This is the dataset of the vertebra segmentation challenge of the CSI 2014 workshop that was held in conjunction with MICCAI 2014. 1-10: Training set, scans of 10 young adult (16-35 years old) 11-15: Test set, scans of 5 young adult (20-35 years old) 16-20: Test set, scans of 5 patients with vertebral compression fractures Scans were acquired at the Department of Radiological Sciences, University of California, Irvine, School of Medicine and were published under the ODC Public Domain Dedication and License on SpineWeb (datasets 2 and 15). The dataset and challenge is further described in this publication: A multi-center milestone study of clinical vertebral CT segmentation The data that is published here has been normalized: Voxel values are Hounsfield values and have been clipped to [-1000, 3095] Image origin has been set to 0,0,0 Image orientation has been standardized to RAI orientation Small islands and other obvious mistakes have been removed Segmentation masks have been smoothed with a 2x2x2 median filter Vertebrae have been anatomically labeled (8 = T1, 9 = T2, ..., 24 = L5) Because not always all visible vertebrae were segmented in the original data, only segmentations of the thoracic and lumbar vertebrae have been retained License This dataset is released under the Open Data Commons Attribution License (which the original license allows me to do). When using this dataset for publication of any kind, please reference the following paper to meet the attribution requirement: Yao J, Burns JE, Forsberg D, Seitel A, Rasoulian A, Abolmaesumi P, Hammernik K, Urschler M, Ibragimov B, Korez R, Vrtovec T, Castro-Mateos I, Pozo JM, Frangi AF, Summers RM, Li S. A multi-center milestone study of clinical vertebral CT segmentation. Comput Med Imaging Graph. 2016; 49:16-28. doi: 10.1016/j.compmedimag.2015.12.006. There is no need to reference this upload, referencing the original authors is sufficient. Notes The dataset contains 2 cases with only 4 lumbar vertebrae in which L5/S1 has not been segmented (i.e., label 25 is missing). The resolution and segmentation quality of the diseased cases (16-20) is quite low.