[Brains-users] Problems with tissue class models and linux paths

[Kevin Berne]

Thank you all for your quick response!

It really seems like the csf/grey/white plugs themselves are the problem, since their volume indeed is zero. As recommended, I have tried to brighten up the T1 set, sadly without any success. Besides me other people here working with Brains2 have noticed that the look of an image set within Brains2 does not necessarily correspond with the look of it in other programs, MRIcro for example. The image math option and adding up an image to itself tends to have different levels of success, as I have experienced it, since some images sometimes get much brighter, whilst other images do not seem to react in any way. With the set I am currently using, it does not react at all. We here have pondered the question wether there might indeed be too bright a spot within the structure of the image itself affecting the dynamic range, but this seems unlikely since none is visible and the image set looks fine when opened with a program like MRIcro. Nevertheless, the fact that images respond so !
 differently to the image math add process hints that this may still have some influence. Generally, the scans do not seem too noisy, so this should also not be the problem. Concerning the youth of the subjects, I have to point out that we are talking abut people between 16 and 40 years of age, which I at least hope is old enough to be segmented by Brains2.
So I think I have determined that the critical point is the plug-picking, which always results in zero volumes, disabling any further computations.
To complete my documentation, I am using the run segment/brains2 option, the resampled T1 (image math added, if possible) and T2 images and the path for saving the plugs is the standard 10_ACPC directory, all of which is automatically recognised by Brains2 (all but the T2, which I have to enter manually). Furthermore, I am using manually traced l_seg_vb.zroi and r_seg_vb.zroi, the bin/talairach/brain_box and a manually created talairach.bnd to fill the fields of thle dialog box. After completing the calculation process I receive the message "failed to create tissue class model" with the following Cygwin shell display:

SIMPLE ERROR: error accessing non-vacous array tree: 0-th PLUG MASK WAS EMPTY.
Module: /home/vince/development/brains2/src/iplObject/obTissueClass.cxx Function: bk_generateClassModel Line: 4107 Code: 123456789

I have tried to use just the T1 or to choose  the image option as the blood method, but nothing resulted in working plugs.
If anyone has experienced this kind of difficulty before, I would be happy to receive any suggestions how to deal with it.

Additionally, we here have experienced a kind of cut-off effect when resampling images to MR5 and then turning them, which often results in the posterior portion of the head and sometimes brain to be cut off, which should give the segmentation and the neural net a really hard time later on, we guess. I have tried to resample the images to larger values than 256x256x192, like for example 356x356x292, which resulted in a perfect image without any cut-offs. We just wondered wether this will have any impact on the calculated volumes later. We would appreciate it if you could answer that question for us, too.

Thank you again for your time!

All the best,

Kevin Berne
______________________________________________________________________
XXL-Speicher, PC-Virenschutz, Spartarife & mehr: Nur im WEB.DE Club!		
Jetzt gratis testen! http://freemail.web.de/home/landingpad/?mc=021130