Curvature based method for selecting features from an electrophysiologic signal for purpose of complex identification and classification

• US 7,289,845 B2
• Filed: 12/12/2003
• Issued: 10/30/2007
• Est. Priority Date: 10/31/2000
• Status: Expired due to Term

• Current Assignee: Cardiac Pacemakers Incorporated (Boston Scientific Corporation)
• Original Assignee: Cardiac Pacemakers Incorporated (Boston Scientific Corporation)
• Inventors: Lovett, Eric G., Sweeney, Robert J.
• Primary Examiner(s): CIRIC, LJILJANA V
• Application Number:
  US10/734,327
• Publication Number:
  US 20040127806A1
• Time in Patent Office: 1,418 Days
• Field of Search: 600/508, 600/513, 600/514, 600/515, 600/516, 600/517, 600/522
• US Class Current: 600/517
• CPC Class Codes: A61B 5/35 by template matching
• 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: A61N 1/362 Heart stimulators heart defibrillators A61N1/39
• 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

A method for curvature based complex identification and classification comprises sensing a cardiac signal and computing curvatures at sample points on the sensed cardiac signal. Then to extract features from the computed curvatures, and compare the extracted features with a set of predetermined templates, and then to classify the sensed cardiac signal based on the outcome of the comparison.

Claims

1. A cardiac rhythm management system, comprising:

at least one electrode;

a signal sensing circuit coupled to the electrode to sense a cardiac signal, the cardiac signal having signal complexes;

a controller coupled to the sensing circuit, wherein the controller receives the sensed cardiac signal, and wherein the controller includes;

memory configured to store signal complexes and to store sets of predetermined templates, the predetermined templates including sets of template zones; and

an analyzing module configured to;

compute curvatures at sample points on the sensed cardiac signal,

extract features from the computed curvatures, and

classify the signal complexes by comparing the extracted features to the template zones of the predetermined templates.

2. The system of claim 1, wherein the template zones include a center amplitude, a time width and an amplitude width.

3. The system of claim 2, wherein the template zones include different centers and widths for different types of cardiac signals.

4. The system of claim 3, wherein the controller further includes a scoring module, wherein the scoring module assigns a score to an extracted feature based on how well the feature matches a set of template zones.

5. The system of claim 4, wherein the template zones are modified over time to improve matching with the extracted features.

6. The system of claim 1, wherein the analyzing module is configured to assign an unknown classification to a signal complex when the extracted features fail to match the sets of templates.

7. A method comprising:

sensing a voltage versus time cardiac signal using an implantable medical device;

converting the voltage versus time cardiac signal into a time versus time signal;

computing curvature at sample points on the time versus time signal;

extracting features from the computed curvature;

comparing the extracted features with a set of predetermined templates; and

classifying the cardiac signal based on an outcome of the comparison.

8. The method of claim 7, wherein extracting features from the computed curvatures includes separating a set of features associated with a first cardiac signal complex upon detecting a subsequent second cardiac signal complex.

9. The method of claim 8, wherein separating the set of features includes identifying a fiducial feature from the set of separated features based on comparing the time when the first signal complex occurs with the time when the separated set of features associated with the first signal complex occurs.

10. The method of claim 7, wherein comparing the extracted features with a set of predetermined templates includes comparing the extracted features with a set of predetermined time versus time templates.

11. A cardiac rhythm management system, comprising:

at least one electrode;

a signal sensing circuit coupled to the electrode to sense a cardiac signal, the cardiac signal having signal complexes;

a controller coupled to the sensing circuit, wherein the controller receives the sensed cardiac signal, and wherein the controller includes;

memory configured to store at least a first signal complex; and

an analyzing module configured to;

compute curvatures at sample points of the sensed cardiac signal,

continuously process the sensed cardiac signal to produce an absolute value of the first derivative of the sensed cardiac signal, and

compare the absolute value of the first derivative to a predetermined decaying threshold value to detect a second subsequent signal complex.

12. The system of claim 11, wherein the analyzing module separates a set of features associated with the first signal complex upon detecting the second signal complex.

13. The system of claim 11, wherein the second signal complex is detected when the absolute value of the first derivative of the cardiac signal exceeds the predetermined decaying threshold value.

14. The system of claim 12, wherein the analyzer module is configured to identify a fiducial feature from the set of separated features associated with the first signal complex.

15. The system of claim 14, wherein the fiducial feature is identified based on comparing a time when the first signal complex occurs with the time when the separated features occur.

16. The system of claim 12, wherein the memory stores the features associated with the first signal complex, and the stored features are replaced with features of the second signal complex when the second signal complex is detected.

17. A method comprising:

sensing a cardiac signal using an implantable medical device;

computing curvature at sample points on the sensed cardiac signal;

computing average curvature between two adjacent sample points by computing an integral of the curvature between the two adjacent sample points;

extracting features from the computed curvatures;

comparing the extracted features with a set of predetermined templates; and

classifying the sensed cardiac signal based on an outcome of the comparison.

18. The method of claim 17, wherein computing the curvature at sample points on the sensed cardiac signal includes computing a curvature versus time curve, and wherein computing an integral of the curvature includes finding the area under a curvature versus time curve.

19. The method of claim 18, wherein extracting features from the computed curvatures includes defining the features by a value of the area under the curvature versus time curve.

20. The method of claim 19, wherein extracting features from the computed curvatures includes comparing the value of the area for the features to an area threshold value and eliminating features with area values less than the threshold area value.