開發頸動脈斑塊與腦內微出血之電腦輔助偵測系統

The Development of CAD system for Cerebral Microbleeds and Carotid Atherosclerosis

楊士奇(Shih-Chi Yang)

指導教授 ： 蘇振隆

中原大學/工學院/生物醫學工程研究所/碩士(2014年)

https://doi.org/10.6840/cycu201400908

摘要

摘　要
腦內微出血對於腦中風的發生及急性腦中風的治療方式都有很密切的關係，然而目前醫師只能透過腦部磁振造影(MRI)之影像進行腦內微出血診斷，對於需緊急治療的患者來說是無法進行的，也因此提高治療併發症發生的機率。本研究目的為開發頸動脈斑塊輔助診斷系統，並從超音波影像中得到腦內微出血的資訊。
本研究利用頸動脈超音波影像以區域成長法對頸動脈進行初步的分割後，利用直方圖等化將影像特徵強化，幫助醫生進行斑塊區域(ROI)的圈選。接著將ROI區域進行灰階分佈的分析，統計有微出血與無微出血之間成份參數的差異性，並挑選出有差異的參數對兩者的影像進行分類。系統之測試以已進行過MRI及超音波檢查之患者的超音波影像為之。共使用20個病患，其中10位經確診具微出血現象，10位為正常。每位患者之影像分為可量測中、內膜厚度(IMT)之部位與斑塊影像，分別為29張與69張。最後，利用支持向量機訓練出分類模型後，將系統分類的結果與醫生透過MRI影像所診斷的結果做比較，驗證系統分類的效能。
研究的結果顯示，在斑塊影像中，血脂比例與鈣化比例對於腦內微出血的症狀具有顯著差異(p<0.05)，使用血脂比例進行分類的效果Kappa值為0.577，使用鈣化比例的效果Kappa值為0.424，而將血脂與鈣化比例共同進行分類之效果為Kappa值0.614，達到中度符合的程度。在IMT影像中，狹窄比例與纖維比例擁有顯著差異，將兩種參數進行分類的結果Kappa值為0.586，達到中度符合的程度。
綜合結果而言，本研究能透過超音波影像輔助醫師判斷腦內微出血的機率，系統分類的結果與醫生透過腦部MRI影像診斷的結果相比具有中度的符合性，因此本系統能夠提供醫生從超音波影像中得到有關腦內微出血的資訊且系統的分類也能作為醫生診斷的參考。

關鍵字

頸動脈斑塊 ； 腦內微出血 ； 電腦輔助偵測系統

並列摘要

Abstract
The purpose of this study is to develop a computer aided system for Carotid plaque which can obtain the cerebral microbleeds information from ultrasound image. A cerebral micro-bleeds (CMBs) has a tight connection between both the stroke and the treatment of Acute Stroke. However, doctors could only currently diagnose the CMBs through the Magnetic resonance imaging (MRI) Scanning, which is impossible for the urgent patients, and thus raise the probability of complication caused by those treatments.
The Carotid Ultrasound(US) image is easy to obtain in non-invasive way, but it is noisy. Histogram equalization method used to enhance the image characteristics after taking the initial segmentation of carotid with the region growing, which offer doctors great conveniences to choose the region of interest (ROI). After the selection, the grayscale distribution of ROI was analyzed, and then statistic of the differences of spectral parameters between CMBs and non-CMBs was calculated. Those different parameters were selected for the image classifications. Totally 20 patients’ US images were used to test this system and are divided into two parts, one stand for the normal group while the other ten are CMBs ones. In each patient’ image was classified as Intimal-Medial Thickness (IMT) region and plaque region which are 29 and 69 regions, respectively. When classification model was trained by Support vector machine (SVM), the system would classified the image and the result would verify by compared with the standard which was diagnostic by doctor through the MRI image.
The result show that, in those plaque region case, two parameters (lipids proportion and calcification proportion) are most significant differences (p<0.05) in CMBs symptom. Kappa values for system with classified by lipids proportion only and calcification proportion only are 0.557 and 0.424, respectively. Yet kappa value will reach the normal consistent as 0.614 when two parameters are both used for classification. In IMT images case, another two parameters (narrow ratio and fiber proportion) have significant differences. Kappa value is 0.586 when both of the parameters were conducted to the classification, which has reached the normal consistent.
Synthesize the results, our system can assist doctors to diagnostics the possibility of CMBs by Ultrasound images and the automated classification effect are normal consistent compare to MRI images. It confirms that our system can provide the information for cerebral microbleeds and the classification result can become a reference to diagnostic by Physician.

並列關鍵字

Computer-Aided Detection System ； Cerebral Microbleeds(CMBs) ； Carotid Plaque

參考文獻

[11] 高醫醫訊月刊, 2010, 三十卷第四期。

[29] 張嘉麟,以影像處理技術評估腦內微出血於電腦斷層影像之研究，中原大學醫學工程系碩士論文，中壢，民國102年七月。

[3] Werring DJ, Frazer DW, Coward LJ, Losseff NA, Watt H, Cipolotti L, et al., “Cognitive dysfunction in patients with cerebral microbleeds on T2*-weighted gradient-echo MRI” Brain,2004; 127:2265-2275.

[4] Leblanc R, Haddad G, Robitaille Y. “Cerebral hemorrhage from amyloid angiopathy and coronary thrombolysis” Neurosurgery, 1992; 31:586-590.

[5] Wong KS, Chan YL, Liu JY, Gao S, Lam WW. “Asymptomatic microbleeds as a risk factor for aspirin-associated intracerebral hemorrhages” Neurology, 2003; 60:511-513.