| Summary: | 碩士 === 國立陽明大學 === 生物醫學影像暨放射科學系暨研究所 === 97 === Purpose： In MR brain perfusion studies, we often use the parametric images to confirm the location of lesion. Due to the abnormal hyperintensity of cerebral spinal fluid (CSF) in mean transit time (MTT) and time to peak (TTP) maps, there are ambiguities in measuring the lesion size. The purpose of this study is removing the CSF pixels on perfusion parametric images by using local adaptive thresholding techniques to improve the clinical diagnosis.
Material and methods： In dynamic susceptibility contrast MR imaging, the longitudinal magnetization is not saturated by radio-frequency pulse on the first image, so the signal intensity of CSF is higher than that on the later images, which had been saturated. We can classify the CSF and brain parenchyma by using the signal intensity difference between the two images. Since the field inhomogeneity is unfavorable for image segmentation, we have to start with correcting the inhomogeneity, then using the global and local threshold methods to find the CSF pixels.
Results： By using image division to eliminate the inhomogeneous field, the CSF pixels can be effectively removed. CSF at the ventricle could be found accurately in all the thresholding methods. The local adaptive thresholding techniques is more suitable to segment the cortical CSF.
Conclusions： We developed the local adaptive thresholding with image division to determine the CSF pixels. It is a proper method to identify the location and area of the lesion in clinical diagnosis.
|
|---|
