| Summary: | 碩士 === 國立中興大學 === 資訊管理學系所 === 100 === This paper proposed a new method to automated segment the blood vessel and the skeleton of the blood vessel in rat’s brain images. Because recognizing the tiny brain vessel by human eyes is difficult task, we propose an automatic image segmentation to assist medical scientists and reduce human judgment errors. However, the segmentation of Blood vessel is a challenge because of the interferences of reflective light from surrounding muscle tissue. Thus, the segmentation method is proposed to solve these problems which are two major experiments (BAS and BVS) in our method.
In the BAS method, we do the pre-processing firstly. In the pre-processing, G channel extracted form RGB color mode is transformed to Gray level image and the area of brain can be enhanced after combining some image texture features of Gray level image. Then, next step is automatic segmentation of brain area of rats. Region labeling is used for each pixel and these pixels with largest label value will be judged as brain pixel. Moreover, the pixels surrounding by brain pixels will be judged as brain pixel after performing closing of morphology method. Hence the area of the brain can be obtained.
The BVS method contains the segmentation of blood vessel of rats and Skeletonizing method. We use Gamma equalization and Otsu’s thresholding from the brain area to derive the blood vessel contour. Then, skeletonizing method is used to segment the skeleton of the blood vessel from rat’s brain images. Finally, the experimental results of the proposed method as well as other common approach for vessel segmentation are demonstrated by calculating the Precision, Recall, and F-measure of results. On average, the Precision, Recall, and F-measure of the proposed method are better than other methods.
|
|---|
