Dog brain atlas generated via spatially constrained spectral clustering

This repository provides a multi-scale functional parcellation atlas of the domestic dog brain derived from resting-state fMRI using a spatially constrained spectral clustering algorithm. Parcellations were generated from an awake rs-fMRI dataset of 27 pet dogs and registered to the stereotaxic, breed-averaged, symmetric T2-weighted canine template space described by Nitzsche et al. (2019), then resampled to 2 mm isotropic resolution. The dataset consists of whole-brain segmentations stored as 3D NIfTI label maps at multiple resolutions ranging from 20 to 300 parcels in steps of 20 (N = 20, 40, …, 300). All segmentations are defined in Nitzsche et al. (2019) template space at 2 mm isotropic voxel size and share its orientation and dimensions; voxel values encode integer parcel IDs, with 0 denoting background/non-brain voxels. The construction of the atlas, as well as its evaluation in terms of cross-dataset and within-participant reproducibility (Dice coefficient, Adjusted Rand Index), within-parcel functional homogeneity, and concordance with the Johnson et al. (2020) anatomical atlas, is detailed in Hernandez-Perez et al., Dog brain atlas generated via spatially constrained spectral clustering (preprint/manuscript linked in this record). These parcellations are intended for use in canine fMRI connectivity and task-based studies, ROI definition, and comparative canine–human neuroimaging. When using this resource, please cite the accompanying article and this Zenodo dataset (DOI) and acknowledge that the segmentations are defined in Nitzsche et al. (2019) template space at 2 mm isotropic resolution.