摘要
为了提高用邻域相关的ICA算法进行脑功能区提取的准确性,先利用区域增长法对数据进行预处理,分别计算出切片中的每一个体元与其空间模型中的其它26个体元的相关系数,若相关系数大于一定阈值的个数小于设定值,则体元判定为非激活区体元。利用健康人手动fMRI数据对两种算法进行实验仿真,预处理可以滤除一些明显不是激活区的点,运行时间缩短为原来的47.3%。结果证明方法有效地保障了结果的准确性,提高了运算效率。
When ICA method based on neighborhood correlation is used to extract the localization of brain activities from the function Magnetic Resonance Images,the result is not very exact.In order to improve the situation,the data are preprocecced using area-growing method before ICA.The correlation coefficients of each pixel to other 26 pixels in a space model are computed respectively.If the number of correlation coefficients which are larger than the threshold is less than a certain value,it is determined that the pixel doesn't belong to the activity area.The experiment whose data contain left hand-moving of healthy people can prove that the area-growing method can cancel out some pixels which doesn't belong to the activity area obviously,and the run time is shortened to 47.3% of the original value.
出处
《计算机仿真》
CSCD
北大核心
2010年第1期222-225,共4页
Computer Simulation