Dear list members,
I have some questions regarding the correct way to analyze data using SPM5
obtained from a sparse-sampling experiment. In this study, 180 MR-volume
scans (echo-planar imaging) were acquired. Trials of all nine auditory
conditions (18 trials and volumes per condition) and 18 additional silent
trials (rest condition) were presented in a pseudo-randomized manner.
Volume scans were acquired using temporal sparse sampling with a
repetition time of 10.0 s, an acquisition time of 2.6 s, and a single
auditory stimulus being presented, in silence, 5 s before the next volume
scan (in other words, for each trial there was 2.6s of acquisition and
7.4s of 'silence').
Images were then corrected for slice timing, realigned, coregistered,
normalized to a T1 template using parameters from gray matter
segmentation, and smoothed (6 mm3 Gaussian kernel). A first question
concerns the need to reslice the functional images. According to a
previous post (http://www.jiscmail.ac.uk/cgi-bin/webadmin?
A2=ind05&L=SPM&P=R64238&I=-3), because slice-timing correction is
performed by interpolating between adjacent scans in the time-series and
that here each scan is 10 seconds apart, the assumptions used in
interpolation are likely to be less valid. Because, images were acquired
in a interleaved, bottom-up, manner (with 40 slices per volume), I
nevertheless resliced the images (TR=2.6s, TA=2.535, number of slices=40,
reference slice=20). Is it correct?
Concerning the design matrix, for each subject, ten regressors of interest
(nine speech conditions plus rest condition) and six realignment
parameters as covariates of non-interest were modeled using a finite
impulse response (window length=2.6s, order=1). The timing parameters
were: interscan interval=2.6s, microtime resolution=40, microtime
onset=20. The events duration for each condition was 0. I used a high-pass
filter of 128s but no global normalization and no serial correlations. Are
these parameters correct? Notably, the interscan interval (if I enter
10s, the regressors in the design matrix are blank) and the events
duration of 0. Also, do I have to use a high-pass filter (here at 128s)
but no serial correlations?
Many thanks in advance for your help.
Marc
****************************************
Marc Sato
GIPSA-LAB, CNRS, University of Grenoble, France
|