Use of Jackknife Resampling Techniques to Estimate the Confidence Intervals of fMRI Parameters

Publisher Website
Google Scholar
Abstract
The objective of this study was to determine the reliability and the confidence intervals of task-activated functional MRI (fMRI) parameters using a computer-intensive resampling technique. The jackknife, a commonly used method for resampling mathematical data, was used to calculate the confidence interval of fMRI parameters for a simple bilateral finger-tapping paradigm. Four healthy test subjects (three men, one woman) were used to test the correlation coefficient and variability in the data. Each subject performed 4.5 cycles, each cycle having 20 s of bilateral finger tapping alternating with rest periods of equal time, producing 90 images. One additional scan of 10 cycles (200 images) was used to test the stability of the method itself. One thousand jackknifed resampled data sets of 85 elements each (from 90 original points) were generated, and the correlation coefficient was determined using an idealized “on/off” box-car reference waveform. Activation maps were generated that had the same confidence intervals at each pixel. These maps were more localized with less extraneous activated pixels than the maps generated with a fixed correlation coefficient threshold. There was no significant difference in the distribution of correlation coefficients between the 85, 90, and 95 element, jackknifed data sets; similar robustness was seen, as well. The jackknife resampling technique for data analysis produced reliable distributions and statistical parameters. The jackknife estimates were shown to be stable, even from a small initial sample size. This method may be used in lieu of test-retest analysis.