Method of detecting P-waves in ECG recordings

• US 4,721,114 A
• Filed: 02/21/1986
• Issued: 01/26/1988
• Est. Priority Date: 02/21/1986
• Status: Expired due to Term

• Current Assignee: Cardiac Pacemakers Incorporated (Boston Scientific Corporation)
• Original Assignee: Cardiac Pacemakers Incorporated (Boston Scientific Corporation)
• Inventors: Wilcox, Ann C., DuFault, Robert A.
• Primary Examiner(s): Kamm, William E.
• Application Number:
  US06/831,519
• Time in Patent Office: 704 Days
• Field of Search: 128/419 PT, 128/695, 128/696, 128/697, 128/702-704, 128/707, 128/708, 128/710, 364/415, 364/417
• US Class Current: 600/509
• CPC Class Codes: A61B 5/35 by template matching

Abstract

A method for automatically detecting P-waves in a continuous electrocardiograph signal stream irrespective of the presence of P-R disassociation wherein the lower energy P-wave can meld into the higher energy R-wave. In accordance with the invention, the ECG signal stream is digitized and fed to a computer programmed to perform signal processing operations thereon. First, a basic QRS morphologic prototype is developed through coherent-averaging of a plurality of QRS complexes appearing in the signal stream. Once the prototype is established, it is aligned with the incoming ECG signal stream where a recognizable similarity exists. The prototype is multiplied by a function comprising a "window" to smooth it to zero at its right and left edges while preserving essential center shape. The resulting windowed prototype is then scaled to match in amplitude or energy content the ECG signal stream at the points of alignment. The thus-modified prototype is then subtracted from the signal stream at those points, forming a residue waveform comprising P-waves, residual R-wave energy and ectopic beats. The residue signal may then be smoothed before comparing it to a predetermined linear threshold. The energy amplitudes in the residue signal exceeding the threshold are identified as P-waves.

Claims

1. A method for automatically detecting P-waves in a continuous electrocardiograph (ECG) signal stream, irrespective of the presence of P-R disassociation, comprising the steps of:

(a) converting said ECG signal stream to a digital representation;

(b) adaptively developing from the digitized ECG signal stream at least one basic QRS morphologic prototype by the coherent averaging of a number of QRS complexes in said ECG signal stream;

(c) aligning said basic QRS prototype with similar QRS complexes in said ECG signal stream by a normalized convolution process;

(d) multiplying said basic QRS prootype by a mathematical window function for tapering said prototype smoothly to zero at its respective edges while preserving its essential shape at the center;

(e) scaling the windowed QRS prototype to provide an amplitude/energy match with said signal stream at the points of alignment;

(f) subtracting the scaled and windowed prototype from said signal stream at the points of alignment to form a residue waveform signal principally made up of P-waves, residual R-wave energy and ectopic beats;

(g) smoothing said residue signal with a linear band pass filter; and

(h) comparing said residue signal to a predetermined linear threshold, the energy amplitudes in said residue signal exceeding said threshold being idenified as P-waves.

2. A method for automatically detecting P-waves in a continuous electrocardiograph (ECG) signal stream, irrespective of the presence of P-R disassociation, comprising the steps of:

(a) converting said ECG signal stream to a digital representation;

(b) adaptively developing from the digitized ECG signal stream at least one basic QRS morphologic prototype by the coherent averaging of a number of QRS complexes in said ECG signal stream;

(c) aligning said basic QRS prototype with similar QRS complexes in said ECG signal stream;

(d) multiplying said basic QRS prototype by a mathematical window function for tapering said prototype smoothly to zero at its respective edges while preserving its essential shap at the center;

(e) scaling the windowed QRS prototype to provide an amplitude/energy match with said signal stream at the points of alignment;

(f) subtracting the scaled and windowed prototype from said signal stream at the points of alignment to form a residue waveform signal principally made up of P-waves, residual R-wave energy and ectopic beats;

(g) developoing a P-wave morphologic prototype by coherent-averaging of P-waves, predetermined to be morphologically characteristic of P-waves in said residual signal;

(h) developing a normalized correlation function between said residue waveform of step (f) and said P-wave morphologic prototype of step (g); and

(i) detecting the P-waves by applying a predetermined threshold to the normalized correlation function of step (h) whereby the largest positive excursions mark the location of the P-waves.