- 學位論文
出血性腦中風之電腦輔助系統開發
Development of Computer-aided Detection System for Hemorrhagic Stroke
摘要
醫師藉由出血性腦中風於電腦斷層影像上的變化，估算出大腦中線偏移量、出血體積、水腫體積作為診斷依據，故本研究之目的為發展一套電腦輔助偵測系統，以量化此三項參數，協助醫師進行病患的治療。 首先利用單一切面，用中值濾波和區與成長法去除雜訊及頭顱骨以外之資訊。第二，藉由腦室和松果體找尋偏移中軸。第三，利用閥值和左右腦對稱概念來分割出血和水腫區域。第四，以各切面的面積推算出體積。經過以上方法後可得中線偏移量、出血體積、水腫體積。最後分別以假體與實體影像8個病例之240張切面影像進行驗證。 假體驗證方面，圖像旋轉找尋偏移軸誤差在3.3%以內；中軸偏移誤差也在5%以下；雜訊濾除效果可將17%誤差降至3.6%；分割後的面積可以完全截取；體積驗證部分，堆積後體積與假體的誤差可在5.5%以內。此外，實體影像驗證方面，以醫師所圈選的出血和水腫面積在240張切面影像中，比較系統所計算的面積和重合度比較，出血區的平均誤差為8.8±3.13%，而重合度為86.4±5.43%，水腫的區域面積偵測平均誤差在14.1±4.48%，重合度在77.4±6.58%。 綜合上述結果，本研究發展一套電腦輔助偵測系統，利用未施打顯影劑之電腦斷層影像進行影像處理，定量出血性中風相關參數，並提供視覺增強影像作為參考影像以助醫生藉由客觀的資訊進行診斷。
並列摘要
Physicians observe the changes on computed tomography of hemorrhagic stroke, and estimate the values of brain midline shift, bleeding volume, and edema volume as a diagnostic basis. Therefore, the purpose of this study is to develop a hemorrhagic stroke computer-aided detection system to quantify these three values, including values of brain midline shift, bleeding volume and edema volume. At first, use single slice, to remove the noise and skull stripping by median filter and region growing. Secondly, detect the brain tissue’s shift axis by detect the ventricle and pineal gland. Third, segment the bleeding region and edema region by thresholding and symmetrical of left and right brain. Fourth, calculate the volume through each slice, after the above method to find the brain midline shift, bleeding volume and edema volume. Finally, verify phantom and real image, 8 cases of the 240 slices images. In phantom verifying stage, the error of midline shift after rotation was less than 3.3%, the error of midline shift was less than 5%. After noise filtering, the area error can reduce to 3.6%. In volume verification, the difference between the volume which system calculate and prosthesis is less than 5.5%. In real image verifying stage, this study use 8 cases including 240 slices to compare the bleeding region and edema region with this system calculated and doctor selected, the results showed that the average error of bleeding region is 8.8±3.13%, the average of degree of coincidence is 86.4±5.43%, and the average error of edema region is 14.1±4.48%, the average of degree of coincidence is 77.4±6.58%. Then, calculate the volume with the areas of each case. Based on the results obtained, the proposed method can clearly benefit not only visual diagnostics, but also quantitative methodologies about hemorrhagic stroke-related parameters, the information will help the doctors to diagnose and then carry out treatments.