Approximating Major Cerebrospinal Fluid Space in a Distance Transformation Based Bayesian Framework from Clinical Non-enhanced Computed Tomography Images
Automatically detecting the abnormality withincerebrospinal fluid space (CSF) from clinical non-enhancedcomputed tomography (NCT) images is significant since it canhelp diagnosis of many neurological diseases such ashydrocephalus and subarachnoid hemorrhage (SAH). However,extracting CSF space from NCT images is not easy, due to suchfactors as small size of CSF, partial volume effect due to largeslice spacing, varied grayscale of CSF especially whenhemorrhage appears in CSF space. In this paper a method isproposed to approximate major CSF space for detectinghemorrhage. The tissues with good contrast in the brain areextracted as anatomical landmarks, followed by extraction offeatures using distance transformation with respect to thelandmarks. By combining kernel density estimation (KDE) andmutual information (MI), discriminative features are selected forBayesian decision based classification. Experiments show that theproposed method can locate the major CSF space.
Liang Zhang Qingmao Hu Yonghong Li
Institute of Computer and Applications, Chinese Academy of Sciences Chengdu, China Shenzhen Institut Shenzhen Institute of Advanced Integration Technology Chinese Academy of Sciences/The Chinese Univer
国际会议
成都
英文
1-4
2010-06-18（万方平台首次上网日期，不代表论文的发表时间）