作者: Yi-Hsuan Kao , James A. Sorenson , Stefan S. Winkler
关键词: Pattern recognition 、 Voxel 、 Dual echo 、 Segmentation 、 Vector decomposition 、 Artificial intelligence 、 Random noise 、 Image processing 、 Mathematics 、 Computer vision 、 Imaging phantom 、 Mr images
摘要: A general model is developed for segmenting magnetic resonance images using vector decomposition and probability techniques. Each voxel assigned fractional volumes of q tissues from p differently weighted (q < or = + 1) in the presence partial-volume mixing, random noise, other tissues. Compared with eigenimage method, fewer are needed tissues, contrast-to-noise ratio calculated improved. The can produce composite tissue-type similar to that methods, by comparing different on each voxel. three-tissue (p 2, 3) illustrated three dual-echo images. It provides statistical analysis algebraic method. three-compartment phantom segmented validation. Two clinical examples presented.