Brian S. Garra

Analysis Of Ultrasound Image Texture Via Generalized Rician Statistics

Tissue signatures are obtained from the first- and second-order statistics of ultrasonic B-scan texture. Laboratory measurements and early clinical results show that the image may be segmented to discriminate between different normal tissues and to detect abnormal conditions based on a multidimensional feature space. These features describe the intrinsic backscatter properties of the tissues imaged and may be used as the basis of an automatic tissue characterization algorithm.

Interleukin-2 induces profound reversible cholestasis: a detailed analysis in treated cancer patients.

Interleukin-2 (IL-2)-based immunotherapy is associated with profound reversible cholestasis and hyperbilirubinemia. We performed a nonrandomized retrospective and prospective analysis to determine the incidence, characteristics, clinical course, and nature of the IL-2-induced liver dysfunction in patients with cancer. Patients received IL-2 at a dose of 20,000 to 100,000 units (U)/kg thrice daily for up to 5 days. Fifty-one patients on adjuvant treatment protocols received a mean of 10.18 +/- 2.38 IL-2 doses and 11.67 +/- 4.16 doses were delivered to 210 patients with advanced disease during this period. Retrospective analysis of all patients receiving this therapy revealed increases in the following liver function tests expressed as median, 25th percentile, and 75th percentile (range): bilirubin (mg/dL) 4.5, 2.6, 6.5 (.4 to 38.5); alkaline phosphatase (U/L) 256, 179, 378 (56-1680); SGOT (U/L) 80, 52, 117 (18 to 483); SGPT (U/L) 91, 64, 132 (20-540); prothrombin time 13.4, 12.8, 14.5 (10.8 to 35.4); and albumin (g/dL) values decreased (trough) slightly 3.0, 2.8, 3.2 (2.3 to 3.8). Multiple regression analysis revealed several factors that were significantly associated with the increase in bilirubin when jointly considered (model P2 less than or equal to .001) including total IL-2 dosage, increase in creatinine, alkaline phosphatase, weight, and SGOT. Similar increases were noted in a prospectively evaluated group of 10 patients. A return to normal levels of bilirubin was noted within 5.6 days of stopping IL-2. Fasting serum cholylglycine increased from a mean of 32.3 +/- 1.6 to a peak of 1556.0 +/- 625.0 mg/mL. Although conventional ultrasound examinations were unrevealing, tissue ultrasound examinations revealed a mean scatterer spacing (MSS) increase compared to baseline of .10 +/- .04 (P less than .02) suggesting hepatic edema or an infiltrative process. Further, computerized hepatobiliary nuclear medicine scans revealed a delay in uptake (2.2 +/- 0.5 fold greater) and excretion (8.0 +/- 5.9 fold greater) of technetium-99m labeled disofenin. These findings support the development of profound reversible cholestasis as the primary basis for the elevated bilirubin in patients undergoing IL-2 treatment and may have implications for understanding the jaundice observed in some patients postoperatively as well as that associated with sepsis and other inflammatory disorders. Specifically, the release of IL-2 or the induction of other factors similarly induced by IL-2 may be responsible for these findings. Tissue ultrasound and computerized hepatobiliary scans provide additional noninvasive assessments of liver function and physiology.

Pattern recognition methods for optimizing multivariate tissue signatures in diagnostic ultrasound

Described is a supervised parametric approach to the detection and classification of disease from acoustic data. Statistical pattern recognition techniques are implemented to design the best ultrasonic tissue signature from a set of measurements and for a given task, and to rate its performance in a way that can be compared with other diagnostic tools. In this paper, we considered combinations of four ultrasonic tissue parameters to discriminate, in vivo, between normal liver and chronic active hepatitis. The separation between normal and diseased samples was made by application of the Bayes decision rule for minimum risk which includes the prior probability for the presence of disease and the cost of misclassification. Large differences in classification performance of various tissue parameter combinations were demonstrated using the Hotelling trace criterion (HTC) and receiver operating characteristic (ROC) analysis. The ability of additional measurements to increase or decrease discriminability, even measurements from other diagnostic modalities, can be evaluated directly in this manner.

Supervised Pattern Recognition Techniques In Quantitative Diagnostic Ultrasound

The methods of statistical pattern recognition are well suited to the problems of in vivo ultrasonic tissue characterization. This paper describes supervised pattern recognition methods for selecting features for tissue classification, calculating decision boundaries within the selected feature space, and evaluating the performance. We address the considerations of dimensionality and feature size which are important in classification problems where the underlying probability distributions are not completely known. Examples are given for the detection of diffuse liver disease in the clinical environment.

Application Of Cluster Analysis And Unsupervised Learning To Multivariate Tissue Characterization

This paper describes a procedure for classifying tissue types from unlabeled acoustic measurements (data type unknown) using unsupervised cluster analysis. These techniques are being applied to unsupervised ultrasonic image segmentation and tissue characteriza-tion. The performance of a new clustering technique is measured and compared with supervised methods, such as a linear Bayes classifier. In these comparisons two objectives are sought: a) How well does the clustering method group the data? b) Do the clusters correspond to known tissue classes? The first question is investigated by a measure of cluster similarity and dispersion. The second question involves a comparison with a supervised technique using labeled data.

A statistical approach to an expert diagnostic ultrasonic system

The techniques of statistical pattern recognition are implemented to determine the best combination of tissue characterization parameters for maximizing the diagnostic accuracy of a given task. In this paper, we considered combinations of four ultrasonic tissue parameters to discriminate between normal liver and chronic hepatitis. The separation between normal and diseased samples was made by application of the Bayes test for minimum risk which minimizes the error rate for classifying tissue states while including the prior probability for the presence of disease and the cost of misclassification. Large differences in classification performance of various tissue parameter combinations were demonstrated by ROC analysis. The power of additional features to classify tissue states, even those derived from other imaging modalities, can be compared directly in this manner.

On the Information Content of Diagnostic Ultrasound

Echo signals in medical ultrasound are rich in information about tissue composition and structure. But not all of this information is readily apparent from the image. For example, the average size, density and organization of microscopic tissue structures which may be obtained from the first- and second-order statistical properties of the data are difficult to visualize directly from the image. Yet these acoustic parameters have been shown to be sensitive indicators of changes brought on by disease processes in the liver and spleen (Fellingham and Sommer 1984, Insana et al. 1987), and therefore make good classifiying features for tissue characterization. Realizing the full potential of these features requires the use of pattern recognition techniques in order to minimize classification errors due to measurement uncertainty and patient variability.