Question,

It is well known that the real-time echoplanar imaging sequence needs several acquisitions to get in the groove and stabilize, such that many sequences in protocols routinely apply several passes of RF to the brain before actually collecting image data to write to disk. In our case, we go a little further and do not start our paradigms for several additional seconds, where the first few volumes collected at the very onset of each run of our paradigm are not included in the concatenated 3d+time dataset.

When I create the first 3d+time dataset for each run with to3d, the application consistently lists sub-briks at the beginning of the run as outliers in the 3doutcount, which is understandable. What I don't get is why this report also almost invariably lists the last several volumes collected as outliers too! It's as if the system knows that the end of the run is coming and decides to get noisy. We usually collect 204-280 volumes in one run.

These outlier sub-briks often begin as soon as 10 volumes (in our case 20 s) from the end of the run-- while the subjects are still performing the task and have to indication it is about to terminate. This early onset, coupled with motion file output from 3dvolreg, does not suggest some systematic head motion in response to noticing the task has concluded.

Why might the system be doing this? We now collect extra volumes at the conclusion of the run and omit them in the concatentation.

Jim B