rs-fMRI analyses ; machine change


I am new in the field of rs-fMRI analyses and need your advice. My problem is about whether my patients are comparable in terms of connectivity due to different scanning parameters.

In more detail:

I’ve done a prospective study with clinical follow-up of subjects with tinnitus. Every subject had to be scanned with 4 rs-fMRI acquisitions within one-year period (T0 – T2 months – T6 months – T12 months). The idea is to compare seed-based fc of the T12-T0 scans, between patients who had good evolution and patients who had bad evolution. Evolution was determined with clinical examination.

Unfortunately, the MRI scanner was replaced during the study. This means that I now have some subjects with all scans acquired on the old machine, some subjects with a mix of data acquired on both the old and new scanner, and some subjects with their scans acquired on the new MRI only.

My question is: can I analyze all subjects regardless of the scanner issue? If yes, how I can account for the scanner effect? Is there a way to account for the MRI machine in the analysis?

Also: is the scanner condition restricting us to perform specific analyses, like seed-based would be less affected than whole-brain ICA for example?

Essentially, what would you recommend as the proper analysis approach for this issue?

Thank you very much for your help,

Marie Detroz, MD, PhD trainee

Hi Marie,

It is hard to say for sure without knowing the details of the scanner and the protocol how much of an effect this would have. However, the most simple way to accommodate for it would be to introduce two columns in your second-level design matrix for scanner_1 and scanner_2, for each row putting a 1 in the columns that corresponds to which scanner the scan came from, and a 0 in the other column.