 |
 |
Hi Matt, Thank you for your reply. The data has already been cleaned a la HCP. It is neonatal data, so in fact, it has been analysed with the developing HCP (dHCP) pipeline, which includes ICA-FIX de-noising. Since the data is from neonates, it has an unusually high level of motion artefacts, even after going through the dHCP pipeline. I would like to compare the effects of 'dHCP-style pre-processing' alone to 'dHCP-style pre-processing + scrubbing'. It is simply exploratory, and scrubbing may not be included in the end, but I would like to experiment with it all the same. I'm sure there is a simpler way to implement scrubbing, than the ways I've already outlined. If a simpler method for scrubbing is known, please let me know. All the best, Luke. > On 4 Apr 2017, at 21:09, Matt Glasser <[log in to unmask]> wrote: > > If you are doing an ICA analysis, I recommend cleaning your data with ICA+FIX if it is HCP-Style (high spatial/temporal resolution) or ICA-AROMA if it isn’t rather than scrubbing. > > Peace, > > Matt. > > From: FSL - FMRIB's Software Library <[log in to unmask] <mailto:[log in to unmask]>> on behalf of lukebaxter <[log in to unmask] <mailto:[log in to unmask]>> > Reply-To: FSL - FMRIB's Software Library <[log in to unmask] <mailto:[log in to unmask]>> > Date: Tuesday, April 4, 2017 at 2:15 PM > To: <[log in to unmask] <mailto:[log in to unmask]>> > Subject: Re: [FSL] How to delete volumes (scrubbing) using fsl? > > Hi Eugene and Matthew, > > Thank you for your helpful feedback. > > Matthew, I won’t be using my resting state data in a feat analysis, so I can’t use that approach unfortunately. I will be inputting my denoised data into an ICA analysis. > > Eugene, I have already denoised the data using melodic and fsl_regfilt. I have then run fsl_motion_outliers on this fully denoised data to find motion artefacts that I can’t seem to remove with the other pre-processing steps. The remaining motion (measured using dvars) is indeed greatly reduced in the fully denoised data compared to the pre-denoised data, but there are still some relatively large spikes remaining. I was hoping to remove these very few troublesome volumes by scrubbing. > > I hope that’s clear. Let me know if not. > > Luke. > > >> On 4 Apr 2017, at 17:10, Eugene Duff <[log in to unmask] <mailto:[log in to unmask]>> wrote: >> >> Hi - >> >> On 4 April 2017 at 17:06, Matthew Webster <[log in to unmask] <mailto:[log in to unmask]>> wrote: >>> Hello, >>> If you will be using your resting-state data in ( e.g. ) a FEAT analysis, then the confound matrix just needs to be added as a confound in the Stats tab in the FEAT GUI to account for the outliers. >>> >> >> If you're not doing a FEAT analysis (e.g. ICA), it may not be useful to remove this volumes. ICA should be able to separate these artefacts itself, and these components may also pick up smaller artefacts than are defined by the scrubbing. >> >> >> Eugene >> >> >> >>> Kind Regards >>> Matthew >>> > On 4 Apr 2017, at 16:46, Luke Baxter <[log in to unmask] <mailto:[log in to unmask]>> wrote: >>> > >>> > Hi, >>> > >>> > I would like to take the output of fsl_motion_outliers and delete those volumes from my resting state data that have been deemed outliers. My resting state scan has 500 volumes, and there are on average 25 outliers per scan to be scrubbed. >>> > >>> > Using fsl_split, manually deleting volumes, then using fsl_merge is quite a long and tedious manual approach. Similarly, using fslroi to isolate chunks of the scan between outliers is not ideal, because the outliers aren't clumping together much, so there would be about 20 chunks. Again, very tedious. >>> > >>> > Is there a way of taking the output of fsl_motion_outliers, and perhaps multiplying this by my resting state nifti, to delete the outlier volumes? Has anyone discovered a simple approach, or is there one already available in fsl? >>> > >>> > Any info would be greatly appreciated. >>> > >>> > Cheers, >>> > Luke. >> >