---
license: odc-by
task_categories:
- image-segmentation
tags:
- medical
- ct
- lung
- covid
- long-covid
- segmentation
pretty_name: LongCIU
dataset_info:
  features:
  - name: slice_index
    dtype: int32
  - name: split
    dtype: string
  - name: image
    dtype: image
  - name: mask_staple
    dtype: image
  - name: overlay_staple
    dtype: image
  - name: mask_annotator1
    dtype: image
  - name: mask_annotator2
    dtype: image
  - name: mask_annotator3
    dtype: image
  - name: ggo_pixels_staple
    dtype: int32
  - name: consolidation_pixels_staple
    dtype: int32
  splits:
  - name: train
    num_bytes: 11546931
    num_examples: 30
  - name: val
    num_bytes: 7131799
    num_examples: 20
  - name: test
    num_bytes: 15555193
    num_examples: 40
  download_size: 34305458
  dataset_size: 34233923
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
  - split: test
    path: data/test-*
---

# Long COVID Iowa-UNICAMP (LongCIU)

Long COVID Iowa-UNICAMP dataset: 90 axial chest CT slices with manual
segmentation of ground glass opacities (GGO) and parenchymal consolidations,
independently annotated by three expert annotator pairs. Provides both the
three individual annotations and a STAPLE consensus recommended as ground
truth.

## Overview
- **Modality:** CT (axial)
- **Body part:** Lung (GGO / consolidation in Long COVID patients)
- **Samples:** 90 axial CT slices (stacked along axis 2 of a single NIfTI volume)
- **Classes:** 0 = background, 1 = GGO, 2 = consolidation

## Files
| File | Description |
|---|---|
| `longciu_img.nii.gz` | Image volume `(512, 512, 90)` — each slice is an independent patient |
| `longciu_STAPLE_tgt.nii.gz` | **Recommended ground truth** (STAPLE consensus of the 3 annotators) |
| `longciu_1_tgt.nii.gz` | Annotator pair 1 mask (for inter-rater study) |
| `longciu_2_tgt.nii.gz` | Annotator pair 2 mask (for inter-rater study) |
| `longciu_3_tgt.nii.gz` | Annotator pair 3 mask (for inter-rater study) |
| `longciu_medpseg_output.nii.gz` | MedPSeg baseline model prediction (for comparison) |
| `longciu_splits.json` | Train / val / test split indices over slice axis |
| `staple_pub_stats.json` | STAPLE statistics published with the paper |
| `README.txt` | Original dataset README from the authors |

## Splits (indices into slice axis of `longciu_img.nii.gz`)
- **train:** 30 slices
- **val:** 20 slices
- **test:** 40 slices

## License
Open Data Commons Attribution License (ODC-By) v1.0.

## Citation
Carmo, D. S., Pezzulo, A. A, Villacreses, P. A., Comellas A. P., Eisenbeisz, M. L.,
Anderson, R. L., Van Dorin, S. E., Rittner, L., Lotufo, R. A., Gerard, S. E.,
Reinhardt, J. M., Comellas, A. P. (2024). Long COVID Iowa-UNICAMP [dataset].
University of Iowa. https://doi.org/10.25820/data.007301

Paper (Scientific Data, Nature, 2025):
https://www.nature.com/articles/s41597-025-04709-2

Source code: https://github.com/MICLab-Unicamp/LongCIU