Digital two-dimensional echocardiography is an ultrasonic imaging technique that is used as an increasingly important noninvasive technique in the comprehensive characterization of the left ventricular structure and function. Quantitative analysis often uses heart wall motion and other shape attributes such as the heart wall thickness, heart chamber area, and the variation of these attributes throughout the cardiac cycle. These analyses require the complete determination of the heart wall boundaries. Poor image quality and large amount of noise makes computer detection of the boundaries difficult. An algorithm to detect both the inner and outer heart wall boundaries is presented. The algorithm was applied to images acquired from animal studies and from a tissue equivalent phantom to verify the performance. Different approaches to exploiting the temporal redundancy of the image data without making use of results from image segmentation and scene interpretation are explored. A new approach to perform image flow analysis is developed based on the Total Least Squares method. The result of this processing is an estimate of the velocities in the image plane. In an image understanding system, information acquired from related domains by other sensors are often useful to the analysis of images. Electrocardiogram signals measure the change of electrical potential changes in the heart muscle an d provide important information such as the timing data for image sequence analysis. These signals are frequently plagued by impulsive muscle noise and background drift due to patient movement. A new approach to solving these problems is presented using mathematical morphology. Experiments addressing various aspects of the problem, such as algorithm performance, choice of operator parameters, and response to sinusoidal inputs, are reported.
Date of this Version