Page MenuHomePhabricator

Tensor recon with negative eig crush
Closed, ResolvedPublic


The procedure of Tensor reconstruction with negative eigenvalue correction is crushing for some dwi files. This affects also some parts in DKFZ MBI code.
I ll try to figure out what is wrong but I am attaching problematic data set if anyone wants to look at it as well.

Event Timeline

Can you add the problematic dataset and describe how to produce this error?

New remote branch pushed: bug-16194-tensor-eigenvalue-input-fix

I examined the problem and it seems to be the deep copy method that is not working in a correct way for a large images.(the size of the image for which method is working correctly is different among diffrent machines so I suspect that it is memmory alocation problem) The best and most efficient idea of solution is to resign from deep copy. I've implemented the code so I wil try to do that.

New remote branch pushed: bug-16194-merge

I will merge this branch and leave some of the old code commented for the time being

[cd5c63]: Merge branch 'bug-16194-cleanUp'

Merged commits:

2013-10-25 11:35:28 Thomas Van Bruggen [c4ba1e]
cleaned code

2013-10-25 11:31:08 Thomas Van Bruggen [25668d]
Merge branch 'bug-16194-merge' into bug-16194-cleanUp

2013-10-18 14:47:58 Thomas Van Bruggen [821347]
added doxygen info

2013-10-15 17:17:27 Thomas Van Bruggen [ced3a8]
provided a GetGradientImagePointer method

2013-10-15 17:16:04 Thomas Van Bruggen [a168b9]
removed another source of memory corruption

2013-10-15 16:20:23 Thomas Van Bruggen [322114]
fixed some out of bound operations on vectors

2013-10-15 14:23:12 Thomas Van Bruggen [433369]
Merge commit '8b7fe4ebe90275935c9becf7b8a4b0fa7c2c8796' into bug-16194-merge

2013-10-15 14:14:18 Thomas Van Bruggen [03c606]
Merge branch 'bug-16194-tensor-eigenvalue-input-fix' into bug-16194-merge

2013-10-15 03:20:39 Franz Binczyk [8b7fe4]
remove of deep copy structures

2013-09-30 08:20:25 Franz Binczyk [91c311]
fixes comparison issue