Devices for accurately classifying cardiac activity

The present invention generally relates to an implantable medical device system that detects and analyzes cardiac signals. More specifically, the present invention relates to implantable medical devices that capture cardiac signals within the body of an implant subject to classify cardiac activity as benign or malignant.

Implantable cardiac stimulation devices typically detect cardiac electrical signals within the implant subject to classify the implant subject's cardiac rhythm as normal / benign or malignant. For example, malignant rhythms can include ventricular fibrillation and / or ventricular tachyarrhythmias. How accurately the implantable medical device analyzes the captured signal determines how well the implantable medical device makes treatment and other decisions.

<p num="0003"> New and / or alternative methods and devices for cardiac signal analysis are desired.</p>

<p num="0004"> One exemplary embodiment relates to the use of correlation analysis to identify overdetection of cardiac events. In one example, a high-low-high correlation pattern is sought for the template. The template may be a static template, the template may be a display of the latest capture event, or the template may be the average of some recent capture events. In another example, multiple boundaries for high correlation are defined, with the first higher boundary (which requires a larger correlation) being based on a smaller set of detection events compared to the second lower boundary. Allows identification of overdetection. In one embodiment, a short high-low-high sequence is sufficient for the first boundary, while for the second boundary, five or more (eg, eight) alternating events are long. A sequence is required. In another embodiment, the high-correlation and low-correlation provisions apply to a particular signal by establishing boundaries using averages for a subset of the correlation of detected events.</p><p num="0005"> In another embodiment, the correlation analysis is performed multiple times for a given template and detection events by shifting the alignment of the template and detection events to maximize the correlation score of the analysis. Such shifts may adjust the alignment by one or more samples away from the identified reference point for analysis. In another embodiment, the stored template is modified to address morphological changes for selected parts of the signal. In yet another embodiment, a plurality of features of the template and / or signal are identified and a plurality of correlation scores are calculated using several different features as alignment points.</p><p num="0006"> Once identified, overdetection can be corrected by modifying the stored data to affect rate analysis. In one such embodiment, data correction is stopped if the interval surrounding the potential overdetection is longer than a predetermined threshold. In one embodiment, the data correction is stopped if the interval analysis for the potential overdetection indicates that the potential overdetection is unlikely to be a particular type of overdetection. In one such embodiment, the interval surrounding the potential overdetection is analyzed to determine if the acceptance formula for estimating the expected QT interval is met, and if not, the method is latent. It is determined that the over-detection is not a T wave, and no data correction is performed.</p>

<figref num="1">FIG. 6 is a block diagram of an exemplary method of identifying overdetection and taking corrective action.</figref><figref num="2">It is a figure which shows an exemplary implantable cardiac stimulation system.</figref><figref num="3A">It is a figure which shows the Example which uses the correlation analysis to identify the over-detection.</figref><figref num="3B">It is a figure which shows the method step for an exemplary embodiment which includes rate correction.</figref><figref num="4">It is a figure which shows the example of the correlation comparison between events.</figref><figref num="5">It is a figure which shows another example of the correlation comparison between events.</figref><figref num="6">It is a figure which shows the analytical method for the correlation analysis of a short series and a long series.</figref><figref num="7A">It is a figure which shows the Example which applies the analytical method of FIG. 6 to the series of correlation analysis.</figref><figref num="7B">It is a figure which shows the Example which applies the analytical method of FIG. 6 to the series of correlation analysis.</figref><figref num="8A">It is a figure which shows the example which adjusts the correlation analysis with respect to the observed correlation level with respect to a template.</figref><figref num="8B">It is a figure which shows the example which adjusts the correlation analysis with respect to the observed correlation level with respect to a template.</figref><figref num="9">It is a figure which shows another method of aligning a captured signal with a correlation analysis template.</figref><figref num="10">It is a figure which shows another method of storing and applying a template for correlation analysis.</figref><figref num="11">It is a figure which shows the method of stopping the correlation analysis identification of over-detection.</figref><figref num="12">It is a figure which shows the method of stopping the correlation analysis identification of over-detection.</figref><figref num="13">It is a figure which shows the further method of stopping the correlation analysis identification of overdetection.</figref><figref num="14A">It is a figure which shows the application of the method shown in FIG.</figref><figref num="14B">It is a figure which shows the application of the method shown in FIG.</figref><figref num="15">It is a figure which shows the shock analysis method which discriminates a shockable detection event and a treatable rhythm.</figref><figref num="16">It is a figure which shows the method of calculating the correlation between a capture signal and a template.</figref>

A detailed description will be given below with reference to the drawings. Drawings that do not necessarily follow a constant scale show exemplary embodiments and are not intended to limit the scope of the invention. Some of the examples and descriptions below include references to issued and pending patent applications. These references are for purposes of illustration and are not intended to limit the invention to the particular methods or structures obtained from the referenced patents and patent applications.

Unless implicitly required or explicitly stated, the following methods do not require any particular order of steps. In the following examples, when we say "current event" in one embodiment, it means that the most up-to-date detected event has been analyzed. However, not necessarily in that situation, certain embodiments perform an analysis that is delayed by only one or more detections or by a period of time. The choices shown regarding the use of rectified / unrectified signals are exemplary only and may be changed at will.

The terms used herein indicate that the signal is detected by an implantable cardiac device system, events are detected within the detection signal, and cardiac activity is categorized by the use of detection events (detections). Rhythm classification includes identification of malignant rhythms, such as ventricular fibrillation or some tachyarrhythmias. Implantable treatment systems make treatment / stimulation decisions based on the classification of cardiac rhythms.

In one embodiment, the detection event is detected by comparing the received signal with the detection threshold defined by the detection profile. Any suitable detection profile may be used. Detected events are separated by intervals. Several intervals are available to generate an average interval over a selected number of intervals, from which the heart rate can be calculated. For example, 4, 8, or 16 intervals may be used to estimate the cardiac event rate as a function of the average interval.

As is well known, the electrocardiographic map is referred to as a number of letters (called "waves") that correspond to a particular physiological event and are represented by letters containing P, Q, R, S, and T. Often). It is common to design a detection algorithm that detects the R wave, but if it is repeatedly detected, any part of the cardiac cycle is used to generate the heart rate. If morphological (shape) analysis is used in addition to heart rate, the system may capture and / or analyze a portion of the period that includes the Q, R, and S waves, called the QRS complex. Other parts of the patient's cardiac cycle, such as P and T waves, are often treated as, but not necessarily limited to, artifacts that are not sought for the purpose of estimating heart rate.

Usually, each cardiac cycle is counted only once to confirm the rate. Overdetection (such as double or triple detection) can occur when declaring more than one detection event within a single cardiac cycle. Overdetection is when two or more parts of a single cardiac cycle are detected, or when no cardiac event has occurred, for example due to external treatment or noise, pacing artifacts, skeletal muscle noise, electrical treatment, etc. It can happen when an event is declared.

Overdetection occurs when one cardiac cycle occurs and the detection algorithm declares multiple detection events. Overcounting occurs if the heart rate is then calculated by counting each of these detections. The calculated heart rate may be used alone or in combination with other factors to classify cardiac rhythms as malignant or benign. Overcounting based on overdetected events can result in erroneously high rate calculations. Miscalculation of heart rate can result in inaccurate rhythm classification and treatment decisions. Some of these concepts are U.S. Patent Application Nos. 12 / 399,914 named "METHODS AND DEVICES FOR ACCURATELY CLASSIFYING CARDICA CACTIVITY" and "ACCURATE CARDIAC EVENT DETECTION IN MEDIA". Further discussed in Application No. 12 / 399,901.

FIG. 1 is a process flow diagram for an exemplary method of identifying overdetection and taking corrective action. An exemplary method begins with event detection 10, the received cardiac signal is captured and compared to the detection threshold, and finally the received signal crosses the detection threshold, resulting in a declaration of the detection event.

Next, the method carries out overdetection identification step 12. This may include one or more of several analysis methods, including morphological analysis 14, interval analysis 16, and wide QRS analysis 18, as illustrated. If one or more overdetections are identified following the overdetection identification step 12, the method corrects the data as shown in 20. In step 20, if no data correction is required, this step may be bypassed.

Finally, the method comprises a treatment decision, as shown in 22. The treatment decision 22 classifies the cardiac rhythm of the implant subject and determines if / when the treatment is delivered. The method then iteratively proceeds to event detection 10.

The treatment decision 22 may include one or more of several forms of analysis. In one embodiment, individual detection events are marked as shockable or non-shockable, and the X-out-of-Y counter indicates whether the overall cardiac rhythm benefits treatment. Is maintained to determine. Marking individual events as shockable or non-shockable may take several forms, including rate-based determination and / or form-based determination or a combination thereof. FIG. 15 below provides an example. A further embodiment is a US patent number 6,75, US patent number 6,75 VENTCULAR AND SUPRAVENTCULAR It is also disclosed in US Pat. No. 7,330,757, entitled "ARRHYTHMIAS".

Treatment decision 22 may also consider the persistence of the malignant state. One example is set forth in U.S. Patent Application Publication No. 2006/0167303, entitled "METHOD FOR ADAPTING CHARGE INITION FOR AN IMPLANTABLE CARDIOVERTER-DEFIBRILLATOR", the disclosure of which is incorporated herein by reference. Other methods may be used as the treatment decision 22.

FIG. 2 shows an exemplary implantable medical device and implant location. More specifically, an exemplary subcutaneous only system is shown in FIG. The subcutaneous system is shown for the heart 40 and includes a canister 42 that connects to the lead 46. The canister 42 preferably houses an actuating circuit element that performs an analysis of cardiac activity and provides a stimulus output. Working circuit elements can include, as is well known, batteries, input / output circuit elements, power capacitors, high voltage charging modules, controllers, memories, telemetry components and the like.

For example, the electrodes 44 on the canister 42 and the electrodes 48, 50, 52 on the lead 46 are located at multiple locations throughout the system. The electrodes 44, 48, 50, 52 may take any suitable form or may be made of any suitable material. For example, the canister electrode 44 may be an insulated button electrode, or the region or surface of the canister 42, and the electrodes 48, 50, 52 on the lead 46 may be coil electrodes, ring electrodes, or It may be another structure known in the art.

Electrodes 44, 48, 50, 52 define a plurality of detection vectors such as V1, V2, V3, and V4. If desired, one or more vectors V1, V2, V3, and V4 may be, for example, US Patent Application Publication No. 45-76 named "SYSTEMS AND METHODS FOR SENSING VECTOR SELECTION IN AN IMPLANTABLE MEDICAL DEVICE". It may be selected as the default detection vector as described in. Other uses of the plurality of vectors are shown, for example, in US Pat. No. 7,392,085, entitled "MULTIPLE ELECTRODE VECTORS FOR IMPLANTABLE CAEDICA TREATMENT DEVICES". Another embodiment is, for example, "SENSING VECTOR SELECTION IN A CARDIAC SIMULUS DEVICE WITH POSTURAL". Attitude is considered in vector analysis, as described in US Patent Application Publication No. 2008-0188901 entitled "ASSESSMENT". A plurality of detection vectors may be analyzed sequentially or in combination, as desired.

Treatment may be applied using any selected electrode pair. One embodiment uses a can electrode 44 and a coil electrode 52 to apply treatment. Other electrode combinations may be used. Treatment may include monophasic or polyphasic defibrillation, cardioversion, and / or pacing.

The present invention is not limited to any particular hardware, implant location, or implant configuration. Instead, the invention is intended as an improvement over any implantable cardiac therapy system. One embodiment is also a monitoring system to control monitoring functions (including advance notice and / or data recording) and / or to test the suitability of data analysis for a particular configuration, condition, or patient. May be used in.

One embodiment is not limited to, for example, device testing, uploading new / revised software, modifying programmable parameters such as detection or treatment settings, device operating status, battery life, or leads. Implantable for a variety of purposes, including determining integrity, enabling or disabling features, and / or downloading data about the implant subject's condition, previous data capture or previous treatment. It may relate to an external programmer 54 configured to communicate with the device. Any suitable communication method, such as various protocols and well-known hardware, may be used.

FIG. 2 omits some anatomical landmarks. The illustrated exemplary system may be implanted outside the thorax of the implant subject and under the skin. An exemplary location is where the canister 42 is placed approximately in the left axilla of the implant subject at the same height as the apex of the heart, with the lead 46 medial to the sword cartilage and then along the left side of the sternum. It extends toward the head of the implant subject. One exemplary embodiment uses the method / system set forth in US Patent Application Publication No. 2006-0122676 assigned to the same assignee named "APPARATUS AND METHOD FOR SUBCUTANEOUS ELECTRODE INSTERION". Other exemplary subcutaneous systems and locations are set forth in US Pat. Nos. 6,647,292, 6,721,597, and 7,149,575, which were assigned to the same assignee.

The present invention may also be embodied, for example, in other systems with various implant configurations, including subcutaneous only, blood vessel only, and / or intravenous implant configurations / locations. The canister 42 is placed in an anterior, lateral, and / or posterior position, including the axillary position, pectoral muscle position, and lower position of the pectoral muscle, without limitation, and on the left side of the torso of the implant subject or It may be placed on the right side and / or in the abdomen. Implantation of the system completely into the blood vessel has also been proposed. The canister 42 and lead 36 may be placed in any of several suitable configurations, including anterior-posterior combinations, anterior-only combinations, transvenous indwelling, or other vascular indwelling. The integrated system may omit the lead 46 and instead include all electrodes on the canister 42.

FIG. 3A shows an example in which correlation analysis is used to identify overdetection. The "Correlation analysis" used herein can take several forms. One embodiment is shown in FIG. With reference to FIG. 16, the captured signal 500 receives an analog-to-digital conversion 502 and, as shown in 504, forms a time-ordered series of samples that form a sampled (and usually digital) display of the signal. It brings {S1, ..., S9}. The example of FIG. 16 is simplified for illustration because the number of samples for a given signal can be greater than nine. For example, in one exemplary embodiment, the capture signal 500 is about 160 ms long and covers 41 samples captured at 256 Hz. Other durations and / or sampling frequencies may be selected. The signal can be windowed to approximately QRS width, but this is not required.

The signal display is compared to the template using correlation analysis 506. The template is shown as containing a series of sample values {T1, ..., T9}. Before or as part of the comparison, the signal display or template is scaled so that the maximum peaks of the two datasets have equal amplitude. One example of correlation analysis is correlation waveform analysis. Other embodiments are widely known in the art.

A simplified version of the correlation analysis is shown in FIG. 16, where the largest sample or peak of the signal display is aligned with the peak of the template and the surrounding samples are compared to each other, as shown in 508. Since the peaks are already scaled to be equal, there is no difference at the peaks, but the surrounding samples may be different. The difference between the signal display and the template is indicated by a crosshatch.

Next, the correlation score may be calculated as shown in 501. The sum of the absolute values of the differences between the (scaled) sample of the signal display and the sample of the template is calculated and divided by the total area under the template. The quotient is subtracted from 1 to give a correlation score of 512. When the correlation score is close to 1, the area of the difference is smaller than the area under the template, indicating a high correlation. Other methods of calculating correlations are known in the art and may be replaced. The method shown in FIG. 16 is merely an example. For example, a weighted CWA may apply a weighted factor to individual sample differences in the manner shown in US Patent Application Publication No. 2008-0077030, which is simultaneously assigned to the same assignee.

With reference to FIG. 3A, individual events are detected by applying the detection profile 70 to the signal 72. The detection profile 70 includes a refractory period 74 followed by a constant threshold period 76 and a decay period 78. Other shapes may be used for the detection profile 70.

The signal 72 emphasizes the R wave and the T wave. In the illustrated embodiment, the T wave is larger than the R wave. The refractory period indicated by the crosshatch across both the R and T waves indicates that each R and T wave is treated as a detection event. As a result, during each cardiac cycle, the detection profile 70 detects two events. This is one example of overdetection.

In the exemplary embodiment, each of the individual detections is also treated for a correlation analysis for an R-wave based template. The result of the correlation analysis is plotted at 80. Plot 80 includes boundaries for "High" and "Low" correlations. In the examples, each "X" indicates a correlation score for each detection event. A high-low-high correlation score pattern occurs as shown by 82. In the examples, each high-low-high sequence leads to the conclusion that detection events with a "low" score are overdetected. As a result, as shown, detection events with a "low" score will be discarded when a high-low-high pattern is found. In a numerical example, "high" is defined as greater than 52% correlation, while "low" is defined as less than 25% when calculated using the formula shown in FIG. Other values and analysis methods may be used.

FIG. 3B shows a method step for an embodiment that includes rate correction. When a morphological overdetection pattern is found, as shown in 90, one or more overdetections are identified, as shown in 92. The event interval and / or rate is then recalculated, as shown in 94.

For example, as shown in 96, a series of detections of R and T waves can result in a set of interval calculations of 225 ms (R to T) and 300 ms (T to R), resulting in an average interval of 263 ms. Be done. An average interval of 263 ms yields a rate of about 229 beats / minute, which is a tachyarrhythmia that can be treated in many patients. However, as shown in 98, when the T wave is identified as overdetected and the spacing on either side of the T wave is combined, the spacing averages 525 ms. The rate can be recalculated to be approximately 114 beats / minute, avoiding possible defibrillation, cardioversion, or pacing that could occur in the absence of data correction.

FIG. 4 shows an example of inter-event correlation comparison. Inter-event comparison is a comparison in which two individual detected events are compared against each other. The comparison may take the form of a correlation analysis, or utilize some other type of analysis, such as wavelet transform, principal component analysis (PCA), to account for similarities between the two detection events. May be good. In the wavelet transform or PCA comparison, the similarity of the results of data compression to the wavelet or PCA output can be compared. For example, the similarity of the eigenvalue output of the PCA and / or the order, or the similarity of the wavelet coefficient obtained from the wavelet transform, can be compared in a qualitative or quantitative manner.

In the example shown in FIG. 4, correlation analysis is performed. In the examples, the correlation score is considered low, medium, or high, as shown in 108. A "high" score zone shows a strong confidence that the compared signals have the same properties (for example, if one event is an R wave, the other event is also an R wave), while "low". Scores indicate that the compared signals are very different from each other. The "intermediate" zone is intended to capture signals that are similar but do not give rise to the strong confidence that the two signals are of the same nature. For example, in patients undergoing rate-dependent morphological changes (such as rate-induced bundle branch block), the captured R-waves do not correlate well with the stored static template, but fall within an intermediate range with respect to the template. there is a possibility. In another embodiment, the monomorphic VT may have a high or intermediate inter-event correlation between the R waves and an intermediate correlation between the T waves, while the polymorphic VT has an intermediate correlation between the R waves. Or it shows a low correlation.

Fuzzy logic may be applied if desired. The use of "intermediate zones" suggests this. For example, rather than the simple "high" and "low" characteristics, additional categories may be provided. In addition, previous measurements may be used to signal barely similar or dissimilar signals to subsequent characterizations.

As shown in 100, a series of events N, N-1, N-2, and N-3 are considered as a group, and the Nth detection is by correlation analysis, N-1, N-2, and Compared to each of N-3. The results of the inter-event comparison and the comparison against the static template are shown in Table 102. The inter-event comparison results are shown in 104 and show the ordered results for the comparison of a given event with the previous three events. Table 102 shows the results for events N, N-1, N-2, and N-3. The inter-event comparison results show that for any given event X, the correlation with X-2 is higher than with X-1 or X-3. This may show a double detection pattern based on the increased correlation between alternating events.

In one embodiment, a comparison to a static normal sinus rhythm template may be performed. An exemplary result is shown in 106. As a result of alternating static templates, low-intermediate-low-intermediate suggests the possibility of overdetection, but strong confidence is only in static templates because the potential R waves do not have a high correlation. Does not occur on the basis. However, when considered in combination with inter-event comparison information, there is significant confidence that some events are over-detected. The set of applicable rules is as follows:

1) Alternating low-high-low and low-high-low for N when compared to N-1, N-2, and N-3, and 2) Alternating low-high-low for N-2 when compared to N-3, N-4, and N-5. Conclusion: Treat N-1 and N-3 as T waves. Further confirmatory rules 3) At least "Medium" correlation for N and N-2 to static templates May be. Another approach is to apply only rules 1) and 3), marking only N-1 as overdetection in response to the set of rules being satisfied. One or more events, when marked as overdetected, may be treated in the manner shown in FIG. 3B above.

FIG. 5 shows another example of inter-event correlation comparison. Here, the captured signal is triple-detected, as shown in 120. In this case, the Nth detection is compared to each of N-1, N-2, N-3, and N-4. The inclusion of four individual comparisons may further aid in distinguishing between triple and double detection, but one embodiment stops at three comparisons.

The results are shown in Table 124. For each set of comparisons, there are three low correlations and one intermediate correlation or one high correlation. Triple detection may result in some detections having a low correlation in each comparison. An exemplary set of rules is as follows.

1. 1. The Nth event has a high correlation with the N-3 event. 2. The N-1 and N-2 events have a low correlation with the N event. 3. 3. The events of N-1 and N-2 have a low correlation with the static template.

If these three conditions are met, N-1 and N-2 may be discarded. Further conditions may be added. For example, the static template characteristics of N and / or N-3 may be considered.

4. The Nth and N-3 events have an intermediate or high correlation with the static template. Then, if all of 1 to 4 are satisfied, N-1 and N-2 are discarded, and the interval from N to N-3 may be calculated and used in the rate analysis.

In a further embodiment, the width of each event may be considered using, for example, this fourth condition. 5. The events of N-1 and N-2 are wider than the width threshold. The width threshold may be set as desired. In one embodiment, the width threshold is in the range of 100 to 140 ms. This width threshold rule may be applied as an additional layer to any determination that an event is discarded as an overdetection. In another embodiment, polarity may be considered.

6. N-1 and N-2 each share the same polarity. Polarity, for example, by reference to most signal samples for an event, as the polarity of the sample with the maximum magnitude in that event, or the maximum positive or minimum positive extremum in the event first. It may be defined by determining if it occurs.

If desired, interval coupling may be added as an alternative condition. 7. Bonding intervals N to N-3 less than a predetermined duration Here, the "predetermined duration" is in the range of 800 to 1200 ms. This condition and its variations are also described in the context of FIGS. 11-13 and 14A-14B below.

FIG. 6 shows an analytical method for correlation analysis of short series and long series. FIG. 6 shows a plot 140 that plots the correlation score for a series of detection events. Correlation scores, shown as X, are plotted against lines 144 and 146, which define the wide band 148, and lines 150 and 152, which define the narrow band 154.

The wide band 148 identifies overdetection when there are two detection events with scores above line 144, separated by one detection event with a score lower than line 146, for example, as shown in FIG. 7A. Applies to The narrow band 154 is applied to identify over-detection (s) when a series of consecutive detections alternates above line 150 and below line 152, as shown, for example, in FIG. 7B. .. Numerical values are shown for each threshold for illustration, and these numerical values may use the correlation as a percentage.

The narrow band 154 applies a less stringent criterion for the correlation score than the wide band 152. Therefore , more events are analyzed before making a decision to discard the events with the lower score. In one exemplary embodiment, the events are not discarded using the narrow band 154, and finally the eight event patterns shown in FIG. 7B are met, at which time one of the low-scoring events The 4 events are discarded and the spacing around each event to be discarded is corrected. After satisfying the pattern in this initial step, only the latest low-scoring event will be discarded. For analytical purposes, previously discarded events are used to determine if the 8-consecutive-outside rule is met, even if excluded from the rate calculation. .. Another embodiment uses only five events and uses a narrow band 154 to look for high-low-high-low-high sequences, and if such a sequence is found, one of the low-scoring events. Or both are discarded.

The examples in FIGS. 6 and 7A-7B show numerical values, with 50% and 20% demarcating the wide band 148 and 40% and 25% demarcating the narrow band 154. These numbers are just examples. In one embodiment, these numbers are applied by scaling the equation shown in Figure 16 to 510 on a percentage basis.

8A-8B show examples of adjusting the correlation analysis for the observed correlation level for the template. With reference to FIG. 8A, a plot of correlation scores for comparing the template to a series of events is shown at 158. An average correlation score is calculated for odd-numbered events to identify double detection. Clustering of odd-numbered events then uses, for example, the standard deviation of the set, or a constant distance, to determine if all odd-numbered events fall within a predetermined distance from their mean. It is analyzed by judging. If all odd-numbered events fall within a predetermined distance from the mean, the mean separation from the lower boundary is calculated. If the separation is greater than a predetermined threshold, it is determined that the odd-numbered event indicates monotonicity in support of the presumption that the odd-numbered event is a QRS complex detection. When the monotonicity of odd-numbered events is identified, one or more even-numbered events below the low threshold are marked as overdetected.

In another embodiment, to determine if clustering of even numbered events has occurred, also using the average of even numbered events, before any event of even numbered events is marked as overdetected. All even-numbered events are analyzed. The separation of even-numbered event means and odd-numbered event means, rather than the separation of odd-numbered event means from the lower boundaries, is calculated to establish event grouping. In this embodiment, the overdetection marker is applied only when sufficient clustering of even numbered events appears.

FIG. 8B shows another embodiment in which the overdetection marking is adjusted for the correlation score for the static template. Here, the average correlation score for a set of 10 events is calculated. A "blank" band is then established around the mean correlation score. For example, the blank band may be defined as +/- 15%. Other "blank band" sizes may be used.

In the embodiment of FIG. 8B, a high score is defined as a score above the blank band and a low score is a score lower than the blank band. Overdetection can be identified if a high-low-high pattern appears around the blank band, and one or more of the low-scoring events are marked as overdetection.

Instead of the static template, the analysis shown by FIGS. 8A-8B may also be applied using the latest detection event as a template for comparison. The analysis described for FIGS. 8A-8B may use the mean / arithmetic mean calculation, or for signals that include mode, median, or other mathematical operations. Other predictors with a center point may be used.

Further use of the inter-event comparisons presented here may be in determining whether a shockable rhythm is occurring. Stimulation delivery is often used to address polymorphic conditions such as polymorphic ventricular tachycardia and ventricular fibrillation. Monomorphic conditions such as monomorphic ventricular tachycardia (MVT) can be treated, but do not always require the most energetic treatment. For example, MVT may be treated with anti-tachycardia pacing (ATP) instead of defibrillation or cardioversion. The reason is that ATP uses less energy and may be less traumatic for the patient. Correlation patterns can be used to distinguish between unimorphic arrhythmias and polymorphic arrhythmias. For example, the ongoing pattern shown in FIG. 7A or 7B or even FIG. 6 in which a high correlation is consistently found can be used to delay treatment if desired.

In another embodiment, the pattern shown in FIG. 8A may be further analyzed by determining the size of the standard deviation for clustered high scores. If the clustered high score is based on a static template and shows a low standard deviation, this can indicate a uniform state. In one embodiment, treatment may be stopped until the monomorphic state is broken down into a more polymorphic state, especially if ATP is available.

In one embodiment, the system uses stepwise correlation analysis to identify treatable arrhythmias. In the embodiment, a simple single event correlation analysis using a static template is performed until the pattern shown in FIG. 8A appears. Such a pattern then triggers a plurality of inter-event comparisons shown in FIGS. 4-5. And if the inter-event comparison shows potential over-detection, the interval data may be corrected. In addition, treatment may be discontinued if the inter-event comparison indicates a monomorphic condition.

FIG. 9 shows a method of aligning and realigning the captured signal to the correlation analysis template. The correlation analysis template is shown in 200 and the signal is shown in 202. The correlation analysis template 200 may be a static template or may represent the average of a single detection event or some of the latest detection events.

As described in FIG. 16, correlation analysis typically uses a reference point as an alignment guide for an ordered set of template values and signal samples. In the embodiment of FIG. 9, the base alignment point is identified as a sample of template 200 and signal 202 with maximum magnitude, respectively. Then, a series of comparisons are made starting with the base-aligned correlation shown in 210, the single sample shift correlation to the right shown in 212, and the single sample shift correlation to the left shown in 214. Since the shift 1 right correlation 212 is worse than the correlation score for the base correlation 210, the result of the shift 1 right correlation 212 is discarded. The shift 1 left correlation 214 results in a higher correlation than the aligned correlation 210, so the result of the base correlation 210 is discarded and another shift left correlation now offsets the alignment point by only two samples to 216. Is calculated. Since the result of 216 shows a lower correlation than the shift 1 left correlation of 214, the process stops and uses the correlation score calculated for the shift 1 left correlation 214 as the correlation score for the signal 202.

When performing a shift to the right and / or left, the scaling of the signal to the template may be modified as well. For example, by comparing the peaks for the signal with the peaks for the template, and then equalizing the two, if scaling is performed first, as soon as the shift occurs, the peaks for the signal instead shift. It may be scaled to the point where the peak aligns in a later template.

The method shown in FIG. 9 may help correct misalignments based on noise or sampling artifacts, slew rates, etc., which can cause the peak alignment points of sample 202 to be suboptimal. The method calculates the correlation score when the reference points are aligned until the maximum correlation score is found, as well as when only one or more samples are misaligned in each of the two directions. Including doing. If desired, there may be a limit on the number of samples that can be shifted to the left or right. In another embodiment, several scores (eg, one base, one to the left, two, and three, two to the right and three) are automatically calculated to give the best value. Be selected.

In another embodiment highlighted in FIG. 9, multiple alignment points may be defined for the template 200. Some examples are QRS start, maximum amplitude, maximum amplitude of opposite polarity of maximum amplitude (maximum amplitude is indicated by the turning point where each maximum amplitude is dV / dt = 0), two major Includes the maximum slope between peaks (indicated as dV / dt = MAX, etc.). By identifying similarities in the signal, the method can determine whether the use of different possible alignment points will provide different correlation analysis results. For example, the default may be to use the maximum amplitude point of the entire signal, but the default is instead to use the maximum tilt point in the monomorphic segment following the maximum amplitude point for some hearts. The event may be able to be aligned.

FIG. 10 shows another method of storing and applying templates for correlation analysis. In this embodiment, the signal forming the basis for the template is shown at 230. In the example example, once the template is formed, an interpolation region is defined between the positive and negative peaks of the signal 230. As a result, the stored template takes the form shown in 240. Template 240 coincides with template signal 230 for the region before the positive peak and the region after the negative peak, but is flexible between the two peaks, as shown by the dotted line 242. The positive peak is the peak of maximum magnitude in the template in the illustrated embodiment and is therefore used to scale the template with respect to the captured signal.

Alignment with respect to sample 232 is then performed as shown in 244. The template is adjusted so that the positive and negative peaks are aligned with the capture signal and have linear interpolation between the positive and negative peaks. Outside the positive and negative peaks, the template continues to match the signal, as shown in 230, but the duration and slope between the positive and negative peaks should match the capture event. Is adjusted to. The adjustment shown in FIG. 10 may avoid the difficulty that the static template has a fixed duration for patients whose QRS complex is affected by the rate. The adjustments made may be limited to avoid over-widening the template.

In another embodiment, three or more template points may be identified and linear interpolation may be used between the template points. For example, a template may consist of five values, each with relative amplitude and relative location. When the detected event is compared to the template, the width and peak amplitude of the detected event are used to scale each of the values in the template, and linear interpolation is applied between the template points.

FIGS. 11-12 show how to stop data correction following identification of potential overdetection. As shown in FIG. 11, the QRS complex occurs at 260, followed by the premature ventricular contraction (PVC) shown in 262, followed by another QRS complex at 264. PVC, in this example, is characterized by a low correlation to the template. Therefore, a high-low-high correlation pattern similar to the pattern shown above in FIG. 3A appears. Therefore, one embodiment would discard the PVC262. However, analytically, it may not be necessary to dispose of PVC262, as it is not actually an overdetection event. In addition, the spacing around PVC262 is both greater than 500 ms. Even without data correction, the average of the two intervals would result in an event rate of approximately 103 beats / minute, a rate that would not be afraid to result in unnecessary treatment. As such, data correction will not improve rhythmic specificity within the device while reducing beat sensitivity.

FIG. 12 shows a method that can avoid discarding the PVC262 shown in FIG. Based on detection event 270, the method determines if a correlation score sequence appears that would support the detection of double detection (DD) or overdetection, as shown in 272. If it does not appear, then no data correction is likely to occur and the method ends. If the result from 272 is "yes", the method then includes determining whether the new interval that will be obtained from the data correction will be greater than a predetermined threshold, as shown in 274. .. In an exemplary example, the threshold is 1000 ms (60 beats / minute), but this number is merely exemplary. Some possible thresholds are in the range of 750 to 1200 ms.

In another example, the order of the analyzes must be reversed and the calculation rate must be high (often above 150 bpm) or the affected intervals must be short enough to pass the applied test. For example, overdetection analysis does not occur. In another embodiment, if the individual intervals are compared to a threshold (eg, in the range of 400-600 ms) and both individual intervals exceed the threshold, no interval coupling occurs. In yet another embodiment, the threshold may be a programmable parameter of the embedded system. In another embodiment, the threshold may be scaled based on a programmable VT parameter used to set the pulsatile rate that the implantable system will treat as the ventricular tachycardia rate.

If the corrected interval is not longer than the threshold, the method continues with the step of combining the intervals, as shown in 276, to correct the overdetected event (s). In step 274, if the corrected interval is longer than the threshold, the method only ends without combining the intervals. In this way, unnecessary correction of the stored data can be avoided.

FIG. 13 shows a further way to stop the correlation analysis after identifying overdetection. The method of FIG. 13 utilizes the known relationship between the QT interval and the RR interval of the physiological cardiac cycle. The exemplary method again begins with the identification of patterns suggesting overdetection, as shown in 300. Then, as shown in 302, the potential overdetection event is treated as a T wave (where the estimation is that the three event patterns are identified and the intermediate event of the three is the potential overdetection. That is, other variants may be used), as shown in 304, the event on either side of the potential overdetection is treated as an R wave.

These "estimated" R and T waves from steps 302 and 304 are then used in step 306 to apply the formula for calculating QT length from RR intervals. In particular, some possible formulas are given in 308. An example is Bazett's formula:

<maths num="1"></maths>Fridericia's formula:

<maths num="2"></maths>And Sagie et al.'S regression formula:

<maths num="3"></maths>including. Sagi et al. Found that A = 0.154.

In each formula, the expected QT is shown as QT (Exp), the value RR is given in seconds, and the value QT is captured between the implant and the programmer during the programming session. QT is captured at a heart rate of 60 beats / minute or adjusted for a heart rate of 60 beats / minute. The RR interval is found at 304 and the measured QT interval is by adding the estimated T wave measurement width to the interval between the first R wave and the estimated T wave. Can be captured.

The expectation is that if the potential over-detection event is an over-detection T wave, given an RR using any formula that applies, with some band tolerating errors, This means that the measured QT interval will match the expected QT value.

If the formula applied in 306 does not result in agreement, then no disposal occurs as shown in 310. Alternatively, if the formula applied at 306 results in a match, the potential overdetection is discarded as shown in 312. When the potential overdetection is discarded at 312, the spacing around the overdetection is combined as previously shown in FIG. 3B. Once again, the order of analysis is reversed in the other examples.

14A-14B show the application of the method shown in FIG. In the examples of FIGS. 14A and 14B, the Fridelicia cube root formula is applied. In each example, the previously measured QT is 400 ms. This value represents the estimated QT interval for a hypothetical patient that would occur at a heart rate of 60 bpm.

With reference to FIG. 14A, given three events X, Y, and Z with a correlation pattern indicating overdetection, the method is applied by presuming that Y is a T wave. The QT interval is measured for X and Y and the RR interval is measured from X to Z as shown. The QT to be measured is also referenced and these values are inserted into the selected formula. In the example shown, using RR = 0.8 seconds, the expected value for QT is 371 ms. Applying a +/- 10% error band for the calculation, the acceptable range is about 334-408 ms for QT. However, as shown, the measurement interval is about 500 ms, which is too long to be the QT interval for a given parameter. As a result, the calculation suggests that the Y detection is not an overdetected T wave and therefore no data correction occurs. A smaller or larger error band may be applied. For example, a band with a +/- 5% error is used in another exemplary embodiment.

Instead, with reference to FIG. 14B, this time the QT interval measured for X and Y is about 370 ms. This value falls within the expected range, so the calculation suggests that the Y detection is an overdetected T wave. Therefore, the Y detection is discarded and the interval data between X and Z is corrected.

In the embodiments of FIGS. 11-13 and 14B, if the potential overdetection is not discarded and results in data correction, the potential overdetection may instead be marked as suspicious detection. In some embodiments, suspicious detection is treated as unreliable both as an indicator of cardiac activity and as an interval endpoint that can be used in rate analysis. If a potential overdetection is marked as a suspicious detection, the suspicious detection and the preceding and subsequent intervals around the suspicious detection are each completely removed from the analysis.

FIG. 15 shows an analytical method for identifying shockable detection events and treatable rhythms. FIG. 15 shows a cross-sectional structure of the analysis method by including a plurality of steps of event detection 402, followed by waveform evaluation 404 and pulsation recognition 406. In particular, event detection 402 will typically include monitoring captured signals to detect signal amplitude changes that indicate cardiac events. At 402, if the cardiac event is captured, waveform evaluation 404 can occur. During waveform evaluation 404, the characteristics of the signal associated with the detection event are analyzed to identify and eliminate detection events that may be caused by noise or artifacts.

The detection event that passes the waveform evaluation 404 then receives a pulsation certification 406, and during the pulsation certification 406, the detection event is analyzed to determine if it exhibits a morphology or interval characteristic that indicates accurate detection. Will be done. This may include the correlation analysis shown above and / or the analysis of the combination of multiple intervals or two intervals, for example, an analysis to eliminate wide and complex double detection is a potential overdetection. The proximity and shape characteristics of the detection event can be used to identify. A further description is given in US Patent Application No. 12 / 399,914 entitled "METHODS AND DEVICES FOR ACCURATELY CLASSIFYING CARDIAC ACTION".

The architecture then relies on a rhythmic classification that can begin with rate considerations at block 408. If the rate is low, the individual detections are marked as "Not Shockable", as shown in 410. Alternatively, the rate is considered to indicate ventricular fibrillation (VF) if it is very high and is therefore marked as "Shockable" as shown in 412. There is a ventricular tachycardia (VT) zone between these low rate bands and the VF rate band, and the rates within the VT zone will be referred to as Detection Enhancement, as shown in 414. Is analyzed using.

Examples of detection improvement are as follows. 1. 1. Compare with static template: if matching, not shockable, otherwise 2. Compare with dynamic template: if it doesn't match, it's shockable, otherwise 3. 3. Compare with QRS Width Threshold: If wide, shockable, otherwise shock impossible. Here, the dynamic template can be any of the following:

a) Mean of several previous detections that correlate with each other, b) A set of individual events, eg {N-1, ..., Ni}, where matching some or all of the individual events counts as matching a dynamic template. c) Templates that are continuously updated.

The QRS complex thresholds mentioned above may be applied in various ways that may be adjusted for the QRS complex measurement method used in a given system and / or for individual patients. Good. In one embodiment, the following rules apply to the QRS complex.

x) The QRS complex is from the start of the longest monotonous segment captured during the refractory period before the reference point during the analysis to the end of the longest monotonous segment captured during the refractory period after the reference point. Calculated as the duration of y) The QRS complex was measured for the patient during the programming session and has a maximum acceptable value of 113 ms. z) The QRS width under analysis is considered to be wide if it is at least 20 ms longer than the QRS width threshold. These rules x), y), and z) are adjusted for one particular embodiment and may vary depending on the system used.

Following marking the event as non-shockable 410 or shockable 412, X / Y counter conditions are applied as shown in 416. The X / Y counter condition analyzes the number X of shockable events marked during the previous set Y of detection events that pass both waveform evaluation 404 and pulsation certification 406. The ratios applied and the set size used are subject to change. In one embodiment, the 18 / 24X / Y counter condition is applied at 416. Other embodiments use ratios as 8 or 9 of 12, 12 or 13, 24/32 of 16 and the like.

If the X / Y condition is not met, no shock will be delivered, as shown in 418. If the X / Y condition is met, the method may proceed to charge confirmation block 420. For example, one embodiment requires that the X / Y ratio / set size be met for a selected number of contiguous events, and this condition may be tested in charge confirmation 420. Another exemplary condition is whether the set N of previous detection events are all shockable, i.e., have intervals short enough to support the conclusion that the detected arrhythmia is in progress. It is to judge whether or not. Other factors should also delay treatment, for example by observing whether overdetection was recently recorded (to ensure that "arrhythmia" is not an indication of overcounting. Charging confirmation (which may suggest that) or by observing whether long, consistent intervals were detected (which may suggest a voluntary shift to normal arrhythmia by the patient). May be applied in. For example, the charge confirmation 420 also has a "Method FOR" whose disclosure is incorporated herein by reference. It may include a method as set forth in US Patent Application No. 11 / 042,911, assigned to the same assignee and simultaneously pending, entitled "ADAPTING CHAREGE INITION FOR AN IMPLANTABLE CARDIOVERTER-DEFIBRILLATOR".

The charge and shock block 422 is reached if the charge confirmation 420 passes. Since the charging process usually takes a period of time, method 400 may be repeated several times before charging is complete. Some or all of the analysis used to reach the initial determination that charging should begin may be repeated during this process. Finally, if the treatable state persists during charging or is identified following charging, a stimulus may be delivered.

Various hardware mechanisms may be incorporated for the embedded system. For example, any suitable chemical cell such as a lithium ion battery may be used. Therapeutic output can be made using a capacitive system to store energy until the stimulation level is reached using one or several capacitors. Charging circuits, such as flyback transformer circuits, can be used to generate therapeutic voltages. Treatment can be delivered using, for example, an H-bridge circuit or an improved version thereof. Dedicated or general purpose circuit elements may be used to perform the analysis function. For example, a dedicated cardiac signal analog-digital circuit and, if desired, a dedicated correlation analysis block may be used, while other functions may be performed by a microcontroller. Static and dynamic memory may be provided and used for any suitable function. All of these elements may be components of the working circuit elements for an implantable cardiac stimulation system.

Those skilled in the art will recognize that the present invention may be in various forms other than the particular embodiments described herein and considered. Therefore, the embodiments may be modified without departing from the scope and gist of the present invention.