[Ultrasonic tissue characterization by spectral analysis of myocardial textural pattern]

Based on the fact that ultrasonic myocardial textural patterns are more irregular in the pathological myocardium than in the normal, evaluation of the myocardial tissue character was attempted in vivo using spectral analysis. Parasternal left ventricular long-axis echocardiograms were obtained from five patients with old myocardial infarction diagnosed by history, electrocardiography and coronary angiography. These echocardiograms were transferred to an image analyzer and digitized (256 x 256 x 8). The waveforms of the gray-scale-changes from the normal myocardium showed periodicity in each 8-pixel cycle, but those from the infarcted myocardium did not. To quantify pattern changes in gray-scale values in the ultrasound beam direction, spectral analysis was performed by the maximum entropy method (MEM). There were four peaks in the MEM spectra both in the normal and infarcted myocardia, but there was a great significance in these patterns: with high, steep peaks in normal MEM spectra, and low, blunt peaks in infarcted ones. By discriminatory analysis of these four peak values, normalized by the whole spatial frequencies as multivariate, the misclassification rate was 4.8-22.7% in anteroseptal infarctions and 5.0-20.0% in posterior infarctions. Thus, spectral analysis of the myocardial textural pattern has advantages for analyzing routine echocardiograms without corrections by any absolute ultrasonic references. Furthermore, the misclassification rate is so low that we are able to characterize myocardial tissue.