Optimal Image Segmentation and Three Dimensional Reconstruction for Functional Magnetic Resonance Imaging
| Autor: | Chung Bin Lu, 盧宗斌 |
|---|---|
| Rok vydání: | 2003 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 91 Recently, the Functional Magnetic Resonance Image (fMRI) is used to evaluation the risk before surgical operation. The MR images processed by image segmentation and three dimensional reconstruction for fMRI, the output can be used to evaluate the functional region to help doctors to evaluate the risk and the plan of the treatment. In this thesis, Pulse Coupled Neural Networks (PCNN) are combined with Expectation Maximum (EM) algorithm to segment MR images. The EM algorithm is used to estimate the distribution parameters of image information to serve as the objective functions. The PCNN is used to segment images into gray matter (GM)、white matter (WM) and cerebrospinal fluid (CSF). The experimental results are compared with Fuzzy C-Mean (FCM) and Bias Corrected Fuzzy C-Mean (BCFCM). Since the PCNN+EM and the BCFCM include spatial factor, the visual effect of their output is better than the FCM. We use Jaccard Similarity index to measure segmentation performance under variour levels of noise or ununiformity. Experimental results show that the FCM is the best and PCNN+EM out performs BCFCM. The major reason of the FCM performing better than the PCNN+EM is that the PCNN+EM will accumulate the error during its iterative procedures. In the future, we can try to embed the EM into the PCNN to provide a more robust segmentation mechanism to provide a more adaptive segmentation output. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|