AUTOMATIC GRAY MATTER AND WHITE MATTER SEGMENTATION ALGORITHM

• Main Authors: Chen Hsuan Huang, 黃晨軒
• Other Authors: S. Y. Tsai
• Format: Others
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/04722495142266157778

碩士 === 長庚大學 === 電機工程學系 === 98 === Brain tissue segmentation for gray matter (GM) and white matter (WM) is helpful on the analysis of brain structure and functional . In this study we investigate the performance of three segmentation strategies on 3D MPRAGE (Magnetization Prepared Rapid Gradient Echo) images including self-developed fuzzy classifier based method (FCM) and two popular free software, SPM (Statistical Parametric mapping) and FSL (FMRIB Software Library). The self-developed FCM can be separate into two steps. Firstly, an automatic threshold based method is used to remove the skull surrounding the brain tissue. Secondly, By the mathematic signal model of MPRAGE we can transform the signal intensity into a psudo-T1 parametric map. Then according the reported T1 relaxation times of GM and WM at 3T, we can do the GM and WM segmentation. Further to account for partial volume effect, we use a T1-based fuzzy classifier to differentiate the region containing both GM and WM. So people may be able to observe the distribution of the gray matter and white matter on MPRAGE using FCM. To compare the performance of these three segmentation methods, two persons were asked to do the GM and WM segmentation manually on upper, middle and lower area of brain on MPRAGE images. Based on the manual segmented results we can evaluate the performance among these three methods. In summary all these three methods have similar performance for general use.