I just added the rCST image

From: Lauren O'Donnell
In case this helps: it's probably that the tensors are not rotated into
RAS space before saving the file.

From: Johnson, Hans J
The file should be in LPS orientation to match the dicom data that it
came from. If conversion needs to Occur, it needs to happen in Slicer.
This does sound like the issue.
Hans

From Johnson, Hans J :
Ryan: Can you please take a first pass at looking at this? I’m really
swamped this week.
Thanks,
Hans

From Ryan - Jan 28, 2014:

Hi Hans,

The problem was in vtk_writer.cc, here is the fix:

================
https://github.com/pnlbwh/ukftractography/blob/master/ukf/vtk_writer.cc

{
eigenVec1 <<
state[_tensor_space (tensorNumber - 1) + _p_m1], state[_tensor_space
(tensorNumber - 1) + _p_m2],
state[_tensor_space * (tensorNumber - 1) + _p_m3];

// Perform ijk->RAS transform on eigen vectors
eigenVec1 = _sizeFreeI2R * eigenVec1;

// Renormalize eigenvectors
eigenVec1.normalize();

}

// Compute the diffusion matrix in RAS coordinate system
// The transformed matrix is still positive-definite
mat33_t I;
I <<
1, 0, 0,
0, 1, 0,
0, 0, 1;

D = ((_p_l1 - _p_l2) eigenVec1 eigenVec1.transpose + _p_l2 I) 1e-6;

From Yogesh - Jan 28. 2014:

Hans,

The problem was that the 2nd and 3rd eigenvectors were not necessarily orthogonal to each other or to the main eigenvector.
The algorithm only estimates the main eigenvector and the other 2 can be easily computed from that one.

This code was in the glyphs section (so, it was not really part of the estimation algorithm, but used only for display).
Hence, Ron was seeing plates instead of ellipsoids in the display.

I hope that the solution we came up with should fix this issue.

Best,
Yogesh

Fixed by https://github.com/pnlbwh/ukftractography/commit/53b72b9a9623a2f02f85208f2cbba501c0ef3ee7