We are currently planning a new event-related fMRI study, which will involve 6 task blocks per session. We would like to perform univariate activation and multivariate (MVPA) decoding analyses, as well as task-related functional/effective connectivity analyses such as PPI and DCM. Now, we are wondering whether it would be better to do continuous scanning (1 run) with 6 blocks that are separated by 20-second rest periods, or to do 6 runs (between which the scanner is stopped)?
Some people I've talked to argue that task-based functional connectivity analyses (e.g. PPI, DCM) should be performed on a continuous timeseries, and concatenating runs would not be good. This speaks for continuous scanning with 6 blocks.
For MVPA decoding, on the other hand, people seem to prefer leave-one-RUN-out cross validation to leave-one-BLOCK-out cross validation to ensure that training and test data are independent. This would speak for 6 runs.
A compromise would be 2 runs with 3 blocks each.
Which option would you prefer?
Thanks in advance and all the best,
Philipp
--
Dr. Philipp Kuhnke
Postdoctoral Researcher
Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany
|