Advanced Support Vector Machines for Image Modeling Using Gibbs-Markov Random Field.
- SourceAvailable from: ritdml.rit.edu[Show abstract] [Hide abstract]
ABSTRACT: "In this thesis, we propose a novel algorithm for detecting needles and their corresponding implanted radioactive seed locations in the prostate. The seed localization process is carried out efficiently using separable Gaussian filters in a probabilistic Gibbs random field framework. An approximation of the needle path through the prostate volume is obtained using a polynomial fit. The seeds are then detected and assigned to their corresponding needles by calculating local maxima of the voronoi region around the needle position. In our experiments, we were able to successfully localize over 85% of the implanted seeds. Furthermore, as a regular part of a brachytherapy cancer treatment, patient's prostate is scanned using a trans-rectal ultrasound probe, its boundary is manually outlined, and its volume is estimated for dosimetry purposes. In this thesis, we also propose a novel semi-automatic segmentation algorithm for prostate boundary detection that requires a reduced amount of radiologist's input, and thus speeds up the surgical procedure. Saved time can be used to re-scan the prostate during the operation and accordingly adjust the treatment plan. The proposed segmentation algorithm utilizes texture differences between ultrasound images of the prostate tissue and the surrounding tissues. It is carried out in the polar coordinate system and it uses three-dimensional data correlation to improve the smoothness and reliability of the segmentation. Test results show that the boundary segmentation obtained from the algorithm can reduce manual input by the factor of 3, without significantly affecting the accuracy of the segmentation (i.e. semi-automatically estimated prostate volume is within 90% of the original estimate)"--Abstract. Typescript. Thesis (M.S.)--Rochester Institute of Technology, 2006. Includes bibliographical references (leaves 48-51).
- [Show abstract] [Hide abstract]
ABSTRACT: Sitting posture analysis is widely applied in many daily applications in biomedical, education, and health care domains. It is interesting to monitor sitting postures in an economic and comfortable manner. Accordingly, we present a textile-based sensing system, called Smart Cushion, which analyzes the sitting posture of human being accurately and non-invasively. First, we introduce the electrical textile sensor and its electrical characteristics, such as offset, scaling, crosstalk, and rotation. Second, we present the design and implementation of the Smart Cushion system. Several effective techniques have been proposed to improve the recognition rate of sitting postures, including sensor calibration, data representation, and dynamic time warping-based classification. Last, our experimental results show that the recognition rate of our Smart Cushion system is in excess of 85.9%.IEEE Sensors Journal 10/2013; 13(10):3926-3934. DOI:10.1109/JSEN.2013.2259589 · 1.85 Impact Factor