Abstract: Several hundred digitized ECG waveforms are averaged together (discarding abnormal waveforms) to form an array of digitized values. Then multiple overlapping subsegments are extracted from this array, processed by a window function in a way that produces a D.C. free result, and transformed into the frequency domain. The logarithm of the square of the amplitude (down to a minimum noise threshold level) of these frequency domain values are then plotted in a three-dimensional plot that graphically illustrates whether or not the patient from whom the waveforms were taken is likely to be at risk of cardiac arrhythmia in the near future. Cross correlation of these functions against one far removed from the QRS complex enables a factor of normality to be computed.

Abstract: Disclosed is a system and a method for computer-supported analysis of arrhythmic potentials in ECG signals, particularly those of late potentials. Interference discrimination advantages of frequency domain analysis are combined with temporal localization advantages of time domain analysis to determine the accurate location of arrhythmic potentials. Several small signal segments are selected in an ECG waveform. A determination is made of parameters corresponding to extended signals which closely match fluctuations of each respective small signal, allowing more information is discerned about the small signals than is possible with more conventional techniques. A comparison is made with respect to extended signals rather than small signals. Two autoregressive models are used, the maximum entropy method and adaptive filter determination. Area integrals of the frequency characteristics of small signal segments are recorded successively with respect to the frequency range of the arrhythmic potentials analyzed.