126 IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 18, NO. 2, FEBRUARY 1999 Detection of Lines and Boundaries in Speckle Images—Application to Medical Ultrasound Richard N. Czerwinski, Member, IEEE, Douglas L. Jones, Senior Member, IEEE, and William D. O’Brien, Jr.,* Fellow, IEEE Abstract— This paper describes an approach to boundary detection in ultrasound speckle based on an image enhancement technique. The enhancement algorithm works by filtering the image with “sticks,” short line segments which are varied in orientation to achieve the maximum projected value at each point. The statistical properties of this approach have been described in an earlier paper; in this work we present three significant extensions to improve the performance of the basic method. First, we investigate the effect of varying the size and shape of the sticks. We show that these variations affect the performance of the algorithm in very fundamental ways, for example by making it more or less sensitive to thinner or more tightly curving boundaries. Second, we present a means of improving the performance of this technique by estimating the distribution function of the orientation of the line passing through each point. Finally, we show that images can be “stained” for easier visual interpretation by applying to each pixel a false color whose hue is related to the orientation of the most prominent line segment at that point. Examples are given to illustrate the performance of the different settings on a single image. Index Terms— Boundary detection, image enhancement, speckle, ultrasound. I. INTRODUCTION M EDICAL ultrasound is a pulse-echo imaging modality capable of quickly producing high-resolution images of soft tissue structures. Because commercial ultrasound systems are used predominantly in real-time diagnostic situations, the images they produce are generally optimized for visual interpretation of qualitative information about the tissue being scanned. Often, however, quantitative information about a scan, such as the size of a macroscopic tissue structure, is also of significant interest, as is the case in fetal maturity estimation. In these cases, the speckle texture which provides diagnostic information to the clinician may corrupt machine estimates of the positions of the boundaries. Manuscript received March 12, 1998; revised October 6, 1998. This work was performed at the Department of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign, and was supported in part by the National Live Stock and Meat Board, by the United States Department of Agriculture, and by the National Cancer Institute, DHHS under Grant PHS Grant 5 T32 CA 09067. The Associate Editor responsible for coordinating the review of this paper and recommending its publication was C. R. Meyer. Asterisk indicates corresponding author. R. N. Czerwinski is with the Lincoln Laboratory, Massachusetts Institute of Technology, Lexington, MA USA. D. L. Jones is with the Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801 USA. *W. D. O’Brien, Jr. is with the Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801 USA (e-mail: wdo@uiuc.edu). Publisher Item Identifier S 0278-0062(99)02776-7. The boundaries of interest in an ultrasound scan correspond to discontinuities between tissue layers, which are large on the scales of both the wavelength of interrogation and the scan line spacing. In the two-dimensional (2-D) scan plane, these three- dimensional (3-D) surfaces take on the appearance of bright streaks against a darker, less densely reflecting background. Features with this appearance are unlikely to occur randomly in speckle noise; speckle’s correlation structure is more likely to give rise to bright spots of characteristic size [1]. Con- ventional edge detection procedures, e.g., Canny, Roberts, or Sobel operators [2], [3] and related techniques such as [4], are ill-suited to detect the boundaries because they are not well modeled as step discontinuities in image intensity. In contrast, we have had success with an approach designed to respond preferentially to line processes [5], [6]. This approach results in an operator that is sensitive even to thin edges, while still providing for speckle reduction. To formalize this idea, we have approached the problem of boundary detection with the techniques of statistical decision theory. This has led to a number of detection rules motivated by a statistical model for the targets and noise. In [7], we derived optimal boundary detection techniques, and tested them in simulated speckle to establish performance bounds for other detectors. We compared several suboptimal detectors of varying complexity and power with the bounds, and showed that a simple suboptimal detector based on the generalized likelihood ratio test (GLRT) is extremely robust in the face of an uncertain or inexactly modeled statistical environment. Furthermore, we were able to quantify the performance lost in using this detector, and identify circumstances in which that loss was negligible [7]. The present work studies that detector in more detail, focusing in part on a number of different parameters which can be changed to alter its properties, such as the length and thickness of the templates used to model the boundaries. This paper also addresses a weakness of the technique in [7], the assumption that all orientation sticks are equally likely at each point. In practice this is not the case, since B-mode ultrasound is most sensitive to the surfaces of structures normal to the beam. To improve upon the performance of the basic technique, the image itself can be used to estimate a distribution on the angle of the lines at each point in the image. This prior information can help to reject unlikely hypotheses. Finally, we discuss the use of false color as a visualization tool to indicate the direction of the most prominent linear image feature at each point. The color can be applied to either the original or a processed image, and represents a 0278–0062/99$10.00  1999 IEEE