Advanced signal processing methodology for the detection, localization and quantification of acute myocardial ischemia

• US 4,974,162 A
• Filed: 03/13/1987
• Issued: 11/27/1990
• Est. Priority Date: 03/13/1987
• Status: Expired due to Fees

• Current Assignee: University of Maryland The A Body Corporate of MD
• Original Assignee: University of Maryland
• Inventors: Nikias, Chrysostomos L., Siegel, John H.
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): BUI, KIM T
• Application Number:
  US07/025,896
• Time in Patent Office: 1,355 Days
• Field of Search: 364/417, 364/413.06, 128/902, 128/703
• US Class Current: 600/509
• CPC Class Codes: A61B 5/349 Detecting specific parameters of the electrocardiograph cycle

Abstract

Monopolar ECG signals are converted into the multichannel spectrum and autocorrelation domains, and different decision variables are identified from the autospectra. Statically validated threshold levels are used for comparison with the decision variables to determine the probability of normal and ischemic conditons.

Claims

1. A method of detecting myocardial ischemia from a time-domain ECG signal X(n);

n=0, 1, . . . , N, where N is an integer, comprising the steps of;

receiving said ECG signal;

generating a plurality of decision variables from said ECG signal;

combining said decision variables to form a combined decision variable in a multivariate probability space;

comparing said combined decision variable to a threshold value to obtain comparison results; and

determining the probability of ischemia based on said comparison results.

2. A method of detecting myocardial ischemia from a time-domain ECG signal X(n);

n=0, 1, . . . , N, where N is an integer, comprising the steps of;

receiving said ECG signal;

generating a decision variable from said ECG signal, said generating step comprising converting said ECG signal to the spectrum domain and determining as said variable the width of said converted ECG signal at 50% below said peak frequency;

comparing said decision variable to a threshold value of said decision variable to obtain comparison results; and

determining a probability of ischemia based on said comparison results.

3. A method of detecting myocardial ischemia from a time-domain ECG signal X(n);

n=0, 1, . . . , N, where N is an integer, comprising the steps of;

receiving said ECG signal;

generating a decision variable from said ECG signal by generating an autocorrelation sequence from said ECG signal, forming an autocorrelation matrix from said autocorrelation sequence, computing eigenvalues from said matrix, calculating the differences between successive ones of said eigenvalues, and selecting a maximum one of said differences as said decision variable;

comparing said decision variable to a threshold value of said decision variable to obtain comparison results; and

determining a probability of ischemia based on said comparison results.

4. A method of detecting myocardial ischemia from a time-domain ECG signal X(n);

n=0, 1, . . . , N, where N is an integer, comprising the steps of;

receiving said ECG signal;

generating a decision variable from said ECG signal by generating an autocorrelation sequence from said ECG signal, calculating the differences between successive values of said autocorrelation sequence, and selecting a maximum one of said differences as said decision variable;

comparing said decision variable to a threshold value of said decision variable to obtain comparison results; and

determining a probability of ischemia based on said comparison results.

5. A method of detecting myocardial ischemia from a time-domain ECG signal X(n);

n=0, 1, . . . , N, where N is an integer, comprising the steps of;

receiving said ECG signal;

generating a decision variable from said ECG signal by generating an autocorrelation sequence from said ECG signal, computing reflection coefficients from said autocorrelation sequence, and calculating as said decision variable the sum of the first L squares of said reflection coefficients, where L is an integer;

comparing said decision variable to a threshold value of said decision variable to obtain comparison results; and

determining a probability of ischemia based on said comparison results.

6. A method as defined in claim 5, wherein said autocorrelation sequence has a plurality of values, and wherein said reflection coefficients are calculated from input data by the Levinson-Durbin algorithm assuming an Mth-order autoregressive (AR) model and using as said input data the first (M+1) values of said autocorrelation sequence, where M is an integer between 6 and 10.

7. A method of detecting myocardial ischemia from a time-domain ECG signal X(n);

n=0, 1, . . . , N, where N is an integer, comprising the steps of;

receiving said ECG signal;

generating a decision variable from said ECG signal by generating an autocorrelation sequence from said ECG signal, said autocorrelation sequence having a plurality of values, and computing as said decision variable the variance of linear prediction errors computed from input data by the Levinson-Durbin algorithm assuming an Mth order autoregressive model, where M is an integer, using as said input data the first (M+1) values of said autocorrelation sequence;

comparing said decision variable to a threshold value of said decision variable to obtain comparison results; and

determining a probability of ischemia based on said comparison results.

8. A method of detecting myocardial ischemia from a time-domain ECG signal X(n);

n=0, 1, . . . , N, where N is an integer, comprising the steps of;

receiving said ECG signal;

generating a decision variable from said ECG signal by generating an autocorrelation sequence from said ECG signal, and computing the entropy of said ECG signal from said autocorrelation sequence as said decision variable;

comparing said decision variable to a threshold value of said decision variable to obtain comparison results; and

determining a probability of ischemia based on said comparison results.

9. A method as defined in claim 8, wherein said step of computing the entropy of said ECG signal comprises the steps of computing error variances from (M+1) successive values of said autocorrelation sequence in accordance with a Levinson-Durbin algorithm, where M is an integer, and computing as said entropy the product of L error variances, where L is an integer between 1 and M.