|
Activity Number:
|
293
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #307301 |
|
Title:
|
Using the Maximum Cross-Correlation Statistic To Find Significant Voxel-Wise Activations in fMRI Experiments
|
|
Author(s):
|
Kinfemichael Gedif*+ and Richard F. Gunst and Qihua Lin and William R. Schucany
|
|
Companies:
|
Southern Methodist University and Southern Methodist University and Southern Methodist University and Southern Methodist University
|
|
Address:
|
12920 Audelia Road, 269, Dallas, TX, 75243,
|
|
Keywords:
|
maximum cross-correlation ; fMRI ; HRF
|
|
Abstract:
|
Most statistical analysis on fMRI experiments have a common objective of identifying activated voxels due to some stimulus. One way this is done is by performing voxel-wise test of the null hypothesis that the observed response is not significantly related to an assigned theoretical hemodynamic response function (hrf). Statistical analysis based on fitting hrf models strongly depends on the adequacy of the fitted model on each voxel's time series. We investigate a technique that does not require fitting an hrf to the voxel time courses. The maximum cross-correlation statistic between the observed response and the ideal stimulus sequence is used to construct a voxel activation map. Resampling can be used to test for significant cross-correlation. Such a method not only avoids fitting an hrf but also handles the fact that the hemodynamic response is temporally blurred.
|
