# Chákṣu IMAGE: 1,345 glaucoma fundus photographs with optic disc and cup segmentations from five experts

## Facts

- Provider: Chákṣu IMAGE authors; mirrored by MedOtter
- License: CC-BY-4.0 (https://api.figshare.com/v2/articles/20123135)
- Size: 12 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/MedOtter/Chaksu
- Access: public, readable without an account
- DOI: https://doi.org/10.6084/m9.figshare.20123135
- Category: Health & Biomedical
- Source: https://doi.org/10.1038/s41597-023-01943-4
- Page: https://scigantic.com/archive/f9fe7432-5be1-57d0-9f1f-c376e4f28950
- Keywords: Health & Biomedical, image segmentation, Parquet, Hugging Face

## About

Chákṣu IMAGE is a glaucoma-specific color fundus image database. This mirror carries 1,345 fundus photographs (train 1,009, test 336) from three cameras: Remidio Fundus-on-Phone (2448x3264, 1,074 images), Bosch (1920x1440, 145) and Forus 3Nethra (2048x1536, 126). Five ophthalmologists annotated the optic disc and optic cup on every image; the STAPLE fusion is the primary label (mask with 0 background, 1 neuroretinal rim, 2 optic cup, plus binary disc and cup masks), with majority and median fusions as secondary columns. Each image also carries a glaucoma decision (normal or glaucoma suspect) by majority vote and the five per-expert decisions, plus disc, cup and rim areas and cup-to-disc measurements. Parquet files under data/ hold images at native resolution with unmodified bytes; about 2 GB (measured 2026-11-17). Use for optic disc and cup segmentation, cup-to-disc ratio estimation and glaucoma screening across devices. Mirrored on Hugging Face by MedOtter from the authors' release, with a card documenting provenance and any normalization. The authors' figshare record 20123135 is CC BY 4.0; cite the Chákṣu IMAGE article in Scientific Data (2023), doi 10.1038/s41597-023-01943-4.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("MedOtter/Chaksu")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `data/test-00000-of-00002.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/f9fe7432-5be1-57d0-9f1f-c376e4f28950
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/f9fe7432-5be1-57d0-9f1f-c376e4f28950
- Over MCP: `get_schema_card` with id `f9fe7432-5be1-57d0-9f1f-c376e4f28950` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Image segmentation datasets](https://scigantic.com/topics/image-segmentation)
- [Health & Biomedical datasets](https://scigantic.com/topics/health-and-biomedical)
