Smart ICM ECG Filtering

Abstract

The present disclosure relates to a method of analyzing an event (E) determined by an implantable device (I), the method including validating the event based at least in part on a cardiac episode associated with the event, the validating being based at least in part on a type (1, 2, 3, 4) of the event.

Description

[Technical field]

[0001] The present invention relates generally to a method for analyzing events determined by an implantable device, a corresponding apparatus, a corresponding system and a corresponding computer program.

[0002] Various methods for determining events related to cardiac activity are currently known in the art. Such methods can be performed by an implantable device that is implanted in a patient to monitor the patient's heart. For example, the implantable device can be configured as an implantable cardiac monitor (ICM), a pacemaker, a defibrillator, etc. Typically, the implantable device can be configured to sense or determine cardiac activity (e.g., electrocardiogram or ECG signals) and, based thereon, determine whether a particular event (e.g., a particular cardiac arrhythmia) is present. For example, an appropriate medical response (e.g., alerting medical personnel, alerting the patient to undergo further medical examination, immediate medical intervention in case of impending cardiac arrest, etc.) can be performed.

[0003] However, currently known methods may not be able to exclude incorrectly determined events (e.g., false positives), which may result in inadequate medical care due to misinformation being conveyed to medical personnel, leading to mistreatment of patients, and may falsely alert patients to adverse cardiac events (which do not actually exist), causing unnecessary patient anxiety.

[0004] Thus, currently known techniques do not always provide optimal analysis of events determined by implantable devices, and there is therefore a need to find ways to improve the analysis of determined events.

[0005] SUMMARY OF THE DISCLOSURE The aspects described herein address, at least in part, the above needs.

[0006] A first aspect relates to a method of analyzing an event determined by an implantable device, the method including validating the event based at least in part on a cardiac episode associated with the event, the validating being based at least in part on a type of the event.

[0007] Thus, an additional verification is performed for the event determined by the implantable device. The (additional) verification can thus serve to double check whether the event has been correctly determined. Moreover, the (additional) verification of the event can be performed depending on the type of the determined event. The inventors have found that this can significantly improve the result of the (additional) verification compared to simply verifying all events by the same verification step. Thus, the present invention can enable verification based on special characteristics of the type of event, in contrast to other general verification steps (e.g., other general verifications can only take into account certain parameters and not the type of event). To illustrate an example of the concept here, if the determined event is of type A, the event is verified by verification A. If the determined event is of a different type B, the event is verified by a different verification B.

[0008] An event may include the occurrence of a statistical anomaly in a patient's heart rate. For example, an event may include a Boolean value associated with the event indicating whether a determination of the event has occurred (e.g., positive / negative, true / false, etc.). For example, an event may be associated with some cardiac arrhythmia (e.g., atrial arrhythmia, ventricular arrhythmia, junctional arrhythmia, etc.). In this regard, a type of event may include a particular type of some cardiac arrhythmia (e.g., a particular type of atrial arrhythmia, a particular type of ventricular arrhythmia, a particular type of junctional arrhythmia, etc.).

[0009] A cardiac episode associated with an event can be cardiac activity over a particular duration. A cardiac episode can also be considered as a time window of cardiac activity during which an event occurs. For example, cardiac activity can include an electrocardiogram or ECG signal (and / or any cardiac vector signal). In this regard, a cardiac episode can include an electrocardiogram or ECG signal (and / or any cardiac vector signal) over a duration during which an event occurs at a particular duration. A cardiac episode can be, for example, a time window of cardiac activity over a particular duration with a start time t0 and an end time t E A cardiac episode may be defined by a duration between 0 and 100 seconds. The particular duration of a cardiac episode may be a clinically relevant duration required to determine an event (and / or a type of event). In another example, the particular duration of a cardiac episode may be selected or adapted for technical purposes to facilitate event determination (e.g., the duration may be fixed at 40-80 seconds, e.g., 50-70 seconds, or 60 seconds, 1 hour, etc., and an event is determined within the fixed duration). In other words, a cardiac episode may be a cardiac activity (e.g., ECG signal) over a period of time during which an event occurs or an event is analyzed. In an example, the method may also include sensing the cardiac activity by an implantable device (e.g., by an implantable sensor).

[0010] In certain examples, the type includes at least one of atrial fibrillation onset, asystole, high ventricular rate, and bradycardia. Atrial fibrillation onset can be considered as a detectable new or initial onset of an atrial arrhythmia. Asystole can be considered as an absence of ventricular contractions and / or a long pause between ventricular contractions. High ventricular rate (HVR) can be considered as an abnormally high ventricular heart rate from a medical perspective. Bradycardia can be considered as an abnormally low heart rate from a medical perspective. In some examples, the type includes at least one member of the group consisting of atrial fibrillation onset, asystole, high ventricular rate, and bradycardia.

[0011] In an example, the validation includes determining whether the event is a true positive and / or whether the event is a false positive. For example, an (initial) determination of an event by the implantable device may be considered to be positive. The (additional) validation may include determining whether the event is indeed a true positive or is in fact a false positive. If the event is validated as a true positive, then the event may be considered to have been correctly determined by the implantable device. However, if the validation result is a false positive, then the event may be considered to have been incorrectly determined by the implantable device.

[0012] In some examples, the verification is performed by a second device, e.g., a device different from the implantable device. Thus, the verification may not necessarily be performed by the implantable device that determined the event. This may allow for improved verification, since limitations of the implantable device may be overcome by the second device. For example, implantable devices are typically limited to a certain size (e.g., narrow size) to ensure optimal medical compatibility after implantation in a patient. However, such size constraints may lead to hardware and power constraints of the implant, which may only allow limited resources in terms of computing power. Such limitations may be overcome by performing the verification by a second device, e.g., which may not be an implantable device. Thus, the second device may be configurable to have higher computing power than the implantable device. In another example, the second device is not limited by power limitations (e.g., battery life) associated with an implantable device in a patient. In one example, the second device may be an external computing device (e.g., a smartphone, a desktop PC, a server (system), a cloud, etc.).

[0013] In an example, the method includes receiving information related to the event from the implantable device. For example, the information can be received by a second device (as outlined herein). The method can further include transmitting the information related to the event, for example, from the implantable device to the second device (e.g., wirelessly and / or directly or indirectly). The information related to the event can include, for example, a cardiac episode (e.g., an ECG signal as outlined herein), a type of event, a Boolean value associated with the event, a notification that the event has been determined, data related to the implantable device (e.g., an identification / serial number and / or model type of the implantable device, patient information (e.g., patient identification number, patient name, etc.)).

[0014] In another example, the verification includes upsampling of the cardiac episode (e.g., upsampling of the ECG signal). The upsampling can include increasing the sample rate of the cardiac episode by at least 2 times, at least 5 times, at least 8 times, such as 8 times or 10 times (e.g., the upsampling can include increasing the sampling rate from 128 Hz to 1024 Hz). For example, the cardiac episode can be received from the implantable device at a relatively low sampling rate and then upsampled (e.g., by a second device) to a relatively high sampling rate. The upsampling can also include applying various interpolation steps (e.g., interpolation filters). In another example, the verification includes anti-aliasing (of the signal) of the cardiac episode.

[0015] In one example, the validation includes detecting the QRS information (e.g., QRS interval and / or QRS complex) of the cardiac episode. In this regard, first, the QRS block (i.e., the QRS candidate) can be detected as a region of interest, where the QRS block can be a time window of the cardiac episode that substantially includes the QRS information. For example, the QRS information can be detected according to the Elgendi algorithm. Then, for each QRS block, the QRS peak (e.g., the R wave peak) can be determined according to the present invention. The QRS peak can be determined based on applying a global filter function to the QRS block to create a squared bandpass signal (which can also be the output of the Elgendi algorithm). Details regarding the Elgendi algorithm can be found in Elgendi M. (2013). Fast QRS detection with an optimized knowledge-based method: evaluation on 11 standard ECG databases. PloS one, 8(9), e73557. However, in contrast to the known Elgendi algorithm, the present invention can include determining the QRS peak based on finding a (local) peak (i.e. a (local) maximum) of the squared bandpass signal (or a (local) maximum of the slope (of the (squared bandpass signal)) that is located furthest to the left within the QRS block (left may refer to an earlier time of the QRS block compared to another time within the QRS block). The present invention can further include subsequently associating the QRS peak with said determined (local) peak of the squared bandpass signal. Thus, a time location within the QRS block and / or a time location within the cardiac episode can be matched to the determined QRS peak.In other words, the method of the present invention can perform the association of the QRS peak with the local peak of the QRS complex, where the local peak is the one with the local maximum peak height (e.g., the maximum difference between the local maximum peak value and the adjacent local minimum value) and / or the one with the (absolute) maximum peak amplitude and / or the one selected as the leftmost local peak of the QRS complex. In contrast, in the Elgendi algorithm, the R peak is detected based on a brute force search that takes into account the frequency band, the event-related duration, and the offset fraction. This leads to a time-consuming algorithm that requires a database for training and optimization. Therefore, the method of the present invention can overcome such drawbacks by considering the local peaks and the respective slopes of the square bandpass filter and choosing the position of the leftmost local peak in the QRS complex (or QRS block) to determine the QRS peak.

[0016] In an example of the method, if the type includes atrial fibrillation onset, the validation includes detecting a (potential) P-wave peak in the cardiac episode by determining a moving linear regression in the vicinity of the (predetermined) QRS peak. The vicinity may be the time region preceding the QRS peak and may be determined based on the time interval between the QRS peak and its preceding QRS peak. The QRS peak (and / or its preceding QRS peak) may be determined as described herein or by any other suitable method known in the art. For example, detecting the (potential) P-wave peak may be based on advanced QRS complex removal. To illustrate an example, first, the vicinity of the QRS peak may be determined based on the detected QRS block (as outlined herein) and the determined QRS peak. The QRS peak may serve as a reference point (i.e., origin) for applying signal processing to the cardiac episode signal in the vicinity of the QRS peak. The vicinity may also be referred to as the region of interest of the P-wave peak. A slope correction can be performed in the neighborhood by moving linear regression, which removes the global slope and (indirectly) emphasizes the local slope. The remaining (local) peaks can then be sorted based on these features to find the (potential) P-wave peaks. For example, the P-wave peaks can be associated with the peaks that have the local maximum peak heights (e.g., the maximum difference between the maximum peak value and the adjacent local minimum), and / or the peaks that have the (absolute) maximum peak amplitudes, and / or the peaks that are furthest from the reference point (i.e., the origin, i.e., the corresponding QRS peak).

[0017] In an example of the method, the validation (of the atrial fibrillation initiating event) is further based at least in part on one of the RR interval, PP interval, RP interval, correlation of QRS complex morphology, and / or ectopy probability in the cardiac episode. For example, the validation may include determining at least one of the RR interval, PP interval, RP interval in the cardiac episode by a suitable method known in the art. For example, the RR interval (e.g., the RR interval of the cardiac episode) may be calculated based on the determined QRS peak (e.g., by calculating the time difference between subsequent QRS peaks). The validation may then include extracting statistical features based on the outlined cardiac intervals (e.g., the RR, PP, RP intervals). The statistical features may relate to statistical features of the cardiac intervals (e.g., average, maximum / minimum, etc.). The validation may also include determining correlation of QRS complex morphology and / or ectopy probability (i.e., ectopic beat probability). To illustrate an example, correlation of QRS complex morphology may include autocorrelation of at least one QRS complex morphology and / or cross-correlation between QRS complex morphologies of different QRS complexes. For example, cross-correlation may be performed on the morphology of the QRS complexes of the same cardiac episode and / or on the morphology of the QRS complexes of the cardiac episode relative to the morphology of a reference QRS complex. Determination of the ectopy probability may be based on the irregularity of the cardiac interval (as outlined herein) compared to one or more characteristics of other cardiac intervals in the cardiac episode (e.g., the ectopy probability may be based on the RR intervals (irregularity of the RR intervals) in the cardiac episode, e.g., the irregularity may be a deviation from a mean value, e.g., by a predetermined threshold). If the validation is performed by the second device, the ectopy probability may be determined according to an algorithm also used to determine the ectopy probability in the implantable device. However, in this case, the ectopy probability used for validation may be based on the QRS peak detected by the second device (e.g., as outlined herein).

[0018] In an example of the method, the validation is performed by an artificial intelligence and / or machine learning system trained with events validated by manual inspection of cardiac episodes. The training can be performed by historical data with known labels (e.g., true positive, false positive). In an example, the artificial intelligence and / or machine learning system can be inputted with one or more statistical features of the cardiac intervals as outlined herein, correlations of QRS complexes as outlined herein, and / or ectopy probabilities as outlined herein, based on which the validation can be performed. The machine learning model can comprise, for example, XGBoost.

[0019] In one example of the method, if the type includes asystole, high ventricular rate and / or bradycardia, the validation includes removing from the cardiac episode RR intervals associated with amplitude clipping (or associated with signal saturation), preferably also removing RR intervals adjacent to at least one RR interval associated with amplitude clipping. In particular, RR intervals that start with amplitude clipping must be removed. RR intervals that are present only at the end of amplitude clipping do not need to be removed and can be further considered.

[0020] Amplitude clipping can be defined by the absolute value of the amplitude of a cardiac episode (e.g., ECG signal) exceeding a threshold. The threshold can, for example, be associated with an abnormally large value that has no medical reason (e.g., amplitude clipping can be caused by technical reasons, e.g., sensing error, etc.). In another example, clipping can be defined by the amplitude of a cardiac episode exceeding a certain percentage of the threshold for the clipping period. For example, the percentage can be X% (e.g., 65%) and the threshold can be C. TH (e.g. 2V), and the clipping period is t clip (e.g. 100 ms), and the amplitude is within the clipping period t clip Over the threshold C THIf the amplitude exceeds X% of the clipping, then clipping of the amplitude may be determined. The RR interval associated with the clipping (i.e., the clipped amplitude) is then removed as outlined herein, and preferably, the RR intervals adjacent to the RR interval associated with the clipping are also removed.

[0021] In one example, the verification (of an asystole event) further includes detecting whether a pause associated with asystole is present in the cardiac episode between subsequent QRS peaks, preferably corresponding to an RR interval that was not removed (as outlined herein). The pause may be considered as an abnormally long pause (e.g., longer than 3 seconds) between subsequent QRS peaks. The pause may be determined based on the QRS peaks detected by the method of the first aspect as outlined herein.

[0022] In a further example, verifying (asystole event) further includes detecting whether a pause associated with asystole exists in the cardiac episode during a subsequent RR interval, preferably during a subsequent RR interval that has not been removed (as outlined herein). The pause can be considered as an abnormally long pause between subsequent RR intervals. The pause can be determined based on the QRS peak detected by the method of the first aspect as outlined herein.

[0023] To illustrate another example, the implantable device is generally configurable to determine pauses according to an asystole algorithm that may be based on a QRS peak detected by the implantable device by an internal QRS method (e.g., a suitable method known in the art). For example, pauses may be determined by the implantable device to determine asystole (or another event). If a verification of asystole is then performed (as outlined herein), the pauses may be determined by the same asystole algorithm (performed by the implantable device). It should be noted that in this case, the invention may include that the determination of pauses is based on a QRS peak detected by the method of the first aspect (as outlined herein) and not by an internal QRS method (performed by the implantable device). Since the computing power of the implantable device may be limited, the internal QRS method may also be limited in terms of algorithm complexity, which may reduce the accuracy of the detection of the QRS peak and thus result in an erroneous determination of pauses (and thus asystole). Thus, the exemplary method outlined reliably determines pauses based on highly accurate QRS peak detection, allowing for a more accurate verification based on QRS peak detection. This allows asystole events to be reliably and securely verified, overcoming limitations of implantable devices.

[0024] In one example, the validation (of the asystole event) further comprises determining that the event is a true positive if a pause is detected, and determining that the event is a false positive if not. Thus, the method of the present invention can enable the determined pause to represent a pause that actually exists by removing the clipped RR interval. Thus, the determined pause can be associated with a real event, and not, for example, a technical malfunction of the implantable device. Then, if a pause is not determined by the method of the first aspect, this may indicate that the implantable device has erroneously determined the event.

[0025] In a further example, the validation does not necessarily depend on the removal of the RR interval. In this case, the method may include detecting whether a pause associated with asystole exists between subsequent RR intervals or between QRS peaks in the cardiac episode, and determining that the event is a false positive if the pause does not exist. This may thus indicate that the implantable device has erroneously detected a pause associated with asystole without removing the RR interval. This may allow for the determination of a false positive event without necessarily determining amplitude clipping in the cardiac episode (as outlined herein).

[0026] In one example of the method, if the type includes high ventricular rate and / or bradycardia, the verification includes calculating a heart rate based on the cardiac episode. The heart rate can include the number of heart contractions per minute. The heart rate can be calculated based on determining the number of QRS peaks in a certain period of time, the results of which can be further processed to obtain the number of QRS peaks per minute (and thus the heart rate per minute). In one example, the heart rate can be calculated based on counting the number of QRS peaks throughout the cardiac episode. It is also contemplated to calculate the heart rate over a certain number of beats or over a predetermined heart rate time window that is smaller than the cardiac episode duration.

[0027] To illustrate another example of the method, the implantable device can generally be configured to calculate the heart rate according to an internal algorithm that may be based on the QRS peaks detected by the implantable device by an internal QRS method (e.g., a suitable method known in the art). For example, the heart rate may be determined by the implantable device to determine an event. Then, when performing the event verification (as outlined herein), the heart rate may be determined by the same internal algorithm. However, in this case, the invention may include calculating the heart rate based on the QRS peaks detected by the method of the first aspect (as outlined herein) rather than the internal QRS method. Since the computing power of the implantable device may be limited, the internal QRS method may also be limited in terms of algorithm complexity, which may reduce the accuracy of the detection of the QRS peaks and thus lead to erroneous determination of the heart rate and related events. Thus, the exemplary method outlined reliably determines the heart rate based on high-precision QRS peak detection, enabling verification based on more accurate QRS peak detection.

[0028] In an example, the verification is based at least in part on a comparison of the calculated heart rate with a predetermined threshold. For example, if the type includes a high ventricular heart rate and the calculated heart rate is above a predetermined threshold, the event can be determined to be a true positive (otherwise a false positive). For example, if the type includes bradycardia and the calculated heart rate is below a predetermined threshold, the event can be determined to be a true positive (otherwise a false positive). The predetermined thresholds can be different for verifying a high ventricular heart rate and for verifying a bradycardia. For example, a first predetermined threshold for verifying a high ventricular heart rate can be 100 beats per minute (bpm), and a first predetermined threshold for verifying a bradycardia can be 60 bpm.

[0029] In another example, the validation is based at least in part on the (calculated) heart rate exceeding (or falling below) a predefined threshold for a particular duration. For example, if the type includes high ventricular rate and the heart rate is above a predefined threshold (e.g., 100 bpm) for a particular duration (e.g., 9 seconds), the event may be determined to be a true positive. For example, if the type includes bradycardia and the heart rate is below a predefined threshold (e.g., 60 bpm) for a particular duration (e.g., 10 seconds), the event may be determined to be a true positive.

[0030] In another example, the validation is based at least in part on the (calculated) heart rate being above (or below) a predefined threshold for a particular number of heart beats. For example, if the type includes high ventricular rate and the heart rate is above a predefined threshold (e.g., 100 bpm) for a particular number of heart beats (e.g., 15 heart beats), the event may be determined to be a true positive. For example, if the type includes bradycardia and the heart rate is below a predefined threshold (e.g., 60 bpm) for a particular number of heart beats (e.g., 10 heart beats), the event may be determined to be a true positive.

[0031] Taking into account a specific duration and / or a specific number of heart beats can ensure a more accurate validation, for example, this allows (medically) unusual heart rates over a relatively short period and / or a relatively small number of heart beats to be excluded due to their low medical relevance.

[0032] A second aspect relates to an apparatus, e.g., a computer or computer system, for analyzing events determined by an implantable device, the apparatus being configured to perform the method according to the first aspect. In one example, the apparatus (e.g., a computer) may be included in a second device. For example, the apparatus may comprise a smartphone, a desktop PC, a server (system), a cloud, etc. Another aspect relates to an implantable device for determining events, the implantable device being configured to perform the method of the first aspect (as outlined herein). The implantable device may be further configured to transmit information related to the events to the computer of the second aspect.

[0033] A third aspect relates to a system for analyzing events determined by an implantable device, the system comprising an apparatus (as generally described herein) and an implantable device.

[0034] A fourth aspect relates to a computer program comprising instructions for, when executed, performing the method according to the first aspect (as outlined herein).

[0035] It should be noted that the method steps described herein, even if not explicitly described as method steps, but rather described with reference to an apparatus (or device or system or computer), may include all aspects described herein. Also, the devices and computer programs outlined herein may include means for performing all aspects outlined herein, even if described in the context of a method step.

[0036] The functions described herein, whether described as method steps, computer programs and / or means, can be implemented in hardware, software, firmware and / or combinations thereof. If implemented in software / firmware, the functions can be stored or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage medium can be any available medium accessible by a general purpose or special purpose computer. By way of example and not limitation, such computer-readable storage media can include RAM, ROM, EEPROM, FPGA, CD / DVD or other optical disk storage, magnetic disk storage or other magnetic storage, or any other medium that can be used to carry or store desired program code means in the form of instructions or data structures and that can be accessed by a general purpose or special purpose computer or a general purpose or special purpose processor. [Brief description of the drawings]

[0037] [Figure 1] 1 is a schematic diagram of an exemplary embodiment of a system including an implantable device and a computer system in accordance with the present invention. [Diagram 2] 1 is a schematic diagram of an exemplary embodiment of a method according to the invention;

[0038] 1 shows a schematic diagram of an exemplary embodiment of a system S according to the present invention, comprising an implantable device I and an apparatus such as a computer (system) 100. It should be noted that an implantable device as referred to herein may relate to a device that is implantable but not yet implanted in a patient, or may relate to a device that is already implanted in a patient.

[0039] For example, the implantable device I may comprise an implantable cardiac monitor ICM, an implantable electrophysiological device, as well as any other implant capable of sensing cardiac activity (e.g., a pacemaker, a defibrillator, an implantable sensor, etc.). For example, if the implantable device I is an ICM, it may be subcutaneously implanted in the patient. The implantable device I may be configured to monitor and / or record cardiac activity of the patient's heart. The cardiac activity may include an electrocardiogram signal, i.e., an ECG signal, or any other cardiac vector signal.

[0040] The system S may further include a computer (system) 100. The computer (system) 100 may constitute a remote service center with various computational nodes (e.g., computers, servers, databases, etc.). In one example, the computer (system) 100 may be a part of the remote service center. The remote service center may enable remote access and data management of information / data acquired by the implantable device I. The computer (system) 100 may also include peripheral interfaces with various devices other than the implantable device I for communicating with the remote service center. The implantable device I and the computer (system) 100 may be communicatively coupled via a communication link 101. The communication link 101 may enable direct communication of information from the implantable device I to the computer (system) 100 and vice versa. It is also conceivable that the communication link 101 may include an indirect communication path via one or more communication nodes (e.g., devices). For example, the implantable device 100 may communicate (e.g., transmit) information to the computer (system) 100 via an intermediate device. Thus, the computer (system) 100 can receive information from the implantable device I via the relay device. In response, the computer (system) 100 can communicate (e.g., transmit) information to the implantable device I via the relay device. Thus, the relay device can function as a relay communication node enabling the communication link 101. The relay device can constitute a programmer device, a cardiac messenger device, a smartphone, etc. For example, the programmer device can be used to adapt the settings of the implantable cardiac monitor (and function as a relay node). The cardiac messenger device can be used to function as a main relay node enabling periodic communication of the patient's cardiac monitoring data from the implantable device I to the computer (system) 100. Thus, the cardiac messenger device can be a portable device carried by the patient.

[0041] The computer (system) 100 can be communicatively coupled via a communication link 102 for external reading, for example, by a medical practitioner C. The communication link 102 can be bidirectional and implemented between the computer (system) 100 and a communication node (not shown) accessible by the medical practitioner C. Thus, the medical practitioner C can access the patient's cardiac activity data (e.g., monitoring data) if the corresponding data is communicated from the implantable device I to the computer (system) 100. The medical practitioner can access, for example, the patient's cardiac activity (e.g., ECG signal), as well as events related to the patient's cardiac episodes (as outlined herein), the cardiac episodes themselves, and / or the type of events. The computer (system) 100 can also receive information from the medical practitioner C via the communication link 102. This can enable the import of clinical data related to the patient and / or the setting of parameters used by the computer (system) 100, as described herein.

[0042] The computer (system) 100 can further be communicatively coupled to various peripheral nodes not shown in Fig. 1. For example, the peripheral node can be a smartphone that can be coupled to various other sensors (which may indicate the patient's medical sensory data). The peripheral node mode can also be an IT support node (for technical coordination of the computer (system) 100), a doctor's web portal, etc.

[0043] In Fig. 2, a schematic diagram of an exemplary embodiment of the method according to the present invention is shown. The method can be executed by a computer (system) 100 (of Fig. 1). First, the implantable device I can be configured to determine an event E associated with a cardiac episode of a patient. A cardiac episode (as outlined herein) can be a cardiac activity (e.g., ECG signal) over a certain duration during which an event occurs in a cardiac episode. The event E can include the occurrence of abnormal or suspicious cardiac activity in the cardiac episode. For example, the event E can include a Boolean value indicating whether abnormal (or suspicious) cardiac activity occurred in the cardiac episode. The presence of the event E (or a corresponding Boolean value) can serve as a flag that some cardiac episode has medical relevance. The implantable device I can further be configured to determine a type 1, 2, 3, 4 of the event E. The type of the event can be some type of cardiac arrhythmia. For example, the implantable device can be configured to determine that the type of the event E is atrial fibrillation onset 1. The implantable device can also be configured to determine that the type of the event E is asystole 2. The implantable device may also be configured to determine that the type of event E is high ventricular rate 3. For example, the implantable device may be configured to determine that the type of event E is bradycardia 4.

[0044] Thus, the event E and the types 1, 2, 3, 4 of the event E are the starting point of the method 200 according to the invention. The method 200 can reliably verify the determined event E and / or the type of the event E. When the method is performed by the computer (system) 100, the method 200 may first include receiving information related to the event E from the implantable device I (e.g., via the communication link 101 as outlined for FIG. 1). For example, the received information may include the event, the type of the event, and the corresponding cardiac episode (e.g., the corresponding ECG signal as outlined herein).

[0045] In other words, method 200 can be used to verify a snapshot (i.e., a certain time window of an ECG signal, i.e., a cardiac episode) in which implantable device I has determined a particular type of event (e.g., atrial fibrillation onset, asystole, high ventricular rate, and / or bradycardia). In the following, for exemplary purposes of describing signal processing, a cardiac episode will often be referred to as an ECG signal of event E. Thus, the verification performed by method 200 can determine event E as a false positive FP or a true positive TP.

[0046] However, if the type of event E is one of types 1, 2, 3, 4, then signal processing and / or filtering specific to the type of event is applied to the ECG signal of event E, e.g., different processing and / or filtering depending on which type the event represents. Thus, the validation is based on the type of event E. First, pre-processing is applied to all ECG signals of event E of types 1, 2, 3, 4. Pre-processing (not shown in FIG. 2) may include single-phase edge smoothing, which may smooth plateaus, steps (e.g., stepped plateaus), single (point) peaks and / or valleys of the ECG signal of event E. Pre-processing may also include adapting the bit depth (i.e., sampling depth) of the ECG signal of event E.

[0047] Then, an upsampling US can be performed. The upsampling US can include an interpolation (e.g., from 128 Hz to 1024 Hz). The upsampling US allows the ECG signal of the event E to be adapted in terms of bit depth and sampling frequency. The interpolation can be performed by a combination of a moving linear regression and (subsequently) a moving average. The interpolated (i.e., upsampled) ECG signal of the event E can also be referred to as the reconstructed ECG signal of the event E. Furthermore, an error correction can be applied. In order to correct the reconstructed ECG signal of the event E, the downsampling performed by the original device (e.g., the implantable device I) that acquired the original ECG signal of the event E can be repeated for the reconstructed ECG signal of the event E. Thus, the error correction can minimize the deviation between the original ECG signal and the downsampled (original) signal.

[0048] Then, a QRS peak detection QD is applied to all ECG signals of event E of type 1, 2, 3, 4. The QRS peak detection QD can be performed as outlined herein. First, a QRS complex detection is performed by a modified Elgendi algorithm. The Elgendi algorithm can detect a region of interest (i.e. a so-called QRS block of a QRS candidate) within which the QRS peak is searched. For the search, a global filter function is applied to the QRS block. The filter function is configured such that a maximum of a squared bandpass signal associated with the QRS block of the ECG signal of event E is found. The squared bandpass signal can be the output of the Elgendi algorithm. In particular, the filter function is configured such that a (local) maximum having a maximum slope and located furthest to the left within the region of interest is found. In this regard, left can refer to an earlier time within the region of interest. The goal is to find a (local) maximum (i.e. a local peak) with a maximum peak height and a maximum signal absolute value, and the (local) maximum located to the left of the region of interest is selected by the filter function. The most suitable peak according to the filter function can be determined as the QRS peak of the ECG signal of the event E. The method can further include detecting all QRS peaks of the cardiac episode (e.g., the ECG signal of the event E) as outlined herein. In other words, it can be said that the QRS peak detection QD is based on maximization of the squared bandpass filtered signal after modification with a slope-aware left-tilted Picard kernel.

[0049] Next, we outline the signal processing for verifying events E of types 1, 2, 3, and 4.

[0050] If the event E is type 1 (i.e., atrial fibrillation onset), a P-wave detection PW is applied to the ECG signal of the event E to find P-wave candidates (which may be related to QRS blocks). In this regard, a region of interest is expanded to search for P-wave candidates. Such expansion can be performed to find cardiac P-wave peaks that usually precede the QRS complex in time. According to the method, the time segment (e.g., RR interval) can be a time window between two subsequent QRS peaks. The two subsequent QRS peaks can be referred to as the first QRS peak and the second QRS peak according to the temporal order. A certain proportion of the time segment constitutes a time region in which P-wave candidates can be assumed. Thus, the time region constitutes a region of interest (ROI) for finding P-wave candidates. The time region is also referred to herein as the vicinity of the QRS peak. The percentage can be variable depending on the method (e.g., the percentage can be at least 10%, 20%, 40%, 50%, etc., for example, the inventors have found that about 30%-50%, e.g., 40%, provides good results in finding P-wave candidates). To illustrate an example, when the percentage is 50%, only the latter half of the time segment is related to the second QRS peak and thus becomes the region of interest for finding P-wave candidates. Thus, in this example, the latter half of the time segment is the time region that directly precedes the second QRS peak. When searching for a P-wave peak, the second QRS peak can serve as the origin (or reference coordinate). Thus, since the region of interest for finding the P-wave peak precedes the second QRS peak (i.e., the reference coordinate), the P-wave peak is searched in a negative time interval. To search for the P-wave peak, a slope correction for finding P-wave peak candidates is applied to the ECG signal of the event E in the region of interest. The slope correction here can be based on a moving linear regression. The moving linear regression may include calculating a slope and one or more regression coefficients. By applying a slope correction, global slopes are removed and local slopes are indirectly emphasized. The peaks obtained after the slope correction may enable a more reliable and certain finding of P-wave peaks. The peaks obtained may be sorted according to their characteristics to determine the P-wave peaks.For example, the peak with the largest peak height, the largest signal value, and the farthest from the second QRS peak (i.e., the origin, i.e., the reference coordinate) can be determined to be the P wave peak. A potential P wave peak can also be the outermost extremum from the slope with the largest slope (with respect to its length). In some instances, the immediate time region preceding the QRS peak may not be considered in the region of interest for finding the P wave peak (e.g., the immediate vicinity may be associated with the Q wave of the heart).

[0051] The determined QRS peak and the determined P wave peak can then be used to determine statistical features associated with the ECG signal of the event E. The QRS peak and the P wave peak can be used to determine heartbeat intervals (e.g., RR interval, PP interval, RP interval, etc.) and other features. The determined features can be used to validate 1000 the Type 1 (onset of atrial fibrillation) event, e.g., whether it is a true positive TP or a false positive FP. Some features are described in more detail below.

[0052] The features can include an RR interval. The RR interval (or simply RR) can be a one-dimensional vector of length N-1, where N is equal to the number of R waves (or R wave peaks) (e.g., in the ECG signal of the event E). The RR interval can be expressed as: RR n =R n+1 -R n (n can be defined as any suitable sequence that numerically indexes the R waves over a cardiac episode, e.g., n can represent an R wave and n+1 can represent a subsequent R wave).

[0053] Another feature can include RR_dist. RR_dist can take into account the variance and correlation of the mean squared differences between RR intervals. RR_dist can be a two-dimensional matrix of size (N-2) x (N-2), where N is equal to the number of R waves (e.g., in the ECG signal of event E). RR_dist can be RR_dist = Cor(Sqrt((RR m -RR m+1 ) 2 +(RR n -RR n+1 ) 2 )) It can be calculated as:

[0054] Further features can include dRR. dRR can take into account the derivative of the RR interval. dRR can be a one-dimensional vector of length N-2, where N is equal to the number of R waves (e.g., in the ECG signal of event E). dRR can be expressed as: dRR n =RR n+1 -RR n where n is a number that allows the RR intervals over a cardiac episode to be numerically indexed in their proper order.

[0055] Further features may include dRR_dist, which is similar to RR_dist, but which considers dRR instead of RR. dRR_dist mn =Cov(Sqrt((dRR m -dRR m+1 )2+(dRR n -dRR n+1 )2)) It can be calculated as:

[0056] Further features can include RR_Rm. RR_Rm can take into account the ratio of two adjacent intervals. RR_RRm can be a one-dimensional vector of length N-2 (N=number of R waves). RR_RRmn can be RR_RRm n =RR n+1 / RR (n being any suitable order capable of numerically indexing the RR intervals over a cardiac episode).

[0057] Further features may include RR_RRm_dist, which may be calculated as RR_dist (as outlined herein) using RR_RRm instead of RR.

[0058] Further features can include the PP interval. The PP interval (or simply PP) can be a one-dimensional vector of length N-1, where N is equal to the number of P waves (or P wave peaks) (e.g., in the ECG signal of the event E). The PP interval can be expressed as the PP n =P n+1 -P n (n can be defined by numerically indexing the P waves in an appropriate order over a cardiac episode, e.g., n can represent a P wave and n+1 can represent a subsequent P wave).

[0059] Further features can include the PR interval (or RP interval), which can be the interval from a detected P wave peak (or a determined P wave candidate) to the closest subsequent R wave peak (e.g., the QRS peak).

[0060] Further features can include a PQ_ratio, which can be a ratio of the number of P-wave candidates found to the number of R-waves found (e.g., the number of QRS peaks found).

[0061] The outlined features related to cardiac episodes can be statistically analyzed. The corresponding statistics can result in the respective statistical features. The statistical features can include at least one of the following: median, mean, standard deviation, ratio of standard deviation to mean, quantiles (e.g., 25% quantile, 50% quantile, 75% quantile), range (e.g., difference between maximum and minimum), kurtosis. The statistical features can also be based on histograms with fixed bin widths. For example, the statistical features can include max_bin (e.g., ratio of maximum bin entry (=maximum) over all entries (fixed bin width)), numbins (e.g., number of bins !=0 (fixed bin width)), frac0bins (e.g., number of bins over all bins==0 (fixed bin width)), n2diff (e.g., mean squared difference from maximum bin count to adjacent bin counts), etc.

[0062] A further feature may be RRautoCovRation, which may be the ratio between the maximum and second largest autocovariance values ​​of RR.

[0063] Further features may be RRIntdInt_NEC and / or RRIntdInt_NEC_Frac as described in Jie Lian, Lian Wang, Dirk Muessig, A Simple Method to Detect Atrial Fibrillation Using RR Intervals, The American Journal of Cardiology, Volume 107, Issue 10, 2011,Pages 1494-1497.

[0064] A further feature may be RRDFA, which may be based on a detrended fluctuation analysis (DFA) of the RR.

[0065] A further feature may be QRS_tmpl_mean_maj which may be the average correlation between the QRS of the main group (from the start of the wave to the end of the wave).

[0066] A further feature may be QRS_tmpl_mean_min which may be the average correlation between the QRS of secondary groups (from the start of the wave to the end of the wave).

[0067] A further feature may be Pcand_mean, which may be the average correlation between P-wave candidates (from the start of the wave to the end of the wave).

[0068] A further feature may be ectopy_frac. The feature ectopy_frac may be the ratio of the number of ectopies to the number of normal QRS (intervals) (and thus may be thought of as an ectopy probability). An ectopy may be counted if the first RR interval is less than or equal to a short threshold and the subsequent second interval is greater than or equal to a long threshold. The long and short thresholds may be updated for each QRS that was not counted as an ectopy.

[0069] In summary, with respect to the verification of type 1 events, the features outlined herein and statistical features (which may be based on the features outlined herein) may be the basis for the verification 1000 of the ECG signal of the type 1 event. As outlined herein, these features may be input to a machine learning model and / or an artificial intelligence system. The machine learning model (and / or the artificial intelligence system) may then perform a binary classification of the type 1 event into a true positive TP or a false positive FP. In one example, the training data for the machine learning model is based on historical data in which medical personnel (e.g., cardiologists) classified the ECG signal of the type 1 event E into false positives and true positives by visual inspection. Thus, accurate verification of the type 1 event E may be ensured since medical expertise is taken into account.

[0070] As shown in FIG. 2, if the event E is type 2 (i.e., asystole), then the QRS peak detection QD is performed, and then the clipping detection CL is performed on the ECG signal of the event E. In some embodiments, if the event E is type 3 or type 4 (i.e., high ventricular rate or bradycardia), then the QRS peak detection QD is performed, and then the clipping detection CL can be performed on the ECG signal of the event E (not shown). In one example, after the clipping detection CL, the verification of the type 2 event can include removing the RR intervals associated with the amplitude clipping (as outlined herein) from the ECG signal of the event E. The clipping detection CL can also be referred to as determining the region with long positive or negative signal saturation. In one example, the RR intervals adjacent to the RR intervals that include the amplitude clipping are also removed. In another example, the RR intervals associated with the clipping (and preferably adjacent RR intervals) are not removed as such from the ECG signal of the event E, but can be selected and / or marked. Thus, the "clipped" RR intervals can still be analyzed, but taking into account the information that the amplitude clipping is present.

[0071] The verification 2000 of a type 2 event E can be based on detecting whether a pause associated with type 2 (i.e., asystole) is present between subsequent QRS peaks in the ECG signal of the type 2 event E, preferably corresponding to an RR interval that is not removed or selected / marked (as outlined herein). A pause can be considered as an abnormally long pause (long RR interval) between subsequent QRS peaks. A pause can be determined based on a QRS peak detected by a method as outlined herein. However, the detection of a long pause can also take into account various other aspects (e.g., features as outlined herein and / or corresponding statistical features). For example, a cardiac episode in which a pause is detected can be considered a true positive TP, and one in which a pause is not detected can be considered a false positive FP. In another example, a cardiac episode with clipping within a detected pause can be classified as a false positive FP, and a cardiac episode without clipping within a pause can be classified as a true positive TP. It is also conceivable that an episode without a pause can be classified as a false positive FP without the need to detect clipping. In summary, validation 2000 of a type 2 event E (as outlined herein) allows determining whether the corresponding ECG signal of event E is a false positive FP or a true positive TP.

[0072] As shown in FIG. 2, if the event E is type 3 (i.e., high ventricular rate), then after QRS peak detection QD is performed, validation 3000 of the corresponding event E is performed as outlined herein. Validation 3000 may include calculating a heart rate based on the ECG signal of the type 3 event E. Furthermore, validation 3000 may include comparing the calculated heart rate with a predefined threshold (as described herein). This may allow determining whether the type 3 event E is a false positive FP or a true positive TP. For example, if the calculated heart rate exceeds a threshold (or a standardized threshold) programmed into the device, the type 3 event E is classified as a true positive TP, otherwise it is classified as a false positive FP.

[0073] As shown in FIG. 2, if the event E is type 4 (i.e., bradycardia), a QRS peak detection QD is performed, followed by a corresponding validation 4000 of the event E, as outlined herein. The validation 4000 may include calculating a heart rate based on the ECG signal of the type 4 event E. Furthermore, the validation 4000 may include comparing the calculated heart rate with a predefined threshold (as described herein). This may allow determining whether the type 4 event E is a false positive FP or a true positive TP. For example, if the calculated heart rate is below a threshold (or a standardized threshold) programmed into the device, the type 4 event E is classified as a true positive TP, otherwise it is classified as a false positive FP.

[0074] For example, the implantable device may be configured to determine additional or fewer types of events.

[0075] For example, in the example of FIG. 2, the implantable device can be configured to determine that the type of event E is type 5, meaning that the event is not type 1, 2, 3, or 4 (i.e., the event is not atrial fibrillation onset, asystole, high ventricular rate, or bradycardia). In the illustrated example, type 5 events are not processed and are automatically treated as related episodes. This results in the type 5 event being determined to be a true positive TP. According to method 200, if the type of the event is type 5, event E is determined to be a true positive TP without further signal processing and / or filtering. In other examples, type 5 events and / or other types of events can also be processed.

Claims

1. 1. A method for analyzing an event (E) determined by an implantable device (I), comprising: the method including validating the event based at least in part on a cardiac episode associated with the event; the verifying is based at least in part on the type (1, 2, 3, 4) of the event; method.

2. 2. The method of claim 1, wherein the type comprises at least one of atrial fibrillation onset (1), asystole (2), high ventricular rate (3), and bradycardia (4).

3. The method of claim 1 , wherein said validating comprises determining whether said event (E) is a true positive (TP) and / or whether said event is a false positive (FP).

4. The method of claim 1 , further comprising receiving information related to the event (E) from the implantable device (I).

5. 2. The method of claim 1, wherein if the type includes atrial fibrillation onset (1), the verifying includes detecting a P-wave peak in the cardiac episode by determining a moving linear regression in the vicinity of a predetermined QRS peak.

6. 6. The method of claim 5, wherein said verifying is further based at least in part on one of RR interval, PP interval, RP interval, correlation of QRS complex morphology and / or ectopy probability in said cardiac episode.

7. 6. The method of claim 5, wherein the verifying is performed by an artificial intelligence system and / or a machine learning system trained with events verified by manual review of the cardiac episode.

8. 2. The method of claim 1, wherein if the type includes asystole (2), high ventricular rate (3) and / or bradycardia (4), the verifying includes removing from the cardiac episode any RR intervals associated with amplitude clipping, and preferably further removing any RR intervals adjacent to at least one RR interval associated with amplitude clipping.

9. The verifying further comprises: Detecting whether, in said cardiac episode, a pause associated with said asystole exists between subsequent QRS peaks, preferably corresponding to a (long) RR interval that has not been eliminated according to claim 8.

9. The method of claim 8, comprising:

10. The verifying further comprises: If a pause is detected, determining that the event (E) is a true positive (TP), otherwise determining that the event is a false positive (FP).

10. The method of claim 9, comprising:

11. 2. The method of claim 1, wherein if the type includes high ventricular rate (3) and / or bradycardia (4), the verifying includes calculating a heart rate based on the cardiac episode.

12. 12. The method of claim 11, wherein the verifying is based at least in part on comparing the calculated heart rate to a predetermined threshold.

13. An apparatus (100) for analyzing events (E) determined by an implantable device (I), the apparatus being configured to perform the method of any one of claims 1 to 12.

14. A system (S) for analyzing an event (E) determined by an implantable device (I), comprising: An apparatus (100) according to claim 13, An implantable device (I); A system comprising:

15. A computer program comprising instructions for carrying out a method according to any one of claims 1 to 12 when the computer program is executed.

Images