AUTOMATIC DETECTION

AUTOMATIC DETECTION OF FEATURES IN ULTRASOUND IMAGES OF EYE

ABSTRACT

In closed angled Glaucoma, fluid pressure in the eye increases because of inadequate fluid flow between the iris and the cornea. One important technique to assess patients at risk of glaucoma is to analyze ultrasound images of the eye to detect abnormal structural changes. Currently theseimages are analyzed manually .This thesis presents an algorithm to automatically identify and measure clinically important features in ultrasound images of the eye. The main challenge is stable detection of features in the presence of ultrasound speckle noise; an algorithm is developed to address this using multiscale analysis and template matching. Tests were performed by comparison of results with eighty images of glaucoma patients and normals against the feature locations identified by a trained technologist. In 5% of cases, the algorithm could not analyze the images; in the remaining cases, features were correctly identified (within 97.5 µm) in 97% of images. This work shows promise as a technique to improve the efficiency of clinical interpretation of ultrasound images of the eye.

INTRODUCTION

Glaucoma is one of the leading causes of blindness. In closed angled Glaucoma, fluid pressure in the eye increases because of inadequate fluid flow between the iris and the cornea. To illustrate the structural changes due to this condition, cross sectional ultrasound images are shown in Figure 1. Figure 1 (a) and (b) show ultrasound images of a healthy and a diseased eye, respectively. Figure 1 (b) shows an eye with a closed-angle as a result of the fluid pressure in the eye that causes damage and eventually death of nerve fibers responsible for vision . One important technique to assess patients at risk of glaucoma is to analyze ultrasound images of the eye to detect the structural changes that reduce the flow of fluids out of the eye. Usually, sequences of ultrasound images of the eye are analyzed manually; a trained technologist determines anatomical feature locations and measures the relevant clinical parameters. We are unaware of any work to develop an automated algorithm to analyze these images. The main features within the eye of clinical interest are: the sclera, a dense, fibrous opaque white outer coat enclosing the eyeball except the part covered by the cornea; the scleral spur, a small triangular region in a meridional section of the sclera tissue with its base along the inner surface of the sclera; the anterior chamber, the region bounded by the posterior surface of the cornea and the central part of the lens; and, the trabecular-iris recess, the apex point between the sclera region and the iris. Manual analysis of eye images is fairly time consuming, and the accuracy of parameter measurements varies between experts. To address these issues, the goal of this thesis is to develop an algorithm to automatically analyze eye ultrasound images and locate all the features of interest within the image. The difficulties in measuring these parameters are associated with noise, poor contrast,poor resolution, and weak edge (boundary) delineation inherently present in ultrasound images. We anticipate that this scheme will reduce the processing time currentlytaken by the technologist to analyze patient images and extract the clinical parameters of interest.

1.1 THESIS CONTRIBUTION

This thesis describes a new method to detect features in ultrasound images, which shows good performance in detection of difficult features. The developed technique makes use of major image processing methods and fundamentals. In order to calculate the clinical parameters of interest, new region classification and segmentation techniques are developed as well as some signal processing to locate the scleral spur. The ultrasound images of the eye are very noisy, with poor resolution and weak edge delineation, which required the development of a three step method to overcome these challenges.

1.2 THESIS OUTLINE

The thesis is organized as follows: Chapter 2 and Chapter 3 present an introduction to Ultrasound imaging and Ultrasound biomicroscopy (UBM) respectively. Chapter 4 describes the Glaucoma disease and presents the major features of interest in the ultrasound image of the eye that are used by the algorithm to compute the trabecular-iris angle. Overviews of the image processing techniques used for feature identification are introduced in Chapter 5. Chapter 6 describes the automated algorithm for feature detection and extraction, including speckle reduction methods, non-linear contrast and edge enhancement, template correlation, regionsegmentation and classification, and computation of clinical parameters. Chapter 7 presents experimental results obtained from testing the algorithm. Finally, Chapter 8 presents a discussion of previous work done in this field and concludes this work.