By default, optseq2 assumes that you want to perform an FIR analysis where you get an average for each post-stimulus time point so that you can create a waveform. This is in contrast to assuming the shape to the hemodyn respnse where you only estimate a single value (the amplitude). By default, the time between FIR waveform points is the TR, but you can perform sub-TR estimation by setting the dPSD to less than the TR. The problem is that every estimate you make reduces the efficienecy. So, assuming a shape is more efficient than an FIR because it only has one estimate. An FIR with dPDS=TR is more efficient than dPSD=TR/2 because there are half as many estimates. Probably when you go to analyze the data you will use an assumed shape, and then you will get the efficiency back.
On 2/21/2020 9:43 AM, Gergely Darnai wrote:
External Email - Use Caution
Dear Developers,
I am planning to run PVT (psychomotor vigilance task) in fMRI using event related design. This is an extremely simple reaction time task: participant has to respond to the appeared geometric shape as quickly as possible. The key feature of this paradigm is that we will have fluctuating and quite rapid event presentation times (ranging between 300 and 500 msec). Another important information is that we will use FSL FEAT for the evaluation. We decided to use opseq2 to optimize the design with the following parameters:
optseq2 --ntp 150 --tr 2 --psdwin 0 20 --ev evt1 1 45 --nkeep 3 --o exp --nsearch 10000 --tnullmin 3 --tnullmax 11 --repvar 10
Although with this design I get quite satisfying efficiency and VRF scores I do not understand that if I decrease dPSD why does it have significant negative effect on efficiency and VRF. Could you explain this? If I understand well, this is the only option to shift the onset of the event from the scanning points, and I would assume that if there is fluctuation in time between scanning points and stimuli presentation, it would help to "catch" the hemodynamic response easier (if the stimulus onset always goes together with the scans, we can always catch the same timepoints of the HRF). Did I misunderstand something? If I use FSL that is based on HRF estimation (and not on FIR), do these parameters (dPSD) and scores (efficiency & VRF) have meaning and function at all? My last question is related to event duration. Although I have fluctuating and short events, as you can see I chose 1 sec (because it has to be the integer multiple of the dPSD). Is it problematic?
Thank you for your suggestions,
Gergely
Gergely Darnai PhD Department of Behavioural Sciences Medical School, University of Pécs Phone: +36/72/536-256 Fax: +36/72/536-257 H-7624 Szigeti u. 12, Pécs, Hungary
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer