Detection of Left Ventricles from Echocardiographic Images
| Autor: | Yi-Yen Chen, 陳逸嬿 |
|---|---|
| Rok vydání: | 2008 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 96 Echocardiography is a noninvasive and popular examination for heart diseases. The capacity of the left ventricle (LV) is a useful estimate of heart functions. Due to speckle noise and low contrast, the detection of left ventricular contours is generally tough. Although many methods have been proposed the results are still not satisfiable. In this thesis, we aim to detect the contours of the left ventricle from a sequence of echocardiographic images automatically. Our method consists of four parts: region of interest (ROI) detection, division of echocardiographic images sequence, LV detection, and temporal adjustment. Echocardiographic images not only have speckle noise but the tissue or other chambers such as right atrium may influence the detection of the left ventricle. Thus in the first part, watershed segmentation was employed to detect the possible region of the left ventricle. In the second part, we use electrocardiograms to divide the images in a full cycle of cardiac motion into both systolic and diastolic phases for improving the accuracy of LV detection. In the third part, an application of active appearance models (AAMs) for automatic border detection in echocardiographic images sequences is reported. In a sequence of echocardiographic images, the changes between consecutive frames are small. We apply temporal smoothness for position adjustment. In the experiments, we applied our method to the image sequences from 47 patients, which are split randomly into a training set (= 31) and a test set (= 16). The experimental results show that our method can detect LV of echocardiographic images effectively. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|