Hi - I guess the calculation of the median, which implies full access
to all data, requires more RAM than more of the other preprocessing
commands.
I'm surprised that it is crashing though, given the amount of RAM and
swap that you have - make sure you definitely are running a 64-bit
version of FSL, and you could try increasing the swap further (and
watching with 'top' whilst it is running to see what is going on in
RAM usage).
However, we never recommend concatenating raw FMRI data across
sessions, but always recommend processing each in its own first-level
analysis, followed by combination in a second-level FE analysis if
necessary, which allows you to contrast across sessions. Is there any
reason why you can't do this?
Cheers, Steve.
On 30 Jul 2009, at 22:38, Nick Wymbs wrote:
> Steve, thank you for the feedback. I was asking because there is a
> nagging error during the prestats that I'm hoping to circumvent.
>
> The problem occurs when handling rather large prefiltered_func files
> (~3.5GB, 6210TRs) and I am getting a consistent error at the line
> where fslstats is called to find the median intensity value of the
> image (fslstats prefiltered_func_data_mcf -k mask -p 50).
>
> Under normal circumstances when the command is finished, the median
> value is entered on the line below in the logfile. With the error
> version, no value is found and this ends up crashing the prestats at
> the smoothing stage.
>
> My first thought was based on the sheer file size, this might be a
> memory load issue. But the command is a simple discrete calculation
> and other stages like motion correction and smoothing are able to
> run, which seem to be more computationally intensive. Given this,
> and that I am working on a machine with 16GB of physical RAM and
> 10GB of swap, it seems that maybe there is something else going on.
> Any thoughts?
>
> Lastly, the prefiltered file is large because it is three training
> sessions, taken from three days, that are concatenated in time. The
> analysis that we are trying to run relies on the sessions being fit
> in the same model. Because each session slightly different scanner
> position of the head, there are different ranges of intensity values
> for each of the sessions. Could this be what is crashing the
> fslstats command?
>
> Cheers
> Nick
>
>
> On Wed, Jul 29, 2009 at 12:49 AM, Steve Smith <[log in to unmask]>
> wrote:
> Hi,
>
> Is it just that you want to avoid using the GUI? If you setup the
> design.fsf file correctly, you can call "feat
> design.fsf" (lowercase) from the command line to do all the
> preprocessing for you.
> If you wanted to avoid that, you'll need to look in detail at the
> TCL scripts, to see exactly where such numbers come from.
>
> Cheers, Steve.
>
>
>
> On 28 Jul 2009, at 20:02, Nick Wymbs wrote:
>
> Dear FSL,
>
> I would like to try running the prestats for my first-level analysis
> from the command line. The most logical thing seems to start off by
> duplicating what is in the report_log file generated previously
> using the GUI.
>
> When looking at the logfile output, it is unclear to me how a some
> of the input numbers are generated. I relaize this is pretty basic
> but it would help a great deal in streamlining my analysis!
>
> Two main burning questions:
>
> How is 653.655 generated for the below command line?
> /usr/local/fsl/bin/susan prefiltered_func_data_thresh 653.655 3.397
> 3 1 1 mean_func 653.655 prefiltered_func_data_smooth
>
>
> How is 11.474 generated for the line below?
> 2. /usr/local/fsl/bin/fslmaths prefiltered_func_data_smooth -mul
> 11.474 prefiltered_func_data_intnorm
>
>
> Cheers
> Nick
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|