Devices for accurately classifying cardiac activity

ES 2 525 691 T3

DESCRIPTION

Devices to accurately classify heart activity

Countryside

The present invention relates generally to implantable medical device systems that detect and analyze cardiac signals. More particularly, the present invention relates to implantable medical devices that pick up cardiac signals in the body of an implant wearer to classify cardiac activity as likely benign or malignant.

Background

Typically, implantable cardiac devices detect cardiac electrical signals in an implant wearer and classify the implant wearer's heart rhythm as normal / benign or malignant. Illustrative malignant rhythms can include ventricular fibrillation and / or ventricular tachyarrhythmia. The precision with which an implantable medical device analyzes captured signals determines how well it performs therapy and makes other decisions.

WO 2004/105871 (A1) discloses a detection architecture for use in implantable heart rhythm devices that provides procedures and devices to distinguish between arrhythmias. Enhanced specificity at the source of the identified arrhythmia is explored to better distinguish between rhythms appropriate for device therapy and those that are not.

New and / or alternative methods and devices for cardiac signal analysis are desirable.

Summary

The present invention relates to an implantable cardiac stimulation device as defined in the claims.

Various illustrative embodiments of the present invention are directed toward greater precision in the analysis of cardiac signals by means of implantable medical devices. Some illustrative embodiments identify an oversensing of cardiac events. Some illustrative embodiments also correct at least some captured data and use the corrected data to make operational decisions. The invention can be implemented in devices.

Brief description of the drawings

FIG. 1 is a block diagram for an illustrative procedure for identifying oversensing and taking corrective action;

FIG. 2 is a block diagram further illustrating an example for identifying oversensing and making therapy decisions;

FIG. 3 shows an illustrative implantable medical device;

FIG. 4 is an illustration of a detection profile that can be used while detecting cardiac events in an implantable medical device;

FIG. 5 is a graphic illustration of a dual detection in which both R and T waves are detected in each cardiac cycle;

FIGURES 6A-6B show an illustrative procedure of a morphological analysis of the detections in FIG. 5, relative to a stored R wave pattern;

FIGURES 7A-7B provide a detailed example of illustrative oversensing identification using morphology analysis;

FIG. 8 shows an illustrative example of the analysis for marking similar and different episodes in FIGS. 7A-7B;

FIG. 9 shows an illustrative oversensed cardiac signal having alternating long-short-long intervals;

ES 2 525 691 T3 FIG. 10 illustrates an analysis of a method of identifying an alternating interval oversensing;

FIG. 11 shows an overdetected broad QRS complex;

FIGURES 12A-12D show an illustrative application of broad complex oversensing identification standards;

FIGURES 13A-13B illustrate the management of standards set analysis results of FIGURES 12A-12D;

FIG. 14 is a process flow diagram for an illustrative broad complex oversensing identification procedure;

FIG. 15 provides a graphical illustration of a data analysis between detections for an illustrative True-False marking;

FIG. 16 shows an illustrative example for integrating a waveform evaluation procedure with morphology, alternating interval, and broad complex oversensing procedures;

FIG. 17 illustrates how modifications to the detection profile may not prevent oversensing in some circumstances;

FIGURES 18-21 provide graphical illustrations of managing suspect markers and oversensing markers in a stream of captured episodes;

FIG. 22 is a process flow diagram for an illustrative loading confirmation procedure; and FIG. 23 shows an illustrative procedure of an analysis.

Detailed description

The following detailed description should be read with reference to the drawings. The drawings, which are not necessarily to scale, show illustrative embodiments and are not intended to limit the scope of the invention.

Some of the following examples and explanations include references to issued patents and pending patent applications. These references are for illustrative purposes and are not intended to limit the present invention to the particular procedures or structures with respect to the referenced patents and patent applications.

Unless implicitly required or explicitly stated, the following procedures do not require any particular order of steps. It should be understood that when the following examples refer to a "current episode", in some embodiments, this means that the most recently detected cardiac event is being analyzed. However, this need not be the case and some embodiments carry out an analysis that is delayed by one or more detections and / or a fixed period of time.

The following illustrative examples use rectified captured signals for event detection purposes, for example, as shown in FIGURES 5, 7A (at 148), 9, 11, 12C-12D, 17 and 18. Some illustrative examples carry out an analysis of characteristics (morphology) of the signals captured using a rectified signal, as shown, for example, by means of FIGURES 6A-6B, 7A, 11, 12A-12D. The options shown regarding the use of rectified / unrectified signals are merely illustrative, and can be changed if desired.

The nomenclature used herein indicates that a signal is detected by means of an implantable heart device system, events are detected in the detected signal, and cardiac activity is classified using the detected events (detections). Rhythm classification includes the identification of malignant rhythms, such as ventricular fibrillation or certain tachyarrhythmias, for example. Implantable therapy systems make therapy / stimulus decisions based on heart rhythm classification.

In an illustrative example, a detected episode is detected by comparing the received signals to a detection threshold, which is defined by a detection profile. FIGURES 4 and 17, below, provide illustrative examples of detection profiles. Some embodiments of the present invention incorporate detection profiles

ES 2 525 691 T3 and an associated analysis as set forth in US Provisional Patent Application No. 61 / 034,938, entitled ACCURATE CARDIAC EPISODE DETECTION IN AN IMPLANTABLE CARDIAC STIMULUS DEVICE, filed March 7, 2008. Any profile can be used adequate detection.

The detected episodes are separated by intervals, for example, as shown in FIG. 18 to 602. Multiple bins can be used to generate an average bin over a selected number of bins. Some examples below use four intervals to calculate a mean interval. Some other number of intervals can be used, as desired. The detected heart rate can then be calculated using the average interval.

An electrocardiogram includes several portions (commonly referred to as "waves") that, according to a well-known convention, are labeled with letters including P, Q, R, S, and T, each of which corresponds to particular physiological episodes. The design of detection algorithms to detect the R wave is typical, although any portion, if detected repeatedly, can be used to generate a beat frequency. Using morphology (shape) analysis, in addition to heart rate, the system can capture and / or analyze the portion of the cycle that includes Q, R, and S waves, called the QRS complex. Other portions of the patient's cardiac cycle, such as the P wave and the T wave, are often treated as non-searchable aberrations for the purpose of estimating heart rate, although this need not be the case.

Typically, for rate determination purposes, each cardiac cycle is only counted once. Oversensing (such as double or triple detection) can occur if the device reports more than one detected event in a single cardiac cycle. Each of FIGS. 5, 7A, 9, 11, 12C-12D, and 17 shows, in one form or another, oversensing. Examples include the detection of both an R wave and a posterior T wave (see FIGURES 5, 7A, 9 and 17) as well as multiple detections of a wide QRS complex (see FIGURES 11, 12C-12D and 17) . These examples are not intended to be exhaustive, and those skilled in the art will understand that detection procedures in implanted devices can be challenged by any number of variations in "normal" cardiac activity. For example, a P wave can be detected and can be followed by a detection of a posterior part of the QRS or a T wave of the same cardiac cycle. Oversensing can also occur if a noise causes an episode to be declared when no cardiac event has occurred, for example due to external noise or therapy, a pacemaker device, skeletal muscle noise, electrotherapy, etc.

Oversensing can lead to an excessive count of cardiac cycles. For example, if a cardiac cycle occurs and a detection algorithm declares multiple episodes detected, an oversensing has occurred. If the heart rate is then calculated by counting each of these detections, oversensing occurs. Heart rates calculated alone or in combination with other factors can be used to classify heart rhythms as malignant or benign. A dependent overcount of over-detected episodes can result in a high frequency miscalculation. An erroneous heart rate calculation can lead to rhythm classification and incorrect therapy decisions. Some embodiments are aimed at identifying oversensing and / or correcting associated data.

FIG. 1 is a process flow diagram for an illustrative procedure of identifying oversensing and taking corrective action. The illustrative procedure begins with an episode detection 10, in which the received cardiac signal is captured and compared to a detection threshold until the received signal crosses the detection threshold, resulting in a declaration of a detected event. . FIGURES 4-5 provide an illustration of the detection step 10. In FIG. 17 an additional example detection profile is also shown.

Next, the method carries out a step 12 of identifying an oversensing. This may include one or more of several analysis procedures including, as illustratively shown, a morphology analysis 14, an interval analysis 16, and a wide QRS analysis 18. FIGURES 6A-6B, 7A-7B, and 8 show an illustrative analysis 14 of morphology as part of identification 12 of oversensing. FIGURES 9-10 show an illustrative interval analysis 16 as part of the identification 12 of oversensing. FIGURES 11, 12A-12D, 13A-13B and 14-15 show an illustrative wide QRS scan 18 as part of an identification 12 of oversensing. Figure 16 shows an example in which a calculated heart rate is used to select from several procedures 14, 16, 18 for identifying oversensing.

After identification 12 of an oversensing, if one or more oversensing are identified, the procedure corrects data, as shown in 20. FIGURES 18-21 show illustrative data correction procedures that can be performed in step 20 If a data correction is not needed in step 20, the procedure can simply proceed to the next step.

ES 2 525 691 T3

Finally, the procedure includes a therapy decision, as shown at 22. A therapy decision 22 can classify a heart rhythm of the implant wearer. The therapy decision 22 may incorporate additional procedures such as a loading confirmation shown in FIG. 22. The procedure then returns to episode detection 10, as indicated by line 24.

The therapy decision 22 may include one or more of several forms of analysis. In an illustrative example, individual sensed episodes are marked shockable or non-shockable and an X of Y counter is maintained to determine whether the overall heart rhythm warrants therapy. The marking of individual episodes as shockable or non-shockable can take various forms, including determinations as a function of frequency and / or as a function of morphology, or combinations thereof. Some illustrative factors and combinations of factors that can be considered are set forth in US Patent No. 6,754,528, entitled APPARATUS AND METHOD OF ARRHYHMIA DETECTION IN A SUBCUTANEOUS IMPLANTABLE CARDOVERTER / DEFIBRILLATOR, and in US Patent No. 7,330,757 entitled METHOD FOR DISCRIMINATING BETWEEN VENTRICULAR AND SUPRAVENTRICULAR ARRHYTHMIAS.

The therapy decision 22 may also take into account the persistence of a malignancy. Some illustrative examples are shown in US Patent Application Publication No. 2006/0167503 entitled METHOD FOR ADAPTING CHARGE INITIATION FOR AN IMPLANTABLE CARDIOVERTER-DEFRIBILLATOR. Other procedures may be used as a part of the 22 therapy decision. Also set forth below is a detailed example that uses multiple frequency zones to identify shockable episodes in therapy decision 22.

The procedure of FIG. 1 includes an oversensing identification 12 and a data correction 20. These stages are designed to improve classification results. The following examples provide details for implementing these steps in some illustrative embodiments.

FIG. 2 is a process flow diagram that further illustrates an example for identifying oversensing and making therapy decisions. Procedure 30 provides an example incorporating each of several different stages of identifying an oversensing, as well as additional analysis of captured data for waveform evaluation. The illustrative procedure begins with the declaration of a new detected episode, as shown at 32 (again reference is made to FIGURES 4-5 and / or 17 to show the use of the detection threshold in step 32).

The detected episode undergoes a waveform evaluation as indicated at 34. Waveform evaluation 34 analyzes the captured data in association with the detected episode to ensure that the detection is cardiac in origin. Waveform evaluation can mark detected episodes that have significant noise as suspicious episodes. For example, noise can be identified by counting the number of crossovers of the signal by zero, or of the first or second derivative of the signal, over a predetermined period of time. US Patent No. 7,248,921, entitled METHOD AND DEVICES FOR PERFORMING CARDIAC WAVEFORM APPRAISAL provides additional detailed examples of a waveform evaluation 34.

If the detected episode does not pass the waveform evaluation 34, it is marked as a suspicious episode and the procedure returns to step 32 and waits for a next crossing of the detection threshold. Once a detected event is captured that passes waveform evaluation 34, procedure 30 proceeds to the steps of analyzing the detections and identifying an oversensing. As shown at 36, illustrative procedure 30 determines whether a morphology pattern exists. A morphology model is a useful data set for a morphological comparison with one or more recently detected events. Morphology models can be comprised of implanted device systems or associated programmers, or can be selected or identified by medical personnel. US Patent No. 7,376,458, entitled METHOD FOR DEFINING SIGNAL TEMPLATES IN IMPLANTABLE CARDIAC DEVICES sets forth some examples of testing and / or modeling. In some examples, modeling is carried out by identifying a representative QRS complex that reflects an average or typical morphology of a cardiac cycle for an implant carrier.

In an illustrative example of an automatic pattern formation, a detected episode is identified and data for the detected episode is stored by means of a device as a preliminary model. In the illustrative example, the preliminary model can be validated by comparing the stored data with captured data for a number of adjacent detected episodes over time. If the set of detected episodes adjacent in time shows a high correlation with each other, the preliminary model is validated and a morphology model is defined using the preliminary model. If the preliminary model cannot be validated, it is discarded. Model formation may fail if the captured signal varies persistently, since high variability may prevent validation of a preliminary model. The query at step 36 determines whether a pattern is available to be used in an identification 38 of an oversensing in morphology.

In some systems, there will always be a morphology model. For example, some embodiments allow a physician to select a representative heartbeat during implantation or during a telemetry session as a

ES 2 525 691 T3 morphology model, or a representative model can be selected from a library of known models. If so, step 36 can be skipped.

At step 38, the morphology of one or more detected episodes is analyzed to determine whether one or more detected episodes are likely to be the result of oversensing. As shown below with reference to FIGURES 6A-6B, 7A-7B and 8, the steps may be carried out as part of step 38. This may include the identification of alternating patterns of morphology that indicate High-Low correlations. Raised to the morphology model.

After step 38 (if there is a stored morphology model) or step 36 (if there is no stored morphology model), the procedure continues at 40, where the measured heart rate of the implant carrier is considered. If the frequency is in an AI range (short for Alternating Range Range) the illustrative example continues with an alternating range oversensing identification as shown at 42. In alternating interval oversensing identification 42, the gaps between detected episodes are analyzed to determine if oversensing is occurring. The alternating interval oversensing identification procedure 42 may include the steps as shown below with reference to FIGS. 9-10.

Returning to step 40, if the implant wearer's heart rate is in the WC interval (wide QRS complex interval), then the wide complex oversensing identification procedure is called, as indicated in 44. The broad complex oversensing identification 44 is designed to identify wide QRS complex oversensing and may include the procedures set forth below with reference to FIGURES 11, 12A-12D, 13A-13B, and 14-15.

The AI ​​interval and the WC interval may be separated from each other, or there may be an overlap of these intervals, such that each of steps 42 and 44 are carried out. A further presentation of the integration of these procedures with reference to FIG. 16, below. In yet another embodiment, each of steps 42, 44 is performed regardless of the calculated heart rate.

In FIG. 2, after the applicable oversensing identification steps 38, 42 and / or 44, a data correction may be invoked, as shown at 46. A data correction 46 is invoked when one or more of the steps 38, 42 and / or Oversensing Identification 44 identifies oversensing. If no oversensing is identified, data correction 46 can be avoided.

In some examples, data correction includes recalculating intervals between detected episodes by removing one or more identified oversensing from the analysis. For example, if an oversensing is identified, then step 46 may manipulate stored data to correct the oversensing and reduce the calculated heart rate. FIGURES 18-21 further illustrate this concept in a particular series of examples.

The examples in FIGURES 18-21 buffer current detection frequency calculations by waiting until an interval between two detections is "certified" before using the interval for a frequency calculation. In some examples, a certified interval is considered if it passes a waveform evaluation 34 and the various stages 38, 42, 44 of identifying an oversensing without being marked as noise or as an oversensed event.

After data correction 46, the method makes a therapy decision 48. If no therapy is necessary, the procedure returns to block 32. If a therapy is indicated at step 48, then the therapy delivery and loading steps can be performed, as shown at 50. Typically, the devices Therapy implants using loading circuitry that takes a period of time to prepare the device for a therapy delivery. The procedure can be iterated multiple times after a load is initiated before a therapy can be delivered. The details of steps 48 and 50 may vary. Once a therapy is indicated at 48, a system can ensure that the therapy remains indicated until it is delivered. US Patent Application Publication No. 2006/0167503 entitled METHOD FOR ADAPTING CHARGE INITIATION FOR AN IMPLANTABLE CARDIOVERTER-DEFRIBRILLATOR provides some illustrative examples of these concepts.

FIG. 3 shows an illustrative implantable medical device and implant location. More particularly, in FIG. 3 an illustrative subcutaneous only system is shown. The subcutaneous system is shown with respect to a heart 60 and includes a receptacle 62 coupled to a wire 66. Preferably, receptacle 62 houses operating circuitry for conducting cardiac circuitry analysis and for providing a therapy output. The operating circuitry can include batteries, input / output circuitry, power capacitors, a controller, memory, telemetry components, etc., as is known in the art.

The electrodes are arranged at locations throughout the system that include, for example, an electrode 64 in the receptacle 62 and electrodes 68, 70, 72 in the wire 66. The electrodes 64, 68, 70, 72 can take any shape.

ES 2 525 691 T3 suitable and may be made of any suitable material. For example, receptacle electrode 64 may be an insulated button electrode or it may be a region or surface of receptacle 62, and electrodes 68, 70, 72 on wire 66 may be wound electrodes, ring electrodes, or other known structures. in technique.

Electrodes 64, 68, 70, 72 define a plurality of detection vectors such as V1, V2, V3, and optionally V4. If desired, one or more vectors V1, V2, V3 and V4 can be chosen as a default detection vector, for example, as set forth in US Patent Application Publication No. 2007-0276445 entitled SYSTEMS AND METHODS FOR SENSING VECTOR SELECTION IN AN IMPLANTABLE MEDICAL DEVICE. Other uses of multiple vectors are shown, for example, in US Patent No. 7,392,085 entitled MULTIPLE ELECTRODE VECTORS FOR IMPLANTABLE CARDIAC TREATMENT DEVICES. Another embodiment considers stance in vector analysis, for example, as set forth in US Patent Application Publication No. 2008-0188901 entitled SENSING VECTOR SELECTION IN A CARDIAC STIMULUS DEVICE WITH POSTURAL ASSESSMENT. Multiple detection vectors can be analyzed, sequentially or in combination, as desired.

Therapy can be applied using any chosen pair of electrodes. An illustrative example uses the canister electrode 64 and coil electrode 72 to deliver therapy. Other combinations of electrodes can be used. Therapy may include monophasic, biphasic, or other multiphasic defibrillation and / or various pacemaker operations.

The present invention is not limited to any particular hardware, implant location, or configuration. Instead, it is intended as an enhancement to any implantable heart system. Some illustrative examples may be associated with an external programmer 74 configured to communicate with the implanted device for various purposes, including, for example and without limitation, one or more of the following: device testing, loading new / revised software; modify perception, sensing, or therapy settings; determining the device's operating status, battery life, or wire integrity; and / or download data related to the condition of the implant carrier, prior to data capture, or treatment. Any suitable communication method can be used, such as various protocols and hardware generally known in the art.

FIG. 3 skips multiple anatomical landmarks. The illustrative system shown may be implanted under the skin outside of the implant wearer's rib cage. The location shown illustratively would place the socket 62 approximately in the implant holder's left axilla, level with the cardiac apex, with the wire 66 extending medially towards the xiphoid and then towards the head of the implant holder on the left side of the sternum. . An illustrative example uses a procedure / system as shown in US Patent Transferred Application Publication No. 2006-0122676 entitled APPARATUS AND METHOD FOR SUBCUTANEOUS ELECTRODE INSERTION. Other illustrative subcutaneous systems and locations are shown in lawfully transferred US Patent Nos. 6,647,292, 6,721,597, and 7,149,575.

The present invention may also be implemented in systems having various implantation configurations including, for example, other implantation configurations / locations only subcutaneous, only vascular and / or transvenous. Receptacle 62 can be placed in anterior, lateral, and / or posterior positions including, without limitation, axillary, pectoral, and subpectoral positions, as well as placements on either the left or right side of the torso and / or the abdomen of the wearer. implant. A completely intravascular implantation of the system has also been proposed. Wire 66 can be placed in any of a number of suitable configurations including anterior-posterior combinations, anterior-only combinations, transvenous placement, or other vascular placements.

FIGURES 4-5 illustrate a detection profile and how its use can, under given circumstances, lead to oversensing. With reference to FIG. 4, a detection profile is shown at 80 that includes a refractory period that is followed by an exponential decay. For illustrative purposes, the height of the refractory period is shown as the "estimated peak". The estimated peak is the implantable system's estimate of the peak amplitude of captured cardiac signals. Using the estimated peak allows the detection profile to be tailored to the amplitude of the captured signals.

The decay slope of the detection profile 80 uses the estimated peak (or, in some embodiments, a percentage of the estimated peak) as its starting point. The decay approaches the detection ground over time. The detection floor can be the maximum or highest sensitivity floor of the system, or it can be set to a predetermined level. Multiple decays can be used, as shown in US Provisional Patent Application No. 61 / 034-938. The decay can be exponential or it can take some other form, such as a straight-line decay, a progressive function, and so on.

FIG. 5 shows an application of the detection profile 80 of FIG. 4 on a captured signal, which is displayed at 104. The refractory periods are shown with a crosshatch at 100, 106, 112, and 118. The decays

ES 2 525 691 T3 exponentials 102, 108, 114 follow each refractory period 100, 106, 112, 118. When the detection profile meets the captured signal 104, a detected episode is declared and a refractory period begins. Therefore, when exponential decay 102 meets captured signal 104, a detected episode is declared and a refractory period 106 begins. In the example shown, oversensing occurs when T waves are detected, as occurs in association with refractory periods 106, 118 in addition to R waves associated with refractory periods 100, 112.

In the illustrative example of Figure 5, the estimated peak is calculated as the average of two previous peaks. As can be seen at 120, the estimated peak (represented as the height of refractory periods 100, 106, 112, 118) falls after the oversensing associated with refractory period 106, since the newly calculated estimated peak is an average of the R wave and T wave amplitudes. This can increase the likelihood of further oversensing by reducing the estimated peak to a level that is closer to more signal peaks that represent potential sources of oversensing.

Identification of a morphology oversensing

Some embodiments of the present invention provide exemplary methods for identifying and correcting oversensing. FIGURES 6A-6B, 7A-7B, and 8 present morphology-dependent approaches to identify oversensing using correlation. For illustrative purposes, these procedures apply to the oversensing shown in FIG. 5.

Some illustrative embodiments of a morphology oversensing identification identify alternating morphology patterns. For example, during oversensing, some episodes may be highly correlated to a stored pattern while other episodes may be poorly correlated (indicating oversensing), in an alternating pattern. When a sequence of comparisons produces High-Low-High correlations, the pattern can be attributed to oversensing. As shown below, Low correlation detected episodes can then be flagged as oversensing. An alternating sequence is a type of pattern, but other patterns can be searched for instead. In another example, a triple detection can be identified by using High-Low-Low triplets, and in yet another example, instead of a stored static morphology model, a series of detections can be compared with each other, making each new detection a separate model. Yet another example uses a dynamic model that changes over time, for example by integrating new detections into the model, or by averaging a plurality of previously detected episodes.

Referring now to FIG. 6A, a waveform correlation analysis is shown. Shown at 130 is a signal portion at refractory period 100 in FIG. 5 in the defined sampling window 132. FIGURES 6A-6B show the unrectified signal, while FIG. 5 shows the rectified signal. The sampling window 132 defines a number of samples 134, which are shown as a solid line for the sake of simplicity, since an understanding of the "sampling" of an analog signal in the digital domain is considered to be within the knowledge of those skilled in the art.

The sampling window also defines the alignment of the samples 134 around a triangulation point (typically the point of maximum amplitude) in the captured signal. Some illustrative procedures for defining a sampling window are set forth in US Patent No. 7,477,935 entitled METHOD AND APPARATUS FOR BEAT ALIGNMENT AND COMPARISON.

Although some embodiments may use the refractory period to define the sampling window 132, other embodiments tailor the sampling window 132 using features of the model 136. In an illustrative embodiment, the model 136 can be formed by analyzing one or more detected episodes to identify points start and end of QRS, as well as its position with respect to a triangulation point (such as the peak during the refractory period). These characteristics can be used to define the stored model so that it approximates the QRS complex. Other types of model can be used, including, for example, data transforms and set reduction techniques. Some models may also rely on multi-channel detection.

The samples in the sampling window 132 are compared to the stored model, which is displayed graphically at 136. The specific mathematical analysis of a model comparison can vary. The morphology analysis may include, for example and without limitation, a correlation waveform analysis (CWA), a reduced data set analysis including an identification and comparison of the location of the peak feature, a transformation of wave trains, a Fourier transform, a signal decomposition such as source separation, or other data analysis procedures, such as compression procedures. For the sake of simplicity, in the following examples, reference is made to the comparison in the form of correlation / CWA, with the understanding that these other analysis procedures may be substituted in other illustrative embodiments. The CWA can use a simplified calculation of the sum of absolute values ​​of differences between a model and a signal being analyzed, or the CWA can use an approach in which they are calculated

ES 2 525 691 T3 squares of the differences between the signal samples and the model samples and are used to find the correlation. Simplified procedures can be used to reduce computational difficulty.

With respect to the comparison in FIG. 6A, it may be noted that, as noted above, the signal at 130 comes from the refractory period at 100 in FIG. 5, which corresponds to the R wave of a cardiac cycle. As a result, there is a good correlation of the R wave with the morphology model.

FIG. 6B shows signal 140 derived from samples adjacent to the peak in refractory period 106 (FIG. 5) that occurs in association with oversensing of a T wave. As can be expected, signal 140, displayed in window at 142, represents visually poor correlation with stored pattern 136. Therefore, accurate detection in FIG. 6A shows a good correlation with the stored model, while oversensing in FIG. 6B visually represents a poor correlation with the stored model. FIGURES 7A-7B and 8 illustrate how these oversensing features can be utilized in FIG. 5 to identify oversensing.

With reference to FIG. 7A, a series of detections and associated refractory periods are displayed at 148, with the signal rectified. An unrectified signal is displayed at 150 including detected episodes 152, 154, 156. For illustration, the episodes are listed as shown at 148: episode 156 is episode N-1, episode 154 is episode N -2, and episode 152 is episode N-3. The most recent episode is displayed on the right side of the Episode Detection Chart 148. Detections 150 correspond to an R wave 152, a subsequent T wave at 154, and another R wave at 156. The sampling windows 160, 162, 164 are defined for each detection 152, 154, 156. In the example, it is displays the triangulation point for each sampling window as a vertical line. The triangulation point is offset to the left of the sample windows 160, 162, 164; you can use an offset triangulation point, but you don't have to.

The signal samples in each sampling window 160, 162, 164 are then compared to a model 172, as shown at 170. The results of the comparisons are shown as percentage correlations, indicated at 174. As shown, the R wave 152 score is high (95%), indicating a strong correlation with model 170. This makes the R wave 152 "similar" to the model, as indicated. Also, the R wave 156 score is high (90%), again indicating a strong correlation with pattern 170, hence the “similar” mark. However, the oversensed T wave 154 does not correlate well with model 170, and has a low (5%) correlation score and is marked "different." The numbers provided in FIG. 7A at 174 are provided for illustration only and are not the result of actual calculations.

After calculation of the scores at 174, the procedure then characterizes each score, as indicated at 182. That shows an illustrative characterization procedure in FIG. 8. With reference to FIG. 8, reference is made to a CWA, with scores provided on a scale from 0-100%. Three comparison zones are shown at 184, 186, and 188. The scores found in the first zone 184 are considered to be different from the stored model, while the scores found in the third zone 188 are considered to be similar to the stored model. The second zone 186 is treated as a hysteresis band in which episodes are marked identically as in the previous episode, for example, an episode that is in the second zone 186 that follows an episode that is in the third zone 188 would be marked "similar". In an illustrative example, the boundary between the first and second zones 184, 186 is set at approximately a 25% correlation, while the boundary between the second and third zones is set at approximately 52% correlation. Other limits and / or forms of this analysis can be used to mark similar and different episodes with respect to a model.

Referring again to FIG. 7A, the comparison scores 174 are characterized as shown at 182. The second detection 154 is a T wave and, due to a low comparison score, it is marked "different", while the other two detections are marked as "similar. ”. The marks "similar" and "different" are used to apply an oversensing comparison rule shown in FIG. 7B. The norms depend in part on the pattern shown in 190, in which the N-1, N-2, N-3 episodes form a similar-different-similar pattern. There are two parts to the oversensing comparison rule:

A) As shown in 192, look for an alternating pattern 190; Y

B) As shown at 194, the N-3 detection should score "high" and above the hysteresis zone 186 in FIG. 8.

As can be seen from the way the episodes are marked, the 194 rule effectively ensures that none of the three detections (N-1, N-2, N-3) has a correlation score that is in the hysteresis zone 186.

ES 2 525 691 T3

As stated in 196, if both standards are met, then the procedure marks one of the episodes (N-2) as morphology oversensing. In the illustrative example, the analysis looks at events N-3, N-2, and N-1. The timing of the analysis (using episodes N-1, N-2, and N-3 but not episode N) is merely illustrative, and the present invention is not limited to any particular architecture with respect to the timing of the analysis of individual episodes. .

The use of the oversensing marker is further discussed below with reference to FIGURES 18-21. In general, the procedure in FIGURES 18-21 does not differentiate morphology oversensing from other detections, however, if desired, treatment One or more overdetections may vary depending on the identification procedure. In another embodiment, data related to what type of oversensing analysis has identified the oversensing (s) can be saved to help analyze device operation, for example, to allow improved detection profile and / or oversensing analysis. .

In addition to morphology analysis, an interval timing can be considered. In one embodiment, morphology analysis oversensing is skipped if the intervals between the three detections are greater than a threshold, such as 500-1000 milliseconds. For example, it may be found that detections that are more than 800 milliseconds apart are unlikely to be the result of oversensing, or that an implantable system is unlikely to make any incorrect therapy decisions based on oversensing, which is what it means. result in intervals of 800 milliseconds (equal to 75 beats per minute).

Alternating Interval Oversensing Identification

As indicated in FIGURES 1-2, another illustrative procedure for identifying oversensing uses episode intervals to identify alternating interval patterns. It is believed that oversensing can be identified by analyzing the intervals between detected episodes. If an analysis of a set of detected episodes indicates an alternating pattern of long-short intervals between episodes, oversensing may be occurring.

FIG. 9 provides an illustration of a pattern of alternating long-short-long intervals. In particular, a captured signal is shown at 200. A detection profile similar to that of FIG. 4 to the captured signal 200. The result is systematic oversensing, with an R-wave detection shown in association with a refractory period at 204, and a T-wave detection shown in association with a refractory period at 206. This pattern repeats with detections associated with refractory periods at 208 and 210.

The intervals between detections are shown and characterized at 212, including short intervals 214 and long intervals 216. In a numerical example, if the refractory periods are approximately 100 ms, then the short intervals may be in the range of 200 ms, while that long intervals are in the range of about 450 ms. This would result in a detected heart rate of approximately 184 beats per minute (bpm) with an actual heart rate of only 92 bpm, the difference being attributed to persistent oversensing. A different duration can be used for the refractory period.

The long-short-long pattern provides another criterion for identifying oversensing. It may be more difficult to differentiate the pattern at higher frequencies, as the difference between long intervals 216 and short intervals 214 becomes less apparent. If desired, and as shown in FIG. 16, analysis of the alternating interval pattern can be omitted when the detected heart rates become relatively high.

FIG. 10 illustrates a pattern of alternating intervals. At 220, a correlation of interval durations is shown, with a center line shown at 222 as the mean interval. Any suitable number of intervals can be used to calculate the mean. An empty band is shown that has high and low limits, with the short interval limit shown at 224 and a long interval limit shown at 226. The empty band is defined by an empty band constant in the example shown. So, for example, if the mean of four intervals at a given point in time is 400 milliseconds (150 bpm), and an empty band constant of approximately 23 milliseconds is used (other empty band constants can be used ), then the limits 224, 226 would be 377 milliseconds and 423 milliseconds (142 and 159 beats bpm), respectively. Instead, a different definition of empty band can be used, eg just +/- 10 bpm, or an offset such as +10 milliseconds, -20 milliseconds.

In the illustrative example, to identify a pattern of alternating intervals, several rules are applied. First, the mean 222 of four intervals must be in a predetermined range, as indicated at 230. Some embodiments omit the standard 230. Second, a specific pattern must be found, as indicated at 232. In the illustrative example, pairs of consecutive intervals are considered, and there must be at least six crosses of the empty band by a line drawn between each interval within the above 8 intervals. The crossings of the empty band are shown in the illustrative example of FIG. 10 and are listed in 234. For example, the In-5 interval is longer than the 226 limit, and the In-4 interval is shorter than the duration defined by the 224 limit. Therefore, the pair In-5, In-4 crosses the empty band, increasing the count of pairs of intervals that

ES 2 525 691 T3 satisfy the specific pattern 232. Other parameters can be used to identify the alternating intervals, if desired.

Another norm is shown at 236 and calls for a Long-Short-Long pattern in the three most recent intervals. With reference to FIG. 9, it can be seen that when a Long-Short-Long pattern is formed due to oversensing of T waves, the short interval probably corresponds to the time from detection of R wave to detection of T wave. Standard 236 requires identifying a set of Long-Short-Long intervals in the intervals N-1 to N-3.

As shown in FIG. 10 at 238, if each rule 230, 232, 236 is met, then the N-2 detection is marked as an alternating interval oversensing. Incorporation of an oversensing flag into a frequency calculation procedure is further illustrated by FIGURES 18-21, below. FIGURES 18-21 illustrate selective data correction for frequency calculation; however, such data correction could interfere with the alternating interval oversensing identification procedure by combining intervals, eliminating short intervals, and avoiding the empty band crossovers that have been noted. Within the alternating interval analysis, the identification of an oversensing does not change the management of detected episodes. Thus, in an illustrative example, alternating interval oversensing procedures use uncorrected raw intervals (based on detections that pass a waveform evaluation) to establish the mean interval and to identify drifts below and above the empty band, instead of using corrected intervals.

The analysis as shown in FIG. 10 is an example of an alternating interval analysis. Other alternating interval analyzes can look for other time- or interval-based patterns in a queue of intervals between detected episodes. Examples include a triple detection (long-short-long triplets), combinations (sets in which three intervals are captured, with the second and third intervals being approximately as long as the first interval, indicating a correct detection followed by a detection pair double), or any other suitable timing-based pattern analysis.

Identification of large complex oversensing

Some embodiments of the present invention are directed toward identifying wide QRS complex oversensing. FIGURES 11, 12A-12D, 13A-13B, and 14 illustrate a broad complex oversensing identification. Broad complex oversensing identification procedures look for detections to occur at short intervals and with predetermined morphology characteristics. If proximity and morphology characteristics are identified, the broad complex oversensing identification procedure determines that oversensing has occurred.

Referring now to FIG. 11, the unrectified signal is shown as signal 290. Signal 290 demonstrates a broad QRS complex. The rectified version of signal 290 is shown at 300. As can be seen at 302, 304, the wide QRS is detected twice. This pattern repeats again in 306, 308.

FIGURES 12A and 12B show combinations of two standards that can be used, either alone or as mutual alternatives, to identify broad complex double detections. In FIG. 12A, at 320, shows N-1 and N-2 detections. For the purposes of applying the set of standards, the positive and negative peaks of each detection are marked as "p +" and "p-", respectively. The positive peak, p +, is marked at the point of maximum (or most positive) signal amplitude, and the negative peak, p-, is marked at the point of minimum (or most negative) signal amplitude during each refractory period.

FIG. 12A shows a first set of broad complex rules. A first rule is shown at 322 and is labeled as a detection range rule. The first rule 322 requires that the interval between detections (displayed as t1 324) be less than a predetermined value, called Standard_Duration_1. The second rule 326 is labeled Peak proximity rule and requires a time t2 328, which is the duration between the last peak (here, p-) of the N-2 detection and the previous peak (here, p +) of the detection N-1, is less than another predetermined value, named Duration_Norma_2 (it should be noted that time to scale is not shown).

In an illustrative example, Standard_Duration_1 is set to approximately 195 milliseconds. In another illustrative example, Standard_Duration_1 is set to the sum of the duration of the refractory period plus approximately 40 milliseconds. In an illustrative example, Standard_Duration_2 is set to approximately 20 milliseconds. Other values ​​can be used for Standard_Duration_1 and Standard_Duration_2. Some examples set the Standard_Duration_1 to a range of 150-240 milliseconds or, in other examples, the refractory duration plus 20-60 milliseconds. Some examples set the Standard_Duration_2 in the range of 10-40 milliseconds. Other formulations can also be used.

ES 2 525 691 T3

FIG. 12B shows a second set of broad complex rules. In FIG. 12B, at 330, shows a set of N-1 and N-2 detections, with positive and negative peaks marked with p + and p- indicators. A first rule at 332 is shown as a detection interval rule in which the interval t1 334 between detections is compared with the Duration_Normal_1. The second standard is shown at 336 and is called a polarity standard. The polarity standard determines whether the N-1 and N2 detections are of opposite “polarity”. For the purposes of the polarity standard, a detection is considered to have a positive polarity if the p + peak occurs before the p- peak; otherwise, the detection is negative. If the polarities of the two detections N-1 and N-2 are not the same, as shown, then the second rule 336 is satisfied.

The signals shown in FIGS. 12A-12B are simplified to highlight the use of the p + and p markers to identify the proximity and polarity of the peaks. FIGURES 12C and 12D provide examples that simulate more realistic signals in which broad QRS complexes are overdetected.

FIG. 12C illustrates an application of a wide complex set of rules to an oversensed signal having a QRS complex. The rectified version of the signal is shown in the upper portion of FIG. 12C to illustrate detections 340 and 342 that occur when the broad complex is over-detected. The unrectified signal is shown at 344. For a one-sided signal, as shown, the negative p-peak can be defined as the lowest amplitude sample. At the first detection 340 p-occurs before p +. By definition, this provides the first detection 340 with a negative polarity.

For the second detection, the p + occurs first, giving the second detection a positive polarity. Because the first detection has a negative polarity and the second detection has a positive polarity, the polarity rule is met. As noted, the detection interval standard is met. As a result, both the first and second standards are satisfied in FIG. 12B by means of the detected pattern of episodes shown in FIG. 12C.

FIG. 12D illustrates the application of another wide complex set of rules to another signal. Again, two detections 350, 352 are shown in rectified form for detection purposes. In FIG. 12D, the alternating polarity standard fails because both the first episode 350 and the second episode 352 have a positive polarity, with the positive peak of each occurring first. Moreover, the detection interval standard is met. In the example, the peak proximity rule is met because p- for the first detection 350 is near the end of the refractory period, while p + for the second detection 352 is near the start of the refractory period.

The sets of standards in each of FIGURES 12C-12D are met. FIG. 13A shows how episodes can be marked for frequency calculation purposes in FIGURES 18-21 using the results of the rule set and other conditions.

As shown in FIG. 13A, the broad complex oversensing identification procedure uses "true" and "false" markers for individual detected events. These markers indicate how confident the system is in individual detections. A marking of "false" indicates a lack of confidence in a given detection, which means that analysis of the identification procedure for a wide complex oversensing has found that the detection of false is likely to be oversensing. A "true" marking indicates that the broad complex analysis has not identified a given detection as likely oversensing. If many detection numbers are flagged false, oversensing is suspected. FIG. 22 provides an example of a loading confirmation procedure that can be used to verify therapy decisions before preparations are made for a therapy delivery if many detection numbers are marked False.

The true-false marking of FIG. 13A can be carried out independently of the heart rate. In an illustrative example, the additional marking shown in FIG. 13B of individual episodes such as suspicious wide complex and / or wide complex oversensing are only carried out while the detected frequency is in a predetermined range (FIG. 16). This heart rate range limit in the wide complex oversensing marking and / or the suspect wide complex marking may be omitted in some embodiments.

With reference to FIG. 13A, the illustrative example shows how the episodes can be marked given the initial circumstances and the results of the rules. For example, as shown in 362, when episode N2 is True and no set of rules is satisfied (FIGURES 12A-12B), then N-2 remains True and N-1 is marked True for the first time. At 364 another circumstance is shown, beginning with N-2 marked True. In this circumstance, a rule has been met, a morphology model is available for the system, and the correlation of the N-1 episode with the morphology model is better (a higher CWA score in the illustrative example) than the correlation of the episode N-2 with the morphology model. In such a 364 circumstance, episode N-2 has its marker changed from True to False, while episode N-1 is marked True. As illustrated by the circumstance in 364, marking an episode as

ES 2 525 691 T3

True is sometimes only a preliminary determination that can be changed later in the analysis.

Then, as shown in 366, in any other circumstance where episode N-2 begins as True and a set of rules is met, the result will be a marker of True for episode N-2 and a marker of False. for episode N-1. Finally, as shown in 368, if the initial circumstance is that N-2 has been marked as False, then the N-1 episode is marked True regardless of the result of applying the sets of standards in FIGURES 12A-12B. .

With reference to FIG. 13B, illustrative handling of True-False markers is shown. Management depends, in part, on the state of the system as shown at 380 and 390. A "pattern found" or "no pattern" state is the result of the identification of a detection pattern that indicates broad complex oversensing. The following are illustrative examples of patterns that can be used to identify "pattern found" and "no pattern" states.

As shown at 380, a first system state is one in which the calculated heart rate is in a predetermined range and a pattern has been found. When in this state, the procedure assigns broad complex oversensing markers to selected episodes. In the illustrative example, when the detected events N-3, N-2, and N-1 form a True-False-True sequence, then a broad complex oversensing marker is assigned to N-2. Otherwise, as shown in 384, no large complex oversensing marker is assigned. The use of the oversensing marker is further explained with reference to FIGURES 18-21.

As shown at 390, a second system state occurs in which the frequency is in range but the system is not in a found pattern state. As shown in 392, when detected episodes N-3, N-2, and N-1 form a True-False-True sequence, episode N-2 is assigned a suspicious episode marker. Again, the use of the suspicious episode marker is further explained with reference to FIGS. 18-21. In any other combination, no suspicious WC marker is assigned, as shown in 394.

As noted at 380, 390, there are defined "pattern found" and "no pattern" states, therefore some illustrative pattern search examples are shown below. In general, the approach is to identify particular characteristics of the overall rhythm, spanning multiple detected episodes, indicating that a broad-complex oversensing pattern appears likely. When such particular characteristics are identified, a "pattern found" state can be invoked, allowing the episodes to be flagged as oversensing.

It has been shown above in FIG. 10 a first example of a pattern that can be used to define "pattern found" and "no pattern" states as a pattern of alternating intervals. Different heart rate ranges can be used for broad complex oversensing analysis and alternating interval analysis, as indicated in FIG. 16, below. Therefore, in the illustrative example, when the frequency is in the wide complex range and the other standards (standards 232, 236) are met for an alternating pattern of FIG. 10, the "pattern found" status is entered.

Other patterns can also be used to establish a "pattern found" status. One example uses alternating markers of Suspicious Broad Complex Episode (Suspicious WC). An alternating pattern could be: [Suspicious WC] - [Not Suspicious] - [Suspicious WC] - [Not Suspicious]. Such a pattern of four episodes may be sufficient to introduce the found pattern state. In an illustrative example, only suspicious markers generated by the broad complex oversensing procedure are used to identify alternating suspicious episode marking. In another example, a larger set of episodes is used to establish the pattern and / or any source of suspicious episode markers can be relied upon to establish the pattern.

FIG. 14 graphically illustrates the transitions between the states of the system. The example in FIG. 14 includes two in-range states and one out-of-range state. In each state, the system performs a True-False dialing as set forth in FIG. 13A. True-False marking can be used in later stages, such as the upload confirmation shown in FIG. 22.

Illustration 400 provides an out-of-range state 402, in which oversensing and suspect broad complex marking is off (WC off). Out of range state 402 is effective when the sensed heart rate is outside of a predetermined range. When the heart rate enters the range, the system leaves the out-of-range state and enters a no-pattern interval 404 state.

Once in the 404 no-pattern interval state, the system begins searching for VFV sequences and if any are found, a suspicious episode marking occurs as shown in 390-392-394 on the

ES 2 525 691 T3

FIG. 13B. The system also looks for patterns that indicate oversensing is occurring. This may include observing a pattern of alternating long-short intervals and / or a pattern of suspicious WC episode markers. In one example, a pattern is searched as shown in FIG. 10. Once the frequency range and a pattern are found, the system then transitions to a state 406 in the found pattern range.

Once in state 406 in the found pattern range, if a VFV pattern is found, the system assigns a broad complex oversensing marker as explained at 380-382-384 in FIG. 13B. A transition from state 406 in the pattern range found to the state in the non-pattern interval 404 may occur if a timeout occurs without any broad complex oversensing marker being assigned. In an illustrative example, if 64 consecutive detected episodes pass without any broad complex oversensing marker being assigned, the pattern is considered lost and the system transitions from state 406 to state 404. The use of N = 64 is merely illustrative , and other thresholds can be used.

In illustration 400, from any state 404, 406, in the range if the calculated frequency is outside the frequency range, the system reverts to state 402 out of range. In an alternative embodiment, the system may wait to start a suspicious or oversensing episode dialing until a pattern has been found in addition to satisfying the frequency range. In yet another embodiment, instead of entering the state in the no-pattern interval 404 when the frequency enters the predetermined interval, the procedure may assume that a pattern exists and enter the state 406 in the pattern interval found immediately after meeting the frequency interval condition.

While in the out-of-range state 402, an illustrative procedure performs neither a suspicious WC nor WC oversensing marking as shown in FIG. 13B. If desired, the True / False dialing can be omitted while in the 402 out of range state. However, in one example, a True / False dialing is performed and can be used, upon transition to the frequency range, to immediately enter state 406 in the found pattern range. In another example, a True / False dialing is performed at all times, and a buffer of True / False markers and episode polarity indications are maintained to provide information to be used in load confirmation procedures displayed in Fig. 22. Correlation and episode width scores can also be performed.

Therefore, FIG. 14 provides an illustration of a system operation integrating the marking of True / False and suspicious and / or oversensing markers of FIGS. 13A-13B, which in turn apply the standards of FIGS. 12A-12D.

The above rules indicate that after a False episode marker, the next True episode is marked (rule 368 in FIG. 13A). However, marking an N-1 episode as “True” is a preliminary indication. During a next iteration of the procedure, an episode that was marked True when it was in the N-1 test slot can be marked false when it is in the N-2 test slot, as shown in the figure below. FIG. fifteen.

FIG. 15 shows an analysis of four episodes, a, b, c and d, which are detected episodes that occur consecutively that have each passed a waveform evaluation. As shown at 450, at time t1, episodes a and b are treated as episodes N-2 and N-1, respectively, for analysis of the norm of FIGS. 12A-12B. As shown in FIG. 13A, at 366, when the norms are met and the previous episode, N-2, does not have a lower correlation with the stored model than the last episode, N-1, the episode N-1 (episode b) is marked as False , while episode N-2 (episode a) is marked as True.

The procedure then returns 452, at which point, at time t2, episodes b and c are treated as N-2 and N1, respectively, and the rules would be applied again. Here, because episode b is already marked as False, episode c is automatically marked as True based on the rule shown in FIG. 13A at 368. If a broad complex frequency range is met, the VFV pattern will result in the N-2 episode (episode b) being marked either as a broad complex oversensing or as a suspicious broad complex, depending on whether there is any vigor, a pattern state found or not. The procedure then returns to 454.

As shown at 454, episodes c and d are treated as N-2 and N-1, respectively, and the rules of FIGURES 12A-12B apply. As noted, one of the sets of standards in FIGS. 12A-12B is again met. In the illustrative embodiment, the correlation to the stored morphology model of the last episode, N-1, is greater than the correlation of the previous episode, N-2. According to standard 366 of FIG. 13A, episode N-2 is marked as False and episode N-1 is marked as True. The result of this analysis is the marking of consecutive episodes b and c as False. It should be noted that at this point, an EOL pattern has developed. No complex or complex oversensing episode marker is applied

ES 2 525 691 T3 broad suspect to episode N-2 (episode c), since this pattern is not one of the marker patterns shown in FIG. 13B. However, consecutive False markers reduce confidence in the accuracy of event detection. FIG. 22 below provides an additional illustration of marking and analysis for a burden confirmation that can add persistence factors to therapy decisions when many episodes are marked False. In yet another embodiment, if desired, an FF match can be analyzed to determine whether the two detections indicate that a triple detection pattern has occurred. For example, consecutive False markers can result in a call to a morphology analysis to determine whether an immediately preceding or following episode corresponds to a stored or dynamic model. Alternatively, if a VFFV sequence is identified, the two detections marked True can be compared to each other; a high correlation may indicate that a triple detection has caused the intermediate False episodes.

Integration, data correction and upload confirmation

FIG. 16 graphically illustrates an integration of various oversensing analysis procedures. In illustrative example 470, an evaluation 472 of the waveform can be enabled on any detected frequency episode, as in a morphology oversensing analysis 474, if a model can be established to be used in the analysis.

As shown at 476, a wide complex oversensing identification scan is enabled in a relatively higher frequency area, with an alternating interval oversensing identification analysis enabled in a lower frequency area at 478. In some embodiments, an upper limit is placed on the 476 wide complex test, for example, in the range of 405 bpm calculated heart rate, and the boundary is set between the 476 wide complex test and the 478 alternating interval test. in the range of approximately 160 bpm.

These variables can change or can be omitted. The upper and / or lower frequency limits in the wide complex analysis 476 may be omitted, for example, as can the upper limit in the alternating interval analysis 478. Furthermore, instead of a strict "boundary", these analysis zones can overlap. Transitions can also take into account various hysteresis factors such as, without limitation, crossing the "limit" by an amount greater than some value (ie 20 ms or 20 bpm beyond the limit) and / or meeting the requirement of a selected number of consecutive detected episodes.

FIG. 17 illustrates the use of an enhanced detection profile. The end of FIG. 17 is to show that, for any given profile, it is likely that it will be possible to identify an implant carrier in which the profile may result in oversensing. The illustrative profile is similar to one shown in US Provisional Patent Application No. 61 / 034,938, entitled ACCURATE CARDIAC EPISODE DETECTION IN AN IMPLANTABLE CARDIAC STIMULUS DEVICE, filed March 7, 2008. As shown above, cardiac cycles labeled "T-wave oversensing" shown at 510, 512 are double counted, since both R waves and subsequent T waves result in detections. Furthermore, as shown above, the profile labeled "broad complex oversensing" also doubly detects the QRS complexes shown at 520, 522. Improving the detection profile may not prevent all oversensing.

FIGURES 18-21 provide illustrations of the management of oversensing and suspicious episode markers from the above illustrative examples. FIG. 18 provides an example of what happens during a “normal” detection, when neither episodes are marked as suspicious nor over detections. At 600 a buffer of detections and associated ranges is shown. The definition of detection and interval is indicated at 602: a crossing of the detection threshold is a detection and an interval is the period of time between consecutive detections.

As shown at 600, detections and intervals occur in an ongoing series, with a more recent detection being shown at 604, separated by a more recent interval 606 from a second more recent detection 608. For illustrative purposes, the examples in the FIGURES 18-21 operate using a delay of at least one episode; Instead, a real-time system could be used to analyze episode 602 as soon as it is defined.

As shown in 610, a scan window is defined to perform a scan on three detected episodes and associated intervals. As the 610 analysis completes, the detected episodes are marked as 612 certified detected episodes. In addition, an interval is marked as a certified interval if no suspicious episode or oversensing markers are applied to episodes that define the interval. The most recently certified interval is entered into a certified interval sequential record buffer (FIFO) 614, as indicated by line 616.

In the illustrative example, four certified intervals 614 are used in the FIFO buffer to calculate a 4RR mean 618, which is used to find a calculated heart rate 620 for the system. In the illustrative example, until the intervals are certified, they are not used in the frequency calculations. The

ES 2 525 691 T3 610 analysis will “certify” ranges to be used in a frequency calculation unless the suspect range is flagged or is combined as a result of a detection that is flagged as oversensing. Analysis 610 may include any of the above analyzes such as a waveform evaluation, a morphology oversensing analysis, an alternating interval oversensing analysis, and / or a broad complex oversensing analysis.

FIG. 19 shows an analysis when a suspicious episode marker is applied. In the above examples a suspicious event marker is shown in the above examples as a possible result of waveform evaluation analysis or broad complex oversensing identification analysis. In the 640 series of detections and intervals, the analysis window is displayed at 642. An episode at 644 is marked as a suspicious episode.

During operation, suspicious episode 644 is known to be unreliable, but it is not known whether suspicious episode 644 is, for example, an R wave masked by parasitic noise, double detection, or detection caused by external noise. Since the source of episode 644 is unclear, as indicated by its marking as suspect, each interval defined by episode 644, including both intervals 646 and 648, is determined to be unreliable for a frequency calculation. The procedure does not pass the slots 646, 648 to the certified slot buffer 650 which is used to generate the average 4RR 652 and therefore the frequency 654. Instead, 650 previous intervals that have already been certified are buffered until a new interval is certified. For example, if none of the detections on either side of interval 656 is marked suspicious or oversensing, then interval 656 will go to buffer 650 once it has advanced through analysis 642, as indicated by 658 .

FIG. 20 shows the treatment of oversensing markers. In the example shown, a persistent oversensing is flagged in the series of detections and intervals 700. The analysis window is displayed at 702, and the oversensing markers are displayed at 704. In the illustrative example, data is corrected when applied an oversensing marker. More specifically, the intervals around an episode that has an oversensing marker are combined, and the episode itself is discarded.

For example, an oversensing marker is applied to the detection at 706. The intervals 708, 710 on either side of the detection 706 are combined into a single interval 712. The detection 706 can also be discarded, for example, by remove it from the estimated peak calculation (s). This combined interval 712 is carried into the certified interval buffer 720. Also, a combined interval at 714 goes into buffer 720. As the analysis proceeds, the combined interval shown at 716 will also be added to the buffer 720 and used to generate the mean 4RR 722 and the frequency 724.

FIG. 20 provides a contrast to FIG. 19. When a detection is marked as suspicious, as in FIG. 19, it is not known whether an associated cardiac cycle has yet been counted. An oversensing 706, when identified, probably corresponds to a cardiac cycle that has already been counted by another detection. Therefore, it is determined that the data correction is appropriate by combining the intervals 708, 710 into the combined interval 712.

It should be noted that for FIG. 19 or FIG. 20, in addition to modifying the calculation of the frequency, the application of an oversensing or suspicious episode marker can also change the calculation of an estimated peak. As noted, oversensing or suspicious events can also sometimes lower the estimated peak and increase the likelihood of further oversensing. When an oversensing or suspicious episode marker is applied, some embodiments exclude one or more detected episodes from the estimated peak calculation. In an illustrative example, if two previous peaks would normally be averaged to calculate the estimated peak, if an oversensing or suspicious episode marker is applied, the larger of the two peaks can be used as the estimated peak.

FIG. 21 combines the above analysis of FIGURES 19-21 by showing a circumstance in which both suspicious episode and oversensing markers are applied. In data 750, a suspicious episode marker has been applied at 752. This results in the intervals shown at 754 being marked suspicious and being treated as unreliable and unusable.

Also in FIG. 21, the detection at 756 is marked as oversensing. Then, this detection 756 is discarded and the associated intervals 758, 760 are combined into a single interval 762. The combined interval 762 is used in the certified interval buffer 764, which is used to calculate the mean 4RR 766 and the heart rate. 768.

FIG. 22 provides an example of upload confirmation analysis. The procedure of FIG. 22 is largely directed toward an analysis of whether or not therapy is appropriate at a time when the heart rhythm is likely malignant. The analysis of FIG. 22 is largely an effort to avoid inappropriate therapy delivery to an implant wearer when oversensing occurs. You can carry out the

ES 2 525 691 T3 method of FIG. 22 as part of a therapy decision, for example, as shown at 22 in FIG. 1 or at 48 in FIG. 2. The analysis shown in FIG. 22 includes a start block that would start with an internal variable, denoted "Tolerance" in Figure, set to zero, and the start block would be called once other factors (such as the persistence and X / And noted above).

FIG. 22 makes use of two additional data sets. First, individual episodes are labeled 0, 1, or 2 using the True-False and polarity designations in FIGURES 12A-12D and 13A, as follows:

If True is checked, the episode is tagged 0.

If False is checked and has a positive polarity, the episode is labeled 1.

If False is checked and has a negative polarity, the episode is labeled 2.

Then, multiple counters are generated from a buffer of the 16 most recently detected episodes that have passed the waveform evaluation (this would include episodes marked as oversensing and / or suspicious by means of procedures other than a waveform evaluation). wave), as follows:

Total_Latidos_WC: number of detections marked as suspicious broad complex or wide complex oversensing in the buffer

Max_Cons_01: maximum number of consecutive combinations of labels 0-1 Max_Cons_02: maximum number of consecutive combinations of labels 0-2

Then these calculated variables are used as shown in FIG. 22.

Starting at a starting block 800, the procedure determines whether the Total_WC_beats is less than six, as shown in 802. If it is not, the procedure checks whether the Total_WC_beats is greater than or equal to eight, as shown in 804. If It is, the procedure determines whether Max_Cons_01 is greater than or equal to three, as displayed by 806. If it is not, the procedure determines whether Max_Cons_02 is greater than or equal to three, as displayed by 808. If it is not, the variable "Tolerance" is compared to five, which is an integer variable created to be used as a persistence factor in the flow chart of FIG. 22. If Tolerance is greater than or equal to five, as shown in 810, the charging confirmation procedure is satisfied, and the procedure returns an indication of the start of charging, as shown in 812, in order to charge charging capacitors. high potency to be used in administration therapy.

Returning to step 810, if Tolerance is not equal to or greater than five, the procedure goes to block 814, where Tolerance is increased and the procedure returns an indication that loading should not be started, as indicated at 816. Setting the Tolerance limit to five is merely illustrative, and larger or smaller settings can be used.

Going back to the procedure for fetching alternate results, if either block 806 or 808 returns a result of Yes, then the procedure sets the Tolerance variable to zero, as shown in 818, and the procedure returns an indication that no It should start loading, as shown in 816. Consecutive pairs of 0-1 or 0-2 that initiate the zeroing of the Tolerance variable from blocks 810 and / or 812 indicate repetitive double detections that have similar morphology over time. It is allowed to reset the Tolerance variable in each of these circumstances at least because a required level of polymorphic behavior does not occur that would be associated with very discordant ventricular fibrillation and / or arrhythmias (such as polymorphic ventricular tachycardia). . Such judgments regarding the heart rhythms to be treated aggressively or not may vary in some embodiments or in response to a physician's preference.

In some illustrative examples, a triple detection identification procedure may be called in addition to other oversensing identification procedures shown herein. The use of the True-False and 0-1-2 markings shown above can provide analytical tools for such triple detection identification. In such an embodiment, triple detection patterns are identified by observing whether a 0-1-2 or 0-2-1 pattern repeats, such as {0-1-2-0-1-2-0 .. .}, and a data correction can be performed to remove each of the detections of 1 and 2. Such an illustrative example may include an analysis of True (0) detections to determine if narrow QRS features can be identified.

At block 802, the Yes result likely indicates a shockable rhythm such as ventricular fibrillation. Therefore, the procedure goes directly to block 816 and returns to a result indicating that loading should begin. In some embodiments, this deviation from the "Tolerance" analysis can be omitted. Finally, if block 804 returns to a result of No, then the checks at 806 and 808 are determined to be unnecessary, and the procedure jumps to block 810 where the Tolerance variable is checked.

ES 2 525 691 T3

In an illustrative example, the loading confirmation procedure in FIG. 22 as a prerequisite for initiating high power capacitor charging in an implantable cardioverter-defibrillator or other implantable therapy delivery system. Once charging of the capacitor begins, in an illustrative example, the procedure of FIG. 22 until a shock is delivered or the episode ends.

FIG. 23 shows an example of an analysis. Some analysis procedures take an approach in which a series of buffers is filled during an analysis leading to a decision to administer therapy to a patient. For example, the heart rate can be measured and, once calculated as a tachyarrhythmic, a counter is started to determine how many consecutive rate calculations are produced that have the tachyarrhythmic rate. Once the tachyarrhythmic rate counter fills, a tachycardia condition is met and the device will perform additional morphology analysis to determine if the patient is displaying a monomorphic rhythm and / or if the patient's individual detected episodes are not correlated with a stored model. In this example, the morphology analysis occurs at the end of the analysis. By using morphology analysis only at the end of the analysis, it is de-emphasized. With morphology only at the end of the analytical procedure, incorrect therapy decisions may not be avoided.

Instead, the procedure of FIG. 23 in 900 using a different order. In particular, the procedure 900 follows a detection 902 of an episode with a 904 evaluation of the waveform, in which the detected episode is analyzed by itself to determine if it is likely to be caused by noise, or is masked by noise. same. As suggested above, the waveform evaluation 904 can take a form as shown in US Patent No. 7,248,921, entitled METHOD AND DEVICES FOR PERFORMING CARDIAC WAVEFORM APPRAISAL. This is followed by a morphology qualification, as shown in 906. The morphology score 906 includes one or more of the dual detection procedures shown above, such as a broad complex oversensing, a morphology oversensing, and an alternating interval oversensing.

The frequency is then estimated, as shown at 908. The frequency is characterized as being in one of three zones: a VF zone, a VT zone, and a Low zone. The VF zone is an elevated zone, typically greater than 180 bpm, and sometimes greater than 240 bpm, for example. The Lower zone is a non-malignant zone, for example, less than 140 bpm, although it possibly reaches up to 170 bpm for some patients, and even more, particularly in younger patients. The VT zone is defined between the Low zone and the VF zone. In this example, "VT" and "VF" are just labels, not diagnostics. Frequency estimation can make use of the procedures shown above that correct data in response to an identified oversensing.

If the rate is low, the detected episode is marked as non-shockable, as indicated by 910. If the rate is in the VT zone, an optional 912 detection enhancement can be called. In an illustrative example, the detection enhancement 912 includes a tiered analysis in which the detected episode is compared being considered with a static model. If the detected episode correlates well with the static model, the detected episode is marked as non-shockable 910. If the episode does not correlate with the static model but correlates well with a dynamic model made up of an average of four captured episodes recently and shows a narrow QRS complex (the combination suggests a monomorphic tachycardia that has narrow complexes), the procedure will also proceed to step 910. Otherwise, if detection enhancement 912 fails, the detected episode is marked as shockable, as shown in 914.

If the rate calculated at 908 is in the VF zone, the detection enhancement 912 can be avoided and the detected episode is marked as shockable, as shown at 914. In one embodiment, the implantable device is programmed to set the VT zone and VF zone limits and / or to bypass the VT zone. In yet another embodiment, the VF zone can be omitted, and all frequencies above the Low zone would be addressed through the detection enhancement 912.

Shockable and non-shockable markings are maintained on an X / Y counter, which provides an initial counter to determine whether therapy is proceeding. If the X / Y counter fails (for example counters such as 12/16, 18/24, or 24/32 can be used), then no therapy is applied and the analysis procedure ends with no shock 918. Then, the system waits to call the procedure again when the next detection occurs. The X / Y counter 916 may also integrate a persistence factor, for example, requiring the X / Y counter condition to be met for a series of consecutive detected episodes.

The illustrative procedure 900 also requires a charge confirmation check, as shown at 920. The charge confirmation check 920 may be as shown above in FIG. 22. If passed, the charge confirmation check results in the charge and discharge decision 922. The charge and discharge 922 can be called while the analysis continues to ensure that the patient's malignant rhythm does not correct itself only. If the patient's malignant rhythm returns to normal before a shock is delivered, the

ES 2 525 691 T3 procedure can finish the loading and unloading sequence 922. If the upload confirmation check 920 fails, the procedure ends at 918 again and waits for the next detected episode.

The procedure shown in Fig. 23 can be separated from the other procedures shown above to identify oversensing and / or to correct data resulting from oversensing.

Additional characteristics

Some embodiments take the form of devices that are directed toward monitoring cardiac activity. An example may be a continuous cycle implantable recorder. With reference to FIG. 1, to monitor performances, instead of a therapy decision 22, a decision may be made to store certain data for later loading instead. For example, some implantable monitors are configured to save data only when the implant makes a decision that abnormal and / or potentially malignant activity is occurring. In some additional embodiments, data can be stored when the captured data requires correction, so that the sensing and sensing characteristics of the system and / or the implant location are analyzed to determine its suitability for long-term use. A monitoring system can also produce a warning if a malignant condition is identified, for example by indicating the implant wearer or by communicating with an external alert system.

The X of Y counter referred to above may be integrated with a persistence factor as in US Patent Application Publication No. 2006/0167503, entitled METHOD FOR ADAPTING CHARGE INITIATION FOR AN IMPLANTABLE CARDIOVERTER-DEFIBRILLATOR. A persistence factor requires the X of Y counter requirement to be met for a predetermined number of consecutive iterations. In the illustrative example, the upload confirmation procedure of FIG. 22 is built in as an additional requirement that follows the persistence factor. That is, in the illustrative example, the load commit would be invoked only after the persistence requirement has been satisfied by meeting the X of Y counter requirement for a predetermined number of consecutive iterations. If / when a nonsustained arrhythmic condition is identified, the persistence factor and / or X / Y condition can be modified as explained in publication 2006/0167503.

The above illustrative examples can be implemented in many suitable ways. Some embodiments will be devices adapted to perform one or more of the procedures set forth above and / or a system that includes implantable devices and associated external programming devices. Some embodiments will comprise controllers / microcontrollers associated with stored sets of instructions to direct the operations of various components in a device according to one or more procedures.

Design details of the operating circuitry contained in a receptacle, such as receptacle 62 of FIG. 3, they can vary a lot. Briefly, an illustrative example may make use of a microcontroller driven system that includes an input switching matrix to select one or more signal vectors as a detection vector. The switching matrix is ​​coupled to filter circuitry and at least one input amplifier. Typically, the amplified filtered signal is supplied to analog-to-digital conversion circuitry. Additional filtering of the incoming signal can be carried out in the digital domain, for example 50/60 Hz band rejection filters. The incoming signal can then be analyzed using the microcontroller and any associated register and suitable logic circuitry. . Some embodiments include, for example, hardware for peak or event detection and measurement, or for morphology analysis such as correlation waveform analysis or wave train transformation analysis.

In various illustrative examples, upon identification of a rhythm that indicates a therapy, a charging operation is performed to charge one or more capacitors to levels suitable for a therapy. A load subcircuit can take any suitable form. An example uses a flyback transformer circuit, a structure well known in the art. Any method and / or circuit can be used that allows relatively low voltage batteries to charge capacitors to relatively high voltages. Some systems also perform an indication and / or communication in response to a detected malignancy, for example, to alert the implant wearer or a medical facility that therapy is imminent or intervention is necessary.

The device may further include output circuitry comprising, for example, an output H-bridge or a modification thereof to control the polarity of the output and a pulse duration from the high power capacitor. Control circuitry associated with the H-bridge may be included, for example, to monitor or control current levels for constant current output signals or to perform diagnostic functions.

The circuitry may be housed in a hermetically sealed receptacle made of any suitable material.

ES 2 525 691 T3

The foregoing description details various oversensing identification procedures and associated data correction procedures. Each of these procedures can be used individually in some embodiments. For example, the broad complex oversensing identification procedures shown below can be used as a stand-alone procedure to identify and, if desired, correct oversensing. In some embodiments, multiple procedures are used in a synchronized manner, for example, each of the morphology oversensing, alternating intervals, and broad complex oversensing procedures can be used together and can analyze individual detected episodes or groups of detected episodes. continually. In still other embodiments, a combination of these procedures is used in response to given conditions.

In addition to selective activation of separate oversensing analysis procedures, there are several ways to integrate the results of an oversensing analysis in addition to those shown by FIGURES 18-21. The following summaries provide alternatives and variants to the illustrative examples shown above . In an illustrative example, the results are integrated as follows:

1. Suspicious waveform evaluation episodes can be used in an oversensing analysis. Any episode marked as oversensing is discarded by any procedure with an associated interval correction, regardless of any Suspicious marker;

two. Any episode marked Suspicious by any procedure and not marked Oversensing by any procedure is Suspicious; Y

3. Any episode not marked Oversensing or Suspicious is considered certified once it is no longer eligible to be marked Oversensing or Suspicious.

This example allows detected events that fail the waveform evaluation to be used in a later identification of oversensing.

Some examples do not allow detected episodes that fail the waveform evaluation to be used in any subsequent analysis. Therefore, in another illustrative example, the outputs are integrated as follows:

1. Any evaluation marking of the waveform of a Suspicious episode prevents marking of that episode by any other procedure, and that episode and associated intervals are marked WA Suspicious;

two. Any episode marked Oversensing and WA Not Suspicious with an associated interval correction is discarded, regardless of any Suspicious marker;

3. Any episode marked Suspicious by any procedure other than Waveform Evaluation is Suspicious unless it has been marked Oversensing by any procedure; Y

Four. Any episode not marked Oversensing or Suspicious is considered certified once it is no longer eligible to be marked Oversensing or Suspicious.

In some embodiments, marking a detected episode as suspicious in a waveform evaluation disables the classification of adjacent episodes by oversensing procedures. This prevents a probable noise detection from causing an actual detection to be discarded. Certain counters can also be retained to avoid a hit by the suspicious WA episode, for example, when a pattern is identified for an alternate interval oversensing identification (or to enable the found pattern state of wide complex oversensing procedures ), suspicious WA episodes and one or more adjacent episodes can be excluded, if desired.

Although voltages and power levels may vary, in one example, an implantable subcutaneous cardioverter-defibrillator 50 includes charging circuitry and capacitors sized to receive and contain power at 1350 volts, and uses a driver / output circuitry that provides an output which produces a delivered load of 80 joules in a biphasic waveform with a slope of approximately 50%. Other voltage, energy and tilt levels (higher and / or lower), and other waveforms can be used, and the charge varies in response to the position and physiology of the electrodes. The output waveform setting need not be static, and any suitable procedure / setting can be used to provide the output (including, without limitation, pre-discharge waveforms, single-phase or multiphase waveforms, a progression or adaptation of the energy level or therapy voltage, changes in duration or polarity, fixed current or fixed voltage, etc.). Some embodiments use progressive therapies that include pacing antitachycardia pacing, as well as cardioversion and / or defibrillation pacing. The foregoing generally assumes two output electrodes (an anode and a cathode), however, it will be understood that other systems may be used including, for example, arrays and / or stimulus systems of three or more electrodes in which a pair or more of electrodes are used in common.

ES 2 525 691 T3

An analysis can take several forms in terms of the inputs taken. For example, a multiple detection electrode system can be configured to select a default detection vector and to use the default vector throughout the analysis. Other systems may prioritize vectors for use in a progressive analysis in which one vector after another is analyzed. Other additional systems can analyze multiple vectors simultaneously.

For conversion purposes in a digital domain, any suitable sample rate can be used. Some examples use 256 hertz; other frequencies can be used as desired. In addition, the illustrative examples shown may vary with respect to particular values, including, without limitation, changes to refractory periods, periods of proximity of episodes and peaks, frequency ranges, frequencies of "shockable" episodes, the number of intervals used to estimate the frequency, and any other values ​​provided. An analysis using "suspect" or "certified" episodes and intervals, an evaluation of the waveform, and other characteristics may vary, and some of these characteristics may be omitted in some embodiments. The completeness of the examples shown is not an indication that all parts are necessary for any given embodiment.

Those skilled in the art will recognize that the present invention may manifest itself in a variety of ways other than the specific embodiments described and contemplated herein. Consequently, modifications in form and in detail can be made without departing from the scope of the present invention.