Diagnostically significant signal features can be identified automatically by computational analysis.
A noninvasive, sensitive method of diagnosing certain pathological conditions of the human heart involves computational processing of digitized electrocardiographic (ECG) signals acquired from a patient at all 12 conventional ECG electrode positions. In the processing, attention is focused on low-amplitude, high-frequency components of those portions of the ECG signals known in the art as QRS complexes. The unique contribution of this method lies in the utilization of signal features and combinations of signal features from various combinations of electrode positions, not reported previously, that have been found to be helpful in diagnosing coronary artery disease and such related pathological conditions as myocardial ischemia, myocardial infarction, and congestive heart failure.
The electronic hardware and software used to acquire the QRS complexes and perform some preliminary analyses of their high-frequency components were summarized in “Real- Time, High-Frequency QRS Electrocardiograph” (MSC- 23154), NASA Tech Briefs, Vol. 27, No. 7 (July 2003), pp. 26-28. To recapitulate, signals from standard electrocardiograph electrodes are preamplified, then digitized at a sampling rate of 1,000 Hz, then analyzed by the software that detects R waves and QRS complexes and analyzes them from several perspectives. The software includes provisions for averaging signals over multiple beats and for special-purpose nonrecursive digital filters with specific low- and high-frequency cutoffs. These filters, applied to the averaged signal, effect a band-pass operation in the frequency range from 150 to 250 Hz. The output of the band-pass filter is the desired high-frequency QRS signal. Further processing is then performed in real time to obtain the beat-to-beat root mean square (RMS) voltage amplitude of the filtered signal, certain variations of the RMS voltage, and such standard measures as the heart rate and R-R interval at any given time.
A key signal feature analyzed in the present method is the presence versus the absence of reduced-amplitude zones RAZs). In terms that must be simplified for the sake of brevity, an RAZ comprises several cycles of a high-frequency QRS signal during which the amplitude of the high-frequency oscillation in portion of the signal is abnormally low (see figure). A given signal sample exhibiting an interval of reduced amplitude may or may not be classified as an RAZ, depending on quantitative criteria regarding peaks and troughs within the reduced-amplitude portion of the high-frequency QRS signal. This analysis is performed in all 12 leads in real time.
Several RAZ criteria are used in this method, including some, heretofore used only by NASA, that involve the calculation of the kurtosis of the high-frequency QRS signal. (Kurtosis is a statistical measure of the degree to which the distribution of data in a given set is peaked or flat, relative to a normal distribution.) In this method, the various RAZ criteria are applied to the ECG signals acquired from various combinations of contiguous conventional ECG electrode positions selected from among all 12 such positions. The sets of criteria and electrode combinations were developed from an analysis of ECG readings from several young and healthy subjects and from several patients with varying degrees of coronary artery disease and/or heart failure.
Again, in terms that must be simplified for brevity, a positive or negative diagnosis is suggested, depending on whether or not one of the RAZ criteria applied to the associated electrode combination is satisfied. For example, the presence of a “kurtosis RAZ” (characterized by a kurtosis value less than about 2.6) in four or more contiguous electrode positions suggests a positive diagnosis. A computer that runs software that implements the sets of RAZ criteria and electrode combinations can be incorporated into an ECG system to construct an automated diagnostic system. Inasmuch as electrocardiography is already a standard part of procedures for diagnosis of cardiac conditions, this method has potential to increase productivity and accuracy at minimal additional cost and without exposing patients to additional risks.
This work was done by Todd T. Schlegel of Johnson Space Center and Brian Arenare (self-employed).
This invention is owned by NASA, and a patent application has been filed. Inquiries concerning nonexclusive or exclusive license for its commercial development should be addressed to the Patent Counsel, Johnson Space Center, (281) 483-0837. Refer to MSC-23449.