Slicer transformation result is different from my vtk codes

Support
1

Recently I used the Transform module in Slicer to apply a linear transformation. Let’s say the translation is [10, 20, 30]. And the model moved to the correct position.

However, when I used my own vtk codes below, the result is not correct.

vtkSmartPointer<vtkTransform> transform = vtkSmartPointer<vtkTransform>::New();
transform->Translate(10, 20, 30);
vtkSmartPointer<vtkTransformPolyDataFilter> transformFilter = vtkSmartPointer<vtkTransformPolyDataFilter>::New();
transformFilter->SetInputData(mesh);
transformFilter->SetTransform(transform);
transformFilter->Update();

First I was thinking it is because of the coordinate system difference (RAS and LPS), so I flipped the two axes:

translationMatrix->SetElement(0, 0, -1.0); 
translationMatrix->SetElement(1, 1, -1.0);

However, it is still not correct. Did I miss some steps used in Slicer? Thanks for any suggestions!

2

You would actually need to change the values of the translations, not just the main diagonal. But in general you should be applying a matrix to do the transform, something like this:

3

Hi pieper, thanks for the reply!
Can I ask what do you mean by “not just the diagonal”?

I have tried to apply the two matrix below but the error still exists:

| 1 | 0 | 0 | 10 |
| 0 | 1 | 0 | 20 |
| 0 | 0 | 1 | 30 |
| 0 | 0 | 0 | 1 |

and

| -1 | 0 | 0 | 10 |
| 0 | -1 | 0 | 20 |
| 0 | 0 | 1 | 30 |
| 0 | 0 | 0 | 1 |

4

This operation from your post only changes the two elements on the diagonal:

What you need to do is multiply your translation matrix by a matrix that is identity but with -1 in the 0,0 and 1,1 locations. The effect in this case will be to negate the first two translation values.

1 Like
5

Hi pieper, I have multiplied the two matrix:

the first is [1, 0, 0, 10; 0, 1, 0, 20; 0, 0, 1, 30; 0, 0, 0, 1]
the second is [-1, 0, 0, 0; 0, -1, 0, 0; 0, 0, 1, 0; 0, 0, 0, 1]
the result is [-1, 0, 0, 10; 0, -1, 0, 20; 0, 0, 1, 30; 0, 0, 0, 1]

But the error still exists.

I was thinking if it is because my mesh model is not in the center of the bounding box. So I output the center coordinates of the mesh before and after the transformation. It’s weird that the coordinates after transformation are not (-10, -20, 30).

My codes below:

vtkSmartPointer<vtkMatrix4x4> translationMatrix = vtkSmartPointer<vtkMatrix4x4>::New();
translationMatrix->SetElement(0, 3, 10); 
translationMatrix->SetElement(1, 3, 20); 
translationMatrix->SetElement(2, 3, 30); 
translationMatrix->SetElement(0, 0, -1.0); 
translationMatrix->SetElement(1, 1, -1.0); 
transform->SetMatrix(translationMatrix);
vtkSmartPointer<vtkTransformPolyDataFilter> transformFilter = vtkSmartPointer<vtkTransformPolyDataFilter>::New();
transformFilter->SetInputData(mesh);

transformFilter->GetOutput()->GetCenter(center); // the center is (0, 0, 0)

transformFilter->SetTransform(transform);
transformFilter->Update();

transformFilter->GetOutput()->GetCenter(center2); // the center is not (-10, -20, 30)
6

There’s more information here: Coordinate systems - Slicer Wiki

You can also search google for LPS and RAS to get more examples.

1 Like
7

You may also find this script useful that shows how to convert ITK transform file to Slicer transform.

1 Like
8

Indeed, even in python image direction/coordinates are different. I tried this code and worked. you may have a look and find out how to convert coordinates in c++:
image=sitk.ReadImage(image)
dirs = np.zeros([3,3])
image.GetIJKToRASDirections(dirs)
dirs=np.multiply(dirs,np.array([[-1,-1,-1],[-1,-1,-1],[1,1,1]]))
new_dirctions=np.reshape(dirs,(1,9)).tolist()
new_origin=label.GetOrigin()

1 Like
9

This worked! Thank you!