請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/55628
標題: | 中風大鼠影像之腦部擷取與腦半球分割技術研發 Brain Extraction and Hemisphere Segmentation in Rat Brain Images after Stroke |
作者: | Min-Yi Chen 陳珉頤 |
指導教授: | 張恆華(Herng-Hua Chang) |
關鍵字: | 磁振影像,缺血型中風,大腦擷取,2,3,5-氯化三苯基四氮唑,顯著區域偵測,腦半球分割,可變形模型, Magnetic resonance imaging (MRI),ischemic stroke,brain extraction (skull stripping),2,3,5-triphenyl tetrazolium chloride (TTC),salient region detection,cerebral hemisphere segmentation,deformable models, |
出版年 : | 2020 |
學位: | 碩士 |
摘要: | 腦血管疾病是國內十大死因的第四名,臨床上,大多數患者中風類型為缺血性中風。為了診斷此疾病,在臨床實驗模型中,大多使用齧齒動物之實驗影像作為研究依據。然而手動分割費時費力,因此本篇論文以大鼠作為實驗動物,使用其影像來開發一自動化演算法,以利醫生進行診斷和研究。 本研究主要以缺血性中風鼠腦為研究對象,針對大腦動脈阻塞的大鼠為實驗對象,以原始腦部磁振影像與經2,3,5-氯化三苯基四氮唑染色的原始大腦影像為研究影像。本論文的演算法分為二部分,首先對原始腦部影像進行大腦擷取,再進行大鼠腦部中線偵測將其分為左右腦半球。大腦擷取的部分是採用基於無邊緣主動輪廓模型的方法和顯著區域之偵測進行背景去除。左右腦半球分割的部分是接續大腦擷取,得到一乾淨的腦部影像後,再來利用邊緣偵測、可變形模型找出中線上的特徵點,再將產生腦部之中線作為分割線,進行左右腦半球的分割。 本篇研究使用擴散權重磁振影像、T2權重影像和 2,3,5-氯化三苯基四氮唑染色影像的腦部中風影像進行實驗。擴散權重磁振影像共有57隻老鼠,結果顯示本研究演算法可以準確分割老鼠大腦(Dice為97.08±0.68%)、右腦半球(Dice為98.36±1.32%)和左腦半球( Dice為98.21±1.53%)。T2權重影像共有65隻老鼠,結果顯示本研究演算法可以準確分割老鼠大腦(Dice為97.16±0.98%)、右腦半球(Dice為98.28±1.08%)和左腦半球( Dice為98.12±1.23%)。2,3,5-氯化三苯基四氮唑染色影像共有44隻老鼠,結果顯示本研究演算法可以準確分割老鼠大腦(Dice為92.33±1.26%)、右腦半球(Dice為96.94±0.83%)和左腦半球( Dice為97.37±0.76%)。 最後,實驗結果顯示我們優於找出鼠腦真實區域的其他方法,並且對於左右腦半球區域之評估都有不錯的成效,因此本研究提出一個全自動的方法,可以成為醫生良好的輔佐工具,進行黃金標準切割。 Cerebrovascular diseases are the fourth domestic cause of deaths. Most of the stroke patients have a ischemic stroke. In order to study this disease, usually different types of imaging modalities of rodents are used as the research basis. However, manual segmentation is time-consuming and laborious. An automated algorithm is developed in this thesis using rats as experimental animals to help doctors conduct diagnosis and investigation. This study focuses on ischemic strokes using rats with cerebral artery occlusion as experimental subjects and examines the images of their brains obtained from staining with 2,3,5-triphenyl tetrazolium chloride (TTC) and magnetic resonance imaging (MRI). The proposed algorithm is divided into two parts. Firstly, the brain is extracted from surrounding regions (skull and non-brain tissues). Secondly, the extracted brain region is used the brain midline detection to divide the brain into left and right hemispheres. The algorithm of the brain segmentation is based on the active contour model without edges in MR images and removes the image background via the salient region detection method in TTC images. To segment the left and right cerebral hemispheres, a clean brain image is processed by edge detection and active contour models (snakes) to find the feature points around the midline. Then, the brain midline is evolved to divide the brain into the left and right brain hemispheres. In this thesis experimental rat imaging with diffusion-weighted magnetic resonance (DWI) and T2-weighted MRI and TTC has been adopted. We used 57 rat brain DWI images. The results showed that the algorithm accurately segmented the rat brain (Dice = 97.08 ± 0.68%) and the right brain hemispheres (Dice = 98.36 ± 1.32%) and the left brain hemispheres (Dice = 98.21 ± 1.53%). We used 65 rat brain T2 images. The results showed that the algorithm accurately segmented the rat brain (Dice = 97.16 ± 0.98%) and the right brain hemispheres (Dice = 98.28 ± 1.08%) and the left brain hemispheres (Dice = 98.12 ± 1.23%). We used 44 rat brain TTC images. The results showed that the algorithm accurately segmented the rat brain (Dice = 92.33 ± 1.26%) and the right brain hemispheres (Dice = 96.94 ± 0.83%) and the left brain hemispheres (Dice = 97.37 ± 0.76%). Finally, the experimental results show that we are superior to other methods to find the real area of the rat brain, and have good results for the evaluation of the left and right brain hemisphere regions. Therefore, this study proposes a fully automatic method, which can become a good auxiliary tool for doctors to perform gold standard segmentation. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/55628 |
DOI: | 10.6342/NTU202002039 |
全文授權: | 有償授權 |
顯示於系所單位: | 工程科學及海洋工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-2907202015390200.pdf 目前未授權公開取用 | 7.55 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。