Real-time nonlinear adaptive filtration of images obtained in a TI628A ultrasonic echo-tomoscope

Abstract

Virtually all modem ultrasonic imaging devices are based on the principle of B-scanning [1]. To obtain high lateral resolution of B-scanning, modem medical diagnostic echo-pulse scanners are equipped with coherent sources of ultrasonic radiation. However, this is accompanied by speckle noise, which spoils the high resolution advantages of coherent radiation [10]. This is explained by the fact that the signal reflected from a continuous medium contains both a coherent component proportional to the local coefficient of reverse scattering and a component determined by a quasi-random distribution of reflectors within the measuring volume formed by the probing pulse. The speckle structure of the image (gain size) appears when individual scattering structures are not resolved, and resulting interference of scattered waves (with random phase) causes fluctuations of resulting high-contrast image intensity. In ultrasonic tomography, such high-contrast noise masks useful structural information of images. This is particularly true in case of small-sized objects.There is now a large number of approaches to the elimination of speckle noise which are based on sophisticated methods of scanning (multi-aspect or multi-frequency) [1, 2, 10]. The physical basis of these methods of processing the echosignals is to obtain several uncorrelated copies of the image of interest and their further averaging intended to form the sought image. The specific digital algorithm is either arithmetic or geometric averaging, an algorithm of minimization, etc., is selected taking into account the efficiency of reduction of the coherence of imaging [6]. However, practical implementation of these algorithms is often difficult because it requires sophisticated scanning and processing equipment [5, 6, 8, 9]. This is particularly true in the case of real-time imaging (RTI) of moving objects, because ff the time of each measurement is large, the resulting image is blurred, lnterframe filtration (averaging by two or three sequential frames) is a compromise solution, because speckle signals are poorly correlated with changes in the measuring volume.The problem of optimal selection between high spatial resolution of diagnostic system and statistical accuracy of attribution of the tissue tested to a given histological class is very important because of statistical uncertainty introduced by the speckle noise artifacts. The importance of this issue has been discussed in the literature [1, 2, 13]. The human visual system responds mainly to the statistical characteristics of an image of the ftrst and second order [2], whereas the use of these orders of the reflected signal statistics allows the relative contributions of both interference and noninterference factors of the object of interest to be determined [13].The approach to elimination of speckle noise of echo-images described in [2, 3] appears very promising. This approach is based on the physical interpretation of the causes of true and false signals. This approach is close to the method of adaptive filtrat...