Detection of cardiac arrhythmias using template matching by signature analysis

• US 5,273,049 A
• Filed: 04/09/1992
• Issued: 12/28/1993
• Est. Priority Date: 04/09/1992
• Status: Expired due to Fees

• Current Assignee: Pacesetter Incorporated (Abbott Laboratories)
• Original Assignee: Telectronics Pacing Systems, Inc (Abbott Laboratories)
• Inventors: Greenhut, Saul E., Steinhaus, Bruce M.
• Primary Examiner(s): Kamm, William E.
• Assistant Examiner(s): Getzow, Scott M.
• Application Number:
  US07/865,320
• Time in Patent Office: 628 Days
• Field of Search: 128/696, 128/702, 128/705
• US Class Current: 600/508
• CPC Class Codes: A61N 1/3621 for treating or preventing abnormally high heart rate
• CPC Class Codes: A61N 1/39622 Pacing therapy
• CPC Class Codes: A61N 1/3987 characterised by the timing or triggering of the shock

Abstract

A method and apparatus for detecting cardiac arrhythmias are disclosed. The detecting method and apparatus sense cardiac electrical signals when the heart is functioning in a known cardiac state, such as a physiologically normal cardiac state, then define and store an array of amplitude windows wherein each sample in the array of amplitude windows corresponds to a sample of the known cardiac signal. Each amplitude window delineates a range of signal amplitudes bracketing the amplitude of the known cardiac electrical signal sample. The array of amplitude windows corresponds in time to periodically occurring known cardiac signal samples occurring within a predetermined time of interest. Subsequently, when the heart is functioning in an unknown cardiac state, the method and apparatus monitor cardiac electrical signals by time aligning samples and comparing, on a sample by sample basis, the amplitude of unknown state cardiac signal samples with the amplitude range within the stored array of amplitude window samples. The relative number of samples falling outside the amplitude windows is the basis for analyzing cardiac signal morphology for the purpose of detecting cardiac arrhythmias.

Claims

1. A method of detecting cardiac arrhythmias in a patient'"'"'s heart, comprising the steps of:

sampling a cardiac electrical signal when the heart is functioning in a known cardiac state;

defining, for each sample of the known state cardiac electrical signal occurring within a predetermined time of interest, an amplitude window of signal amplitudes bracketing the amplitude of the known state cardiac electrical signal sample;

storing a template sequence of amplitude window samples encompassing the predetermined time of interest;

monitoring a time sequence of cardiac electrical signal samples when the heart is functioning in an unknown cardiac state;

time aligning samples of the unknown cardiac state sequence with respect to samples of the template sequence of amplitude window samples;

comparing, sample by sample, the samples of the unknown cardiac state sequence to the template sequence of amplitude window samples to determine the relative proportion of unknown cardiac state samples with amplitudes falling outside the amplitude window in comparison to unknown cardiac state samples with amplitudes falling inside the amplitude window; and

classifying the functioning of the patient'"'"'s heart as other than the known cardiac state when the determined relative proportion meets a predetermined criteria and, otherwise, classifying the functioning of the patient'"'"'s heart as the known cardiac state.

2. A method according to claim 1, further comprising the step of normalizing the amplitudes of the unknown cardiac state sequence samples so that the average unknown cardiac state sequence sample amplitude lies within a range determined by the template sequence of amplitude window samples.

3. A method according to claim 2, wherein said time aligning step comprises the sub-steps of:

defining an alignment pointer for equating in time a sample in the unknown cardiac state sequence with a sample in the template sequence of amplitude window samples;

deriving a manifestation of the sample by sample differences between the unknown cardiac state and the template sequence of amplitude window samples;

shifting said alignment pointer with respect to one sequence of the unknown cardiac state and the template sequences of amplitude window samples;

repeating said deriving and shifting sub-steps until said deriving step has been performed with respect to multiple alignment pointers; and

comparing each of said derived manifestations to determine the alignment pointer associated with the best match between the unknown cardiac state sequence and the template sequence of amplitude window samples.

4. A method according to claim 3, wherein said manifestation of the sample by sample differences is derived by summing, over the length of the sequences, the absolute values of the differences between each of the template sequence of amplitude window samples and its alignment pointer-corresponding unknown cardiac state sequence sample.

5. A method according to claim 3, wherein said manifestation of the sample by sample differences is derived by comparing, sample by sample, the samples of the unknown cardiac state sequence with the template sequence of amplitude window samples to determine the relative proportion of unknown cardiac state samples having amplitudes falling outside the amplitude window in comparison to unknown cardiac state samples having amplitudes falling inside the amplitude window.

6. A method according to claim 1, wherein said time aligning step comprises the sub-steps of:

defining an alignment pointer for equating in time a sample in the unknown cardiac state sequence with a sample in the template sequence of amplitude window samples;

deriving a manifestation of the sample by sample differences between the unknown cardiac state and the template sequence of amplitude window samples;

shifting said alignment pointer with respect to one sequence of the unknown cardiac state and the template sequence of amplitude window samples;

repeating said deriving and shifting sub-steps until said deriving step has been performed with respect to multiple alignment pointers; and

comparing each of said derived manifestations to determine the alignment pointer associated with the best match between the unknown cardiac state sequence and the template sequence of amplitude window samples.

7. A method according to claim 6, wherein said manifestation of the sample by sample differences is derived by summing, over the length of the sequences, the absolute values of the differences between each of the template sequence of amplitude window samples and its alignment pointer-corresponding unknown cardiac state sequence sample.

8. A method according to claim 6, wherein said manifestation of the sample by sample differences is derived by cardiac state sequence with the template sequence of amplitude window samples to determine the relative proportion of unknown cardiac state samples having amplitudes falling outside the amplitude window in comparison to unknown cardiac state samples having amplitudes falling inside the amplitude window.

9. A method of detecting cardiac arrhythmias in a patient'"'"'s heart, comprising the steps of:

sensing cardiac electrical signals when the heart is functioning in a known cardiac state;

characterizing cardiac electrical signal samples occurring within a predetermined time of interest when the heart is functioning in said known cardiac state;

storing said known cardiac state samples as a time sequence of template samples;

defining an array of amplitude ranges bracketing said timing sequence of template samples;

storing said array of amplitude ranges in a time sequence of window ranges corresponding sample by sample to said time sequence of template samples;

subsequently monitoring a time sequence of cardiac electrical signal samples when the heart is functioning in an unknown cardiac state;

reconstructing a template sequence of amplitude window samples from said stored time sequence of template samples and said time sequence of window amplitude ranges;

time aligning samples of said unknown cardiac state sequence with respect to said reconstructed template sequence of amplitude window samples;

comparing time aligned samples of said unknown cardiac state sequence with said reconstructed template sequence of amplitude window samples;

ascertaining the number of the compared time aligned samples of said unknown cardiac state sequence having amplitudes outside of said reconstructed template sequence of amplitude window samples;

comparing said ascertained number of unknown cardiac state sequence samples having amplitudes outside said template sequence of amplitude window samples with a threshold value; and

classifying said unknown cardiac state sequence within said known cardiac state when the ascertained number of monitored samples having amplitudes outside said template sequence of amplitude window samples is less than said threshold value, and classifying said unknown cardiac state as other than said known cardiac state when the ascertained number of monitored samples having amplitudes outside said template sequence of amplitude window samples is greater than said threshold value.

10. A method according to claim 9, further comprising the step of normalizing the amplitudes of the unknown cardiac state sequence samples, wherein said normalizing step comprises the sub-steps of:

summing the absolute values of the amplitudes of the time sequence of template samples;

subsequently summing the absolute values of the amplitudes of the unknown cardiac state sequence samples;

determining the difference between said template sequence sum and said unknown cardiac sequence sum;

ascertaining a normalization factor from this difference value for multiplying the unknown cardiac state sequence sum to place this sum within a predetermined range of said template sequence sum; and

multiplying each sample of the unknown cardiac state sequence by said normalization factor.

11. A method according to claim 10, wherein said time aligning step comprises the sub-steps of:

defining an alignment pointer for equating in time a sample in the unknown cardiac state sequence with a sample in the time sequence of template samples;

deriving a manifestation of the sample by sample differences between the unknown cardiac state sequence and the time sequence of template samples;

shifting said alignment pointer with respect to one sequence of the unknown cardiac state sequence and the time sequence of template samples;

repeating said deriving and shifting sub-steps until said deriving step has been performed with respect to multiple alignment pointers; and

comparing each of said derived manifestations to determine the alignment pointer associated with the best match between the unknown cardiac state sequence and the time sequence of template samples.

12. A method according to claim 11, wherein said manifestation of the sample by sample differences is derived by summing, over the length of the sequences, the absolute values of the differences between each of the time sequence of template samples and its alignment pointer-corresponding unknown cardiac state sequence sample.

13. A method according to claim 9, wherein said time aligning step comprises the sub-steps of:

defining an alignment pointer for equating in time a sample in the unknown cardiac state sequence with a sample in the time sequence of template samples;

deriving a manifestation of the sample by sample differences between the unknown cardiac state sequence and the time sequence of template samples;

shifting said alignment pointer with respect to one sequence of the unknown cardiac state sequence and the time sequence of template samples;

repeating said deriving and shifting sub-steps until said deriving step has been performed with respect to multiple alignment pointers; and

comparing each of said derived manifestations to determine the alignment pointer associated with the best match between the unknown cardiac state sequence and the time sequence of template samples.

14. A method according to claim 13, wherein said manifestation of the sample by sample differences is derived by summing, over the length of the sequences, the absolute values of the differences between each of the time sequence of template samples and its alignment pointer-corresponding unknown cardiac state sequence sample.

15. A method according to claim 13, wherein said manifestation of the sample by sample differences is derived by comparing, sample by sample, the samples of the unknown cardiac state sequence with the template sequence of amplitude window samples to determine the relative proportion of unknown cardiac state samples having amplitudes falling outside the amplitude window in comparison to unknown cardiac state samples having amplitudes falling inside the amplitude window.

16. A cardiac arrhythmia detector for detecting cardiac arrhythmias in a patient'"'"'s heart, comprising:

means for sampling a cardiac electrical signal when the heart is functioning in a known cardiac state;

means for defining, for each sample of the known state cardiac electrical signal occurring within a predetermined time of interest, an amplitude window of signal amplitudes bracketing the amplitude of the known state cardiac electrical signal sample;

means for storing a template sequence of amplitude window samples encompassing the predetermined time of interest;

means for monitoring a time sequence of cardiac electrical signal samples when the heart is functioning in an unknown cardiac state;

means for time aligning samples of the unknown cardiac state sequence with respect to samples of the template sequence of amplitude window samples;

means for comparing, sample by sample, the samples of the unknown cardiac state sequence to the template sequence of amplitude window samples to determine the relative proportion of unknown cardiac state samples with amplitudes falling outside the amplitude window in comparison to unknown cardiac state samples with amplitudes falling inside the amplitude window; and

means for classifying the functioning of the patient'"'"'s heart as other than the known cardiac state when the determined relative proportion meets a predetermined criteria and, otherwise, classifying the functioning of the patient'"'"'s heart as the known cardiac state.

17. A detector according to claim 16, further comprising a means for normalizing the amplitudes of the unknown cardiac state sequence samples so that the average unknown cardiac state sequence sample amplitude lies within a range determined by the template sequence of amplitude window samples.

18. A detector according to claim 16, wherein said time aligning means further comprises:

means for defining an alignment pointer for equating in time a sample in the unknown cardiac state sequence with a sample in the template sequence of amplitude window samples;

means for deriving a manifestation of the sample by sample differences between the unknown cardiac state and the template sequence of amplitude window samples;

means for shifting said alignment pointer with respect to one sequence of the unknown cardiac state and the window window sequence of amplitude window samples;

means for repeating the operations of said deriving and shifting means until said deriving means has acted with respect to multiple alignment pointers; and

means for comparing each of said derived manifestations to determine the alignment pointer associated with the best match between the unknown cardiac state sequence and the template sequence of amplitude window samples.

19. A detector according to claim 18, wherein said deriving means further comprises:

means for subtracting unknown cardiac state sequence samples from time aligned samples of said template sequence of amplitude window samples on a sample by sample basis;

means for determining the absolute values of each difference sample determined by said subtracting means; and

means for summing said absolute values over the length of the sequences.

20. A detector according to claim 18, wherein said time aligning means further comprises:

means for defining an alignment pointer for equating in time a sample in the unknown cardiac state sequence with a sample in the template sequence of amplitude window samples;

means for deriving a manifestation of the sample by sample differences between the unknown cardiac state and the template sequence of amplitude window samples;

means for shifting said alignment pointer with respect to one sequence of the unknown cardiac state and the template sequence of amplitude window samples;

means for repeating the operations of said deriving and shifting means until said deriving means has acted with respect to multiple alignment pointers; and

means for comparing each of said derived manifestations to determine the alignment pointer associated with the best match between the unknown cardiac state sequence and the template sequence of amplitude window samples.

21. A detector according to claim 20, wherein said deriving means further comprises:

means for subtracting unknown cardiac state sequence samples from time aligned samples of said template sequence of amplitude window samples on a sample by sample basis;

means for determining the absolute values of each difference sample determined by said subtracting means; and

means for summing said absolute values over the length of the sequences.

22. A cardiac arrhythmia detector for detecting cardiac arrhythmias in a patient'"'"'s heart, comprising:

means for sensing cardiac electrical signals when the heart is functioning in a known cardiac state;

means for characterizing cardiac electrical signal samples occurring within a predetermined time of interest when the heart is functioning in said known cardiac state;

means for storing said known cardiac state samples as a time sequence of template samples;

means for defining an array of amplitude ranges bracketing said time sequence of template samples;

means for storing said array of amplitude ranges in a time sequence of window ranges corresponding sample by sample to said time sequence of template samples;

means for subsequently monitoring a time sequence of cardiac electrical signal samples when the heart is functioning in an unknown cardiac state;

means for reconstructing a template sequence of amplitude window samples from said stored time sequence of template samples and said time sequence of window amplitude ranges;

means for time aligning samples of said unknown cardiac state sequence with respect to said reconstructed template sequence of amplitude window samples;

means for comparing time aligned samples of said unknown cardiac state sequence with said reconstructed template sequence of amplitude window samples;

means for ascertaining the number of the compared time aligned samples of said unknown cardiac state sequence having amplitudes outside of said reconstructed template sequence of amplitude window samples;

means for comparing said ascertained number of unknown cardiac state sequence samples having amplitudes outside said template sequence of amplitude window samples with a threshold value; and

means for classifying said unknown cardiac state sequence within said known cardiac state when the ascertained number of monitored samples having amplitudes outside said template sequence of amplitude window samples is less than said threshold value, and classifying said unknown cardiac state as other than said known cardiac state when the ascertained number of monitored samples having amplitudes outside said template sequence of amplitude window samples is greater than said threshold value.

23. A detector according to claim 22, further comprising a means for normalizing the amplitudes of the unknown cardiac state sequence samples, wherein said normalizing means further comprises:

means for summing the absolute values of the amplitudes of the time sequence of template samples;

means for subsequently summing the absolute values of the amplitudes of the unknown cardiac state sequence samples;

means for determining the difference between said template sequence sum and said unknown cardiac sequence sum;

means for ascertaining a normalization factor from this difference value for multiplying the unknown cardiac state sequence sum to place this sum within a predetermined range of said template sequence sum; and

means for multiplying each sample of the unknown cardiac state sequence by said normalization factor.

24. A detector according to claim 22, wherein said time aligning means further comprises:

means for defining an alignment pointer for equating in time a sample in the unknown cardiac state sequence with a sample in the time sequence of template samples;

means for deriving a manifestation of the sample by sample differences between the unknown cardiac state and the time sequence of template samples;

means for shifting said alignment pointer with respect to one sequence of the unknown cardiac state and the time sequence of template samples;

means for repeating the operations of said deriving and shifting means until said deriving means has acted with respect to multiple alignment pointers; and

means for comparing each of said derived manifestations to determine the alignment pointer associated with the best match between the unknown cardiac state sequence and the time sequence of template samples.

25. A detector according to claim 24, wherein said deriving means further comprises:

means for subtracting unknown cardiac state sequence samples from time aligned samples of said time sequence of template samples on a sample by sample basis;

means for determining the absolute values of each difference sample determined by said subtracting means;

means for summing said absolute values over the length of the sequences.

26. A detector according to claim 24, wherein said time aligning means further comprises:

means for defining an alignment pointer for equating in time a sample in the unknown cardiac state sequence with a sample in the time sequence of template samples;

means for deriving a manifestation of the sample by sample differences between the unknown cardiac state and the time sequence of template samples;

means for shifting said alignment pointer with respect to one sequence of the unknown cardiac state and the time sequence of template samples;

means for repeating the operations of said deriving and shifting means until said deriving means has acted with respect to multiple alignment pointers; and

means for comparing each of said derived manifestations to determine the alignment pointer associated with the best match between the unknown cardiac state sequence and the time sequence of template samples.

27. A detector according to claim 26, wherein said deriving means further comprises:

means for subtracting unknown cardiac state sequence samples from time aligned samples of said time sequence of template samples on a sample by sample basis;

means for determining the absolute values of each difference sample determined by said subtracting means; and

means for summing said absolute values over the length of the sequences.

28. A detector according to claim 26, wherein said deriving means further comprises:

means for comparing, sample by sample, the samples of the unknown cardiac state sequence with the template sequence of amplitude window samples to determine the relative proportion of unknown cardiac state samples having amplitudes falling outside the amplitude window in comparison to unknown cardiac state samples having amplitudes falling inside the amplitude window.

29. A method of detecting cardiac arrhythmias in a patient'"'"'s heart, comprising the steps of:

selecting samples of the patient'"'"'s physiologically normal cardiac electrical signal;

defining for each of the selected samples an amplitude window delineating a range of signal amplitudes bracketing the amplitude of the normal cardiac electrical signal sample;

subsequently monitoring the patient'"'"'s cardiac electrical signal when the heart is operating in an unknown cardiac state by sampling said cardiac electrical signals, comparing said samples to time-corresponding defined amplitude windows and counting the relative number of samples wherein the sample amplitude falls outside the amplitude window; and

classifying said unknown cardiac state as either a normal state or an arrhythmia state as a function of said relative number of samples.

30. A cardiac arrhythmia detector for detecting cardiac arrhythmias in a patient'"'"'s heart, comprising:

a sensor for sensing, amplifying and digitizing cardiac electrical signals, said cardiac electrical signals being indicative of a normal physiological state at certain times and being indicative of an arrhythmia cardiac state at other times, and a controller, said controller further comprising;

means for selecting samples of the patient'"'"'s physiologically normal cardiac electrical signal;

means for defining for each of the selected samples an amplitude window delineating a range of signal amplitudes bracketing the amplitude of the normal cardiac electrical signal sample;

means for subsequently monitoring the patient'"'"'s cardiac electrical signal when the heart is operating in an unknown cardiac state by sampling said cardiac electrical signals and by calculating a signature function by comparing said samples to time-corresponding defined amplitude windows and counting the relative number of samples wherein the sample amplitude falls outside the amplitude window; and

means for classifying said unknown cardiac state as either a normal state or an arrhythmia state as a function of said signature function.