Hi all, I have the same "problem", so I split the 4800 volumes in half and used only 2400 data points so that I have at least two 15-min sessions (2x 1200) with both phase encoding directions (I'm using the node timeseries txt-files). I also thought about looking at the other 2400 data points and compare the difference between the first two and the last two session. As I understood, only using 1200 volumes which is only one phase encoding direction, seems to be problematic.
What do you guys think? Is it ok to do it this way? greetings David 2015-01-23 16:54 GMT+01:00 Glasser, Matthew <[email protected]>: > The function wb_command -cifti-correlate allows you to conserve RAM when > doing a pairwise correlation. Also I doubt doing it one run at a time > would reduce the RAM requirements much since the main array is the 91282 X > 91282 dense connectome. > > Peace, > > Matt. > > On 1/23/15, 9:08 AM, "Chao Liu" <[email protected]> wrote: > > >Dear Matt., > > > >Thanks for your reply. I have only one machine with 16GB RAM and the > >others with 8GB. for a dtseries.nii file, if I calculate pairwise > >correlation for a 91282 by 1200 matrix. It would cost 60GB RAM in matlab. > >Although we find a way to calculate a sparse pairwise correlation matrix, > >we still can not handle a 91282 by 4800 (or more) matrix easily. > > > >Now that using all four files is essential, I plan to first calculate the > >pairwise correlation matrix for each run, then average the correlation > >matrix from 4 runs for each subject. Is this ok? > > > >Many thanks! > >Chao > > > >> On 23 Jan 2015, at 13:57, Glasser, Matthew <[email protected]> > >>wrote: > >> > >> How much RAM do you have? The CIFTI files use far less RAM than the > >> volume files, but contain all the grey matter information represented > >>in a > >> way that makes combining across subjects more accurate. If you don¹t > >>use > >> all the data your estimates of the subject¹s functional connectivity > >>will > >> be noisier and generally less stable (one 15 min run looks more > >>different > >> than another than if you compared two sets of 1 hour¹s worth of data). > >> Also there are some left/right asymmetries in the data induced by the > >> LR/RL phase encoding directions, which some people perceive as a problem > >> (though if we had used AP/PA there would still be asymmetries > >> posterior/anterior). > >> > >> Peace, > >> > >> Matt. > >> > >> On 1/23/15, 6:44 AM, "Chao Liu" <[email protected]> wrote: > >> > >>> Dear HCP Experts, > >>> > >>> I am inspecting the HCP rfMRI data. For each subject, there are two > >>> sessions with 2 runs (LR&RL) in each session, yielding 4 files in > >>>total. > >>> Although we can concatenate all the four time series and do the > >>> correlation, I found the memory needed is more than the memory I have > >>>on > >>> lab computer, leading to a very slow calculation. So I am wondering the > >>> difference between using only one run and using all the four runs. Will > >>> it give wrong results if I use only one run? > >>> > >>> Many thanks! > >>> Chao > >> > >> > >> ________________________________ > >> The materials in this message are private and may contain Protected > >>Healthcare Information or other information of a sensitive nature. If > >>you are not the intended recipient, be advised that any unauthorized > >>use, disclosure, copying or the taking of any action in reliance on the > >>contents of this information is strictly prohibited. If you have > >>received this email in error, please immediately notify the sender via > >>telephone or return mail. > >> > >> -- > >> This message has been scanned for viruses and > >> dangerous content by MailScanner, and is > >> believed to be clean. > >> > > > > > ________________________________ > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
