基于非参数Bootstrap检验的fMRI脑激活区域检测

Brain Activation Based on Bootstrap Test for fMRI

统计学上由多重校验问题所带来的高假阳性率一直困扰着fMRI研究人员,SPM等主流fMRI数据处理软件也一直没有有效的解决方法。由于数据采集、被试个体差异等造成的问题,使得数据不一定完全符合参数方法中总体分布上的假设,而非参数方法不需要对数据分布进行假设,可能比参数检验方法更可靠。基于Bootstrap检验提出一种新的非参数体素激活检测方法,该方法从小样本抽样出大样本来推断数据分布形态,并设定更合适的阈值进行校正,从而尽可能地减少体素假阳性率。

The high false positive rate caused by multiple calibration problems,which is a statistic problem,has always plagued fMRI researchers.There is no effective solution in the mainstream fMRI data processing software such as SPM.Due to the problems caused by data acquisition or individual differences of the subjects,etc.,the data may not be completely consistent with the assumption of the overall distribution in the parameter method.So the non-parametric test method may be more reliable than the parameter test method.Based on the non-parametric Bootstrap test,proposes a new non-parametric voxel activation detection method.Large samples are sampled from small samples to infer the distribution of data,setting a more appropriate threshold for correction,and then reducing the false positive rate of voxels as much as possible.