English  |  正體中文  |  简体中文  |  Items with full text/Total items : 26988/38789
Visitors : 2356360      Online Users : 32
RC Version 4.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Adv. Search
LoginUploadHelpAboutAdminister

Please use this identifier to cite or link to this item: http://ntour.ntou.edu.tw:8080/ir/handle/987654321/35694

Title: 應用卡爾曼濾波法於3D MR影像腦組織分類之研究
Tissue Classfication of 3D Brain MR Images by Kalman Filtering Approach
Authors: Shang-Yi Lin
林尚毅
Contributors: NTOU:Department of Communications Navigation and Control Engineering
國立臺灣海洋大學:通訊與導航工程學系
Keywords: MR影像;形態學影像處理;獨立成份分析;K-means;最大可能性法;卡爾曼濾波法
MR image;morphological image processing;independent component analysis;K-means;maximum likelihood estimation;Kalman filtering
Date: 2012
Issue Date: 2013-10-07T02:57:57Z
Abstract: 本論文針對磁共振(Magnetic Resonance;MR)影像以多頻譜影像處理技術開發MR切片影像與3D影像腦組織分類法,以提高腦內部灰、白質組織的正確辨識率之外,並能有效分類出其他文獻無法準確分類的腦脊髓液組織。首先對MR切片影像提出二階段腦組織分類法,先透過形態學影像處理去除外腦殼部份,接著以第一階段之獨立成份分析法(ICA)萃取組織特徵,再透過群聚法K-means分類得到腦內部灰、白質與腦脊髓液等組織的初步分類結果。此階段之分類結果顯示腦內部灰質與膠質組織易被誤判。第二階段則透過各組織模型的建構,並依最大可能性之決策函數將第一階段類灰質的結果進一步分為灰質與膠質兩類。接著以卡爾曼濾波法對3D MR影像進行腦內部組織分類,此法以MR切片影像之分類結果作初始狀態,透過卡爾曼濾波法遞迴地預測與追蹤每一張切片的腦組織狀態。模擬結果證明本論文之3D腦組織分類法能有效得到各類組織的分類結果。以白質為例,在無雜訊時分類準確率可達97%。即使在有雜訊干擾的情況下,透過去雜訊之前處理,此分類法之TI值也能比未去雜訊前提升約6%。
In order to enhance the classification accuracy of brain tissues such as gray matter, white matter and cerebrospinal fluid, the thesis propose an efficient classification method for Magnetic Resonance (MR) slices and 3D images by using multispectral image processing and Kalman filtering approach. First, a two-stage classification of brain tissue in MR slice is developed. Before the two-stage classification, skull of the brain MR image is removed by using morphological image processing. In the first stage classification which is non-supervised, Independent Component Analysis (ICA) method is applied to enhance the image features, and then the brain tissues are categorized through K-means clustering method. The classification results in the first stage show that the gray matter and the glia matter are easily to be misjudged. In the second stage, according the Likelihood Ratio Test (LRT) of Maximum Likelihood (ML) estimation, we propose a supervised classification to further recognized gray matter from the classification results in the first stage by modeling the probability distribution of each tissue. Concerning the information correlations between MR slices, the author then applies Kalman filter to estimate the brain tissues within 3D MR images. The classification results in MR slice image are used as the initial state in Klaman filtering algorithm. Simulation results demonstrate that Kalman filter can track and update the tissue states in every MR slice images effectively. For example, the classification accuracy of White matter is above 97% in noise free images. Even in noisy MR image, the proposed method can increase around 6% TI values of accuracy by the image de-noising preprocess compared with the results without de-noising.
URI: http://ethesys.lib.ntou.edu.tw/cdrfb3/record/#G0019967009
http://ntour.ntou.edu.tw/handle/987654321/35694
Appears in Collections:[通訊與導航工程學系] 博碩士論文

Files in This Item:

File Description SizeFormat
index.html0KbHTML101View/Open


All items in NTOUR are protected by copyright, with all rights reserved.

 


著作權政策宣告: 本網站之內容為國立臺灣海洋大學所收錄之機構典藏,無償提供學術研究與公眾教育等公益性使用,請合理使用本網站之內容,以尊重著作權人之權益。
網站維護: 海大圖資處 圖書系統組
DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback