الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 131 |  |  |
| | | from | 131 | | |
Abstract
Two-dimensional (2-D) echocardiographic images of the left ventricular are commonly used to diagnose cardiac mechanical function. Quantification procedures using this imaging technique entail inaccuracies, mainly due to relatively poor image quality and the absence of a definition of the relative position of the probe and the heart. The poor quality dictates subjective determination of the myocardial edges, while the absence of a position vector increases the errors in the calculations of the left ventricular volume, and ejection fraction. In this thesis, a procedure is proposed for boundary detection of the 2D- echocardiographic images. The procedure includes speckle noise reduction and contour detection. Thirty 2-D echocardiograms were obtained from Mansoura University Hospital and Mahallah Cardiology Center for 14 subjects and 3 different views. The images were digitized producing images of 256x256 pixels. The performance of four speckle noise reduction techniques is investigated. These are: Median Filter, Wavelet-Based Speckle Reduction approach, Multiscale Nonlinear Thresolding algorithm with adaptive weighted median filter and a Multiscale Nonlinear Thresolding algorithm without adaptive weighted median filter. The results have shown that Multiscale Nonlinear Thresolding algorithm with adaptive weighted median filter gives the best performance. As for edge detection, difficulties arise from the fact that the observed time trajectories of cardiac borders sometimes present discontinuities and may not always correspond to well-defined edges. Two edge detection algorithms have been investigated. The first includes a Median filtering followed by a Laplacian operator and then a Sobel Operator. The second technique is based on gray-scale morphological operations for the localization of the echocardiographic borders. The results obtained show that the edge detection based on morphological operations gives more definite borders and a continuous contour. Some important measurements including global left ventricular systolic function such as, the left ventricular volume, Ejection Fraction and Fractional Shortening were calculated from the processed images. The values obtained for those parameters are found to be in good agreement with those assessed by the trained cardiologist. Further investigation is required to extend the capability of the present methodology toward fully automation.