DIFF_PREP Update Notes
April 15th, 2008
SVN
Subversion control repository is created in "/raid2/home/irfanogo/svn/repos"
. We will use this repository to keep control over the versions of the software. There are three subfolders:
- Trunk: For the stable, release versions
- Branches: For non-stable, in-development version.
- Tags: For versions with an achievement done, to tag the versions.
The version in trunk is called version v1 and it corresponds to the old version 1.4.3 before the use of SVN.
It would be better to set up the repository in a better, more available folder and to setup a SVN web server to have access to the versions notes over the web. To do this we need to contact Jay.
Update Notes
Revision 4
Added the support for systems where FSLOUTPUTTYPE environmental variable is different than ANALYZE.
Revision 5
Registration settings moved to; the NIH_DTI_WORK folder in user's home directory. The registration files in the settings folder in the main software directory are kept just as samples for copy and paste purposes.
Revision 6
Extra registration settings file for step 5 is no longer needed. The code changes the settigns as necessary as needed.
Revision 7
Image resample tag is removed from the registration settings file.
The "ddd.txt" file coming from FSL masking library file is no longer generated for Linux 32 and Linux 64 bit versions.
It is still generated by Mac version.
Revision 8
Upsampling changed to neasest neighbor and bicubic interpolation.
The registration settings file has a new tag for it called "upsampling_method"
Automatic noise computation code was made much more robust.
Philips PAR/REC support improved.
Conversion from HDR/IMG tensor to VTK tensor improved.
Extra redundant registration of structural to structural was removed from step 5.
May 20th, 2008
Update Notes
Revision 9
Read analyze routine was fixed to account for negative resolution images. Now the images will be read and displayed with the correct orientation.
Slight change to automatic noise computation.
Revision 10
PAR/REC conversion now supports more than one trace weighted image in the REC file.
PAR/REC conversion now supports old version of PAR format, ie V4. The user needs to provide a gradient file as an additional parameter if that is the case.
Revision 11
Cropping and bounding box operations improved. Instead of just copying the brain locations obtained with FSL (to the new image with desired FOV), the code finds the center off mass of the brain and copies as many voxels as possible around the center to fit the new image.
A new tag added to data desciption file called "is_brain_data". If set to yes, FSL is used to do skull stripping for cropping and automatic noise computation. If set to no, cropping is based on the center of mass of the image (not brain) and automatic noise computation is just histogram based.
Revision 12
B-matrix rotation is added to PAR/REC conversion based on the angulation of the brain.
Guan's noise estimation code is included in the project but not yet active.
Revision 13
GUI for data properties file creation is included.
Some bugs corrected.
Guan's noise included in automatic noise estimation for Linux.