Chan, Sheila Oi Yip (2007) Fuzzy-based echocardiogram boundary enhancement. Masters thesis, Universiti Malaysia Sabah.
|
Text
mt0000000248.pdf Download (2MB) | Preview |
Abstract
In cardiology, ultrasound is rapidly becoming one of the most sought after tool to investigate cardiac performance. Echocardiogram, or cardiac ultrasound, has many advantages and plays an important role in studies concerning heart disease. Nevertheless, speckle noise, poor contrast, and artifacts in echocardiogram hamper human interpretation and impede automated analysis. The image quality degradation has rendered echocardiogram boundary enhancement necessary. However, edges in echocardiogram are identified with ambiguous spatial and intensity information that pose challenges for methods that depend on conventional gradient approximations. Although many other techniques have been proposed to improve speckle tarnished images, most of them are problem-oriented and hold certain compromise. Many techniques commonly take tentative solutions by making trade-off between noise suppression and edge detection while some others are motivated by statistics and subject to noise modeling error. Apart from that, there are also techniques available which require time consuming manual definition and are liable to observer variability. Therefore, two different approaches incorporating fuzzy-based methods are proposed to automate the boundary enhancement process and reduce the commonly encountered drawbacks aforementioned. The first approach employs a multiscale scheme to resolve the uncertainty in the information obtained across scales so as to localize the features at low scales and minimize noise at higher scales. In the proposed multiscale analysis, fuzzy reasoning provides a mean to combine the derivative information of various scales stored in a filter bank. By doing so, the proposed approach reduces the trade-off between noise suppression and edge detection that is inherent in fixed scale operators. On the other hand, the second approach attempts to take advantage of the highly intuitive and appealing aspect of edge definition in echocardiogram by manipulating the local image characteristics. Edges in noise tarnished echocardiogram manifest themselves differently from the standard definition that typically characterizes edges in an image as abrupt changes in gray-level. Nevertheless, the ambiguous edge definition allows the second fuzzy-based approach to work on different edge notions that are defined by operators based on local statistics in the image. All the proposed methods are comprised of a comprehensive series of noise suppression, fuzzy reasoning and boundary extraction operations. Some of the proposed boundary enhancement methods are also optimized by a neuro-fuzzy system with learning capability. The results of the proposed methods are compared with each other and to that of a conventional method for performance evaluation. The results are compared subjectively by visual observation for qualitative performance analysis. In addition, the results are measured using a standard performance index for quantitative comparison.
Item Type: | Thesis (Masters) |
---|---|
Keyword: | ultrasound, cardiac performance, Echocardiogram, quantitative comparison, standard performance index |
Subjects: | R Medicine > RC Internal medicine |
Department: | SCHOOL > School of Engineering and Information Technology |
Depositing User: | SITI AZIZAH BINTI IDRIS - |
Date Deposited: | 18 Nov 2013 11:01 |
Last Modified: | 12 Oct 2017 10:50 |
URI: | https://eprints.ums.edu.my/id/eprint/7506 |
Actions (login required)
![]() |
View Item |