Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
The dataset viewer is not available for this split.
Server error while post-processing the rows. Please report the issue.
Error code:   RowsPostProcessingError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

CoMI course CT pack v1

A small, ready-to-use CT teaching set for the CoMI course (Computational Medical Imaging / AI in Medicine, University Hospital Bonn and University of Bonn). It is a re-packaged subset of the public TotalSegmentator dataset: 40 thorax–abdomen–pelvis CT scans with masks of 117 anatomical structures, resampled so a whole class can download it in a few minutes and run the labs on a laptop.

It is downloaded automatically by the course's data loader (comi_data.fetch_ct_pack()), which checks the size and SHA-256 below — you do not need to download it by hand.

Contents

comi_ct_pack_v1.zip (459,035,216 bytes) unpacks to ct_pack_v1/:

Path What
sXXXX/ct_3mm.nii.gz, sXXXX/labels_3mm.nii.gz 40 subjects: CT (int16 HU) and one label map with all 117 structures, 3 mm isotropic
sXXXX/ct_1.5mm.nii.gz, sXXXX/labels_1.5mm.nii.gz the same at the native 1.5 mm for 7 subjects (incl. the registration pair)
example/ the TotalSegmentator example CT used on the lecture slides
meta.csv the source metadata (age, sex, institute, study type, scanner, kVp, pathology) as released with the small subset
pack_subjects.csv the 40 subjects, their sex × age-band cell and their role (e.g. reg_fixed s1233 / reg_moving s0649)
volumes.csv, coverage.csv structure volumes (ml, from the native 1.5 mm masks) and which organs are completely in the field of view
manifest.json every file with shape, spacing and MD5; the label list (index → name)

Selection. From the 102 subjects of the official small subset: scans in which liver, spleen and both kidneys are complete; stratified by sex × age band (< 50, 50–64, 65–74, ≥ 75; up to 6 per cell; seed 2026). Result: 18 women, 22 men, age 30–91, Siemens 38 / GE 2, 5 institutes. There is no eligible woman under 50.

Processing. CT resampled to an exact 3 mm grid (linear); labels by majority vote in 2 × 2 × 2 blocks; 117 binary masks merged into one integer label map (index = TotalSegmentator v2 class id). No other changes.

Checksums. SHA-256 c0105626abfc21bcd96f92cf4590a90d23d019d9c968091f896c551e1a591937 · MD5 e28a4576082978b05ae076ba61bb95f0.

Source and licence

Derived from the TotalSegmentator dataset v2.0.1, small subset (Zenodo record 10047263, file Totalsegmentator_dataset_small_v201.zip), released under CC BY 4.0. This pack is shared under the same licence. Please cite the original work:

Wasserthal J, Breit HC, Meyer MT, Pradella M, Hinck D, Sauter AW, Heye T, Boll DT, Cyriac J, Yang S, Bach M, Segeroth M. TotalSegmentator: robust segmentation of 104 anatomic structures in CT images. Radiology: Artificial Intelligence 2023;5(5):e230024. doi:10.1148/ryai.230024

@article{wasserthal2023totalsegmentator,
  title   = {TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images},
  author  = {Wasserthal, Jakob and Breit, Hanns-Christian and Meyer, Manfred T. and Pradella, Maurice and Hinck, Daniel and Sauter, Alexander W. and Heye, Tobias and Boll, Daniel T. and Cyriac, Joshy and Yang, Shan and Bach, Michael and Segeroth, Martin},
  journal = {Radiology: Artificial Intelligence},
  volume  = {5}, number = {5}, pages = {e230024}, year = {2023},
  doi     = {10.1148/ryai.230024}
}

Intended use and caveats

  • Teaching only. Built for course exercises (windowing, segmentation metrics, registration, subgroup and scanner-shift analysis). Not for clinical use, and not a benchmark.
  • Metadata. The small subset's meta.csv differs from the full v2.0.1 release (ages differ by up to ±4 years in 79 of 102 rows; institute letters are re-coded). This pack keeps the small subset's own file.
  • Head coverage. 10 of the 40 scans include part or all of the head (non-zero skull / brain masks: s0727, s0763, s0777, s0970, s1123, s1124, s1206, s1369, s1372, s1384); a 3D rendering may show facial features. These scans are identical in content to the public source release; do not attempt to re-identify anyone.
  • Coarse labels. Pathology labels are coarse (several "unclear"); do not use them as diagnostic targets.

Course

CoMI course and notebooks: Albarqouni Lab, University Hospital Bonn — https://albarqouni.github.io/teaching/comi-lecture/

Downloads last month
35