 |
 |
Hi Ella, I am no expert and just an undergrad student have been working with ICA for a while so I think I can help you a little bit. 1. Have you try to run ICA (by using FEAT) with the spatial filtering option first then using ICA-AROMA on the filtered data (with the -md option point to the ICA result directory to avoid ICA-AROMA run ICA again. This will reduce the number of components. Base on your experimental design and what you expect, you can change the value of FWHM but for me, I usually use 5mm. 2. I haven't found any paper mention the effect of reducing number of components but I will not advise you to do so, some of my useful information appear in IC number 78 (over 112 ICs). As long as you want to retain as much as information, try not to reduce number of IC. 3. I dont know which machine you are using but I had use an 2010 Macbook (8GB RAM, Core Duo 2, which is very slow, of course) to run fsl_regfilt for a data with had 278 ICs so you better check again, maybe, just maybe, the memory is not the problem here. Feel free to tell me if I can help you with any information. Best regards, Khoi On Thu, May 5, 2016 at 7:01 PM, Ella Hinton <[log in to unmask]> wrote: > Hi > I have run ICA-AROMA on my fast TR data - unfortunately possibly due to > the large number of components that are identified (~299) the final > fsl_regfilt command cannot run as it requires an inordinately large amount > of memory to run. > > I noticed the -dim option to limit the number of components that ICA-AROMA > identifies, and thought that this might be a good option to increase the > likelihood of the process completing, and yet still produce motion > corrected data without too much signal lost. > > However I can't find any information on the impact of limiting the > components, or indeed what to limit them to - I've read both Pruim papers, > and searched elsewhere. I found one message on this list suggesting its not > a great idea (as can't get an idea of data quality), but given ICA AROMA > doesn't run without limiting the components, I think I have to limit them > (if I am to use ICA AROMA, which does sound sensible!). > > Do any FSLers have any advice? > With thanks in advance, > Ella >