To create an MRI-based 3D brain atlas for the dwarf cuttlefish, ex vivo magnetic resonance imaging of 8 adult dwarf cuttlefish brains (4 males, 4 females) at 50 µm isotropic resolution was performed. Deep learning techniques were used to improve manually-annoted brain masks that were then used to extract the brain regions for subsequent diffeomorphic registration using the ANTs toolbox. Prior neuroanatomical descriptions were used to guide the annotation of brain lobes, which were co-registered to create a merged, annotated, brain template representative of the 8 cuttlefish subjects used in this study. This repository contains sourcecode for boldened steps outlined below used to create the MRI-based 3D brain atlas:
Step 0. NIFTI conversion, bias field correction, isotropic upsampling to 50 micron resolution
Step 1. Manual Brain Mask Annotation
Step 2. Brain Mask Improvement using DL-BET (see https://github.com/SAIL-GuoLab/DL-BET)
Step 3. Brain mask used to extract brain from MRI scan for each animal
Step 4. Diffeomorphic Registration using ANTs toolbox to build MRI atlas
Step 5. Segmentation of Brain ROIs
Step 6. Transformation of segmented ROI labels to template space
Step 7. Majority voting of labels in template space for MRI atlas