Hello,
To add to the question of low-passing in MELODIC ... there was a question that was
posted a while ago and somehow got lost, so I'll take another shot at posting the
question. Any suggestions would be quite useful. Thanks -Ruchika
Hello,
I've been using group MELODIC tensor pica on a block design that presents 20 second
blocks of faces followed by 20 secs of scrambled images, followed by 20 secs of houses,
X 3, for 3 blocks of faces and 3 blocks of houses, alternating with scrambled image
baseline.
I've run it both with a lowpass filter of 9.32 seconds fwhm, and without lowpass filtering.
As you might expect there are usually half the amount of components when I run on
lowpassed data. What I've found, as well, is that with the lowpassed data I get one
clearly delineated component for face (FFA and some pfc regions), one for houses (PPA),
and one for the default mode, and the remaining components are either noise or task-
irrelevant networks. When I run on the un-lowpassed data, and get many more
components, within the first 10 components in addition to the first face and place
components, there are 2 to 3 duplicate FFA and PPA related components with slight
variations on either space or time mode compared to the first of each.
In terms of the "purity" of the component for capturing a signal of interest in the data, is
this a case for when lowpassing may be okay? This is a unique position since the FFA and
PPA are so robust, and it is a block design, so I have a very clear idea of what to expect
in time and space. Or, with lowpassing am I letting in more noise to my first components
of interest?
thanks,
Michelle
|