 |
 |
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.