Method and apparatus for ECG signal analysis and cardiac arrhythmia detection

• US 5,271,411 A
• Filed: 09/10/1991
• Issued: 12/21/1993
• Est. Priority Date: 09/21/1990
• Status: Expired due to Term

• Current Assignee: Omron Healthcare Company Limited (Omron Corporation)
• Original Assignee: Colin Electronics Co., Ltd.
• Inventors: Ripley, Kenneth L., Kaplan, Daniel T.
• Primary Examiner(s): Kamm, William E.
• Assistant Examiner(s): SCHAETZLE, KENNEDY
• Application Number:
  US07/757,307
• Time in Patent Office: 833 Days
• Field of Search: 128/702, 128/696, 128/703, 364/413.06
• US Class Current: 600/515
• CPC Class Codes: A61B 5/366 Detecting abnormal QRS complex, e.g. widening

Abstract

A method and apparatus for arrhythmia detection comprise steps and means for acquiring at least one continuous analog signal produced by an ECG system, producing at least one digital signal based on the analog signal, producing a plurality of scalar signals from the at least one digital signal, extracting features from the scalar signal, and plotting the extracted features in a feature space having a number of dimensions equal to the number of extracted features. A normal QRS complex is identified based on the population of QRS complexes located within clusters of QRS features within the feature space. Subsequent QRS complexes acquired after identification of the normal QRS complex are labeled based on a plurality of rules and the location of each subsequent QRS complex with respect to both prior and subsequent normal QRS complexes.

Claims

1. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said determining step comprises;

generating a plurality of scalar signals based on said at least one digital signal;

locating said time of occurrence of said possible QRS complex by detecting a peak value within at least one of said plurality of scalar signals; and

determining said noise attribute within said at least one scalar signal within a noise time window about said time of occurrence according to said threshold crossing criteria.

2. The method of claim 1, wherein said step of determining a noise attribute according to said threshold crossing criteria comprises:

varying a candidate noise threshold from a first adaptively determined threshold to a second determined threshold, in threshold steps, wherein said second threshold is greater than said first threshold;

for each threshold step, determining a candidate noise attribute based on said candidate noise threshold and at least one of said plurality of scalar signals, said candidate noise attribute comprising one of not noisy and noisy;

declaring said noise attribute to be not noisy when said candidate noise attribute is not noisy; and

declaring said noise attribute to be noisy when said candidate noise threshold has been varied from said first threshold to said second threshold and said candidate noise attribute is noisy at each previous threshold step.

3. The method of claim 1, wherein said step of extracting comprises identifying and removing a bad QRS complex prior to said plotting step.

4. The method of claim 1, wherein said identifying step comprises identifying one of said clusters having the largest population of said possible first QRS complexes as a normal cluster so as to designate each of said first QRS complexes located within said normal cluster as a normal QRS complex.

5. The method of claim 1, wherein said identifying step comprises labeling a first one of said clusters as said normal cluster, and labeling a second one of said clusters as a PVC cluster based on populations of said first and second clusters and a timing between QRS complexes in said first and second clusters.

6. The method of claim 1, wherein said step of identifying further comprises characterizing a normal QRS complex according to features characterizing said normal cluster.

7. The method of claim 1, wherein said labeling step further comprises labeling each of said plurality of possible second QRS complexes based on at least one of said noise attribute and an average normal QRS complex characterized according to features of said normal cluster.

8. The method of claim 1, wherein said labeling step comprises labeling each of said second QRS complex, the based on said N features of said each QRS complex, the timing of said each QRS complex, the QRS complexes located within said normal cluster and the cluster within which said each QRS complex is located.

9. The method of claim 1, wherein said labeling step comprises the steps of:

annotating each of said second QRS complexes according to a plurality of rules based on said N features of said each QRS complex, the timing of said each QRS complex, the QRS complexes located within said normal cluster and said one cluster within which said each QRS complex is located, said one cluster having a cluster label.

10. The method of claim 1, wherein said method further comprises the step of:

evaluating said labeled QRS complexes to produce a signal indicative of a character of a rhythm state within said at least one digital signal.

11. The method of claim 10, wherein said method further comprises the steps of:

displaying selected arrhythmic portions of said indicative signal.

12. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said extracting step further comprises filtering said at least one digital signal through a filter selected from a plurality of alternative filters according to said noise attribute.

13. The method of claim 12, wherein said extracting step further comprises:

determining a width threshold according to said filtered at least one digital signal and said threshold crossing criteria;

determining an onset time according to said filtered at least one digital signal and said width threshold;

determining an offset time according to said filtered at least one digital signal and said width threshold; and

determining a QRS width according to said onset time and said offset time.

14. The method of claim 13, wherein said extracting step further comprises:

determining a QRS complex center according to said onset time, said offset time and said filtered at least one digital signal;

determining a plurality of pre-areas arranged in time before said center; and

determining a plurality of post-areas arranged in time after said center.

15. The method of claim 14, wherein said plurality of pre-areas comprises a signed pre-area and an unsigned pre-area and said plurality of post-areas comprises a signed post-area and an unsigned post-area.

16. The method of claim 14, wherein said step of determining a QRS complex center determines said center so that one of said plurality of pre-areas has a magnitude equal to a magnitude of one of said plurality of post-areas.

17. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said step of extracting comprises identifying and removing a bad QRS complex prior to said plotting step, and said step of identifying a bad QRS complex comprises comparing against a predetermined number, a number of threshold crossings resulting from said threshold crossing criteria of said step of producing a value of a noise attribute when said noise attribute value is not noisy, and declaring the beat bad when said comparison indicates the number of threshold crossings is less than said predetermined number.

18. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said at least one cluster comprises a plurality of clusters, and wherein said identifying step further comprises merging a first cluster having a largest population of said plurality of clusters and a second cluster being within a predetermined distance from said first cluster within said N-dimensional space to produce at least one merged cluster.

19. The method of claim 18, wherein said at least one merged cluster comprises a plurality of merged clusters, and wherein said identifying step further comprises:

determining an average backwards coupling interval for each of a largest and a second largest cluster of said plurality of merged clusters;

declaring bigeminy based on populations of said largest and second largest merged clusters and coupling intervals of said largest and second largest merged clusters; and

labeling said largest merged cluster as normal and said second largest merged cluster as PVC when bigeminy is declared.

20. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said step of identifying further comprises characterizing a normal QRS complex according to features characterizing said normal cluster, and said step of characterizing comprises initializing normal QRS complex morphology parameters.

21. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said step of identifying further comprises characterizing a normal QRS complex according to features characterizing said normal cluster, and said step of characterizing comprises initializing normal QRS complex timing parameters.

22. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said step of identifying further comprises determining normal QRS complex parameters bounds.

23. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said plurality of possible second QRS complexes comprise both QRS complexes and false QRS complexes which are not QRS complexes,

said labeling step further comprises annotating at least one of said plurality of possible second QRS complexes according to a plurality of rules based on said normal cluster and said N features of a selected plurality of possible QRS complexes selected from said plurality of possible second QRS complexes, and

said selected plurality of possible QRS complexes comprises at least one of said possible second QRS complexes having a time of occurrence later in time than the time of occurrence of said annotated at least one of said plurality of possible second QRS complexes.

24. A method for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system, the method comprising the steps of:

determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to produce at least one cluster from said plurality of possible first QRS complexes;

identifying a normal cluster based on said at least one cluster;

second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said determining step, extracting step and plotting step to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said step of labeling further comprises;

identifying said patient rhythm state from said plurality of possible second QRS complexes; and

reinitializing cluster labels when said patient rhythm state is identified as bizarre for a period of time.

25. An arrhythmia detector for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system comprising a processor having an input circuit for receiving said at least one digital signal and including:

means for determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

means for extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

means for plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

means for first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said means for determining, extracting and plotting to produce at least one cluster from said plurality of possible first QRS complexes;

means for identifying a normal cluster based on said at least one cluster;

means for second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said means for determining, extracting and plotting to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

means for labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said means for determining further comprises;

means for generating a plurality of scalar signals based on said at least one digital signal,

means for locating said time of occurrence of said possible QRS complex by detecting a peak value within said at least one scalar signal, and

means for determining said noise attribute within said at least one scalar signal within a noise time window about said time of occurrence according to said threshold crossing criteria.

26. The arrhythmia detector of claim 25, wherein said means for extracting comprises means for identifying and removing a bad QRS complex prior to plotting by said means for plotting.

27. The arrhythmia detector of claim 25, wherein said means for labeling labels each of said plurality of possible second QRS complexes based on at least one of said noise attribute and an average normal QRS complex characterized according to features of said normal cluster.

28. The arrhythmia detector of claim 25, wherein said means for labeling labels each of said second QRS complexes based on said N features of said each QRS complex, the timing of said each QRS complex, the QRS complexes located within said normal cluster and the cluster within which said each QRS complex is located.

29. The arrhythmia detector of claim 25, wherein said means for labeling annotates each of said second QRS complexes according to a plurality of rules based on said N features of said each QRS complex, the timing of said each QRS complex, the QRS complexes located within said normal cluster and said one cluster within which said each QRS complex is located, said one cluster having a cluster label.

30. The arrhythmia detector of claim 25, wherein said detector further comprises:

means for evaluating said labeled QRS complexes to produce a signal indicative of a character of a rhythm state within said at least one digital signal.

31. The arrhythmia detector of claim 30, wherein said detector further comprises:

means for displaying selected arrhythmic portions of said indicative signal.

32. The arrhythmia detector of claim 28, wherein said means for determining a noise attribute according to said threshold crossing criteria comprises:

means for varying a candidate noise threshold from a first adaptively determined threshold to a second determined threshold, in threshold steps, wherein said second threshold is greater than said first threshold;

for each threshold step, said means for determining determines a candidate noise attribute based on said candidate noise threshold and one of said plurality of scalar signals, said candidate noise attribute comprising one of not noisy and noisy;

wherein said noise attribute is declared to be not noisy when said candidate noise attribute is not noisy; and

wherein said noise attribute is declared to be noisy when said candidate noise threshold has been varied from said first threshold to said second threshold and said candidate noise attribute is noisy at each previous threshold step.

33. An arrhythmia detector for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system comprising a processor having an input circuit for receiving said at least one digital signal and including:

means for determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

means for extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

means for plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

means for first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said means for determining, extracting and plotting to produce at least one cluster from said plurality of possible first QRS complexes;

means for identifying a normal cluster based on said at least one cluster;

means for second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said means for determining, extracting and plotting to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

means for labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said means for extracting further comprises means for filtering said at least one digital signal through a filter selected from a plurality of alternative filters according to said noise attribute.

34. The arrhythmia detector of claim 33, wherein said means for extracting further comprises:

means for determining a width threshold according to said filtered at least one digital signal and said threshold crossing criteria;

means for determining an onset time according to said filtered at least one digital signal and said width threshold;

means for determining an offset time according to said filtered at least one digital signal and said width threshold; and

means for determining a QRS width according to said onset time and said offset time.

35. The arrhythmia detector of claim 34, wherein said means for extracting further comprises:

means for determining a QRS complex center according to said onset time, said offset time and said filtered at least one digital signal;

means for determining a plurality of pre-areas arranged in time before said center; and

means for determining a plurality of post-areas arranged in time after said center.

36. The arrhythmia detector of claim 35, wherein said plurality of pre-areas comprises a signed pre-area and an unsigned pre-area and said plurality of post-areas comprises a signed post-area and an unsigned post-area.

37. The arrhythmia detector of claim 35, wherein said means for determining a QRS complex center determines said center so that one of said plurality of pre-areas has a magnitude equal to a magnitude of one of said plurality of post-areas.

38. An arrhythmia detector for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system comprising a processor having an input circuit for receiving said at least one digital signal and including:

means for determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

means for extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

means for plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

means for first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said means for determining, extracting and plotting to produce at least one cluster from said plurality of possible first QRS complexes;

means for identifying a normal cluster based on said at least one cluster;

means for second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said means for determining, extracting and plotting to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

means for labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said means for extracting comprises means for identifying and removing a bad QRS complex prior to plotting by said means for plotting, and said means for identifying a bad QRS complex comprises means for comparing against a predetermined number, a number of threshold crossings resulting from said threshold crossing criteria of said means for producing a value of a noise attribute when said noise attribute value is not noisy, and means for declaring the beat bad when said comparison indicates the number of threshold crossings is less than said predetermined number.

39. An arrhythmia detector for detecting arrhythmia in at least one digital signal produced from at least one analog ECG signal provided by an ECG system comprising a processor having an input circuit for receiving said at least one digital signal and including:

means for determining a noise attribute within said at least one digital signal based on a time of occurrence of a possible QRS complex according to a threshold crossing criteria, said noise attribute comprising one of not noisy and noisy;

means for extracting a number N of features from said at least one digital signal based on said time of occurrence and said noise attribute;

means for plotting in an N-dimensional feature space said possible QRS complex according to said extracted N features;

means for first group plotting each of a plurality of possible first QRS complexes by determining, extracting and plotting according to said means for determining, extracting and plotting to produce at least one cluster from said plurality of possible first QRS complexes;

means for identifying a normal cluster based on said at least one cluster;

means for second group plotting each of a plurality of possible second QRS complexes by determining, extracting and plotting according to said means for determining, extracting and plotting to perform at least one of adding possible QRS complexes to said at least one cluster, and producing at least one additional cluster; and

means for labeling each of said plurality of possible second QRS complexes within said at least one digital signal based on said normal cluster and said N features extracted from each of said possible second QRS complexes;

wherein said plurality of possible second QRS complexes comprise both QRS complexes and false QRS complexes which are not QRS complexes,

said means for labeling annotates at least one of said plurality of possible second QRS complexes according to a plurality of rules based on said normal cluster and said N features of a selected plurality of possible QRS complexes selected from said plurality of possible second QRS complexes, and

said selected plurality of possible QRS complexes comprises at least one of said possible second QRS complexes having a time of occurrence later in time than the time of occurrence of said annotated at least one of said plurality of possible second QRS complexes.