Hi @zivy I think the question was not clear. I am using the transform matrix learned using least squares on MR image. So far, I haven’t used registration. It’s just transforming the image using the affine(4x4) matrix to see how transformation works. I am using the affine matrix

```
array([[ 0.94969925, -0.07489753, -0.09085982, 2.34559412],
[ 0.0128945 , 0.97870216, -0.01826539, -1.60140903],
[ 0.04813273, -0.02432601, 1.06080813, -5.17650667],
[ 0. , 0. , 0. , 1. ]])
```

to transform the MR image and after resampling using the above matrix the get this resultant image. When I use the inverse transform on the point `(50.16, -31.71, 34.84)`

, the result is `(47.81, -30.10, 40.01)`

. However, when I see compare the first point in the source image and the second point in the transformed image, they seem to be different. I was expecting them to locate to similar place across the images since the same matrix is used in transforming from one point to the other.

I’ve used resample function from this notebook.