This is a great example of why you don't want to interpolate your data (on
the scanner or otherwise) for tractography. Your data is stored in float 32
format, which means 32 bits per pixel (or 4 bytes). To find at least how
much RAM you would need to run probtrackx multiply by the dimensions of the
image: 4*256*240*140*50=1.65GB. Now, there are six "merged" files you have
to load for probtrackx, so the total RAM that you have to have just to
complete the loading process is 9.89GB (if you don't include any other use
for the probtrackx exe itself, or the operating system). If your data were
not interpolated, the footprint would be 1.24GB. I believe it is possible
to "deinterpolate" data, but I don't have experience with that. Although
files are smaller on your disk because of compression, they must be loaded
into RAM uncompressed.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of James Sheehan
Sent: Tuesday, March 11, 2008 10:19 PM
To: [log in to unmask]
Subject: Re: [FSL] ProbtrackX Memory Allocation
Here is the result for th1samples and th2samples.
Thank you!
# fslinfo /root/Desktop/control/subject100.bedpostX/merged_th1samples.nii.gz
data_type FLOAT32
dim1 256
dim2 240
dim3 144
dim4 50
datatype 16
pixdim1 1.0000000000
pixdim2 1.0000000000
pixdim3 1.0000000000
pixdim4 1.0000000000
cal_max 0.0000
cal_min 0.0000
file_type NIFTI-1+
#
# fslinfo /root/Desktop/control/subject100.bedpostX/merged_th2samples.nii.gz
data_type FLOAT32
dim1 256
dim2 240
dim3 144
dim4 50
datatype 16
pixdim1 1.0000000000
pixdim2 1.0000000000
pixdim3 1.0000000000
pixdim4 1.0000000000
cal_max 0.0000
cal_min 0.0000
file_type NIFTI-1+
#
|