Attention-driven multi-channel deformable registration of structural and microstructural neonatal data

Author: Irina Grigorescu | irina[dot]grigorescu[at]kcl[dot]ac[dot]uk

This is the companion code for the following publications:

• Grigorescu, I. et al. (2021). Uncertainty-Aware Deep Learning Based Deformable Registration. UNSURE 2021. LNCS (Springer)

  • doi.org/10.1007/978-3-030-87735-4_6

• Grigorescu, I. et al. (2022). Attention-Driven Multi-channel Deformable Registration of Structural and Microstructural Neonatal Data. PIPPI 2022. LNCS (Springer)

  • doi.org/10.1007/978-3-031-17117-8_7

Example train/validation/test file (one subject / line, see example_csv_files folder)

t2w,lab,dti,fa,ga,as,gender
subj1_T2_img.nii.gz,subj1_tissue_labels.nii.gz,subj1_DTI.nii.gz,subj1_FA.nii.gz,41.0,41.14285714,Male
subj2_T2_img.nii.gz,subj2_tissue_labels.nii.gz,subj2_DTI.nii.gz,subj2_FA.nii.gz,40.14285714,40.28571429,Female

Also, bare in mind that the code currently expects the suffix _T2_img.nii.gz for the images, the suffix _tissue_labels.nii.gz for the tissue maps, the suffix _DTI.nii.gz for the diffusion tensor images, or _FA.nii.gz for the fractional anisotropy maps. It also expects _ic_m.nii.gz for the internal capsule label.