My preprocessing also takes quite long. As far as I know this depends on several factors:
1. The architecture one uses. I've heard that Intel processors uses a BLAS (mathematical routines of matlab) that can use both cores of a dual core processor, while AMD cannot still use both cores. I tried to compile a new version of BLAS without success. It's also not clear to me if Matlab is using both cores or not. It just saies 50% in task manager. For this reason I open two matlab instances and run different subjects on them.
2. The number of voxels you use to reslice. If you reslice in dimensions 2x2x2 it will take definitely longer than 3x3x3. Also statistical analysis will take longer. However it seems that for GLM is better to have more voxels for more accurate results.
3. The order of polynomials during realignment and reslicing (ie. normalization). For example I used 3rd order for realignment and 4rth order for reslicing. This is quite much and takes longer than its deafult 2nd and 3rd order respectively.
4. Number of images you have for each subject. I have around 1700.
5. Probably (I am not sure) it may help changing the default value of maximum memory SPM can use in "spm_defaults.m".
defaults.stats.maxmem = 2^30; (it uses 1 GB)
Hope this helps. Others may know better if adding RAM can improve performance significantly.
Dorian.
2009/1/23 s yq
<[log in to unmask]>Dear users,
The pre processes of my current experiment is really time consuming. More than two hours are need for each subject. So I wonder to know if the ram could be improved from
1GB to 2GB , how much benifit would be? Or is there any other good way to increase the speed?
Thank you very much :)
Best regards
Yiquan
