Hi - I think your main two choices are whether to run FIX on each 5min run
separately, or to preprocess and concatenate each pair of scans from each
session and run FIX for each of the 4 paired datasets. You could try FIX both
ways on a few subjects and decide which is working better.
Cheers.
In the context of the long resting state runs that we have available, I would
argue that throwing in additional possible confounds is the appropriate thing
to do. Are you suggesting that sex, age, age^2, sex*age, sex*age^2, brain
size, head size, and average motion shouldn’t all be included?
Certainly one difference is that HCP (i.e., Steve) tends to take the more
conservative approach of regressing a *lot* of potential confounds, which
tends to result in a lower prediction values. You can see that without
confound regression, Steve's prediction is 0.21 versus 0.06.
Regards,
Thomas
Dear Experts:
We have acquired rfMRI dataset with A-P and P-A phase encoding direction and
the data was acquired in eight 5-minute runs split across four imaging
sessions. We have processed the data using HCP pipelines (e.g., PreFreeSurfer,
FreeSurfer, PostFreeSurfer, fMRIVolume, fMRISurface
I’ll note that when I do variance normalization, I do it with just the
unstructured noise map which is available in the HCP release (though it will
need to be resampled to MSMAll space if you are using the MSMAll data).
Peace,
Matt.
From:
I’m not sure what you are looking for, beyond what is in the FAQ.
For a given voxel/grayordinate/parcel, if
M = mean_over_time
S = std_over_time
and X(t) is your time-series
then demeaning is just: X(t) - M
and variance normalization is: (X(t) - M)/S
cheers,
-MH
--
Michael Harms, Ph.D.
Re (2) (expanding on Matt’s response): Demeaning and variance normalizing a
parcellated timeseries (or equivalently the time series for a single ROI), and
then concatenating those, is not the same as demeaning and variance normalizing
the dense time series, concatenating those, and then
1. Yes
2. I haven’t tried it on parcellated timeseries, but suspect that would be
fine too.
Matt.
From:
>
on behalf of hercp >
Date: Thursday, October 5, 2017 at 2:12 PM
I am extracting time series from regions of interest. Matt Glasser suggested
that I mean/variance-correct and concatenate the RL and LR phase encoded time
series. I still have a couple of questions.
1. Is the concatenation over time? If so, doesn’t this introduce temporal
discontinuity?
Basically you can think of it as an SNR bias.
Peace,
Matt.
From:
>
on behalf of "Elam, Jennifer" >
Date: Wednesday, October 4, 2017 at 4:17 PM
To: Romuald Janik
Indeed I think we would need to know what you needed the distance for to know
how best to compute it. For things like MR artifacts, a 3D distance might be
most appropriate. For something like smoothing, a geodesic distance would be
appropriate. For something neurobiological, the tractography
Right we will recommend using the areal classifier to find these areas rather
than the group parcellation once the areal classifier is available.
Peace,
Matt.
From:
>
on behalf of Timothy Coalson
12 matches
Mail list logo