File size: 9,321 Bytes
5f43746 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 | ---
license: cc-by-4.0
tags:
- medical-imaging
- anatomy
- ultrasound
- computed-tomography
- image-classification
task_categories:
- image-classification
---
# OrganScan training corpus — recipe and provenance
This repository publishes the **recipe**, not the pixels.
Medical imaging corpora carry licences that differ source by source, and several forbid
redistribution outright. More importantly, frames rendered from DICOM can carry burned-in patient
identifiers even after tag-level de-identification. Rather than ship images that are easy to misuse,
this repo contains everything needed to rebuild the corpus from the pinned upstream releases.
## What is here
| File | Purpose |
|---|---|
| `sources.json` | Every source with its pinned revision, checksum, licence, access status, and training decision |
| `anatomy-tiers.json` | The three-level taxonomy, the organ -> region parent map, and which classes are actually trained |
| `anatomy-mappings.json` | Explicit decision for every upstream class and DICOM tag value; no default branch |
| `canonical-frame-row.schema.json` | The row schema every source is normalised to |
| `preprocessor.json` | Exact pixel handling, so a rebuild is byte-comparable |
| `build-manifest.json` | Counts by split, source, modality, organ — and **patients**, not just frames |
## Corpus shape
| Split | Frames | Patients |
|---|---:|---:|
| train | 82,354 | see `build-manifest.json` |
| validation | 10,280 | see `build-manifest.json` |
| test | 15,395 | see `build-manifest.json` |
| Modality | Frames |
|---|---:|
| CT | 83,318 |
| MR | 1,880 |
| US | 14,907 |
| XR | 7,924 |
## Why patient counts are printed next to frame counts
300 consecutive frames from one cine loop carry roughly the information of one sample. MedMNIST
OrganAMNIST is 58,830 images from **201** CT scans — 293 images per patient. Training on the raw
count inflates every metric and teaches the model scanner identity. `sampling_policy` in
`sources.json` caps frames per patient before anything is written, and the builder enforces it.
## Redistribution status by source
| Source | Licence | Access | Redistributed here |
|---|---|---|---|
| `medmnist-plus-organamnist-224` | CC-BY-4.0 | open | recipe only |
| `medmnist-plus-organcmnist-224` | CC-BY-4.0 | open | recipe only |
| `medmnist-plus-organsmnist-224` | CC-BY-4.0 | open | recipe only |
| `medmnist-plus-breastmnist-224` | CC-BY-4.0 | open | recipe only |
| `medmnist-plus-chestmnist-224` | CC-BY-4.0 | open | recipe only |
| `msu-abdominal-ultrasound` | CC-BY-4.0 | open_manual_browser_download | recipe only |
| `fetal-planes-db` | CC-BY-4.0 | open | recipe only |
| `totalsegmentator-ct-sample` | CC-BY-4.0 | open | recipe only |
| `totalsegmentator-ct-v300` | CC-BY-4.0 | open | recipe only |
| `amos22` | CC-BY-4.0 | open | recipe only |
| `medical-segmentation-decathlon` | CC-BY-SA-4.0 | open_but_unresolved_host | recipe only |
| `rocov2` | CC-BY-NC-4.0 | open | **no** — licence forbids |
| `radimagenet` | undefined | registration_and_agreement_required | **no** — licence forbids |
| `camus` | undefined | registration_required | **no** — licence forbids |
| `echonet-dynamic` | Stanford EchoNet-Dynamic Dataset Research Use Agreement | signed_research_use_agreement | **no** — licence forbids |
| `echonet-lvh` | Stanford EchoNet-LVH Dataset Research Use Agreement | signed_research_use_agreement | **no** — licence forbids |
| `busi-original` | undefined | mirror_only | **no** — licence forbids |
| `tn3k` | undefined | third_party_drive | **no** — licence forbids |
| `ddti` | undefined | source_site_unavailable | **no** — licence forbids |
| `mmotu` | undefined | third_party_drive | **no** — licence forbids |
| `pocus-covid19-ultrasound` | mixed per-file (CC-BY-4.0, CC-BY-NC-4.0, all-rights-reserved) | open | **no** — licence forbids |
| `biomedclip` | MIT tag, but the model card restricts use | open | **no** — licence forbids |
| `usfm` | CC-BY-NC-4.0 | third_party_drive | **no** — licence forbids |
## Rebuild it
```bash
git clone https://github.com/palarnab/organscan && cd organscan
python downloaded-training-data/download.py # pinned revisions, checksum-verified
python scripts/build_frame_dataset.py
python scripts/validate_frame_dataset.py # leakage, mask coherence, coverage
```
The build fails rather than silently dropping data when an unmapped upstream class appears, and
validation fails on patient leakage across splits, identical frames in two splits, or a weak label
found in an evaluation split.
## Attribution
Every source below requires credit, and every one of them was modified — sliced from volumes or cine
loops, windowed, resized to 224×224, capped per patient, and relabelled into the OrganScan taxonomy.
Full per-source detail, including the exact modifications, is in `sources.json` under each entry's
`attribution` block.
| Source | Creators | Licence | Frames used |
|---|---|---|---:|
| [TotalSegmentator CT dataset v3.0.0](https://zenodo.org/records/22688904) | Jakob Wasserthal et al. | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 61,902 |
| [AMOS22: A Large-Scale Abdominal Multi-Organ Benchmark for Versatile Medical Image Segmentation](https://zenodo.org/records/7262581) | Yuanfeng Ji et al. | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 11,480 |
| [FETAL_PLANES_DB: Common maternal-fetal ultrasound images](https://zenodo.org/records/3904280) | Xavier P. Burgos-Artizzu et al. | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 10,748 |
| [MedMNIST v2 (ChestMNIST, 224px)](https://zenodo.org/records/10519652) | Jiancheng Yang et al. | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 7,924 |
| [MedMNIST v2 (OrganSMNIST, 224px)](https://zenodo.org/records/10519652) | Jiancheng Yang et al. | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 4,000 |
| [MedMNIST v2 (OrganAMNIST, 224px)](https://zenodo.org/records/10519652) | Jiancheng Yang et al. | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 3,914 |
| [MedMNIST v2 (OrganCMNIST, 224px)](https://zenodo.org/records/10519652) | Jiancheng Yang et al. | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 3,902 |
| [Abdominal Ultrasound Image Dataset for Organ Classification and Disease Detection](https://scholarsjunction.msstate.edu/research-data/5/) | Sifat Zina Karim | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 3,380 |
| [MedMNIST v2 (BreastMNIST, 224px)](https://zenodo.org/records/10519652) | Jiancheng Yang et al. | [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) | 779 |
```
OrganScan is trained on the following datasets, all of which were modified
(sliced to 2D, windowed, resized to 224px, and relabelled):
- TotalSegmentator CT dataset v3.0.0
Jakob Wasserthal, Hanns-Christian Breit, Manfred T. Meyer, Maurice Pradella, Daniel Hinck, Alexander W. Sauter, Tobias Heye, Daniel T. Boll, Joshy Cyriac, Shan Yang, Michael Bach, Martin Segeroth - CC-BY-4.0
https://zenodo.org/records/22688904
- AMOS22: A Large-Scale Abdominal Multi-Organ Benchmark for Versatile Medical Image Segmentation
Yuanfeng Ji, Haotian Bai, Chongjian Ge, Jie Yang, Ye Zhu, Ruimao Zhang, Zhen Li, Lingyan Zhang, Wanling Ma, Xiang Wan, Ping Luo - CC-BY-4.0
https://zenodo.org/records/7262581
- FETAL_PLANES_DB: Common maternal-fetal ultrasound images
Xavier P. Burgos-Artizzu, David Coronado-Gutierrez, Brenda Valenzuela-Alcaraz, Elisenda Bonet-Carne, Elisenda Eixarch, Fatima Crispi, Eduard Gratacós - CC-BY-4.0
https://zenodo.org/records/3904280
- MedMNIST v2 (ChestMNIST, 224px)
Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, Bingbing Ni - CC-BY-4.0
https://zenodo.org/records/10519652
- MedMNIST v2 (OrganSMNIST, 224px)
Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, Bingbing Ni - CC-BY-4.0
https://zenodo.org/records/10519652
- MedMNIST v2 (OrganAMNIST, 224px)
Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, Bingbing Ni - CC-BY-4.0
https://zenodo.org/records/10519652
- MedMNIST v2 (OrganCMNIST, 224px)
Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, Bingbing Ni - CC-BY-4.0
https://zenodo.org/records/10519652
- Abdominal Ultrasound Image Dataset for Organ Classification and Disease Detection
Sifat Zina Karim - CC-BY-4.0
https://scholarsjunction.msstate.edu/research-data/5/
- MedMNIST v2 (BreastMNIST, 224px)
Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, Bingbing Ni - CC-BY-4.0
https://zenodo.org/records/10519652
```
## Author
Arnab Pal — [LinkedIn](https://www.linkedin.com/in/pal-arnab) · [GitHub](https://github.com/palarnab)
<!-- patients by source: {"amos22": 360, "fetal-planes-db": 1792, "medmnist-plus-breastmnist-224": 3, "medmnist-plus-chestmnist-224": 3, "medmnist-plus-organamnist-224": 3, "medmnist-plus-organcmnist-224": 3, "medmnist-plus-organsmnist-224": 3, "msu-abdominal-ultrasound": 3380, "totalsegmentator-ct-v300": 1936} -->
|