I acquired multiband BOLD data and I want to use topup to correct for distortions. I have plenty of options. I acquired: 

1- 1 measurement of BOLD with phase encoding P to A
2- 1 measurement BOLD A to P
3- 530 measurements of BOLD P to A (run 1)
4- 530 measurements of BOLD A to P (run 2)
5- 530 measurements of BOLD P to A (run 3)
6- 530 measurements of BOLD A to P (run 4)
7- 530 measurements of BOLD P to A (run 5)
8- 530 measurements of BOLD A to P (run 6)
9- 530 measurements of BOLD P to A (run 7)

The session lasted about 90 minutes. Although people didn't move too much during the runs, it feels likely that head position changed somewhat from beginning to end. People tended to move lower body parts during breaks. A few took a break outside the scanner between run 4 and run 5. 

My question: Instead of creating a single topup image from acquisition 1 & 2, I was wondering if there could be benefits to picking 2 images that are close in time to the to-be corrected data. To do this, I could pick the last measurement of the previous run (including the 1 measurement of BOLD P to A, #2 above), and create an image with the first measurement of the run (e.g. run 1). The topup image would be applied to that run (e.g. run 1). So each run would get its own 2 images (last measurement of the previous run + first measurement of the run). 

This is not the only option of course, and I am interested in your thoughts regarding what you think is optimal.

Thank you,