We are currently early in the process of planning a fMRI study that will include a baseline measurement, as well as 4 testing sessions. Of course, our protocol will be time-constrained and we have been wondering about the need to acquire anatomical images (T1w and T2w) at each study visit (we are not interested in and do not expect changes in anatomy). It would be really handy to only take anatomical images at baseline and have 10 min of additional time for tasks during the testing sessions.
The only study that used multiple measurement days that I have been involved in repeated the anatomical aquisitions at each study day. I noted that during preprocessing with fMRIprep, these were simply averaged and the average used for coregistration. I am therefore wondering whether in our case (1) it would suffice to have anatomical acquisitions at baseline only, or (2) whether taking an anatomical image at each study visit is best-practice and (3) if yes, why that is?
We’re only collecting one set of anatomical images per scanning wave for the DMCC project and it is working well. We have three fMRI scanning sessions per wave, with several days to weeks between sessions, but only collect the anatomicals in the first session unless their quality is poor.
(Shameless plug: the DMCC55B subset of the data and a lot more detail and pointers about the project is in an (accepted but not yet online) preprint (https://doi.org/10.1101/2021.05.28.446178); dataset is at openneuro.org/datasets/ds003465/.)
Edited to add: The underlying logic of this practice of using one anatomy for several sessions is that the anatomy should not change substantially between sessions. I believe this is a reasonable assumption for a project like the DMCC, in which the population is healthy young to middle aged adults, and sessions separated by weeks or months, not years. (We do repeat the anatomical scans for later waves, so could evaluate the magnitude of any changes, but have not done so (yet).) The calculation would likely be different with populations in which changes are expected to be rapid (perhaps very young people, or people with tumors or hydrocephalus-type conditions), or perhaps extremely high resolution imaging.
It is an interesting question! You are right, if there are multiple T1w available in your data set for one subject, either within one session or within session, here is what fmriprep will do:
In the case of multiple T1w images (across sessions and/or runs), T1w images are merged into a single template image using FreeSurfer’s mri_robust_template. This template may be unbiased , or equidistant from all source images, or aligned to the first image (determined lexicographically by session label). For two images, the additional cost of estimating an unbiased template is trivial and is the default behavior, but, for greater than two images, the cost can be a slowdown of an order of magnitude. Therefore, in the case of three or more images, fMRIPrep constructs templates aligned to the first image, unless passed the
--longitudinal flag, which forces the estimation of an unbiased template.
This means that for 3 or more T1w images, fmriprep will by default use the 1st T1w image and register all the other on it. So in you case, if you were to use fmriprep to preprocess your dataset, the subsequent anatomical images will not really be useful and you can acquire just once your T1w and T2w images. As written above it is still possible however to estimate an unbiased template that would be used to register all the bold images in case you acquire anatomical images for each session.
My personal opinion: For use as a target for bold registration, only one anatomical session is sufficient, considering the gain in time you get for each session without anatomical acquisition.
It depends on what you are doing. If you are acquiring 3T data at conventional resolutions and bringing everything to a common space, then one anatomical should be enough, and its nice to learn from others that they have tried this and its working well!
However, extending on jaetzel’s answer, if you are acquiring high res data (in particular at 7T or above) and analysing each session on the subjects anatomical space, then it’s certainly safer to acquire an anatomical at each session. This is because for high res imaging different distortions from gradient nonlinearity and B0 may change and they can have a detrimental effect on your registration later on. Also, anatomicals do have distortions too (hence people use multi-echo MPRAGEs with different readout polarities at each readout), and they can change based on scanner position and day to day physiological fluctuations including hydration (Brain morphometry reproducibility in multi-center 3 T MRI studies: A comparison of cross-sectional and longitudinal segmentations - ScienceDirect).
Another reason to acquire anatomicals at every session is because subjects may move during your scan and you may not have time in a session to acquire an extra anatomical scan. Maybe that’s ok if you are doing 4 sessions so you can try again at a different session, but maybe those other sessions are time-starved too, so having planned an anatomical ahead of time would avoid issues.