急性缺血性腦中風的治療方式是藉由施打血栓溶解劑(Tissue Plasminogen Activator, tPA)來緩解病情。而中風時間超過三小時之病患,評估其施打血栓溶解劑與否主要根據CT Perfusion影像中假影區(Penumbra)的面積多寡。故本研究目的在於發展一套以電腦斷層影像為基礎的評估系統,針對急性缺血性腦中風於影像上的變化進行偵測與分析,利用具統計意義之參數來計算假影區的面積,進而提供醫師一套有效之評估系統。 首先利用影像前處理技術去除雜訊及頭顱骨以外之資訊,並將所得之腦組織影像進行對稱軸偵測,以分割左右腦半球。接著系統對於二度中風之案例進行系統校正,校正後將進行兩階段分類,第一階段依據左右腦平均灰階參數,來分辨異常的腦半球,同時截取異常腦半球中的腦中風病變區域;第二階段利用訓練組所得之左右腦灰階差異參數作分類,以辨別壞死區與假影區的部位並計算其切面面積。最後提供一視覺強化之影像給醫師作為診斷的參考影像。本研究使用之案例有10組(5組左腦異常,另外5組右腦異常),一組案例有5張CT影像及驗證用的5張CT Perfusion影像共100張,CT影像分為5個案例25張影像之訓練組及5個案例25張影像之測試組。 實驗結果顯示,第一階段辨別病變位置的準確率高達1,而第一階段與CT Perfusion的面積比較誤差在9.9%以下,可看出平均灰階對於分辨急性缺血性腦中風病變區域有顯著的效果;第二階段相較於CT Perfusion所提供的面積其誤差於8.6%以下,可看出電腦斷層影像確實能提供足夠的資訊輔助醫師診斷。經由醫師與數位不熟悉系統操作之使用者測試系統之處理速度,二度中風之案例需要手動圈選故需時20秒,其餘一般案例平均為5秒,故得知本系統處理速度是符合臨床上第一線診斷的要求。 總言之,本研究利用未施打顯影劑之電腦斷層影像進行假影區之計算,同時提供一視覺增強影像作為參考影像。本系統之完成對於改善電腦斷層影像敏感度低的問題,有正面的幫助。
Treating acute ischemic stroke usually injects tPA (Tissue Plasminogen Activator) to soothe symptoms of disease. But if patients stroke are over three hours, assess injecting tPA or not according to the penumbra area in Perfusion CT images. The purpose of this study was to develop a valid assessment system for physicians. And this system was established by detecting and analyzing of image change for acute ischemic stroke CT image. The parameters of statistical significance were used to calculate the area of penumbra. First of all, image processing techniques were used to remove the noise and image outside the skull tripping, and the symmetry axis of the brain tissue image was detected and split the left and right brain. Then, our system corrected the case of secondary stroke and classified two stages. In first stage, according to the average gray-scale, the abnormal area of brain was distinguished and intercepted. In second stage, the parameters of grayscale differences about right and left brain which were from training group were classified in order to distinguish dead region and penumbra region and calculate the cross-sectional area. Finally, a visual reinforcement images could be provided to the physician as a diagnostic reference. During study, we used 10 cases (five cases left brain abnormal, another 5 cases right brain abnormal) to analysis, one of cases included 5 CT images and 5 Perfusion CT image as test, total of 100 images. And CT images were divided into 5 cases 25 images of training group and 5case 25 images of testing group. The results show that the lesion location was identified that accuracy was up to 100% in the first stage, and the error which compared with Perfusion CT was less than 9.9%. There was significant effect about the average gray applied for distinguishing the region of acute ischemic stroke. The area comparing with Perfusion CT, its error was less than 8.6% in the second stage, so computed tomography could really provide enough information to assist physician diagnosis. For system processing time tests by physicians and users who were unfamiliar with our system, it took 20 seconds for the case of secondary stroke because it needed to manually circle, but for the general cases, it took 5 seconds average. Therefore, our system performance was consistent with clinical diagnostic requirements. In summary, this study calculated the penumbra region with non-contrast CT images and provided visual-enhancing images as reference images. This system was positively helpful to improve low sensitivity of CT images.