search

cassianobecker / dnn

MIT License
4 stars 0 forks source link

Add registration and warping to hcp_data.py #21

Closed lindenmp closed 4 years ago

lindenmp commented 4 years ago

The first version of process_dti only runs dtifit and then builds the tensor image using these outputs which are in subject's T1 space. Subject's dti outputs need to be in a standard space otherwise the DNN will be training on voxels that are not spatially matched across subjects.

lindenmp commented 4 years ago

I have written code that: 1) takes outputs from dtifit and converts them to be ANTs compliant 2) registers FA map to an FA template and moves the DTI image as well (ANTs) 3) reorients the warped DTI image (ANTs) 3) re-extracts the eigenvectors and eigenvalues from the warped/reoriented DTI (ANTs) 4) masks the eigenvectors and eigvenlues using an FA mask that matches the FA template

Need to include an ANTs install for each person We can change the template/mask combination any time we want, let's discuss.