论文部分内容阅读
在分析功能磁共振成像(fMRI)试验数据时,为了在不同被试之间比较实验结果、增加实验数据的可靠性,需要对所有被试的脑解剖学影像在三维空间空间内归一到一个共同的标准模板影像,使得它们之间的解剖结构差异达到最小。过去被广泛使用的基于体素的自动化归一化方法虽然很精确,但很耗时。本文章发展了可以在高速度下达到高精确度归一的基于体素的3D归一化方法,该方法采用了余弦离散变换(DCT),克服了原来方法需要较长计算时间的缺点,在现有普通微机上2分钟之内完成对128×128×30体素尺寸下的三维脑图像归一化,文章对方法和结果进行了讨论。
In analyzing fMRI test data, in order to compare the experimental results between different subjects and increase the reliability of the experimental data, all the brain anatomical images of the subjects need to be normalized in a three-dimensional space Common standard template images minimize the difference in anatomy between them. Voxel-based automated normalization methods, widely used in the past, are time-consuming, though accurate. This paper develops a voxel-based 3D normalization method that can achieve high-precision normalization at high speed. This method uses the cosine discrete transform (DCT) to overcome the shortcomings of the original method requiring a long calculation time. The existing normal computer completes the normalization of three-dimensional brain images of 128 × 128 × 30 voxel size in less than 2 minutes. The methods and results are discussed in this paper.