Mouse thorax segmentation dataset

This repository contains reference segmentations of the heart, spinal cord, right lung and left lung for native and contrast-enhanced mouse CT images. The CTs were drawn from a publicly available preclinical micro-CT database¹. Annotations are provided for the entire native CT dataset (140 images) and for a subset (35 images) of the contrast-enhanced CT dataset. Annotations by a second observer are also available for 35 native CTs and contrast-enhanced CTs.

This dataset was used to train a nnU-Net 3d_fullres model for automated mouse thorax segmentation. The pre-trained model can be downloaded from https://doi.org/10.5281/zenodo.5786839 and more details about the dataset and training can be found in the publication. In case you find these annotations useful for your research, please cite the original work:

Malimban, J., Lathouwers, D., Qian, H. et al. Deep learning-based segmentation of the thorax in mouse micro-CT scans. Sci Rep 12, 1822 (2022). https://doi.org/10.1038/s41598-022-05868-7

¹Rosenhain S, Magnuska Z A, Yamoah G G, Rawashdeh W A, Kiessling F and Gremse F 2018 A preclinical micro-computed tomography database including 3D whole body organ segmentations Online: https://springernature.figshare.com/collections/A_preclinical_micro-computed_tomography_database_including_3D_whole_body_organ_segmentations/4224377/1