Improving Neuropathological Reconstruction Fidelity via AI Slice Imputation - UW Dataset

Overview

This record contains one of the datasets used in the manuscript "Improving
Neuropathological Reconstruction Fidelity via AI Slice Imputation"
(https://arxiv.org/abs/2602.00669). It comprises 3D reconstructions of
post-mortem brain dissection photographs from the University of Washington
(UW) cohort, organised in BIDS layout, together with the isotropic volumes
derived from them.

Contents and Format

28 specimens, each at three coronal slab spacings: 4 mm as acquired, and
8 mm and 12 mm obtained by subsampling the 4 mm photographs by factors of
2 and 3 respectively. 

The dataset is organised following the BIDS specification. Reconstructions with our proposed imputation method, segmentation and surface reconstructions are located in the /derivatives folder.

All image files are NIfTI-1 (.nii.gz)

Related Links

Source photographs: previously released with Gazula et al. (2024), eLife 12,
RP91398, doi:10.7554/eLife.91398 [accession and URL]. This record
redistributes that material in BIDS layout and adds the derived isotropic
volumes. Please cite both the original release and this record.

Analysis code: https://github.com/marinacrsp/photo-recon

Method implementation, distributed with FreeSurfer:
https://surfer.nmr.mgh.harvard.edu/fswiki/mri_3d_photo_recon