Method and apparatus for cardiac arrhythmia classification using sample entropy

• US 7,480,529 B2
• Filed: 06/13/2005
• Issued: 01/20/2009
• Est. Priority Date: 06/13/2005
• Status: Expired due to Fees

• Current Assignee: Cardiac Pacemakers Incorporated (Boston Scientific Corporation)
• Original Assignee: Cardiac Pacemakers Incorporated (Boston Scientific Corporation)
• Inventors: Li, Dan
• Primary Examiner(s): Layno; Carl H
• Assistant Examiner(s): LEE, YUN H
• Application Number:
  US11/151,102
• Publication Number:
  US 20060281999A1
• Time in Patent Office: 1,317 Days
• Field of Search: 600515-518
• US Class Current: 600/515
• CPC Class Codes: A61B 5/35 by template matching
• CPC Class Codes: A61B 5/363 Detecting tachycardia or bradycardia
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

An implantable medical device includes an arrhythmia detection and classification system that classifies an arrhythmia episode based on an irregularity parameter and/or a complexity parameter. The arrhythmia episode is detected from a cardiac signal. The irregularity parameter is indicative of the degree of cycle length irregularity of the cardiac signal and the complexity parameter is indicative of the degree of morphological complexity of the cardiac signal. One example of the irregularity parameter is an irregularity sample entropy, or a parameter related to the irregularity sample entropy, computed to indicate the cycle length irregularity. One example of the complexity parameter is a complexity sample entropy, or a parameter related to the complexity sample entropy, computed to indicate the morphological complexity. In one embodiment, the detected arrhythmia episode is classified using both the irregularity parameter and the complexity parameter.

Claims

1. A system for classifying cardiac arrhythmias, the system comprising:

a signal input to receive a cardiac signal indicative of a detected arrhythmia episode;

an irregularity analyzer coupled to the signal input, the irregularity analyzer adapted to produce an irregularity parameter being an irregularity sample entropy (SampEn_I) indicative of a degree of cycle length irregularity of the cardiac signal or a parameter having a predetermined mathematical relationship with SampEn_I;

a complexity analyzer coupled to the signal input, the complexity analyzer adapted to produce a complexity parameter being a complexity sample entropy (SampEn_C) indicative of a degree of morphological complexity of the cardiac signal or a parameter having a predetermined mathematical relationship with SampEn_C; and

an arrhythmia classifier coupled to the irregularity analyzer and the complexity analyzer, the arrhythmia classifier adapted to classify the detected arrhythmia episode based on the irregularity parameter and the complexity parameter.

2. The system of claim 1, wherein the arrhythmia classifier comprises:

an irregularity comparator adapted to compare the irregularity parameter to a predetermined irregularity threshold and to indicate a detection of irregularity based on an outcome of the comparison between the irregularity parameter and the predetermined irregularity threshold; and

a complexity comparator adapted to compare the complexity parameter to a predetermined complexity threshold and to indicate a detection of complexity based on an outcome of the comparison between the complexity parameter and the predetermined complexity threshold.

3. The system of claim 2, wherein the arrhythmia classifier is adapted to classify the detected arrhythmia episode as fibrillation if at least one of the detection of irregularity and the detection of complexity is indicated and to classify the detected arrhythmia episode as tachycardia if none of the detection of irregularity and the detection of complexity is indicated.

4. The system of claim 3, further comprising:

a sensing circuit to sense an electrogram indicative of an atrial tachyarrhythmia episode; and

a tachyarrhythmia detector, coupled to the sensing circuit, to detect the atrial tachyarrhythmia episode from the electrogram, and

wherein the arrhythmia classifier is adapted to classify the detected atrial tachyarrhythmia episode as atrial fibrillation if at least one of the detection of irregularity and the detection of complexity is indicated and to classify the detected atrial tachyarrhythmia episode as atrial flutter if none of the detection of irregularity and the detection of complexity is indicated.

5. The system of claim 3, further comprising:

a sensing circuit to sense an electrogram indicative of a ventricular tachyarrhythmia episode; and

a tachyarrhythmia detector, coupled to the sensing circuit, to detect the ventricular tachyarrhythmia episode from the electrogram, and

wherein the arrhythmia classifier is adapted to classify the detected ventricular tachyarrhythmia episode as ventricular fibrillation if at least one of the detection of irregularity and the detection of complexity is indicated and to classify the detected ventricular tachyarrhythmia episode as ventricular tachycardia if none of the detection of irregularity and the detection of complexity is indicated.

6. The system of claim 1, wherein the irregularity analyzer comprises an irregularity entropy computer adapted to compute SampEn_I, and the complexity analyzer comprises a complexity entropy computer adapted to compute SampEn_C.

7. The system of claim 6, wherein the arrhythmia classifier comprises:

an irregularity comparator adapted to compare SampEn_I to a predetermined irregularity entropy threshold (θ

_I) and indicate a detection of irregularity if SampEn_I exceeds θ

_I; and

a complexity comparator adapted to compare SampEn_C to a predetermined complexity entropy threshold (θ

_C) and indicate a detection of complexity if SampEn_C exceeds θ

_C; and

wherein the arrhythmia classifier is adapted to classify the detected arrhythmia episode as fibrillation if at least one of the detection of irregularity and the detection of complexity is indicated and to classify the detected arrhythmia episode as tachycardia if none of the detection of irregularity and the detection of complexity is indicated.

8. The system of claim 1, wherein the irregularity analyzer comprises an irregularity entropy computer adapted to compute the irregularity parameter being an irregularity parameter Γ

_I related to SampEn_I, wherein SampEn_I=−

InΓ

_I, and the complexity analyzer comprises a complexity entropy computer adapted to compute the complexity parameter being a complexity parameter Γ

_C related to SampEn_C, wherein SampEn_C=−

InΓ

_C.

9. The system of claim 8, wherein the arrhythmia classifier comprises:

an irregularity comparator adapted to compare Γ

_I to a predetermined irregularity threshold (γ

_I) and indicate a detection of irregularity if Γ

_I is below γ

_I; and

a complexity comparator adapted to compare Γ

_C to a predetermined complexity threshold (γ

_C) and indicate a detection of complexity if Γ

_C is below γ

_C, and

wherein the arrhythmia classifier is adapted to classify the detected arrhythmia episode as fibrillation if at least one of the detection of irregularity and the detection of complexity is indicated and to classify the detected arrhythmia episode as tachycardia if none of the detection of irregularity and the detection of complexity is indicated.

10. The system of claim 8, wherein the irregularity entropy computer comprises:

a cycle length detector, coupled to the signal input, to measure n₁ successive cycle lengths, d(1), d(2), . . . d(n₁), from the cardiac signal, where n₁ is a predetermined number; and

a first sample entropy computer coupled to the cycle length detector, and

wherein the complexity entropy computer comprises;

a feature extractor, coupled to the signal input, to extract p morphological features, f₁(i), f₂(i), . . . f_p(i), associated with each heart beat i of n₂ successive heart beats in the cardiac signal for the n successive heart beats, where p and n₂ are predetermined numbers, and 1≦

i≦

n₂;

a feature vector generator, coupled to the feature extractor, to produce n₂ feature vectors each associated with one of the n₂ successive heart beats, f(1), f(2), . . . f(n₂), where f(i)=[f₁(i), f₂(i), . . . f_p(i)] for each i; and

a second sample entropy computer coupled to the feature vector generator.

11. The system of claim 10, wherein the first sample entropy computer and the second sample entropy computer each comprise a sample entropy computer including:

a sample input to receive n samples of a signal;

u(1), u(2), . . . u(n), the n samples including one of u(1)=d(1), u (2)=d(2), . . . u (n₁)=d(n₁), where n=n₁, and u(1)=f(1), u(2)=f(2), . . . u (n₁)=f(n₂), where n=n₂;

a signal segment generator coupled to the sample input, the signal segment generator adapted to produce n−

m+1 signal segments each including m successive cycle lengths, x_m(i)=[u(i), u(i+1), . . . u(i+m−

1)], where 1≦

i≦

n−

m+1, and m is a predetermined number smaller than n; and

a vector matching module coupled to the signal segment generator, the vector matching module adapted to produce vector match scores D_m(i, j) each being a measure of similarity between a pair of x_m(i) and x_m(j), where 1≦

j≦

n−

m+1, and i≠

j, and

wherein the sample entropy computer is adapted to compute a parameter Γ

using D_m(i, j), Γ

related to an irregularity sample entropy (SampEn) indicative of at least one of irregularity and complexity of the signal, wherein Γ

=Γ

_I if u(i)=d(i), and Γ

=Γ

_C if u(i)=f(i).

12. The system of claim 11, wherein the sample entropy computer is further adapted to compute SampEn by using an equation:

SampEn=−

InΓ

, wherein SampEn=SampEn_I if the n samples include d(1), d(2), . . . d(n₁), where n=n₁, and SampEn=SampEn_C if the n samples include f(1), f(2), . . . f(n₂), where n=n₂.

13. A method for classifying cardiac arrhythmias, the method comprising:

detecting an arrhythmia episode;

receiving a cardiac signal indicative of the arrhythmia episode;

computing an irregularity parameter being an irregularity sample entropy (SampEn_I) indicative of a degree of cycle length irregularity of the cardiac signal or a parameter having a predetermined mathematical relationship with SampEn_I;

computing a complexity parameter being a complexity sample entropy (SampEn_C) indicative of a degree of morphological complexity of the cardiac signal or a parameter having a predetermined mathematical relationship with SampEn_C; and

classifying the arrhythmia episode based on the irregularity parameter and the complexity parameter.

14. The method of claim 13, further comprising sensing an electrogram indicative of an atrial tachyarrhythmia episode, and wherein classifying the arrhythmia episode comprises discriminating between atrial flutter and atrial fibrillation.

15. The method of claim 13, further comprising sensing an electrogram indicative of a ventricular tachyarrhythmia episode, and wherein classifying the arrhythmia episode comprises discriminating between ventricular tachycardia and ventricular fibrillation.

16. The method of claim 13, wherein classifying the arrhythmia episode comprises:

comparing the irregularity parameter to a predetermined irregularity threshold;

indicating a detection of irregularity based on an outcome of the comparison between the irregularity parameter and the predetermined irregularity threshold;

comparing the complexity parameter to a predetermined complexity threshold;

indicating a detection of complexity based on an outcome of the comparison between the complexity parameter and the predetermined complexity threshold; and

classifying the arrhythmia episode as fibrillation if at least one of the detection of irregularity and the detection of complexity is indicated and as tachycardia if none of the detection of irregularity and the detection of complexity is indicated.

17. The method of claim 13, wherein computing the irregularity parameter comprises computing SampEn_I,and computing the complexity parameter comprises computing SampEn_C.

18. The method of claim 17, wherein classifying the arrhythmia episode comprises:

comparing SampEn_I to a predetermined irregularity entropy threshold (θ

_I);

indicating a detection of irregularity if SampEn_I exceeds θ

_I;

comparing SampEn_C to a predetermined complexity entropy threshold (θ

_C);

indicating a detection of complexity if SampEn_C exceeds θ

_C; and

classifying the arrhythmia episode as fibrillation if at least one of the detection of irregularity and the detection of complexity is indicated and as tachycardia if none of the detection of irregularity and the detection of complexity is indicated.

19. The method of claim 13, wherein computing the irregularity parameter comprises computing the irregularity parameter being an irregularity parameter Γ

_I related to SampEn_I, wherein SampEn_I=−

InΓ

_I, and computing the complexity parameter comprises computing the complexity parameter being a complexity parameter Γ

_C related to SampEn_C, wherein SampEn_C=−

InΓ

_C.

20. The method of claim 19, wherein classifying the arrhythmia episode comprises:

comparing Γ

_I to a predetermined irregularity threshold (γ

_I);

indicating a detection of irregularity if Γ

_I is below γ

_I;

comparing Γ

_C to a predetermined complexity threshold (γ

_C);

indicating a detection of complexity if Γ

_C is below γ

_C; and

classifying the arrhythmia episode as fibrillation if at least one of the detection of irregularity and the detection of complexity is indicated and as tachycardia if none of the detection of irregularity and the detection of complexity is indicated.

21. The method of claim 19, wherein computing the irregularity parameter comprises:

measuring n₁ successive cycle lengths, d(1), d(2), . . . d(n₁), wherein n₁ is a predetermined number; and

computing Γ

_I using d(1), d(2), . . . d(n₁), and

wherein computing the complexity parameter comprises;

extracting p morphological features, f₁(i), f₂(i), . . . f_p(i), associated with each heart beat i of n₂ successive heart beats in the cardiac signal for the n₂ successive heart beats, wherein p and n₂ are predetermined numbers, and 1≦

i≦

n₂;

producing n₂ feature vectors each associated with one of the n₂ successive heart beats, f(1), f(2), . . . f(n₂), where f(i)=[f₁(i), f₂(i), . . . f_p(i)]; and

computing Γ

_C using f(1), f(2), . . . f(n₂).

22. The method of claim 21, wherein computing Γ

_I and computing Γ

_C each comprise;

receiving n samples of a signal;

u(1), u(2), . . . u(n), the n samples including one of u(1)=d(1), u (2)=d(2), . . . u (n₁)=d(n₁), where n=n₁, and u(1)=f(1), u(2)=f(2), . . . u (n₁)=f(n₂), where n=n₂;

producing n−

m+1 signal segments each including m successive cycle lengths, x_m(i)=[u(i), u(i+1), . . . u(i+m−

1)], where 1≦

i≦

n−

m+1, and m is a predetermined number smaller than n;

producing vector match scores D_m(i, j) each being a measure of similarity between a pair of x_m(i) and x_m(j), where 1≦

j≦

n−

m+1, and i≠

j; and

computing a parameter (Γ

) using D_m(i, j), Γ

related to an irregularity sample entropy (SampEn) indicative of at least one of irregularity and complexity of the signal, wherein Γ

=Γ

_I if u(i)=d(i), and Γ

=Γ

_C if u(i)=f(i).

23. The method of claim 22, wherein computing the irregularity parameter further comprises computing an irregularity sample entropy (SampEn_I) indicative of the cycle length irregularity of the cardiac signal, wherein SampEn_I=−

InΓ

_I, and computing the complexity parameter further comprises computing a complexity sample entropy (SampEn_C) indicative of the morphological complexity of the cardiac signal, wherein SampEn_C=−

InΓ

_C.