Dear all,
For some yet unexplained reason, a 0.5 Hz scanner artefact appears in our 800
volume PRESTO SENSE timeseries (TR 0.609). This noise sums nearly linearly on
top of existing BOLD signal intensity. Analysis shows that it can be
eliminated during postprocessing using user-specified regressors in SPM5
(with delta functions every 4th scan) as follows:
Regressor1:
1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 ... etc: complete 800 volumes.
Regressor2:
0 1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 ... etc: complete 800 volumes.
Regressor3:
0 0 1 0 0 0 1 0 0 0 1 0 0 0 1 0 ... etc: complete 800 volumes.
This yields 3 PE images for these 3 regressors with almost 100% noise
elimination with good residual SD.
Since I want to use MELODIC for further study of these data (and since I'm
forever hooked to FSL), I'd like to do something similar in FSL. I tried
MELODIC denoising, however MELODIC does not find the relevant components.
Instead, nearly all components are contaminated with the 0.5 Hz artefact
(possibly because of its regular nature?). Thus, I think I may first need to
do a model-based analysis in all individuals (as in SPM5, see above) using
custom EVs with delta functions, and then somehow filter out the beta images
from the filtered_func_data (using avwmaths?). Or is there perhaps a more
sound way of dealing with this problem?
Any help would be greatly appreciated,
Cheers,
Rutger Goekoop.
|