MORE: Multi-Organ Medical Image REconstruction Dataset

Venue: ACM Multimedia 2025 Project page: https://more-med.github.io/
Paper: https://doi.org/10.1145/3746027.3758233 Dataset on HF: https://huggingface.co/datasets/WSKINGDOM/MORE

---

Dataset Summary

MORE is a comprehensive benchmark for computed tomography (CT) reconstruction across multiple anatomical regions and lesion types. Compared to prior resources, MORE emphasizes scale, diversity, and clinical realism to better reflect real-world reconstruction scenarios and enable rigorous evaluation of modern methods (traditional IR/MBIR, deep learning, and diffusion-based approaches).

• Scale: 135 CT studies totaling 65,575 scans/slices.
• Multi-Organ: 9 anatomical categories.
• Multi-Lesion: 15 lesion types.

MORE supports common reconstruction settings (e.g., sparse-view, limited-angle, and low-dose), and is accompanied by a baseline/benchmark protocol and code to facilitate reproducible evaluation.

---

Use Cases and Supported Tasks

• Primary task: CT image-to-3D reconstruction from projection/slice data.
• Secondary tasks:
  • Data-driven priors for classical MBIR/IR pipelines
  • Denoising/deartifacting for low-dose or sparse-view recon
  • Cross-anatomy generalization studies and transfer evaluation

task_categories:
  - image-to-3d
tags:
  - medical-imaging
  - CT
  - tomographic-reconstruction
  - low-dose
  - sparse-view
  - multi-organ
  - multi-lesion
  - benchmark

---

Intended Uses & Limitations

Intended Uses

• Research and development of CT reconstruction algorithms under clinically relevant variations (organ diversity, lesion presence, acquisition differences).
• Robustness and generalization studies across anatomy and pathologies.

Limitations

• MORE is a research dataset; results may not directly translate to clinical outcomes.
• Labels and metadata granularity vary by study; always cross-check assumptions.
• If using simulated measurement protocols, clearly document geometry and noise models.

---

Ethical & Responsible Use

• Use for research and educational purposes.
• Do not attempt to identify individuals; adhere to all applicable privacy regulations.
• When publishing results, include task setting, data split, and evaluation protocols for transparency and reproducibility.
• If creating derivative datasets or releasing trained models, ensure license compliance and responsible disclosure of potential risks.

---

License

license: cc-by-nc-4.0

---

How to Cite

If you use MORE in your research, please consider citing our paper:

ACM MM 2025

Shaokai Wu, Yapan Guo, Yanbiao Ji, Jing Tong, Yuxiang Lu, Mei Li, Suizhi Huang, Yue Ding, and Hongtao Lu. 2025. MORE: Multi-Organ Medical Image REconstruction Dataset.
In Proceedings of the 33rd ACM International Conference on Multimedia (MM '25). Association for Computing Machinery, New York, NY, USA, 12890–12896. https://doi.org/10.1145/3746027.3758233

BibTeX:

@inproceedings{10.1145/3746027.3758233,
author = {Wu, Shaokai and Guo, Yapan and Ji, Yanbiao and Tong, Jing and Lu, Yuxiang and Li, Mei and Huang, Suizhi and Ding, Yue and Lu, Hongtao},
title = {MORE: Multi-Organ Medical Image REconstruction Dataset},
year = {2025},
isbn = {9798400720352},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3746027.3758233},
doi = {10.1145/3746027.3758233},
abstract = {CT reconstruction provides radiologists with images for diagnosis and treatment, yet current deep learning methods are typically limited to specific anatomies and datasets, hindering generalization ability to unseen anatomies and lesions. To address this, we introduce the Multi-Organ medical image REconstruction (MORE) dataset, comprising CT scans across 9 diverse anatomies with 15 lesion types. This dataset serves two key purposes: (1) enabling robust training of deep learning models on extensive, heterogeneous data, and (2) facilitating rigorous evaluation of model generalization for CT reconstruction. We further establish a strong baseline solution that outperforms prior approaches under these challenging conditions. Our results demonstrate that: (1) a comprehensive dataset helps improve the generalization capability of models, and (2) optimization-based methods offer enhanced robustness for unseen anatomies. The MORE dataset is freely accessible under CC-BY-NC 4.0 at our project page https://more-med.github.io/.},
booktitle = {Proceedings of the 33rd ACM International Conference on Multimedia},
pages = {12890–12896},
numpages = {7},
keywords = {benchmark, computed tomography, ct reconstruction, medical image reconstruction, multi-organ dataset},
location = {Dublin, Ireland},
series = {MM '25}
}

---

Related Links

• Project page: https://more-med.github.io/
• Paper: https://doi.org/10.1145/3746027.3758233
• Appendix: See project website
• Baselines/Code: code.zip in the dataset repository (see project page for updates)

---

Contributions

We welcome issues and pull requests that improve documentation, loaders, and benchmarks (e.g., additional baselines, standardized evaluation scripts, or organ-wise reports).
Maintainers: Shaokai Wu, Yapan Guo*, Yanbiao Ji, Jing Tong, Yue Ding, Yuxiang Lu, Mei Li, Suizhi Huang, Hongtao Lu*

---

Changelog

• v1.0: Initial public release with multi-organ, multi-lesion CT studies and baseline protocols.
• v1.1+ (planned): Extended MRI benchmarks, additional utilities, and community submissions.

---