between the unknown signal acquired during the post-pacing time interval of the VCS pacing pulse to be classified and each template beat signal in the established template beat signal. As described above, the unknown signal and the template beat signal can be time-aligned starting from the first sample point of the corresponding signal segment (which can be a pre-pacing sample point), and the baseline of each signal in the signal can be zeroed by zeroing the first sample point of each signal before determining the corresponding template difference signal.

[0201] Each template difference signal can be provided as an input signal to the PCC 51. Fig.12 As shown, a high output template difference input 462, an intermediate output template difference input 466, and a low output template difference input 468 may be provided as inputs to the PCC 51. In one example, the high output template beat signal is obtained during VCS pacing delivered at the highest test pacing pulse output (e.g., between 5 volts and 8 volts pulse amplitude) used during VCS pacing capture threshold testing of the IMD 14. The low output template beat signal is obtained during VCS pacing delivered at the lowest test pacing pulse output (e.g., between 0.5 volts and 1.5 volts pulse amplitude) used during VCS pacing capture threshold testing. The intermediate template beat signal may be determined at a pacing pulse output intermediate the high output and the low output.

[0202] In this example, the PCC 51 is trained to determine a capture type classification output 470 based on four inputs 452, 462, 466, and 468. In some examples, one, two, or all three of the template beat signals used to determine the template difference inputs 462, 466, and 468 may be provided as inputs to the PCC 51 in addition to the template difference inputs. When the PCC 51 is trained to use at least one template difference input to generate the capture type classification output 470, any differences between the waveform morphology of the unknown input signal to be classified and the established template beat signals may be learned more quickly or more efficiently according to a machine learning algorithm during the training of the PCC 51. The difference between the unknown signal input 452 and any one of the template beat signals established at a plurality of different pacing pulse amplitudes is provided directly to the PCC 51 as a template difference input.

[0203] As described above, any of the signal inputs 452, 462, 466, and / or 468 may be pre-pended or supplemented with additional data, such as an associated pacing voltage amplitude, pacing pulse width, specific signal characteristics, and / or patient information (e.g., age, gender, etc.), as described in any of the examples given above. In the illustrated example, each signal input 452, 462, 466, and 468 is pre-pended with the VCS pacing pulse voltage amplitude 482, 484, 486, and 488 encoded in the corresponding signal input 452, 462, 466, and 468, respectively. In other examples, the PCC 51 may receive additional pacing pulse-related data and / or signal characteristic-related data as separate inputs without being pre-pended or supplemented to the signal inputs. Furthermore, it should be understood that any of the examples of one or more unknown signal inputs, one or more unknown signal derivative inputs, one or more template beat inputs, one or more template difference inputs, one or more VCS pacing pulse related data inputs, and / or one or more signal feature data inputs may be received by the PCC 51 in any selected combination during a machine learning training session and subsequently during PCC operation for classifying VCS pacing pulses. Figure 4 , Fig. 6A , Figure 6B , Figure 8 , Fig. 9 , Fig.11 and Fig.12The conceptual diagram of PCC 51 presented in is illustrative in nature and is not intended to limit the specific types and combinations of inputs that PCC 51 may receive during training and during subsequent VCS pacing pulse classification operations. The various input signals or channels shown or described herein may be received by PCC 51 in combinations different than the specific combinations shown in the figures and described herein, and / or one or more input signals shown in the illustrative examples may be omitted.

[0204] In addition, it should be understood that various operations can be performed on one or more unknown input signals and / or one or more template beat signals by the medical device system processing circuit system to obtain input signals provided to PCC 51, and as examples, these operations can include addition, subtraction, determination of integrals and / or determination of derivatives of cardiac electrical signal segments. In some cases, two or more operations can be performed. For example, a template difference signal can be determined between the unknown signal input and the established template beat signal, and a derivative signal can be determined from the template difference signal to provide the template difference derivative input to PCC51.

[0205] Fig.13 4 is a flow chart of a method for generating a VCS pacing pulse capture type classification during a pacing capture threshold test according to some examples. At block 402, IMD 14 may perform a capture threshold test. Control circuitry 80 may control therapy delivery circuitry 84 to use, for example, Figures 1A to 2B The VCS pacing pulse is delivered to the His-Purkinje pacing site with a selected pacing electrode vector of any one of the examples. The therapy delivery circuit 84 delivers the VCS pacing pulse at a plurality of pacing pulse output settings. For convenience, the capture threshold search is described as being performed by changing the pacing pulse amplitude using a fixed pacing pulse width. However, it should be appreciated that the capture threshold search may be performed by changing the pacing pulse width using a fixed pacing pulse amplitude, or by changing both the amplitude and the pulse width. In one example, one or more VCS pacing pulses are delivered at each of a plurality of pacing pulse amplitudes that start from a relatively high pulse amplitude (e.g., 5.0 volts or higher) and are reduced to a relatively low pulse amplitude (e.g., 1.0 volts or lower). As an example, the VCS pacing pulse amplitude may be reduced in steps of 0.25 volts, 0.5 volts, or 1 volt. In other cases, the pacing pulse amplitude may start at a low amplitude and be increased or randomly changed.

[0206] At block 404, the processing circuitry implementing the PCC 51 receives capture threshold test data. For convenience, the external device processor 52 is described as receiving capture threshold test data. However, in other examples, the control circuitry 80 may be configured to detect the capture threshold test data. Fig.13The analysis of the capture threshold test data is performed by a process of the IMD 14. During the capture threshold search performed by the IMD 14, the sensing circuit 86 communicates one or more EGM signals to the control circuit 80, which can be transmitted to the processor 52 via telemetry. At box 404, the processor 52 can receive the capture threshold test data by receiving at least one EGM signal (e.g., an FF EGM signal) along with associated VCS pacing pulse timing markers and pacing pulse output data, which can be transmitted in real time by the IMD 14. Alternatively, a segment of the FF EGM signal representing the FF EGM signal during at least one post-pacing time interval following a VCS pacing pulse delivered at each pacing voltage amplitude applied during the capture threshold test can be stored in the IMD memory 82 and subsequently transmitted to the external device 50.

[0207] The external device processor extracts the inputs required by PCC 51 from the received capture threshold test data. For example, at block 406, external device processor 52 may create one or more template beat signals from the EGM signals received from IMD 14 for one or more preselected pacing pulse voltage amplitudes applied during the capture threshold test. The template beat signals may be combined with the above-described Fig.12 Generally built as described.

[0208] At box 408, processor 52 may extract a first unknown signal input to be classified by PCC 51. The unknown signal input may be extracted from the EGM signal received from IMD 14 during at least one post-pacing time interval after capturing the test pacing pulse. At box 410, processor 52 may determine a template difference input for each template beat signal determined at box 406. Processor 52 may determine the difference between each time-aligned sample point of the unknown signal input and the template beat signal to determine the template difference input. As described above, multiple template difference inputs may be determined and provided to PCC 51 in a multi-channel input. At box 412, the unknown signal input and the determined template difference input are provided as inputs to PCC 51.

[0209] The external device processor 52 may receive or determine any other PCC input signals required for capture type classification based on input signals previously trained on the PCC 51. According to any of the above examples, the PCC input signals may be received or determined by the processor 52 to be provided as input to the PCC 51.

[0210] At box 412, the processor 52 determines a capture type classification by applying the machine learning model of the PCC 51 to the unknown input signal and any other input signals. The PCC 51 can generate a capture type classification output for the unknown signal input (and the associated VCS capture test pacing pulse). The capture type classification output by the PCC 51 can include a confidence level for the capture type classification. In some examples, the PCC 51 can output a confidence level for each possible type of capture classification that the PCC 51 is trained to identify.

[0211] At block 414, the external device processor 52 receives the PCC output and may store the VCS pacing pulse capture type classification and the associated pacing pulse amplitude for use in generating an annotated display of the capture threshold test results. If another VCS pacing pulse remains to be classified from the capture threshold test (as determined by the processor 52 at block 416), the processor 52 may return to block 408 to obtain the PCC input signal required to determine the capture type classification for the next VCS pacing pulse by applying the machine learning model of the PCC 51 until all VCS pacing pulses of the capture threshold test have been classified.

[0212] At box 418, the external processor 52 can use the PCC output classification and the stored EGM signal data and / or VCS pacing pulse output data to generate a representation of the capture threshold test results. The external processor 52 can display at least one cardiac electrical signal and / or VCS pacing marker channel to present the results of the capture threshold test to a clinician or other user. The EGM signal received from the IMD 14 for obtaining input to the PCC 51 can be displayed and annotated with a capture type classification along with VCS pacing markers labeled according to the pacing pulse output (e.g., pulse amplitude and / or pulse width). Additionally or alternatively, any other ECG and / or EGM signals sensed during the capture threshold test and received by the processor 52 can be annotated with the capture type classification or displayed in time alignment with the capture type classification.

[0213] In some examples, all capture threshold test pulses are classified by the processor 52, and after the capture type classification of all capture threshold test VCS pacing pulses is completed, the capture test results are generated for display at block 418. In other examples, an annotated EGM and / or ECG signal may be displayed when the capture test classification is determined from the output of the PCC 51. In various examples, the display unit 54 may display the results of the capture threshold test in a rolling time-based signal, in a table, and / or in a graph, e.g., listing and / or plotting each VCS pacing pulse amplitude and the associated capture type classification.

[0214] Fig.14 5 is a conceptual diagram 500 of cardiac electrical signals and marker channels that may be generated by an external device processor 52 for display by a display unit 54 using a capture type classification output by a PCC 51, according to one example. Diagram 500 may represent a GUI that may be displayed by the display unit 54. The PCC 51 may be trained to output a capture type classification for a VCS pacing pulse according to any of the examples described above. The processor 52 and the display unit 54 may, for example, be displayed via Figure 1A The interface 55 shown in receives one or more ECG signals 502 for displaying ECG signals sensed simultaneously with the EGM signals sensed by the IMD 14. Atrial EGM signals 504 and ventricular EGM signals 510 can be sensed by the IMD sensing circuit 84 and transmitted from the IMD 14 to the external device 50 via telemetry. The atrial EGM signal 504 can be sensed by the RA lead electrodes 20 and 22. The ventricular EGM signal 510 can be a FF EGM signal (e.g., a tip-to-shell signal or a ring-to-shell signal as shown in this example) or a NF EGM signal (e.g., a tip-to-ring signal). The ventricular EGM signal 510 can be a signal sensed by the sensing circuit 86 using a sensing electrode vector, which is used to obtain an unknown signal input and one or more template beat signals for obtaining an input signal received by the PCC 51. However, external device 50 may receive and display cardiac electrical signals, including ECG and / or EGM signals, in addition to or in addition to the signals used by external device 50 or processing circuitry of IMD 14 to derive the signal input received by PCC 51 .

[0215] The marker channel 506 includes an atrial pacing pulse (AP) marker 520 and a VCS pacing pulse (VP) marker 522 to mark the relative timing of the respective atrial pacing pulses and VCS pacing pulses delivered by the IMD 14. Each VCS pacing pulse can be delivered with a shortened AV interval from the atrial pacing pulse during the capture threshold test (to pace the ventricle earlier than any inherently conducted depolarization). Each VCS pacing pulse marker 522 can be annotated with a pacing pulse amplitude (A), a pulse width (PW), a capture type classification output by the PCC 51, and a confidence level expressed as a percentage (if available). The external device processor 52 can generate a capture type annotation 524 as an output for display to the user in response to the capture type classification (and corresponding confidence level) made by the PCC 51, which can be combined with at least one time-aligned cardiac electrical signal.

[0216] In some examples, external device processor 52 can determine a VCS pacing capture threshold based on the PCC output. For example, processor 52 can determine the lowest VCS pacing pulse amplitude that occurs for each type of capture type classification output by PCC 51 during the capture threshold test. In the illustrated example, NSHP capture is determined for pacing pulse amplitudes of 2.5 volts and 2.25 volts, according to annotation 524. When the pacing pulse amplitude is reduced to 2.0 volts, VMO capture is determined, as indicated by annotation 524. The lowest pulse amplitude for determining VMO capture is not shown in FIG. Fig.14 However, based on the PCC output during the capture threshold test, the external device processor 52 may determine the SHP capture threshold, the NSHP capture threshold, and / or the VMO capture threshold. In the hypothetical example shown, the processor 52 may determine the NSHP capture threshold to be 2.25V and the VMO capture threshold to be 1.5V (in Fig.14 The determined capture threshold may be reported in a capture threshold display window 530.

[0217] The determined capture threshold may be the lowest VCS pacing pulse amplitude at which at least X of Y capture classifications are the same and the confidence level is at least a predetermined threshold percentage. For example, during the capture threshold test, each test pacing output may be delivered for at least 3 VCS pacing pulses, 5 VCS pacing pulses, or another selected number of test pacing pulses. Processor 52 may determine a capture type classification that is the same as the capture type classification of the test pacing pulse output when at least 2 / 3 or 3 / 5 or other selected portion of the VCS pacing pulses delivered with the test output are classified as the same and each of the same classifications has a confidence level of at least 70% (or any other predetermined threshold percentage). The lowest test output that results in the same capture type classification may be determined as the capture threshold for that capture type.

[0218] Depending on the progression of capture types as pacing pulse output increases or decreases, processor 52 may not determine one or more capture types based on the output of PCC 51. For example, in some cases, NSHP capture may occur at the highest pacing pulse output, and because pacing pulse output decreases before VMO capture is lost, capture of the His-Purkinje system may be lost. In this case, PCC 41 may output NSHP capture classification, VMO capture classification, and LOC classification, but not SHP classification. Processor 52 may determine NSHP capture thresholds and VMO capture thresholds, such as Fig.14However, in other examples, ventricular myocardial capture may be lost prior to selective capture of the His-Purkinje system. In this case, processor 52 may determine the NSHP capture threshold and the SHP capture threshold, and the VMO capture threshold may be undeterminable because capture of both the His-Purkinje system and the ventricular myocardium occurs together at or above the NSHP capture threshold, and only SHP capture may occur below the NSHP capture threshold, where LOC occurs at a pacing pulse amplitude less than the SHP capture threshold.

[0219] In other examples, SHP capture may occur at all VCS pacing pulse outputs until LOC is determined, such that only the SHP capture threshold is determined. In other examples, NSHP capture may occur at all VCS pacing pulse outputs until LOC is determined, such that only the NSHP capture threshold is determined. Thus, the capture thresholds of different capture classification types, which may or may not be determined, may vary in different patients and / or over time.

[0220] The external device processor 52 may be configured to output a recommended operational VCS pacing output setting based on the capture type classification output by the PCC 51, which may be displayed in a recommended settings window 532. The processor 52 may determine the recommended operational VCS pacing output setting to be a pacing pulse amplitude and / or width that is a safety margin, e.g., 0.25 volts to 2.0 volts greater than a determined SHP or NSHP capture threshold. Fig.14 In the illustrative example of , the processor 52 can generate a recommended operational VCS pacing amplitude setting of 2.75 volts, which is a safety margin of 0.5 volts greater than the NSHP capture threshold of 2.25 volts. In some examples, the user can review and accept or confirm the recommended operational VCS pacing output setting. For example, the user can click on the "Confirm" button of the recommended settings window 532 to program the recommended operational settings in the IMD 14, which can be transmitted by the external device telemetry unit 58 for receipt by the IMD 14. In other cases, the user can select and program the operational VCS pacing pulse amplitude and / or pulse width based on the visual representation of the cardiac electrical signal and the capture type classification presented in the GUI by the display unit 54. The IMD therapy delivery circuit 84 can then be controlled to deliver VCS pacing pulses according to the programmed operational VCS pacing pulse output.

[0221] Fig.15 600 is a flow chart of a VCS capture monitoring method that may be performed by a processing circuit system of a medical device system according to some examples. For simplicity, Fig.15The process of flowchart 600 may be performed in whole or in part by an external computing system such as a processor, processor, or processor 10. The process of flowchart 600 may be performed by a processor 100, processor 101, processor 102, processor 103, processor 104, or processor 105. The process of flowchart 600 may be performed by a processor 100, processor 104, processor 106, processor 107, processor 108, processor 109, processor 110, or processor 111. The process of flowchart 600 may be performed in whole or in part by an external computing system such as a processor 100, processor 104, processor 105, processor 106, processor 107, processor 108, processor 109, processor 109, processor 101, processor 109, processor 101, processor 104, processor 105, processor 106, processor 108, processor 109, processor 109, processor 101, processor 109, processor 101, processor 109, processor 101, processor 10 Figure 1A The external device 50 shown in FIG. 1 is executed.

[0222] At block 601, control circuit 80 may determine whether it is time to check for VCS capture. Control circuit 80 may be configured to check for VCS capture on a regular or scheduled basis, which may be once a day, once an hour, once a minute, or other scheduled frequency. In one example, control circuit 80 may be configured to perform the process of flowchart 600 once a day, for example, at a scheduled time during the night hours (which may be between midnight and 4 a.m., as an example) or at a time when the patient is expected to rest. In other examples, control circuit 80 may be configured to check for VCS capture relatively more frequently, for example, after every predetermined number of delivered VCS pacing pulses, for example, after every 5 to 120 VCS pacing pulses. In some examples, control circuit 80 may be configured to determine VCS capture on a beat-to-beat basis, although the processing power required to classify the type of VCS capture using PCC 51 on a beat-to-beat basis may shorten the life of IMD power supply 98 in an unacceptable manner.

[0223] Additionally or alternatively, the control circuit 80 may perform a VCS capture check on a triggered basis, e.g., in response to detecting a change in a lead impedance measurement, a pacing mode switch, or in response to detecting a change in a post-pacing QRS morphology or in response to any other detected triggering event that may indicate a possible change in the type of VCS pacing pulse capture. In some examples, the control circuit 80 may determine that it is time to check VCS capture based on a determination of a change in a post-pacing QRS signal morphology or signal characteristic (e.g., peak amplitude, peak slope, activation time, signal width, or other signal characteristic), which may be monitored on a beat-to-beat basis or on a relatively more frequent basis than performing a VCS capture check. At block 601, in response to detecting a change in a post-pacing EGM signal characteristic or QRS morphology compared to a previously determined post-pacing EGM signal characteristic or QRS morphology or compared to a predetermined or previously established threshold applied to the post-pacing EGM signal characteristic, the control circuit 80 may determine that it is time to perform a VCS capture check.

[0224] When it is determined at block 601 that it is time to check for VCS capture, at block 602, the control circuit 80 may wait for the next VCS pacing pulse delivered by the therapy delivery circuit 84. The VCS pacing pulse may be delivered at the current pacing pulse output settings (e.g., the currently programmed pulse amplitude and pulse width). In some examples, the control circuit 80 may control the therapy delivery circuit 84 to deliver the VCS pacing pulse with a shortened pacing interval (e.g., with a shortened AV delay or a shortened ventricular pacing rate interval) to avoid a fusion beat or intrinsic ventricular depolarization from occurring prior to the delivered VCS pacing pulse.

[0225] At block 604, the control circuit 80 obtains input signals required for the PCC 51, which correspond to input signals used to train the PCC 51. As described above, the PCC inputs include at least one unknown signal input obtained from a sensing electrode vector signal sensed during at least one post-pacing time interval following the VCS pacing pulse delivered at block 602. The control circuit 80 obtains any additional input signals required, which may include a plurality of unknown signal inputs obtained from EGM signals sensed during the post-pacing time interval, one or more derivatives of the unknown signal inputs, one or more template beat inputs, and / or one or more template difference inputs. In addition, the control circuit 80 may obtain additional input data related to the associated VCS pacing pulse output and / or a particular signal feature determined from any of the input signals as described in the foregoing examples herein.

[0226] At box 606, the PCC 51 receives the input signal and outputs a capture type classification based on the machine learning model of the PCC 51 applied to the input signal. At box 608, the control circuit 80 determines whether the output capture type is the expected capture type. For example, the control circuit 80 can determine whether the output capture type is a SHP capture type or a NSHP capture type. In some examples, the expected output capture type can be a selective or non-selective capture of the targeted portion of the His-Purkinje system, for example, a selective or non-selective LBB capture without RBB block. In some cases, the expected output capture type can be a capture type previously determined to occur when a VCS pacing pulse is delivered using the same pacing pulse output setting as the VCS pacing pulse delivered at box 602. When the capture classification is the expected capture type, the control circuit 80 can return to box 601 to wait for the next VCS capture check. When the output capture type classification received from the PCC 51 is not the expected capture type (e.g., not SHP capture or NSHP capture), LOC capture or VMO capture can occur. As such, control circuit 80 may proceed to block 610 to perform a capture threshold test.

[0227] At block 610, the control circuit 80 may control the therapy delivery circuit 84 to deliver a VCS pacing pulse at each of a plurality of different pacing pulse outputs, perform a capture threshold test (e.g., as described above in conjunction with Fig.13 86 and obtain the necessary input signals and data for the PCC 51 after each VCS pacing pulse. The control circuit 80 may store the PCC output capture type classification associated with each VCS pacing pulse tested for the capture threshold in the memory 82. For example, the therapy delivery circuit 84 may deliver VCS pacing pulses at each pacing voltage amplitude within a test range, such as within a range starting from 5 volts down to 1 volt or until the LOC is the determined capture type output by the PCC 51. The therapy delivery circuit 84 may deliver one or more VCS pacing pulses per pacing pulse amplitude. For example, three to five VCS pacing pulses may be delivered per pacing pulse amplitude. The capture type classification may be output by the PCC 51 for each VCS pacing pulse and may include a confidence level for the capture type classification.

[0228] The control circuit 80 may store each capture type classification, associated confidence level, and associated VCS pacing pulse amplitude (and / or pulse width) as capture threshold test result data in the memory 82. At block 612, the control circuit 80 may be configured to determine a capture threshold for one or more capture types based on the stored capture threshold test result data. In various examples, when more than one capture type classification is stored in the memory 82 for each VCS pacing pulse output, the control circuit 80 may determine a capture type for the VCS pacing pulse output based on at least a threshold number percentage of the capture types being classified as one type and / or having at least a threshold confidence level.

[0229] For example, if the therapy delivery circuit 84 delivers three VCS pacing pulses at each pacing pulse amplitude, the control circuit 80 may require that at least two of the capture type classifications output by the PCC 51 be the same classification in order to determine the capture type for the associated VCS pacing pulse amplitude. The control circuit 80 may determine the capture type for a given pacing pulse output when the same capture type classification is made for at least 50%, 60%, 75%, 100%, or other threshold percentage of the capture type classifications output by the PCC 51 for the given pacing pulse output. In some examples, the control circuit 80 may additionally require that the capture type classifications meet a threshold confidence level. For example, the control circuit 80 may determine when at least 50% (or a higher percentage) of the capture type classifications output by the PCC 51 for multiple VCS pacing pulses with the same pacing pulse output are the same capture type, and each of the matching capture types has a confidence level of at least 80% or at least 90%. A matched capture type classification that satisfies the threshold confidence level may be stored in memory 82 as the capture type for the associated pacing pulse output. Control circuitry 80 may apply other criteria or thresholds to the PCC 51 output for determining and storing the capture type associated with each VCS pacing pulse output delivered during the capture threshold test.

[0230] Once the control circuit 80 determines the capture type for each pacing pulse output based on the PCC 51 output, the control circuit 80 may determine a capture threshold for each capture type at block 612. The capture threshold may be the lowest pacing pulse amplitude for which a particular capture type is determined (when a fixed pulse width is used). In various examples, the control circuit 80 may determine the NSHP capture threshold, the SHP capture threshold, and / or the VMO capture threshold, as generally described above, based on the capture type determined for each pacing pulse output used during the capture threshold test. As previously indicated herein, the capture types detected during the capture threshold test may vary between patients and / or over time, and not all capture types and associated capture thresholds may be determined.

[0231] At box 614, the control circuit 80 can select an operational pacing pulse output based on the determined capture threshold. The selected operational pacing pulse output can be a safety margin higher than the capture threshold of the desired capture type. At box 614, the control circuit 80 can adjust the pacing pulse amplitude and / or the pacing pulse width to the selected operational pacing pulse output. In some examples, the control circuit 80 can set the operational pacing pulse amplitude to an amplitude greater than the SHP capture threshold or the NSHP capture threshold by a safety margin. In some examples, when the SHP capture threshold is less than the NSHP capture threshold, the control circuit 80 can set the operational pacing pulse amplitude to an amplitude greater than the SHP capture threshold but less than the NSHP capture threshold. In yet other examples, the control circuit 80 can set the operational pacing pulse amplitude to a safety margin greater than the NSHP capture threshold or the SHP capture threshold (whichever is lower) to facilitate capture of at least a portion of the His-Purkinje system and conserve the IMD power supply 98. In yet other examples, control circuit 80 may set the operational pacing pulse output to be greater than the NSHP capture threshold by a safety margin to minimize the likelihood of complete LOC of both the His-Purkinje system and the ventricular myocardium.

[0232] The operational pacing pulse output may be adjusted by the control circuit 80 at block 614 and used by the therapy delivery circuit 84 to deliver subsequent VCS pacing pulses in accordance with the pacing therapy or pacing mode, at least until the next VCS capture check is performed or a programmed change is made by the user. The control circuit 80 returns to block 601 to await the next time a VCS capture check is performed. The capture threshold test data, results, and / or adjusted pacing pulse output may be stored in the memory 82 and subsequently transmitted by the IMD telemetry circuit 88, e.g., for receipt by the external device 50, for display in the GUI, and for review by a clinician.

[0233] It should be understood that, depending on the example, certain actions or events of any of the methods described herein may be performed in a different order, may be added, combined, or omitted entirely (e.g., not all of the actions or events described are necessary to practice the method). In addition, in some examples, actions or events may be performed simultaneously, for example, through multi-threaded processing, interrupt processing, or multiple processors, rather than sequentially. In addition, although for clarity, certain aspects of the present disclosure are described as being performed by a single circuit or unit, it should be understood that the techniques of the present disclosure may be performed by a combination of circuits or components associated with, for example, a medical device.

[0234] In one or more examples, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored as one or more instructions or codes on a computer-readable medium and executed by a hardware-based processing unit. A computer-readable medium may include a computer-readable storage medium, which corresponds 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 desired program code in the form of instructions or data structures and can be accessed by a computer).

[0235] Instructions may be executed by one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, application specific integrated circuits (ASICs), field programmable logic arrays (FPLAs), or other equivalent integrated or discrete logic circuit systems. Thus, the term "processor" as used herein may refer to any of the foregoing structures or any other structure suitable for implementing the techniques described herein. Additionally, these techniques may be fully implemented in one or more circuits or logic elements.

[0236] Therefore, medical devices have been presented in the foregoing description with reference to specific examples. It should be understood that the various aspects disclosed herein may be combined in combinations different from the specific combinations presented in the drawings. It should be understood that various modifications may be made to the referenced examples without departing from the scope of the present disclosure and the following claims.

[0237] The following examples are a non-limiting list of clauses in accordance with one or more techniques of the present disclosure.

[0238] Embodiment 1. A medical device system comprises: a memory configured to store a first cardiac signal sensed after delivery of a ventricular conduction system (VCS) pacing pulse; a processing circuit system configured to: receive the first cardiac signal sensed after delivery of the VCS pacing pulse; apply a pacing capture classification machine learning model to at least the first cardiac signal; determine a capture type classification of the VCS pacing pulse from a plurality of capture types based on the applied pacing capture classification machine learning model; and generate an output based on the capture type classification; and a user interface configured to present, in response to the generated output, a representation of the capture type classification associated with the delivered ventricular conduction system pacing pulse.

[0239] Example 2. A medical device system according to Example 1, wherein: the memory is further configured to store a first template beat signal corresponding to a first VCS pacing pulse output; and the processing circuit system is further configured to: input at least the first template beat signal and the first cardiac signal into the pacing capture classification machine learning model; and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the first template beat signal and the first cardiac signal.

[0240] Example 3. A medical device system according to any one of Examples 1 to 2, wherein the memory is further configured to store a first template beat signal corresponding to a first VCS pacing pulse output; and the processing circuit system is further configured to: determine a first template difference signal from the first cardiac signal and the first template beat signal; input at least the first template difference signal and the first cardiac signal into the pacing capture classification machine learning model; and determine the capture type classification based on the pacing capture classification machine applied to at least the first template difference signal and the first cardiac signal model.

[0241] Embodiment 4. The medical device system of any one of embodiments 1 to 3, wherein: the memory is further configured to store a plurality of template beat signals, wherein each template beat signal corresponds to one of a plurality of VCS pacing pulse outputs. And the processing circuit system is further configured to: input each of the plurality of template beat signals and the first cardiac signal to the pacing capture classification machine learning model; and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the plurality of template beat signals and the first cardiac signal.

[0242] Embodiment 5. A medical device system according to any one of Embodiments 1 to 4, wherein: the memory is further configured to store a plurality of template beat signals, wherein each template beat signal corresponds to a VCS pacing pulse output among a plurality of VCS pacing pulse outputs; and the processing circuit system is further configured to: determine a plurality of template difference signals by determining a template difference signal from each template beat signal among the plurality of template beat signals and the first cardiac signal; and input the plurality of template difference signals and the first cardiac signal into the pacing capture classification machine learning model; and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the plurality of template difference signals and the first cardiac signal.

[0243] Example 6. A medical device according to any one of Examples 4 to 5, wherein the memory is further configured to store the multiple template beat signals by storing a first template beat signal corresponding to a first VCS pacing pulse output, a second template beat signal corresponding to a second VCS pacing pulse output that is smaller than the first VCS pacing pulse output, and a third template beat signal corresponding to a third VCS pacing pulse output between the first VCS pacing pulse output and the second VCS pacing pulse output.

[0244] Example 7. A medical device system according to any one of Examples 1 to 6, wherein the sensing circuit system is further configured to: determine a derivative signal from the first cardiac signal; input the first cardiac signal into the pacing capture classification machine learning model by inputting at least the derivative signal; and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the derivative signal.

[0245] Embodiment 8. The medical device system of any one of Embodiments 1 to 7, wherein: the memory is further configured to store a second cardiac signal sensed after delivery of the VCS pacing pulse, wherein the first cardiac signal is sensed by a first sensing electrode vector, and the second cardiac signal is sensed by a second sensing electrode vector different from the first sensing electrode vector. And the processing circuit system is further configured to: input each of the first cardiac signal and the second cardiac signal to the pacing capture classification machine learning model; and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the first cardiac signal and the second cardiac signal.

[0246] Embodiment 9. The medical device system of any one of Embodiments 1 to 8, wherein: the memory is further configured to store a pacing pulse output of the delivered VCS pacing pulse. And the processing circuit system is further configured to: input at least the pacing pulse output and the first cardiac signal to the pacing capture classification machine learning model; and determine the capture type classification based on the pacing capture classification machine learning model applied to at least the pacing pulse output and the first cardiac signal.

[0247] Example 10. A medical device system according to any one of Examples 1 to 9, wherein the processing circuit system is further configured to: determine a characteristic of the first cardiac signal; input the characteristic of the first cardiac signal and the first cardiac signal into the pacing capture classification machine learning model; and determine the capture type classification based on the characteristic of the first cardiac signal and the pacing capture classification machine learning model applied to at least the first cardiac signal.

[0248] Example 11. A medical device system according to any one of Examples 1 to 10, wherein the processing circuit system is further configured to: receive a training cardiac signal data set obtained from multiple patients, the training cardiac signal data set comprising multiple training cardiac signals, each training cardiac signal being sensed after delivering a VCS pacing pulse, wherein the VCS pacing pulse associated with the multiple training cardiac signals comprises VCS pacing pulses delivered with multiple different pacing pulse outputs; train the pacing capture classification machine learning model with the training cardiac signal data set according to a machine learning algorithm; and apply the pacing capture classification machine learning model trained with the training cardiac signal data set to at least the first cardiac signal.

[0249] Example 12. A medical device system according to any one of Examples 1 to 11, wherein the processing circuit system is further configured to: receive multiple cardiac signals including the first cardiac signal, each of the multiple cardiac signals being associated with a VCS pacing pulse, wherein the VCS pacing pulse associated with the multiple cardiac signals includes VCS pacing pulses delivered with multiple different pacing pulse outputs; input each of the multiple cardiac signals into the pacing capture classification machine learning model; determine a capture type classification of each of the VCS pacing pulses delivered with the multiple different pacing pulse outputs associated with the multiple cardiac signals based on the pacing capture classification machine learning model; and determine a capture threshold of at least one of the multiple capture types based on the capture type classification.

[0250] Embodiment 13 The medical device system of Embodiment 12, wherein the user interface is further configured to present the capture threshold of the at least one capture type.

[0251] Example 14. The medical device system of any one of Examples 12 to 13, wherein the processing circuit system is further configured to determine an operational pacing pulse output based on the capture threshold determined for the at least one capture type.

[0252] Example 15. The medical device system of Example 14, further comprising a therapy delivery circuit configured to generate a VCS pacing pulse based on the operational pacing pulse output.

[0253] Example 16. A medical device system according to any one of Examples 1 to 15, wherein the processing circuit system is further configured to classify the capture type as one of: selective His-Purkinje capture without ventricular myocardial capture, non-selective His-Purkinje capture with ventricular myocardial capture, ventricular myocardial capture without His-Purkinje capture, or loss of capture.

[0254] Embodiment 17. A method, comprising: storing in a memory a first cardiac signal sensed after delivery of a ventricular conduction system (VCS) pacing pulse; receiving, by a processing circuit system, the first cardiac signal sensed after delivery of the VCS pacing pulse; applying a pacing capture classification machine learning model to at least the first cardiac signal; determining, based on the applied pacing capture classification machine learning model, a capture type classification of the VCS pacing pulse from a plurality of capture types; generating, by the processing circuit system, an output based on the capture type classification; and presenting, by a user interface, in response to the generated output, a representation of the capture type classification associated with the delivered VCS pacing pulse.

[0255] Example 18. According to the method described in Example 17, the method further includes: storing a first template beat signal corresponding to a first VCS pacing pulse output in the memory; inputting at least the first template beat signal and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the first template beat signal and the first cardiac signal.

[0256] Example 19. According to the method described in any one of Examples 17 to 18, the method further includes: storing a first template beat signal corresponding to a first VCS pacing pulse output in the memory; determining a first template difference signal from the first cardiac signal and the first template beat signal by the processing circuit system; inputting at least the first template difference signal and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the first template difference signal and the first cardiac signal.

[0257] Example 20. A method according to any one of Examples 17 to 19, further comprising storing a plurality of template beat signals in the memory, wherein each template beat signal corresponds to a VCS pacing pulse output among a plurality of VCS pacing pulse outputs; inputting each of the plurality of template beat signals and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the plurality of template beat signals and the first cardiac signal.

[0258] Example 21. A method according to any one of Examples 17 to 20, further comprising storing a plurality of template beat signals in the memory, wherein each template beat signal corresponds to a VCS pacing pulse output among a plurality of VCS pacing pulse outputs; determining a plurality of template difference signals by determining a template difference signal from each template beat signal among the plurality of template beat signals and the first cardiac signal; inputting the plurality of template difference signals and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the plurality of template difference signals and the first cardiac signal.

[0259] Example 22. According to the method described in any one of Examples 20 to 21, the method further includes storing multiple template beat signals by storing a first template beat signal corresponding to a first VCS pacing pulse output, a second template beat signal corresponding to a second VCS pacing pulse output that is smaller than the first VCS pacing pulse output, and a third template beat signal corresponding to a third VCS pacing pulse output between the first VCS pacing pulse output and the second VCS pacing pulse output.

[0260] Example 23. According to the method described in any one of Examples 17 to 22, the method further includes: determining a derivative signal from the first cardiac signal; inputting the first cardiac signal into the pacing capture classification machine learning model by inputting at least the derivative signal; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the derivative signal.

[0261] Example 24. A method according to any one of Examples 17 to 23, further comprising: storing a second cardiac signal sensed after delivery of the ventricular conduction system (VCS) pacing pulse, wherein the first cardiac signal is sensed by a first sensing electrode vector and the second cardiac signal is sensed by a second sensing electrode vector different from the first sensing electrode vector; inputting each of the first cardiac signal and the second cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the first cardiac signal and the second cardiac signal.

[0262] Example 25. A method according to any one of Examples 17 to 24, further comprising: storing the pacing pulse output of the delivered VCS pacing pulse in the memory; inputting at least the pacing pulse output and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the pacing capture classification machine learning model applied to at least the pacing pulse output and the first cardiac signal.

[0263] Example 26. According to the method described in any one of Examples 17 to 25, the method further includes: determining a characteristic of the first cardiac signal; inputting the characteristic of the first cardiac signal and the first cardiac signal into the pacing capture classification machine learning model; and determining the capture type classification based on the characteristic of the first cardiac signal and the pacing capture classification machine learning model applied to at least the first cardiac signal.

[0264] Example 27. A method according to any one of Examples 17 to 26, the method further comprising: receiving, by a processing circuit system, a training cardiac signal data set obtained from a plurality of patients, the training cardiac signal data set comprising a plurality of training cardiac signals, each training cardiac signal being sensed after delivery of a VCS pacing pulse, wherein the VCS pacing pulse associated with the plurality of training cardiac signals comprises VCS pacing pulses delivered with a plurality of different pacing pulse outputs; training the pacing capture classification machine learning model with the training cardiac signal data set according to a machine learning algorithm; and applying the pacing capture classification machine learning model trained with the training cardiac signal data set to at least the first cardiac signal.

[0265] Example 28. A method according to any one of Examples 17 to 27, the method further comprising: receiving, by the processing circuit system, a plurality of cardiac signals including the first cardiac signal, each of the plurality of cardiac signals being associated with a VCS pacing pulse, wherein the VCS pacing pulse associated with the plurality of cardiac signals comprises a VCS pacing pulse delivered with a plurality of different pacing pulse outputs; inputting each of the plurality of cardiac signals into the pacing capture classification machine learning model; determining, based on the pacing capture classification machine learning model, a capture type classification of each of the VCS pacing pulses delivered with the plurality of different pacing pulse outputs associated with the plurality of cardiac signals; and determining, based on the capture type classification, the capture threshold of at least one of the plurality of capture types.

[0266] Embodiment 29. The method according to Embodiment 28 further comprises presenting, by the user interface, a capture threshold of the at least one capture type.

[0267] Example 30. A method according to any one of Examples 28 to 29, further comprising determining an operational pacing pulse output based on the capture threshold determined for the at least one capture type.

[0268] Example 31. A method according to Example 30, further comprising generating a VCS pacing pulse based on the operating pacing pulse output.

[0269] Example 32. A method according to any one of Examples 17 to 31, further comprising classifying the capture type as one of: selective His-Purkinje capture without ventricular myocardial capture, non-selective His-Purkinje capture with ventricular myocardial capture, ventricular myocardial capture without His-Purkinje capture, or loss of capture.

[0270] Embodiment 33. A non-transitory computer-readable storage medium storing a set of instructions that, when executed by a processing circuit system of a medical device system, causes the medical device system to: store a cardiac signal sensed after delivery of a ventricular conduction system (VCS) pacing pulse in a memory of the medical device system; apply a pacing capture classification machine learning model to at least the cardiac signal; determine a capture type classification of the VCS pacing pulse from a plurality of capture types based on the applied pacing capture classification machine learning model; generate an output based on the capture type classification; and present, through a user interface of the medical device system, a representation of the determined capture type associated with the delivered ventricular conduction system pacing pulse.

[0271] Embodiment 34. A medical device system, the medical device system comprising: a sensing circuit configured to sense a cardiac electrical signal; a therapy delivery circuit configured to deliver a ventricular conduction system (VCS) pacing pulse; and a processing circuit system configured to: receive the cardiac electrical signal; apply a machine learning model to at least the sensed cardiac electrical signal, training the machine learning model to determine a pacing capture type using multiple cardiac signal data sets; determine a pacing capture type for the VCS pacing pulse based on the applied machine learning model; and a user interface configured to present a representation of the determined pacing capture type for the VCS pacing pulse.

[0272] Embodiment 35. A medical device system, the medical device system comprising: a sensing circuit configured to sense at least one cardiac electrical signal; a therapy delivery circuit configured to deliver a ventricular conduction system (VCS) pacing pulse; a memory configured to store a cardiac signal segment sensed by the sensing circuit during at least one post-pacing time interval after delivery of a first VCS pacing pulse delivered by the therapy delivery circuit; and a processing circuit system configured to: apply a machine learning model to at least the first cardiac signal segment; determine a pacing capture type of the first VCS pacing pulse from a plurality of capture types based on the machine learning model; select an operational pacing pulse output based on at least the determined pacing capture type; wherein the therapy delivery circuit is further configured to deliver a second VCS pacing pulse based on the selected operational pacing pulse output.

Claims

1. A medical device system, comprising: a memory configured to store a first cardiac signal sensed after delivery of a ventricular conduction system (VCS) pacing pulse; processing circuitry configured to: receiving the first cardiac signal sensed after delivery of the VCS pacing pulse; applying a pace capture classification machine learning model to at least the first cardiac signal; determining a capture type classification of the VCS pacing pulse from a plurality of capture types based on the applied pacing capture classification machine learning model; as well as generating an output based on the capture type classification; and A user interface is configured to present, in response to the generated output, a representation of the capture type classification associated with the delivered ventricular conduction system pacing pulse.

2. The medical device system of claim 1, wherein: The memory is further configured to store a first template beat signal corresponding to a first VCS pacing pulse output; and The processing circuit system is further configured to: inputting at least the first template beat signal and the first cardiac signal into the pace capture classification machine learning model; and The capture type classification is determined based on the pacing capture classification machine learning model applied to at least the first template beat signal and the first cardiac signal.

3. The medical device system according to any one of claims 1 to 2, wherein: The memory is further configured to store a first template beat signal corresponding to a first VCS pacing pulse output; and The processing circuit system is further configured to: determining a first template difference signal from the first heart signal and the first template beat signal; inputting at least the first template difference signal and the first cardiac signal into the pace capture classification machine learning model; as well as The capture type classification is determined based on the pacing capture classification mechanism applied to at least the first template difference signal and the first cardiac signal model.

4. The medical device system according to any one of claims 1 to 3, wherein: The memory is further configured to store a plurality of template beat signals, wherein each template beat signal corresponds to one VCS pacing pulse output among a plurality of VCS pacing pulse outputs; and The processing circuit system is further configured to: inputting each of the plurality of template beat signals and the first cardiac signal into the pace capture classification machine learning model; and The capture type classification is determined based on the pacing capture classification machine learning model applied to at least the plurality of template beat signals and the first cardiac signal.

5. The medical device system according to any one of claims 1 to 4, wherein: The memory is further configured to store a plurality of template beat signals, wherein each template beat signal corresponds to one VCS pacing pulse output among a plurality of VCS pacing pulse outputs; and The processing circuit system is further configured to: determining a plurality of template difference signals by determining a template difference signal from each template beat signal of the plurality of template beat signals and the first heart signal; as well as inputting the plurality of template difference signals and the first cardiac signal into the pacing capture classification machine learning model; as well as The capture type classification is determined based on the pacing capture classification machine learning model applied to at least the plurality of template difference signals and the first cardiac signal.

6. The medical device according to any one of claims 4 to 5, wherein the memory is further configured to store the multiple template beat signals by storing a first template beat signal corresponding to a first VCS pacing pulse output, a second template beat signal corresponding to a second VCS pacing pulse output that is smaller than the first VCS pacing pulse output, and a third template beat signal corresponding to a third VCS pacing pulse output between the first VCS pacing pulse output and the second VCS pacing pulse output.

7. The medical device system of any one of claims 1 to 6, wherein the processing circuit system is further configured to: determining a derivative signal from said first cardiac signal; inputting the first cardiac signal to the pace capture classification machine learning model by inputting at least the derivative signal; and The capture type classification is determined based on the pace capture classification machine learning model applied to at least the derivative signal.

8. The medical device system according to any one of claims 1 to 7, wherein: The memory is further configured to store a second cardiac signal sensed after delivery of the VCS pacing pulse, wherein the first cardiac signal is sensed by a first sensing electrode vector and the second cardiac signal is sensed by a second sensing electrode vector different from the first sensing electrode vector; and The processing circuit system is further configured to: inputting each of the first cardiac signal and the second cardiac signal into the pace capture classification machine learning model; and The capture type classification is determined based on the pacing capture classification machine learning model applied to at least the first cardiac signal and the second cardiac signal.

9. The medical device system according to any one of claims 1 to 8, wherein: The memory is further configured to store a pacing pulse output of the delivered VCS pacing pulse; and The processing circuit system is further configured to: inputting at least the pacing pulse output and the first cardiac signal into the pacing capture classification machine learning model; and The capture type classification is determined based on the pacing capture classification machine learning model applied to at least the pacing pulse output and the first cardiac signal.

10. The medical device system of any one of claims 1 to 9, wherein the processing circuit system is further configured to: determining a characteristic of the first cardiac signal; inputting the features of the first cardiac signal and the first cardiac signal into the pace capture classification machine learning model; and The capture type classification is determined based on the features applied to at least the first cardiac signal and the pacing capture classification machine learning model of the first cardiac signal.

11. The medical device system of any one of claims 1 to 10, wherein the processing circuit system is further configured to: receiving a training cardiac signal data set obtained from a plurality of patients, the training cardiac signal data set comprising a plurality of training cardiac signals, each training cardiac signal being sensed after delivery of a VCS pacing pulse, wherein the VCS pacing pulses associated with the plurality of training cardiac signals comprise VCS pacing pulses delivered at a plurality of different pacing pulse outputs; According to a machine learning algorithm, training the pacing capture classification machine learning model with the training cardiac signal data set; as well as The pacing capture classification machine learning model trained with the training cardiac signal dataset is applied to at least the first cardiac signal.

12. The medical device system of any one of claims 1 to 11, wherein the processing circuit system is further configured to: receiving a plurality of cardiac signals including the first cardiac signal, each cardiac signal of the plurality of cardiac signals being associated with a VCS pacing pulse, wherein the VCS pacing pulses associated with the plurality of cardiac signals include VCS pacing pulses delivered at a plurality of different pacing pulse outputs; inputting each cardiac signal of the plurality of cardiac signals into the pace capture classification machine learning model; determining, based on the pacing capture classification machine learning model, a capture type classification for each of the VCS pacing pulses delivered with the plurality of different pacing pulse outputs associated with the plurality of cardiac signals; as well as Based on the capture type classification, a capture threshold for at least one capture type of the plurality of capture types is determined.

13. The medical device system of claim 12, wherein the processing circuitry is further configured to determine an operational pacing pulse output based on the capture threshold determined for the at least one capture type.

14. The medical device system of any one of claims 1 to 13, wherein the processing circuit system is further configured to determine the capture type classification as one of: selective His-Purkinje capture without ventricular myocardial capture, non-selective His-Purkinje capture with ventricular myocardial capture, ventricular myocardial capture without His-Purkinje capture, or loss of capture.

15. A non-transitory computer-readable storage medium storing a set of instructions that, when executed by processing circuitry of a medical device system, causes the medical device system to: storing, in a memory of the medical device system, cardiac signals sensed after delivery of a ventricular conduction system (VCS) pacing pulse; applying a pace capture classification machine learning model to at least the cardiac signal; determining a capture type classification of the VCS pacing pulse from a plurality of capture types based on the applied pacing capture classification machine learning model; generating an output based on the capture type classification; and A representation of the determined capture type associated with the delivered ventricular conduction system pacing pulse is presented via a user interface of the medical device system.

Images