I have some very naive questions. I am applying DCM to task fMRI data. However I have some very basic questions

1.    I have extracted the complete time series from VOIs. My task of interest only lasts for 30 seconds within the total 300 second time series. Stimuli of interest was presented in a blocked design with 4 unique stimuli per block for three blocks. The stimuli of interest within the time series was randomised for each participant to avoid any bias.  Now my question is, how do I enter the timings of my task of interest inside my DCM design ?

 

2. Since it was part of a large experiment, during rest of the time series, several other tasks were performed. How do I separate the baseline and task of interest from these unrelated tasks ?

 

With fMRI, we don’t just model one particular experimental condition – unlike EEG, the slow dynamics of BOLD means that we can’t chop up the timeseries. So, you should specify a GLM containing all your conditions of interest, trying to explain as much variance as possible. When extracting timeseries, you can regress out conditions of no interest, by excluding them from your ‘effects of interest’ f-contrast. In practice, this is generally used to remove nuisance regressors such as the mean of the signal and movement. Some or all of the GLM regressors are then used in the DCM as driving inputs on regions or modulatory inputs on connections.

 

Feel free to give more detail on your experimental design and hypotheses if you’d like more guidance on how to structure the model.

 

Best

P