| Summary: | 碩士 === 國立陽明大學 === 生物醫學影像暨放射科學系 === 100 === Abstract
Purpose:
In the traditional technique of brain CT perfusion, the head couch causes beam hardening and back scattering artifact and leads to the inhomogeneity on images. This phenomenon causes difficulty on separate the white and gray matter on these images. This study uses a segmentation process to improve the correction rate on separation white and gray matter.
Material and methods:
We remove skull bone and applied anisotropic diffusion filter on the brain CT perfusion image series. The mean images were first calculated. After the calculation CBV, CBF, MTT images, pixels with CBV value higher than 6 ml/100g were removed as vessels. The CSF threshold was choose manually, remove those pixel with signal intensity lower than the threshold. In the next step, we use both local and Global Otsu’s method to separate gray and white matter. In the final step, we draw ROI on the hemodynamic images to measure CBV, CBF and MTT..
Result:
The local Otsu’s method performs better than global method on separated hemodynamic parameters on all four regions (normal white, lesion white, normal gray and lesion gray). The hemodynamic parameter acquired from local Otsu’ method shows more accurate than global Otsu’s method. In addition, we observed that in the stenosis area, CBF in gray matter decrease but CBV increase, and it is more significant than in white matter.
Discussion:
The Local Otsu method in this study has a better performance than both Global Otsu and traditional separation, and the hemodynamic parameters acquired from it can provide a new standard for clinical diagnosis in prevention brain stroke.
|
|---|
