Find masked data voxel coordinates in the original Nifiti structure

, ,
1

Dear all,

I computed a vector that each element in the vector corresponding to each voxel in my mask. I want to find out a way to put those values back to Nifiti data structure in the original mask file space.

The total number of elements in the vector is exactly the same as the number of voxels in my mask. I used “nib.load” from the nibabel package to get the data for the whole brain time series. I load the mask file using “image.load_img” from the nilearn package, then mask the whole brain time series using “apply_mask” from the nilearn package. After masking, I got a 2D matrix with each column in the matrix corresponding to a voxel in the mask region. Then I did some computations to this 2D matrix that reduced this matrix into a vector, which the index of each element correlates with the original 2D matrix column index.

I am trying to find whether there is a command that allows me to find the coordinates in Nifiti structure data for each column (that corresponds to the voxel) in the 2D matrix (that I got from “apply_mask”).

I also tried the command “new_img_like”, but failed as I don’t know how to find the element coordinates to the original 3D numpy array.

Thanks in advance!

2

Hi, you can use NiftiMasker.transform and NiftiMasker.inverse_transform: nilearn.maskers.NiftiMasker - Nilearn

3

Dear Jeromedockes,

Thank you so much to provide this fuction. It is working with my vector. It put back to my mask file space and inside the mask.

Though I failed to varify that using NiftiMasker.inverse_transform put every element of my vector in the right place. The vector is generated from apply_mask. If you could provide insigts about validate the result. That is deeply appreciated.

Thank you!