Friday, October 14, 2016

an update and some musings on motion regressors

In case you don't follow practiCal fMRI (you should!), his last two posts describe a series of tests exploring whether or not the multiband (AKA simultaneous multi-slice) sequence is especially liable to respiration artifacts: start here, then this one. Read his posts for the details; I think a takeaway for us non-physicists is that the startlingly-strong respiration signal I (and others) have been seeing in multiband sequence timecourses and motion regressors is not from the multiband itself, but rather that respiration and other motion-type signals are a much bigger deal when voxels are small (e.g., 2 mm isotropic).

This week I dove into the literature on motion regression, artifact correction, etc. Hopefully I'll do some research blogging about a few papers, but here I'll muse about one specific question: how many motion regressors should we use (as regressors of no interest) for our task GLMs? 6? 12? 24? This is one of those questions I hadn't realized was a question until running into people using more than 6 motion regressors (the 6 (x,y,z,roll,pitch,yaw) come from the realignment during preprocessing; transformations of these values are used to make the additional regressors).

Using more than 6 motion regressors seems more common in the resting state and functional connectivity literature than for task fMRI (Power et al. 2015, and Bright & Murphy 2015 , for example). I found a few (only a few) task papers mentioning more than 6 motion regressors, such as Johnstone et al. 2006, who mention testing "several alternative covariates of no interest derived from the estimated motion parameters", but they "lent no additional insight or sensitivity", and Lund et al. 2005, who concluded that including 24 regressors was better than none.



Out of curiosity, we ran a person through an afni TENT GLM (FIR model) using 6 (left) and 24 (right) motion regressors. This is a simple control analysis: all trials from two runs (one in blue, the other orange), averaging coefficients within my favorite left motor Gordon parcel 45 (there were button pushes in the trials). It's hard to tell the difference between the model with 6 and 24 regressors: both are similar and reasonable; at least in this test, the extra regressors didn't have much of an effect.

My thinking is that sticking with the usual practice of 6 regressors of no interest is sensible for task fMRI: adding regressors of no interest uses more degrees of freedom in the model, risks compounding the influence of task-linked motion, and hasn't been shown superior. But any other opinions or experiences?

9 comments:

  1. Hi Jo, Ged Ridgway ran the time series from my blog through the 12dof (affine) realignment. See this tweet and tweets beneath: https://twitter.com/RidgwayGR/status/786881456840769536 and the plots here: https://spideroak.com/browse/share/RidgwayGR/Temp/Share/motion_traces/ At first inspection I am struck by the way the scale and skew parameters capture the deep breaths for the MB6 data (less well for the ep2d data, perhaps because of the lower brain coverage). The way scale & skew capture these episodes seems to be similar for the conventional foam restraint and for the custom head restraint, suggesting to me that it is doing a good job of capturing the modulation of B0 across the head, as distinct from the direct mechanical motion. Intriguing result.

    ReplyDelete
    Replies
    1. correction looks good - but I'd still prefer good signal in as affine transform won't capture subtle non-linear effects due to motion/field interaction (yes we can also do field unwarping)

      Delete
    2. Let's assume good head restraint, as with just the custom holder, such that the majority of remaining apparent motion is actually modulation of magnetic field from breathing. Would you then prefer affine over rigid body realignment?

      Delete
    3. In case any readers are confused, this is a different issue than what I blogged about above: the post talks about using more than 6 motion regressors in the GLM, while these comments are about using a realignment algorithm that allows more than 6 rigid transformations.

      Delete
  2. For task based fMRI, some (I do) now use motion censoring that was introduced for rs-fmri. This paper https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3895106/ showed it was better. I even don't thing about it anymore, it's automatic. I wrote a routine for it here https://github.com/CPernet/spmup/blob/master/spmup_first_level_qa.m

    ReplyDelete
    Replies
    1. But if the head restraint is improved as with the custom system then censoring then likely redundant, no?

      Delete
    2. We are starting to investigate include censoring in our task fMRI as well. Not interpolation(as is sometimes done with resting state), but simply flagging timepoints to omit. This potentially introduces other issues (such as un-balancing previously balanced tasks), but seems prudent, particularly for people that are generally quite low motion, but have a few brief gross head movements. Though, it's possible that motion censoring is adding yet more experimenter degrees of freedom, particularly until we reach some sort of consensus on how much is reasonable for different task/acquisition schemes.

      Delete
  3. With (Bayesian) variable selection it is possible to have many covariates, and learn from the data which covariates to include. I'm working on that right now, and will hopefully soon publish a preprint related to head motion.

    ReplyDelete
    Replies
    1. Request: please test using data from well restrained heads! Either bite bar, custom printed case or some other way of restraining so that respiratory effects now become the major issue. I'd really like to know what we're up against as a fundamental limit today, since breathing will continue to be non-optional in most fMRI experiments!

      Delete