Dear SPMers,
I am trying to figure out how to apply DCM for the following experiment:
The overall goal is to explore the processing of visually presented
sentences of different syntactic complexity (e.g. easy, moderate, and hard
sentences). The experiment is of block design, it has 4 experimental
conditions: Easy, Moderate, Hard, and Baseline.
The blocks of these conditions are presented in a randomized fashion within
a single fMRI scan.
The DCM goals are:
-Estimate the effective connectivities within the language processing network
-Characterize changes in connectivity induced by the different types of
processed sentences
-Eventually, to compare connectivities and their induced changes for the two
groups of subjects (normal and special population group).
It looks like there are two possible ways to do the DCM analysis:
1. Split the the experiment into three separate time series, corresponding
to Easy, Moderate, and Hard conditions (should the baseline images be
included in the all three of the time series?). Then a simple DCM without
contextual inputs can be constructed separately for each of these time
series. The estimates of the intrinsic connections should then reflect the
condition-specific connections.
2. Use the full time course of the experiment in a single DCM model, but
create a special design matrix (SPM.mat) that differs from the design used
for the standard SPM analysis:
Specify a “Sentence” regressor for the blocks of all types of sentences, and
separate regressors for each type of the sentences. Then use the “Sentence”
regressor as a stimulus-bound input, and the three sentence-specific
regressors as contextual inputs.
It looks like something similar to the second approach was used in the paper
recently published in Cerebral Cortex (A. Mechelli, C. J. Price, K. J.
Friston and A. Ishai. Where Bottom-up Meets Top-down: Neuronal Interactions
during Perception and Imagery).
I would greatly appreciate any advice.
Thanks in advance,
Vladimir Cherkassky
Center for Cognitive Brain Imaging
Psychology Department
Carnegie Mellon University
Pittsburgh, PA 15213
email: [log in to unmask]
fax: (412) 268-2804
tel: (412) 268-3379
|