Here we present a dataset of synthetic, high-resolution MRI scans with labels for the hypothalamic subregions, which can be used to develop segmentation methods based on synthetic images. The dataset is composed of label maps built from publicly available ultra-high resolution ex vivo MRI from 10 whole hemispheres, using a combination of manual labels for the hypothalamic nuclei and automated segmentations for the rest of the brain, obtained with unsupervised clustering. The final label maps are mirrored to simulate an entire brain. We will also provide the pre-processed ex vivo scans, once the dataset supports an extension to other structures by manually tracing them.
The original ex vivo images are public available on DANDI Archive (https://dandiarchive.org/dandiset/000026)The codes to generate new label maps can be found at https://github.com/liviamarodrigues/hsyn... along with the codes for the automated hypothalamus segmentation method.
This dataset is under CC-BY license.