Is this in the nightly or in the 4.5 release (or both?)

Actually is in the nightly. There was no selector for nightly. Last stable version of Slicer works fine

Regression most likely introduced by the recent update of ITK/BRAINS.

I was able to reproduce the problem.
Note: I edited the description to provide more details.

To reproduce, open the python interpreter and copy the following:

<pre>
import SampleData
from SampleData import SampleDataLogic
logic = SampleDataLogic()

fixed = logic.downloadMRBrainTumor1()
moving = logic.downloadMRBrainTumor2()

slicer.util.selectModule("BRAINSFit")

widget = slicer.modules.brainsfit.widgetRepresentation()
cmdLineNode = widget.currentCommandLineModuleNode()

cmdLineNode.SetParameterAsNode("fixedVolume", fixed)
cmdLineNode.SetParameterAsNode("movingVolume", moving)

outputTransform = slicer.vtkMRMLLinearTransformNode()
slicer.mrmlScene.AddNode(outputTransform)

cmdLineNode.SetParameterAsNode("linearTransform", outputTransform)

outputVolume = slicer.vtkMRMLScalarVolumeNode()
slicer.mrmlScene.AddNode(outputVolume)

cmdLineNode.SetParameterAsNode("outputVolume", outputVolume)

cmdLineNode.SetParameterAsBool("useRigid", True)

widget.apply()
</pre>

Commit r24875 introduced the regression.

See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&amp;revision=24875

Will work on a fix later.

Without this commit .. i get the following output:

Note: I added few extra print statements

SetVTKBSplineParametersFromITKGeneric - bsplineItk NULL
SetVTKBSplineParametersFromITKGeneric - bsplineItk NULL
SetVTKLinearTransformFromITK - transformItk_LPS:VersorRigid3DTransform
SetVTKLinearTransformFromITK - dlt:VersorRigid3DTransform
CreateVTKTransformFromITK - SetVTKLinearTransformFromITK conversionSuccess:1

The issue is that the corresponding static cast didn't account for all type of linear transform.

Issue fixed in the following topic: https://github.com/jcfr/Slicer/tree/4139-fix-crash-vtkITKTransformConverter

Now integrating the python snipped in a test and will integrate.

Fix integrated in r24902
See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&amp;revision=24902

For future reference, a simpler version of the python code has been
used in the associated self-test:

<pre>
import SampleData
from SampleData import SampleDataLogic
sampleDatalogic = SampleDataLogic()

fixed = sampleDatalogic.downloadMRBrainTumor1()
moving = sampleDatalogic.downloadMRBrainTumor2()

outputTransform = slicer.vtkMRMLLinearTransformNode()
slicer.mrmlScene.AddNode(outputTransform)

outputVolume = slicer.vtkMRMLScalarVolumeNode()
slicer.mrmlScene.AddNode(outputVolume)

parameters = {
'fixedVolume' : fixed,
'movingVolume' : moving,
'linearTransform' : outputTransform,
'outputVolume' : outputVolume,
'useRigid' : True
}
cmdLineNode = slicer.cli.runSync(slicer.modules.brainsfit, parameters=parameters)
</pre>