Method of processing cardiovascular sound signals

• US 9,364,184 B2
• Filed: 05/14/2015
• Issued: 06/14/2016
• Est. Priority Date: 03/18/2002
• Status: Active Grant

• Current Assignee: AusculSciences, Inc.
• Original Assignee: SonoMedica, Inc.
• Inventors: Figgatt, Alan M., Mulholland, Daniel J., Holford, Warren
• Primary Examiner(s): GETZOW, SCOTT M
• Application Number:
  US14/711,931
• Publication Number:
  US 20150238147A1
• Time in Patent Office: 397 Days
• Field of Search: 600/513, 600/514, 600/528, 600/586, 600/493
• US Class Current: 1/1
• CPC Class Codes: A61B 5/02 Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
• CPC Class Codes: A61B 5/0205 Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
• CPC Class Codes: A61B 5/0245 by using sensing means generating electric signals, i.e. ECG signals
• CPC Class Codes: A61B 5/318 Heart-related electrical modalities, e.g. electrocardiography [ECG]
• CPC Class Codes: A61B 5/322 Physical templates or devices for measuring ECG waveforms, e.g. electrocardiograph rulers or calipers by template matching A61B5/35
• CPC Class Codes: A61B 5/7203 for noise prevention, reduction or removal
• CPC Class Codes: A61B 5/7246 using correlation, e.g. template matching or determination of similarity
• CPC Class Codes: A61B 5/725 using specific filters therefor, e.g. Kalman or adaptive filters A61B5/316 take precedence
• CPC Class Codes: A61B 5/7257 using Fourier transforms
• CPC Class Codes: A61B 5/7278 Artificial waveform generation or derivation, e.g. synthesising signals from measured signals A61B5/327 takes precedence
• CPC Class Codes: A61B 5/7282 Event detection, e.g. detecting unique waveforms indicative of a medical condition A61B5/4094, A61B5/4818, A61B5/349, A61B5/0823 take precedence
• CPC Class Codes: A61B 7/026 comprising more than one sound collector
• CPC Class Codes: A61B 7/04 Electric stethoscopes microphones, acoustic transducers therefor H04R
• CPC Class Codes: G16H 50/30 for calculating health indices; for individual health risk assessment

Abstract

A math model envelope of a single heartbeat between primary and secondary peaks of an autocorrelation of a received cardiovascular sound signal comprises an S1 lobe and a like-polarity S2 lobe. A bootstrap filter envelope is generated by averaging data from a plurality of heart cycles within heart cycle boundaries located responsive to a convolution of the envelope with the cardiovascular sound signal. Start points of a plurality of heart cycle signals are located from a convolution of the bootstrap filter envelope with the cardiovascular sound signal. The S1 and S2 regions are located within a cardiovascular sound signal responsive to peak and valley locations thereof and responsive to a phase window signal generated by filtering an average of a plurality of heart cycle signals, wherein the peak and valley locations are located responsive to a second derivative of the phase window signal and responsive to the start points.

Claims

1. A method comprising:

a. receiving a cardiovascular sound signal;

b. generating an autocorrelation signal from an autocorrelation of said cardiovascular sound signal;

c. generating a math model envelope of a single heart cycle extending between primary and secondary peaks of said autocorrelation signal, wherein said math model envelope comprises a first lobe corresponding to an S1 region of said single heart cycle and a second lobe corresponding to an S2 region of said single heart cycle, and said first and second lobes are each of like polarity relative to a common reference;

d. generating a bootstrap filter envelope by averaging data of said cardiovascular sound signal from a plurality of corresponding heart cycles within a corresponding plurality of heart cycle boundaries so as to generate averaged heart cycle data, wherein the operation of averaging of said data is over a period of one heart cycle, and said plurality of heart cycle boundaries are located within said cardiovascular sound signal responsive to a convolution of said math model envelope with said cardiovascular sound signal; and

e. locating a set of start points of each of a plurality of heart cycle signals in said cardiovascular sound signal responsive to a convolution of said bootstrap filter envelope with said cardiovascular sound signal.

2. A method as recited in claim 1, further comprising smoothing said cardiovascular sound signal prior to the operation of generating said autocorrelation signal.

3. A method as recited in claim 2, wherein the operation of smoothing said cardiovascular sound signal comprises operating on said cardiovascular sound signal with at least one first set of operations, at least one second set of operations, and at least one third operation, wherein said at least one first set of operations is selected from filtering, smoothing and decimating, said at least one second set of operations comprises subtracting a mean value so as to generate a zero mean first intermediate signal and generating an absolute value second intermediate signal, and said at least one third operation comprises convolving an intermediate signal responsive to said cardiovascular sound signal with an averaging or low-pass filtering window signal.

4. A method as recited in claim 3, wherein said averaging or low-pass filtering window signal comprises either a unity averaging window, a Blackman window, a Gaussian window, or a Kaiser-Bessel window.

5. A method as recited in claim 1, further comprising determining an estimate of a heartbeat duration responsive to a period of time between said primary and secondary peaks of said autocorrelation signal.

6. A method as recited in claim 5, further comprising:

a. determining a parsing score responsive to at least one of the group selected from a first condition, a second condition, and a third condition, wherein said first condition is indicative of said heartbeat duration of at least one heart cycle being less than a minimum heartbeat duration threshold, said second condition is indicative of a variation of said heartbeat duration being greater than a heartbeat duration variation threshold, and said third condition is indicative of the occurrence of unexpected peaks in said set of start points; and

c. if said parsing score is less than a maximum possible parsing score, then repeating the step of locating said set of start points of each of said plurality of heart cycle signals in said cardiovascular sound signal using data of a subsequent heart cycle of said cardiovascular sound signal instead of said bootstrap filter envelope, so that the locations of said start points of said set of start points of each of said plurality of heart cycle signals in said cardiovascular sound signal is responsive to a convolution of said data of said subsequent heart cycle with said cardiovascular sound signal.

7. A method as recited in claim 1, wherein each of said first and second lobes comprises a corresponding half-cosine envelope.

8. A method as recited in claim 1, wherein the operation of generating said math model envelope comprises:

a. determining a plurality of peaks and valleys of said autocorrelation signal responsive to a second derivative of said autocorrelation signal between said primary and secondary peaks; and

b. determining a location and a period of each of said first and second lobes responsive to said plurality of peaks and valleys of said autocorrelation signal.

9. A method as recited in claim 1, wherein the operation of generating said bootstrap filter envelope further comprises:

a. convolving a Kaiser-Bessel weighting window with said cardiovascular sound signal so as to generate a gain normalization signal; and

b. dividing values of said cardiovascular sound signal by corresponding values of said gain normalization signal so as to generate a corresponding automatic gain control signal, wherein said plurality of heart cycle boundaries are responsive to values of a convolution signal responsive to said convolution of said math model envelope with said cardiovascular sound signal, divided by corresponding values of said automatic gain control signal.

10. A method as recited in claim 1, wherein said plurality of corresponding heart cycles are selected from said cardiovascular sound signal responsive to a corresponding magnitude of said convolution.

11. A method as recited in claim 1, further comprising suppressing data points beyond said S2 region of said averaged heart cycle data using a Kaiser filter.

12. A method as recited in claim 1, wherein the operation of locating said set of start points of each of said plurality of heart cycle signals in said cardiovascular sound signal comprises:

a. convolving said bootstrap filter envelope with said cardiovascular sound signals so as to generate a cross-correlation signal;

b. generating an any of correlation peaks from said cross-correlation signal; and

c. multiplying said array of correlation peaks by said cardiovascular sound signal so as to determine a corresponding array of peak values, wherein said set of start points of each of said plurality of heart cycle signals is responsive to said any of peak values.

13. A method as recited in claim 12, wherein the operation of generating said array of correlation peaks is responsive to non-negative values of a sign of difference of a sign of difference of successive points of said cross-correlation signal.

14. A method as recited in claim 12, wherein said array of peak values are of at least one peak of said cross-correlation signal each having a value in excess of a threshold, and said threshold is selected so that a count of said peak values in said array of peak values is not less than a maximum number of peaks of said cross-correlation signal.

15. A method as recited in claim 12, further comprising determining a duration of at least one heart cycle from a period of time between adjacent-in-time peak values in said array of peak values.

16. A method as recited in claim 1, further comprising:

a. convolving S1 regions of a plurality of adjacent heart cycles with one another so as to generate a convolution signal;

b. determining a delta shift of a corresponding peak value of one of said plurality of adjacent heart cycles responsive to said convolution signal; and

c. shifting said corresponding peak value in time responsive to said convolution signal.

17. A method, comprising:

a. receiving either a cardiovascular sound signal or a corresponding smoothed cardiovascular sound signal;

b. if said cardiovascular sound signal is not said smoothed cardiovascular sound signal, then smoothing said cardiovascular sound signal so as to generate said smoothed cardiovascular sound signal;

c. if said cardiovascular sound signal is not a narrowband cardiovascular sound signal, then either generating said narrowband cardiovascular sound signal from said cardiovascular sound signal, OR receiving said narrowband cardiovascular sound signal that was previously generated from said cardiovascular sound signal, wherein said narrowband cardiovascular sound signal has a bandwidth greater than 130 Hz and less than 460 Hz;

d. locating a set of start points of each of a plurality of heart cycle signals in said smoothed cardiovascular sound signal;

e. averaging said plurality of heart cycle signals in said smoothed cardiovascular sound signal so as to generate an average envelope signal;

f. filtering said average envelope signal so as to enhance S1 and S2 regions and so as to generate a corresponding phase window signal;

g. locating a plurality of peaks and valleys in said average envelope signal responsive to a second derivative of said phase window signal;

h. determining peak and valley locations of said S1 and S2 regions in said cardiovascular sound signal responsive to corresponding peaks and valleys of said plurality of peaks and valleys in said average envelope signal and responsive to said set of start points of each of said plurality of heart cycle signals in said smoothed cardiovascular sound signal; and

i. determining indices of said S1 and S2 regions within said smoothed cardiovascular sound signal responsive to said peak and valley locations of said S1 and S2 regions in said smoothed cardiovascular sound signal and responsive to said phase window signal.

18. A method as recited in claim 17, further comprising:

a. bandpass filtering said narrowband cardiovascular sound signal so as to generate a bandpass-filtered narrowband signal;

b. determining locations of peaks and valleys of a signal responsive to an absolute value of said bandpass-filtered narrowband signal; and

c. responsive to said locations of said peaks and valleys of said signal responsive to said absolute value of said bandpass-filtered narrowband signal, determining at least one of an index of an S3 region within said smoothed cardiovascular sound signal or an index of an S4 region within said smoothed cardiovascular sound signal.

19. A method as recited in claim 17, wherein the operation of smoothing said cardiovascular sound signal comprises operating on said cardiovascular sound signal with at least one first set of operations, at least one second set of operations, and at least one third operation, wherein said at least one first set of operations is selected from filtering, smoothing or decimating, said at least one second set of operations comprises subtracting a mean value so as to generate a zero mean first intermediate signal and generating an absolute value second intermediate signal, and said at least one third operation comprises convolving an intermediate signal responsive to said cardiovascular sound signal with an averaging or low-pass filtering window signal.

20. A method as recited in claim 19, said averaging or low-pass filtering window signal comprises either a unity averaging window, a Blackman window, a Gaussian window, or a Kaiser-Bessel window.

21. A method as recited in claim 17, wherein the operation of generating said narrowband cardiovascular sound signal comprises operating on said cardiovascular sound signal with at least one first set of operations and at least one second set of operations, wherein said at least one first set of operations is selected from filtering, smoothing or decimating, said at least one second set of operations comprises subtracting a mean value so as to generate a zero mean first intermediate signal and generating an absolute value second intermediate signal.

22. A method as recited in claim 17, wherein the operation of locating said set of start points of each of said plurality of heart cycle signals in said smoothed cardiovascular sound signal comprises:

a. generating an autocorrelation signal from an autocorrelation of said smoothed cardiovascular sound signal;

b. generating a math model envelope of a single heart cycle extending between primary and secondary peaks of said autocorrelation signal, wherein said math model envelope comprises a first lobe corresponding to an S1 region of said single heart cycle and a second lobe corresponding to an S2 region of said single heart cycle, and said first and second lobes are each of like polarity relative to a common reference;

c. generating a bootstrap filter envelope by averaging data of said smoothed cardiovascular sound signal from a plurality of corresponding heart cycles within a corresponding plurality of heart cycle boundaries so as to generate averaged heart cycle data, wherein the operation of averaging of said data is over a period of one heart cycle, and said plurality of heart cycle boundaries are located within said smoothed cardiovascular sound signal responsive to a convolution of said math model envelope with said smoothed cardiovascular sound signal; and

d. locating said set of start points of each of a plurality of heart cycle signals in said smoothed cardiovascular sound signal responsive to a convolution of said bootstrap filter envelope with said smoothed cardiovascular sound signal.

23. A method as recited in claim 17, further comprising receiving an ECG signal associated and synchronized with said cardiovascular sound signal, wherein the operation of locating said set of start points of each of said plurality of heart cycle signals in said smoothed cardiovascular sound signal is responsive to said ECG signal.

24. A method as recited in claim 17, wherein the operation of filtering said average envelope signal comprises convolving said average envelope signal with a Finite Impulse Response filter signal.

25. A method as recited in claim 17, wherein the operation of filtering said average envelope signal comprises enhancing frequencies of said average envelope signal between 20 and 60 Hertz.

26. A method as recited in claim 17, further comprising removing a low-pass filter offset from said phase window signal.