search

m-qiang / CoTAN

[MICCAI 2023] Conditional Temporal Attention Networks for Neonatal Cortical Surface Reconstruction
Apache License 2.0
10 stars 1 forks source link

wrong registration of dHCP T2 #1

Open romainVala opened 11 months ago

romainVala commented 11 months ago

Hello thanks for providing this code, I am very exited to test it. (nice work !)

I just tried on a dHCP subject, but the registration.py you provide failed to find the appropriate scaling ... why did you choose an adult size mni template ?

I am surprise since you train and evaluated on dHCP dataset. did you register the sub-CC01234BN14_ses-155230_T2w.nii.gz to the template in your code base only with the register.py script ?

Many thanks for your help Romain

m-qiang commented 11 months ago

Hi Romain,

Thanks very much for your interest in our work.

The register.py script is based on ANTsPy and may fail on some subjects. My suggestion is to change the _affmetric parameter in ants.registration and re-run the script for the failed subjects. Alternatively, you can run the affine registration between the binary brain masks instead of the MRI intensity. Hope these can help.

We chose the MNI-152 template so that CoTAN can be easily generalized and trained on adult datasets as well. Sorry for the inconvenience caused.

Qiang