M Tristam's research while affiliated with Institute of Cancer Research and other places

A basic assumption of `ultrasonic tissue characterization' is that quantitation of the interactions between ultrasound and tissue will provide information which is more tissue specific than that obtained by current visual observation of pulse-echo images. Many approaches to this goal are being assessed; for example the measurement of propagation properties such as ultrasound speed, attenuation coefficient or nonlinearity parameter. In their laboratory the authors have tended to concentrate on quantifying tissue characteristics which are associated with morphology rather than chemical composition, as assessed by the ultrasound scattering-related properties of tissues. Some of the methods which they have developed, particularly those which they feel have potential for clinical application are described

An analysis is made of the kinetics of human liver parenchyma in response to mechanical impulses arising in the heart and aorta, and the results are applied to predicting the time course of the correlation between two time-separated A-scans derived from various regions of the liver. Such predictions are found to correspond well with data derived clinically, both from volunteers and from patients with liver metastases, using a commercial, real-time sector scanner. On the basis of Fourier spectral features of the clinically derived correlation patterns, a clear quantitative separation was demonstrated between the kinetic response of three classes of tissue: normal liver in volunteers, metastatic deposits in liver of cancer patients, and histologically normal liver regions in the same patients.

A method is described for quantifying tissue movement in vivo from the computation of correlation coefficient between pairs of A-scans with appropriate time separation. The method yields quantifiable and repeatable secondary patterns of soft tissue movement in response to primary cardiac movement in a given subject, shows consistently different results as between normal livers and a variety of abdominal tumours, and is sensitive to either progress or therapeutically-induced regression of malignant disease. While the results reported here have been obtained using somewhat simple and crude equipment, the method is well suited to implementation on a commercial real-time scanner.

As part of a general programme of research which aims to develop methods of ultrasonic B-scan texture analysis for remote tissue classification, we are studying the influence of a number of system and processing variables. Some of this work (dealing with the effects of system gain, TGC rate and beam profile) was reported at a previous meeting in this series (Nassiri et al., 1982). In this paper we concentrate on the important question: “with what spatial resolution might it be possible to achieve a useful classification of tissue structures”. To begin to answer this question we are at present pursuing a study of the interrelationships between the size and shape of a region of interest (the region to be analyzed), the ultrasonic beam and pulse characteristics, and the ability of the texture features to discriminate between images of different object structures. This is being accomplished by the use of theoretical models (computer simulations), physical models (tissue phantoms) and some real (fixed) tissues. Images of each object, either simulated or real, are generated for a number of different pulse and beam widths, and wavelengths. Data will be presented on information content of texture features as function of both size of the region of interest and number of scans analysed.

In a living organism, soft tissues remain in constant motion which, when voluntary movements are controlled, results from mechanical action of the cardiovascular system. In many diseases, elastic properties of tissues are pathologically altered; this will affect tissue response to a mechanical stimulus and, hence, its movement. Thus, analysis of tissue movement will play a potentially useful role in clinical tissue differentiation.