中国医物理学杂志.pdfVIP

2002 10 19 4 Spatial Normalization Method Based On Discrete Cosine Transform 1 2 1 LUO Qing-fei , CHEN Guo-yue , BAO Shang-lian (1. The Institute of Heavy Ion Physics, the Research Center for Tumor Diagnosis and Therapeutical Physics, and Beijing Key Laboratory: Medical Physics Engineering, College of Physics, Beijing king University, Beijing 100871.China; 2. Department of Electronics and Information Systems, Akita Prefectural University, Honjo City, Akita, 015-0055, Japan) Abstract: In order to do the inter-comparison between individual volunteer subjects for human brain functional magnetic resonance imaging experiments, one has to do the spatial normalization to optimize the anatomical consistence between individual brain images by warping them to match both of the size and features to a template brain image. Among the most used methods of spatial normalization: label based and non-label (signal intensity) based, the signal intensity based mehtod is more widely used because it can be done automatically at higher precision. The disadvantage of the signal intensity based method is its heavy computer time. In this paper, a 3D signal intensity based method with discrete cosine transform (DCT) was introduced to solve the problem at efficient computer time. The DCT method was tested on a normal equipped PC computer with less than 2 minutes in 3D frame for image matrix size of 128×128×30 with high precision. Key words: spatial normalization; discrete cosine transform; polynomial basis function; Gauss-Newton iteration CLC number: R81