ABSTRACTS, ULTRASONIC IMAGING AND TISSUE CHARACTERIZATION SYMPOSIUM medium. Acoustic parameters such as the scatterer size (or correlation length), the product of their number density times their impedance difference, and the integrated backscatter coefficient were estimated over small volumes and displayed in an image format. When these techniques are used on in vitro tissue samples, we believe we may be obtaining information about the amount and spatial organization of collagen and fat. We are currently studying the correlation between our parameter estimates for in vitro tissue samples and histological observations from the same regions of interest. For this study, freshly excised surgical and autopsy tissue samples are scanned in normal saline. Several adjacent tissue slices were processed by the pathologist; half of these were stained for collagen, the other half for fat. The size of structures within histological samples and their number density were estimated. Several methods for making this estimation and performing the correlation will be discussed. This work was supported in part by the West Trust and the Whitaker Foundation. 1.3 IN VIVO TISSUE CHARACTERIZATION FOR ABDOMINAL PARENCHYMAL ORGANS; COMPARISON OF ATTENUATION AND BACKSCATTER ANALYSIS WITH B- SCAN IMAGE ANALYSIS, M. Fein, I. Zuna, U. RBth, G. Layer, G. van Kaick and W.J. Lorenz, Institute of Radiology and Pathophysiology, German Cancer Research Center, Im Neuenheimer Feld 280, D-6900 Heidelberg, FRG. The advantages and limitations of rf signal analysis versus B- scan image analysis were studied for different examination modalities with in vivo imaging of the liver, spleen, right and left kidney and the psoas muscle. Forty six healthy volunteers were examined. The averaging of at least three images of one patient for each task minimizes the variance due to intra- and interindividual differences, guarantees examination conditions similar to phantom measurements and a close relation to the clinical practice. Our online ultrasonic system with a 3.5 MHZ phased array sector scanner provides the rf signal sampled with 16 MHz and B-scan lines with 2 MHz. For preprocessing, an axial and lateral diffraction correction and the detection of nonergodic structure was done. Eight different methods are used for calculating the attenuation in time and frequency domain. A high sampled envelope was reconstructed by a Hilbert transform and several backscatter characteristics, e.g., the average scatterer spacing were examined. New developed procedures that describe the scatterer contributions were tested. The result for rf signal analysis were compared with our standard calculation of texture parameters. The texture parameters are much more stable than the rf parameters except for 'central frequency' and 'bandwidth', that describe transducer characteristics. Although particularly the first order texture parameters show significant differences between the organs and no difference with one organ, there is no significant difference for, e.g., the average scatterer spacing and most of the rf parameters. The size of the region of interest imitates the in vivo calculation of attenuation to the liver and the spleen. It is remarkable that the attenuation coefficient is not correlated with the mean grey level or other B-scan parameters. The results indicate that B-scan lmage analysis led to reprodicable results in clinical examination and describes the differences of image quality in different organs better than rf signal parameters. 1.4 IN VIVO STUDY OF QUASI REAL-TIME LIVER TISSUE CLASSIFICATION BASED ON POWER SPECTRUM ANALYSIS, M. Lang,l H. Ermert' and L. Heuser=, =Institut far Hoch-und HBchstfrequenztechnik, Ruhr- Universitat Bochum, P.O. Box 10 21 40, D-4630 Bochum 1, FRG and 124