I have an event-related paradigm with variable length, and I would like to get a functional activity parcellation ROI measurement for each trial individually in order to enter into a reinforcement learning model at a trial level. 

Can I extract ROIs for parcellations directly from preprocessing output (e.g., fmcpr.sm5.fsaverage.rh.nii.gz) created by preproc-sess? I would then apply an HRF gamma function myself to each time series and then match each trial to the activity at that time point.

I initially tried creating paradigm files for mkanalysis-sess where every trial was essentially its own condition. But this seems clumsy and might not work well because there's a differing number of 'conditions' (trials) in different subjects and runs.

So now I'm wondering if there's a way to extract time courses associated with each cortical parcellation? So, for instance with the Destrieux Atlas there are 58 parcellation units, so for each session I would be ultimately looking for a 58*t matrix where t is the number of time points in that session. 

I've already run the anatomical parcellations, so I'd think the next step is to match those to the preprocessed functional data like fmcpr.sm5.fsaverage.rh.nii.gz, then extract across the time course.