harvardairobotics/Harvard-FairSeg

---
license: cc-by-nc-nd-4.0
task_categories:
- image-segmentation
modality:
- image
language:
- en
tags:
- medical
- ophthalmology
- fairness
- segmentation
- fundus
- glaucoma
- disc-cup
pretty_name: Harvard-FairSeg
size_categories:
- 10K<n<100K
---

# Dataset Card: Harvard-FairSeg

## Dataset Summary

Harvard-FairSeg is a large-scale ophthalmology dataset for studying **fairness in medical image segmentation**. It contains 10,000 SLO fundus images with pixel-wise optic disc and cup segmentation masks, paired with comprehensive demographic annotations. The dataset is designed to benchmark and improve demographic equity in segmentation models, including foundation models such as SAM (Segment Anything Model).

This dataset was introduced at ICLR 2024: [FairSeg: A Large-Scale Medical Image Segmentation Dataset for Fairness Learning Using Segment Anything Model with Fair Error-Bound Scaling](https://openreview.net/pdf?id=qNrJJZAKI3).

## Dataset Details

### Dataset Description

| Field | Value |
|------------------|-------|
| **Institution** | Department of Ophthalmology, Harvard Medical School |
| **Task** | Optic disc and cup segmentation |
| **Modality** | Scanning Laser Ophthalmoscopy (SLO) fundus images |
| **Scale** | 10,000 patients, 10,000 images |
| **Annotation** | Pixel-wise disc and cup masks |
| **License** | [CC BY-NC-ND 4.0](https://creativecommons.org/licenses/by-nc-nd/4.0/) |

- **Curated by:** Yu Tian, Min Shi, Yan Luo, Ava Kouhana, Tobias Elze, Mengyu Wang
- **License:** [CC BY-NC-ND 4.0](https://creativecommons.org/licenses/by-nc-nd/4.0/) — non-commercial research only
- **Paper:** [ICLR 2024](https://openreview.net/pdf?id=qNrJJZAKI3)
- **Contact:** [e-mail], [e-mail]

### Data Fields

Each subject is stored as a `.npz` file (`data_00001.npz` … `data_10000.npz`) containing:

| Field | Description |
|------------------|-------------|
| `slo_fundus` | Scanning Laser Ophthalmoscopy (SLO) fundus image |
| `disc_cup_mask` | Pixel-wise optic disc and cup segmentation mask |
| `age` | Patient age |
| `gender` | `0` = Female, `1` = Male |
| `race` | `0` = Asian, `1` = Black, `2` = White |
| `ethnicity` | `0` = Non-Hispanic, `1` = Hispanic, `-1` = Unknown |
| `language` | `0` = English, `1` = Spanish, `2` = Other, `-1` = Unknown |
| `maritalstatus` | `0` = Married/Partnered, `1` = Single, `2` = Divorced, `3` = Widowed, `4` = Legally Separated, `-1` = Unknown |

A metadata overview is provided in `data_summary.csv` under the `ReadMe` folder.

### Demographics

6 demographic attributes are annotated per patient: age, gender, race, ethnicity, preferred language, and marital status.

## Uses

### Direct Use

Fairness benchmarking for medical image segmentation models, including evaluation of demographic disparities in optic disc and cup segmentation. Intended for use with segmentation architectures such as SAMed and TransUNet.

### Out-of-Scope Use

Clinical decisions, patient care, or any commercial application. This dataset shall not be used for clinical decisions at any time.

## Citation

**BibTeX:**

```bibtex
@inproceedings{tianfairseg,
title={FairSeg: A Large-Scale Medical Image Segmentation Dataset for Fairness Learning Using Segment Anything Model with Fair Error-Bound Scaling},
author={Tian, Yu and Shi, Min and Luo, Yan and Kouhana, Ava and Elze, Tobias and Wang, Mengyu},
booktitle={The Twelfth International Conference on Learning Representations}
}
```

**APA:**

Tian, Y., Shi, M., Luo, Y., Kouhana, A., Elze, T., & Wang, M. (2024). FairSeg: A Large-Scale Medical Image Segmentation Dataset for Fairness Learning Using Segment Anything Model with Fair Error-Bound Scaling. *The Twelfth International Conference on Learning Representations (ICLR 2024)*.