Method and apparatus for observation of ventricular late potentials

• US 5,341,811 A
• Filed: 03/26/1991
• Issued: 08/30/1994
• Est. Priority Date: 03/26/1991
• Status: Expired due to Fees

• Current Assignee: Allegheny-Singer Research Institute
• Original Assignee: Allegheny-Singer Research Institute
• Inventors: Cano, Gerald G.
• Primary Examiner(s): Kamm, William E.
• Assistant Examiner(s): SCHAETZLE, KENNEDY
• Application Number:
  US07/675,996
• Time in Patent Office: 1,253 Days
• Field of Search: 128/696, 128/901, 128/702, 128/708, 364/413.06, 364/413.05
• US Class Current: 600/508
• CPC Class Codes: A61B 5/30 Input circuits therefor
• CPC Class Codes: A61B 5/308 for electrocardiography [ECG]
• CPC Class Codes: A61B 5/366 Detecting abnormal QRS complex, e.g. widening

Abstract

The present invention is directed to a method and apparatus for enhancing high resolution ECG signals acquired with electrodes on the surface of a patient'"'"'s thorax so that ventricular late potentials can be detected and observed. The ECG signals are detected, measured and digitized and are then processed to remove low frequency components and preferably, common mode signals. Enhancement is then accomplished by filtering out the remaining electrical interference caused by underlying muscle tissue, nerve tissue and environmental noise with a particular adaptive filtering technique.

Claims

1. A method of detecting late potentials at the surface of a patient comprising the step of:

(a) acquiring two or more ECG signals at a plurality of external locations on the surface of a patient;

(b) filtering the surface ECG signals with a bandpass filter;

(c) digitizing the acquired ECG signals by continuous sampling at a rate equal to or greater than the Nyquist rate;

(d) storing the digitized ECG signals;

(e) processing the digitized ECG signals to reduce low frequency signal components;

(f) selecting a reference signal with the remaining ECG signals referred to as input signals and selecting a filter window in an ECG cycle from one or more of the digitized ECG input signals and defining filter intervals within said filter window;

(g) calculating a feedback coefficient u_i associated with an adaptive filter algorithm for each filter interval within a filter window in an ECG cycle from one or more of the digitized ECG input signals;

(h) initializing a bias weight associated with an adaptive filtering algorithm for each filter interval by computing an initial value for the bias weight associated with the filter interval;

(i) thereafter adaptively filtering in time-sequenced manner selected filter intervals of the input signals; and

(j) displaying or storing the results of the adaptive filtering step.

2. The method of claim 1 wherein the ECG signals are acquired using two or more concentric electrodes.

3. The method of claim 1 wherein the bandpass filter has a lower cut-off frequency of 0.5 Hz and an upper cut-off frequency of about 400 Hz so that each output after filtering has a bandwidth of approximately 0.5 Hz to 400 Hz.

4. The method of claim 1 wherein the ECG signals are digitized at a rate of 1000 Hz.

5. The method of claim 1 wherein the processing of the digitized ECG signals to reduce low frequency signal components utilizes a multiple pole, high-pass zero-phase shift filter algorithm.

6. The method of claim 5 wherein the cut-off frequency of the multiple pole, high-pass zero phase shift filter algorithm is selectable.

7. The method of claim 1 wherein the digitized ECG signals are processed to reduce common mode signals.

8. The method of claim 7 wherein the step of processing the digitized ECG signals to reduce common mode signals utilizes a 60 Hz Harmonic Reduction Algorithm where:

space="preserve" listing-type="equation">sine=A*sin(2*pi*N*60*i/f.sub.s +pi/2), and (1)

space="preserve" listing-type="equation">cosine=A*cos(2*pi*N*60*i/f.sub.s +pi/2), (2)

are two representations of a 60 Hz harmonic, the representations being 90 degrees out of phase;

the algorithm to reduce the 60 Hz harmonic is implemented by the following equations on consecutive input signal samples;

space="preserve" listing-type="equation">ecg.sub.f (i)=ecg(i)-w.sub.sine *sine-w.sub.cosine *cosine;

(3)

space="preserve" listing-type="equation">w.sub.sine next=w.sub.sine +2*u.sub.harmonic *ecg.sub.f (i) * sine;

(4)

space="preserve" listing-type="equation">w.sub.cosine next=w.sub.cosine +2*u.sub.harmonic *ecg.sub.f (i)*cosine; and

(5)

space="preserve" listing-type="equation">increment i which moves the process to the next sample point;

return to equation (1) or stop if at end of signal;

(6)

where pi is 3.1416, N is a whole number determining the 60 Hz harmonic, f_s is a sampling frequency of the ECG signals, ecg(i) is an acquired and digitized ECG signal, ecg_f (i) is an ECG signal after filtering to reduce 60 Hz harmonic, w_sine is an adaptive filter weight, w_cosine is an adaptive filter weight, u_harmonic is a feedback coefficient controlling convergence rate of the filter.

9. The method of claim 1 wherein the step of calculating the feedback coefficient u_i for any filter interval in the filter window utilizes a Feedback Approximation Algorithm as defined by:

space="preserve" listing-type="equation">u.sub.i β

*[ecg.sub.1 (i).sub.ave.sup.2 +σ

.sub.noise.sup.2 ].sup.-1 *(2*n+1).sup.-1,

where ecg₁ (i)_ave² is a squared value of the average ecg₁ (i) computed from several beats n, σ

_noise is a standard deviation of noise in the acquired signal estimated over a diastolic segment of the ECG, and β

is a positive fraction.

10. The method of claim 1 wherein the means for initializing the bias weight includes a Bias Weight Initialization Algorithm as defined by:

space="preserve" listing-type="equation">w.sub.b (i)=ecg.sub.f (i).sub.beat j, such that Magnitude [ecg.sub.f (i).sub.beat j [<

Magnitude [ecg.sub.f (i).sub.beat k ], for any cycle k considered after the filter has satisfactorily converged,

where w_b (i) is an initial value for the bias weight for each point i in the filter interval, ecg_f (i) is an ECG signal after filtering, beat j is a first beat in the filter interval, beat k is a last beat selected for inclusion in the average, and

where w_b (i) is calculated after adaptive filtering has been applied to several ECG signals, the number of said several ECG signals being identified as the number necessary for said filter to converge to a desired level of satisfactory performance.

11. The method of claim 1 wherein the step of adaptively filtering in a time-sequenced manner filter intervals within a filter window utilizes a Time-Sequenced Adaptive Filtering Algorithm where:

1) a first iteration in a next cycle to filter calculates error(i)'"'"', W(i)'"'"' and Bias Weight as;

space="preserve" listing-type="equation">error(i)'"'"'=ECG.sub.2 (i-1)-ECG.sub.1 (i).sup.T *W(i)-w.sub.b (i);

space="preserve" listing-type="equation">W(i)'"'"'=W(i)+2*u.sub.i *error(i)'"'"'*ECG.sub.1 (i); and

space="preserve" listing-type="equation">Bias Weight=w.sub.b (i)'"'"'=w.sub.b (i)+2*u.sub.b *error(i)'"'"';

where error (i)'"'"' is a computed error of the first iteration, ECG₂ (i-1) is a reference input at i-1, ECG₁ (i)^T is a primary input matrix tranpose, w_b (i) is the bias weight, W(i)'"'"' is a computed weight matrix of the first iteration, W(i) is a starting weight computed in the previous cycle, u_i is a feedback coefficient, ECG₁ (i) is a primary input matrix, w_b (i)'"'"' is a bias weight for the first iteration, and u_b is a feedback coefficient for modifying the bias weight,

2) a second iteration in said cycle calculates error(i)", W(i)" and w_b (i)" as;

space="preserve" listing-type="equation">error(i)"=ECG.sub.2 (i+1)-ECG.sub.1 (i).sup.T *W(i)'"'"'-w.sub.b (i)'"'"';

space="preserve" listing-type="equation">W(i)"=W(i)'"'"'+2*u.sub.i error(i)"*ECG.sub.1 (i); and

space="preserve" listing-type="equation">w.sub.b (i)"=w.sub.b (i)'"'"'+2*u.sub.b *error(i)";

where error*(i)" is a compound error of the second iteration, ECG₂ (i+1) is a reference input at i+1, W(i)" is a computed weight matrix at the second iteration, and w_b (i)" is a bias weight for the second iteration,

3) a filtered output value of an i^th point is determined as the produce of a matrix transpose of the signal at a point i times the adaptive filter weight matrix at point i plus the bias weight value at point i as follows;

space="preserve" listing-type="equation">ecg.sub.f (i)=ECG.sub.1 (i).sup.T *W(i)"+W.sub.b (i)",

space="preserve" listing-type="equation">error(i)=ECG.sub.2 (i)-ecg.sub.f (i),

space="preserve" listing-type="equation">W(i).sub.next cycle =W(i"+2*u.sub.i *error(i)* ECG.sub.1 (i); and

space="preserve" listing-type="equation">w.sub.b (i).sub.next cycle =w.sub.b (i)"+2*u.sub.b *error(i)'"'"',

where ecg_f (i) is a filtered output, error(i) is the error at input i, W(i)_next cycle is a weight matrix to use in a first iteration of the next cycle, w_b (i)_next cycle is a bias weight to use in the first iteration of the next cycle.

12. The method of claim 11 wherein the adaptively filtered filter window includes all or a portion of one or more ECG cycles.

13. The method of claim 12 wherein the adaptively filtered filter window includes a portion of one or more ECG cycles, the portion being about and containing the QRS complex.

14. The method of claim 13 wherein the adaptively filtered filter window is about 400 milliseconds in duration.

15. The method of claim 1 wherein the adaptive filtering within each filter interval includes the steps of:

(a) selecting a last computed value for the bias weight for the filter interval;

(b) conducting a first updating of both a weight matrix associated with the adaptive filter algorithm and the bias weight for the input signal within the filter interval, using the reference signal for the preceding filter interval;

(c) conducting a second updating of the weight matrix and the bias weight for the input signal of the filter interval using the reference signal of the subsequent filter interval;

(d) calculating an output signal by multiplying the second updated weight matrix of the filter interval by the associated input signal, and adding the bias weight;

(e) conducting a third updating of the weight matrix and bias weight of step (d) associated with the filter interval using the output calculated in step (d), and storing the third updating of the weight matrix and bias weight for use in a subsequent cycle.

16. The method of claim 15 wherein the filter interval is one sample point in duration and the weight matrix is a five element weight matrix.

17. The method of claim 1 wherein the display includes computation of predefined measurements of the patient'"'"'s electrocardiographic activity either in fully processed individual beats or in averages of selected fully processed beats, wherein the number of beats is selectable.

18. The method of claim 17 wherein the display includes computation of the root mean squared amplitude of the last 40 milliseconds of a QRS complex within the filter interval.

19. The method of claim 17 wherein the display includes computation of the duration of QRS complex within the filter interval.

20. The method of claim 17 wherein the display includes computation of a low amplitude signal duration.

21. An apparatus for detecting late potentials at the surface of a patient comprising:

(a) a means for acquiring two or more ECG signals at a plurality of external locations on the surface of a patient;

(b) means for digitizing the acquired ECG signals by continuous sampling at a rate equal to or grater than the Nyquist rate;

(d) a means for storing the digitized ECG signals;

(e) a means for processing the digitized ECG signals to reduce low frequency signal components;

(f) means for selecting a reference signal with the remaining ECG signals referred to as input signals and a means for selecting a filter window in an ECG cycle from one or more of the digitized ECG input signals and defining filter intervals within said filter window;

(g) a means for calculating a feedback coefficient u_i associated with an adaptive filter algorithm for each filter interval within a filter window in an ECG cycle from one or more of the digitized ECG input signals;

(h) a means for initializing a bias weight associated with an adaptive filtering algorithm for each filter interval by computing a initial value for the bias weight associated with the filter interval;

(i) a means for thereafter adaptively filtering in time-sequenced manner selected filter intervals of the input signals; and

(j) a means for displaying or storing the results of the adaptive filtering step.

22. The apparatus of claim 21 wherein the ECG signals are acquired using two or more concentric electrodes.

23. The apparatus of claim 21 wherein the bandpass filter has a lower cut-off frequency of 0.5 Hz and an upper cut-off frequency of about 400 Hz so that each output after filtering has a bandwidth of approximately 0.5 Hz to 400 Hz.

24. The apparatus of claim 21 which further comprises a means for digitizing the ECG signals at a rate of 1000 Hz.

25. The apparatus of claim 24 wherein the means for processing the digitized ECG signals to reduce low frequency signal components utilizes a multiple pole, high-pass zero-phase shift filter algorithm.

26. The apparatus of claim 25 which further comprises a means to select the cut-off frequency of the multiple pole, high-pass zero phase shift filter algorithm.

27. The apparatus of claim 21 which further comprises a means to reduce common mode signals in the digitized ECG signals.

28. The apparatus of claim 27 which further comprises a means for processing the digitized ECG signals to reduce common mode signals, wherein said means utilizes a 60 Hz Harmonic Reduction Algorithm where:

space="preserve" listing-type="equation">sine=A*sin(2*pi*N*60*i/f.sub.s) and (1)

space="preserve" listing-type="equation">cosine=A*cos(2*pi*N*60*i/f.sub.s +pi/2), (2)

are two representations of a 60 Hz harmonic, the representations being 90 degrees out of phase;

the algorithm to reduce the 60 Hz harmonic is implemented by the following equations on consecutive input signal samples;

space="preserve" listing-type="equation">ecg.sub.f (i)=ecg(i)-w.sub.sine *sine-w.sub.cosine *cosine;

(3)

space="preserve" listing-type="equation">w.sub.sine next =w.sub.sine +2*u.sub.harmonic *ecg.sub.f (i)*sine;

(4)

space="preserve" listing-type="equation">w.sub.cosine next =w.sub.cosine next =w.sub.cosine +2*u.sub.harmonic * ecg.sub.f (i)*cosine; and

(5)

space="preserve" listing-type="equation">increment i which moves the process to the next sample point;

return to equation (1) or stop if at end of signal;

where pi is 3.1416, N is a whole number determining the 60 Hz harmonic, f_s is a sampling frequency of the ECG signals, ecg(i) is an acquired and digitized ECG signal, ecg_f (i) is an ECG signal after filtering to reduce 60 Hz harmonic, w_sine is an adaptive filter weight, w_cosine is an adaptive filter weight, u_harmonic is a feedback coefficient controlling convergence rate of the filter.

29. The apparatus of claim 21 which further comprises a means for calculating the feedback coefficient u_i for any filter interval in the filter window wherein said means utilizes a Feedback Approximation Algorithm as defined by:

space="preserve" listing-type="equation">u.sub.i =β

*[ecg.sub.1 (i).sub.ave.sup.2 +σ

.sub.noise.sup.2 ].sup.-1 *(2*n+1).sup.-1,

where ecg₁ (i)_ave² is a squared value of the average ecg₁ (i) computed from several beats n, σ

noise is a standard deviation of noise in the acquired signal estimated over a diastolic segment of the ECG, and β

is a positive fraction.

30. The apparatus of claim 21 wherein the means for initializing the bias weight utilizes a Bias Weight Initialization Algorithm as defined by:

space="preserve" listing-type="equation">w.sub.b (i)=ecg.sub.f (i).sub.beat j,

such that Magnitude [ecg_f (i)_beat j ]<

Magnitude [ecg_f (i)_beat k ], for any cycle k considered after the filter has satisfactorily converged,

where w_b (i) is an initial value for the bias weight for each point i in the filter interval, ecg_f (i) is an ECG signal after filtering, beat j is a first beat in the filter interval, beat k is a last beat selected for inclusion in the average, and

where w_b I is calculated after adaptive filtering has been applied to several ECG signals, the number of said several ECG signals being identified as the number necessary for said filter to converge to a desired level of satisfactory performance.

31. The apparatus of claim 21 which further comprises a means for adaptively filtering in a time-sequenced manner filter intervals within a filter window, where said means utilizes a Time-Sequenced Adaptive Filtering Algorithm where:

a first iteration in a next cycle to filter calculates error(i)'"'"', W(i)'"'"' and Bias Weight as;

space="preserve" listing-type="equation">error(i)'"'"'=ECG.sub.2 (i-1)-ECG.sub.1 (i).sup.T *w.sub.b (i);

space="preserve" listing-type="equation">W(i)'"'"'=W(i)+2*u.sub.i * error(i); and

space="preserve" listing-type="equation">Bias Weight=w.sub.b (i)'"'"'=w.sub.b (i)+2*u.sub.b *error(i)'"'"';

where error (i)'"'"' is a computed error of the first iteration, ECG₂ (i-1) is a reference input at i-1, ECG₁ (i)^T is a primary input matrix transpose, w_b (i) is the bias weight, W(i)'"'"' is a computed weight matrix of the first iteration, W(i) is a starting weight computed in the previous cycle, u_i is a feedback coefficient, ECG₁ (i) is a primary input matrix, w_b (i)'"'"' is a bias weight for the first iteration, and u_b is a feedback coefficient for modifying the bias weight,

2) a second iteration in said cycle calculates error(i)", W(i)" and w_b (i)" as;

space="preserve" listing-type="equation">error(i)"=ECG.sub.2 (i+1)-ECG.sub.1 (i).sup.T *W(i)'"'"'-w.sub.b (i)'"'"';

space="preserve" listing-type="equation">W(i)"=W(i)'"'"'+2*u.sub.i *error(i)"*ECG.sub.1 (i); and

space="preserve" listing-type="equation">w.sub.b (i)"=w.sub.b (i)'"'"'+2*u.sub.b *error(i)";

where error(i)" is a computed error of the second iteration, ECG₂ (i+1) is a reference input at i+1, W(i)" is a computed weight matrix at the second iteration, and w_b (i)" is a bias weight for the second iteration,

3) a filtered output value of an i^th point is determined as the product of a matrix transpose of the signal at a point i times the adaptive filter weight matrix at point i plus the bias weight value at point i as follows;

space="preserve" listing-type="equation">ecg.sub.f (i)=ECG.sub.1 (i).sup.T *W(i)"+W.sub.b (i)",

space="preserve" listing-type="equation">error(i)=ECG.sub.2 (i)-ecg.sub.f (i),

space="preserve" listing-type="equation">W(i).sub.next cycle =W(i)"+2*u.sub.i *error(i)* ECG.sub.1 (i); and

space="preserve" listing-type="equation">w.sub.b (i).sub.next cycle =w.sub.b (i)"+2*u.sub.b *error(i)'"'"',

where ecg_f (i) is a filtered output, error(i) is the error at input i, W(i)_next cycle is a weight matrix to use in a first iteration of the next cycle, w_b (i)_next cycle is a base weight to use in a first iteration of the next cycle.

32. The apparatus of claim 31 wherein the adaptively filtered filter window includes all or a portion of one or more ECG cycles.

33. The apparatus of claim 32 which further comprises a means for detecting a QRS complex and adaptively filtering the filter window which includes a portion of one or more ECG cycles, the portion being about and containing the QRS complex.

34. The apparatus of claim 33 wherein the adaptively filtered filter window is about 400 milliseconds in duration.

35. The apparatus of claim 21 wherein the adaptive filtering within each filter interval includes:

(a) a means for selecting a last computed value for the bias weight for the filter interval;

(b) a means for conducting a first updating of both a weight matrix associated with the adaptive filter algorithm and the bias weight for the input signal within the filter interval, using the reference signal for the preceding filter interval;

(c) a means for conducting a second updating of the weight matrix and the bias weight for the input signal of the filter interval using the reference signal of the subsequent filter interval;

(d) a means for calculating an output signal by multiplying the second updated weight matrix of the filter interval by the associated input signal, and adding the bias weight;

(e) a means for conducting a third updating of the weight matrix and bias weight of step (d) associated with the filter interval using the output calculated in step (d), and storing the third updating of the weight matrix and bias weight for use in a subsequent cycle.

36. The apparatus of claim 35 wherein the filter interval is one sample point in duration and the weight matrix is a five element weight matrix.

37. The apparatus of claim 21 wherein the display includes a means for computation of predefined measurements of the patient'"'"'s electrocardiographic activity either in fully processed individual beats or in averages of selected fully processed beats, wherein the number of beats is selectable.

38. The apparatus of claim 37 wherein the display includes a means for computation of the root mean squared amplitude of the last 40 milliseconds of a QRS complex within the filter interval.

39. The apparatus of claim 37 wherein the display includes a means for computation of the duration of a QRS complex within the filter interval.

40. The apparatus of claim 37 wherein the display includes a means for computation of a low amplitude signal duration.

41. A method of detecting low level biophysical signals of a cyclic nature at the surface of a patient comprising the steps of:

(a) acquiring two or more biophysical signals at a plurality of external locations on the surface of a patient;

(b) filtering the surface biophysical signals with a bandpass filter;

(c) digitizing the acquired biophysical signals by continuous sampling at a rate equal to or greater than the Nyquist rate;

(d) storing the digitized biophysical signals;

(e) processing the digitized biophysical signals to reduce low frequency signal components;

(f) selecting a reference signal with the remaining biophysical signals referred to as input signals and selecting a filter window in a biophysical cycle from one or more of the digitized biophysical input signals and defining filter intervals within said filter window;

(g) calculating a feedback coefficient u_i associated with an adaptive filter algorithm for each filter interval within a filter window in a cycle of the biophysical signal from one or more of the digitized biophysical input signals;

(h) initializing a bias weight associated with an adaptive filtering algorithm for each filter interval by computing an initial value for the bias weight associated with the filter interval;

(i) thereafter adaptively filtering in time-sequenced manner selected filter intervals of the input signals; and

(j) displaying or storing the results of the adaptive filtering step.

42. An apparatus for detecting low level biophysical signals of a cyclic nature at the surface of a patient comprising:

(a) means for acquiring two or more biophysical signals at a plurality of external locations on the surface of a patient;

(b) means for bandpass filtering the surface biophysical signals;

(c) a means for digitizing the acquired biophysical signals by continuous sampling at a rate equal to or greater than the Nyquist rate;

(d) a means for storing the digitized biophysical signals;

(e) a means for processing the digitized biophysical signals to reduce low frequency signal components;

(f) a means for selecting a reference signal with the remaining biophysical signals referred to as input signals and a means for selecting a filter window in a biophysical cycle from one or more of the digitized biophysical input signals and defining filter intervals within said filter window;

(g) a means for calculating a feedback coefficient u_i associated with an adaptive filter algorithm for each filter interval within a filter window in a cycle of the biophysical signal from one or more of the digitized biophysical input signals;

(h) a means for initializing a bias weight associated with an adaptive filtering algorithm for each filter interval by computing an initial value for the bias weight associated with the filter interval;

(i) a means for thereafter adaptively filtering in time-sequenced manner selected filter intervals of the input signals; and

(j) a means for displaying or storing the results of the adaptive filtering step.