This module turns the tensor eigenvectors in a tensor image so that the tensors are oriented according to the world coordinate system of the image, and not according to the local image coordinate system.

This is needed by modules like TensorTractography which do fiber tracking in world coordinates. - They don't do the rotation internally so that the transformation can be cached and used for other purposes, e.g. color-coded visualization.


Normally this module detects the needed rotation from the voxel-to-world-matrix of the input image. This can be overwritten by a user-supplied matrix.


This module supports tensor vectors supplied in the u-dimension of an image. The size in the u-dimension must be a multiple of 3 (every 3 values are treated as one directional vector). The output image has the same format as the input image.


The example network supplied with this module demonstrates why TensorTransform is needed when the tensor image is not aligned with the world coordinate system.


Default Panel


Input Fields


name: input0, type: Image

The tensor image.

Output Fields


name: output0, type: Image

The transformed tensor image.

Parameter Fields

Visible Fields

Use external matrix

name: useExternalMatrix, type: Bool, default: FALSE

Set this to true if you don't want to use the voxel-to-world-matrix of the input image but want to supply your own.

Transformation Matrix

name: transformation, type: Matrix, default: 1 0 0 0, 0 1 0 0, 0 0 1 0, 0 0 0 1

Set this matrix to override the voxel-to-world-matrix of the input image and set useExternalMatrix to true.