Iım guessing those are an interaction between subject motion and the
multi-band (because multiple slices are acquired at the same time, motion
will produce a banding artifact like that. You should be denoising your
subjects individually, however, not in the group ICA stage.
Peace,
Matt.
On 6/25/14, 12:21 PM, "Matthias Heil"
<[log in to unmask]> wrote:
>I have a question on some Melodic IC-Maps. The attached maps show - in
>my opinion - mostly multi-slice acquisition related artifacts.
>
>The data set contains 700 * 14 = 9.800 data points (10 minutes each with
>0.841s T_R)
>
>The analysis is done with most of the Melodic default settings or those
>which reach better quality (BBR for registration, no resampling).
>
>I also choose the number of components manually to 30 (automatic
>estimation would have been something between 200 and 500)
>
>The highest outliers (3 datapoints) you see in the time course fit to
>movement artifacts.
>
>There are also 2 other components looking like that. DMN, visual and
>motor (left+right is 1 component) networks are easy to find too.
>
>
>
>So...
>
>Are those maps containing data or noise or both or undistingushable in
>this analysis settings? (I also have similar maps with automatic
>estimation of components... I will have a look how different they are.)
>
>Can we cleanup our data using such maps? Or, what would you do with such
>findings?
>
>
>I'm grateful for any comments and ideas!
>
>Matthias
|