請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/22854
標題: | 電腦斷層掃描影像之自動化冠狀動脈血管分割與斑塊偵測演算法 Automatic Segmentation of Coronary Vessel and Plaque Detection Algorithm in CT Image |
作者: | Chien-Hsuan Wang 王建琁 |
指導教授: | 陳中明 |
關鍵字: | 多切片電腦斷層掃描,冠狀動脈,血管樹重建,斑塊,影像分割, multi-slice computed tomography,coronary artery,reconstruction of vascular tree,plaques,image segmentation, |
出版年 : | 2011 |
學位: | 碩士 |
摘要: | 近十年來,心血管疾病在國人十大死因上一直佔據前三名的位置。而多切片電腦斷層掃描(MSCT)是目前非侵入性的冠狀動脈疾病影像學檢查中,最為重要的影像方法之一。然而,目前臨床上,MSCT冠狀動脈血管樹(Vessel tree)與斑塊的商用分析軟體,對於斑塊的定量分析以及潛在的危險性,所提供的資訊十分有限。雖然已有相當多的研究嘗試以自動或半自動的方式分割出MSCT冠狀動脈血管樹與斑塊,或是嘗試預測斑塊的危險性,其成效仍是未完善。
本研究主要目的是發展出一套較為自動化的工具,分割血管中造成心臟疾病的斑塊,提供完善的資訊幫助醫師診斷與治療。研究主要分成三步驟:第一步驟使用本實驗室在2010年發展的三維電腦斷層影像冠狀動脈追蹤與重建演算法,取得冠狀動脈走向,第二步驟使用重建出的血管當作初始輪廓(initial contour),使用本研究改良的主動輪廓線模型(Active contour model),稱為拉普拉斯強化(Laplacian-enhanced)之多相位等位函數(3D multiphase level set method)勾勒出多切片電腦斷層掃描冠狀動脈影像之冠狀動脈之輪廓,以利進行斑塊分割與後續危險性之評估與預測。 在本研究的第三步驟會完成兩種斑塊的邊界自動化分割演算法。斑塊分割部分本研究使用Vladimir與Vadim在2005年所提出的GrowCut影像分割技術做為分割斑塊的主要架構,在影像中給予斑塊種子點(stroke),以定義斑塊位置與灰階值特色,同時也給予背景種子點,並在剩餘區域競爭得到分割結果。而鈣化斑塊的偵測先使用最大事後機率評估法(Maximum a posteriori,MAP)模擬血管的灰階值直方圖,再定義斑塊之VOI(Volume of interest);非鈣化斑塊的部份做法有二,第一是針對非鈣化區域與血管大量出現在同一張影像上之案例,在VOI中做等位函數模型及二次Fuzzy C Mean以篩選非鈣化斑塊之位置;做法二是針對其餘的案例,使用最大密度投影(Maximum intensity projection,MIP)將影像每個軸向的最大灰階值投影在三個平面上,血管邊緣的非鈣化斑塊在投影之後,便會出現在其中兩個軸向的投影上,再利用取交集的做法即可取得部分鈣化斑塊中非鈣化區域之種子點。 本研究之影像皆由台大醫院影像醫學部提供之多切面胸腔電腦斷層影像(MSCT),本實驗的演算法可以分割出鈣化與部分鈣化兩種類型的斑塊以供之後危險性之評估與預測。 Cardiovascular disease has ranked as one of the top three leading causes of death in Taiwan for the past decade. Currently, multi-slice computed tomography (MSCT) is one of the most important methods for noninvasive imaging examination of coronary artery diseases. However, the commercial software that is presently available for the analysis of MSCT coronary arteries and plaques provides very limited information for quantitative analysis of plaques and their potential risk. Although many researchers have tried to develop a measure to automatically or semi-automatically segregate the coronary arteries and plaques from the MSCT or to predict the potential risk of a plaque, their results are far from clinically useful. The main purpose of this study is to develop a more automated tool to segregate the heart disease-causing vascular plaques and to provide more comprehensive information to assist physicians in diagnoses and treatments. The investigation consists of three steps. The first step is to use the three microcomputer tomography coronary tracking and reconstruction algorithms, developed by our team last year, to obtain the coronary arteries. The second step is to utilize the vascular trees as the initial contour and identify the vessel boundaries by using an improved active contour model named Laplacian-Enhanced 3D multiphase level set method to enhance the prediction of the plaques and the risk assessment . In the third step of this study, two kinds of coronary artery plaque border automated segmentation algorithm were developed. The main method to segregate plaques is GrowCut, which is an image segmentation technology proposed by Vladimir and Vadim in 2005. The user-defined plaque stroke in the images will locate the plaques and identify the intensity feature. Detection of calcified plaques is forged by first utilizing Maximum a Posteriori to simulate the vascular histogram followed by defining the volume of interest. There are two methods for detecting non-calcified plaques. The first is to use the Level Set method and Fuzzy C Mean to find a suitable stroke for the non-calcified plaques, and this algorithm is used when a large number of plaques and blood vessels are present in an image. The second method is for all other cases. Maximum Intensity Projection is used to project three directional images, and non-calcified plaques near vascular edges will reveal on two of the images. And the Intersection is our goal. Our raw MSCT images were provided by National Taiwan University Hospital Department of Medical Imaging. We have proposed two automatic segmentation algorithms for boundary delineation of the calcified plaques and mixture plaques in the coronary arteries to facilitate the risk assessments and predictions. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/22854 |
全文授權: | 未授權 |
顯示於系所單位: | 醫學工程學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-100-1.pdf 目前未授權公開取用 | 4.3 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。