OCTAVE: 3D SD-OCT Retinal Segmentation Dataset

198 annotated 3D SD-OCT volumes (3,762 B-scans) with pixel-level labels for 13 anatomic and pathological retinal features: retina, choroid, vitreous, RPE, hyaloid, ERM, fluid, subretinal material, hypertransmission defects, and more. 4 additional external validation sets (221 volumes). nnU-Net-compatible format. IOVS 2025.

At a glance

Field	Value
Short name	octave
Full name	OCTAVE: 3D SD-OCT Retinal Segmentation Dataset
Primary category	oct
Contained modalities	oct
Tasks	segmentation
Samples	198
Classes	13 (RET, CHO, VIT, RPE, HYA, RHS, ERM, SES, ART, HRM, FLU, SRM, HTD)
Splits	train
Size	15.0 GB
Source-stated terms	CC BY-SA 4.0
Normalized terms	cc-by-sa
Descriptive screening label	Standard label without an explicit NC clause; not a permission finding
Terms scope	dataset_files
Access friction	anonymous_direct
Route backend	Zenodo
Availability	available (checked 2026-07-21)
Acquisition support	standard_platform_supported
Legacy sample-loader status	Standard loader included

Notes

nnU-Net layout: nnUNet_raw/Dataset001_OCTAVE/imagesTr+labelsTr/. External test sets under nnUNet_raw/external_tests/. 19 standardised B-scans per volume after preprocessing. GitHub: https://github.com/Translational-Biophotonics-Laboratory/octvision3d

Access preflight and acquisition

CLI

# Read-only preflight
eyehub download octave --data-dir ./data --dry-run --json

# Explicit transfer, only when preflight reports supported behavior
eyehub download octave --data-dir ./data

Python

from eyedatahub.acquisition import preflight_dataset
from eyedatahub.datasets.registry import REGISTRY

ds = REGISTRY.get_dataset('octave')
print(preflight_dataset(ds, './data'))  # no transfer

Upstream page: zenodo.org/records

Source-term evidence: zenodo.org/records

Loader example

This entry includes a standard DatasetSample loader.

from pathlib import Path
from eyedatahub.datasets.registry import REGISTRY

data_dir = Path('~/.eyedatahub/data').expanduser()
ds = REGISTRY.get_dataset('octave')
samples = ds.load(data_dir, split='train')
for s in samples[:5]:
    print(s.sample_id, s.label, s.image_path)

Citation

BibTeX

@misc{octave,
  title  = { OCTAVE: 3D SD-OCT Retinal Segmentation Dataset },
  note   = { Kermany DS et al., 'Identifying Retinal Features Using a Self-Configuring CNN for Clinical Intervention', Investigative Ophthalmology & Visual Science 66(6):55 (2025). doi:10.1167/iovs.66.6.55 — Zenodo: https://zenodo.org/records/14580071 },
  year   = { 2025 },
  url    = { https://zenodo.org/records/14580071 },
}

Plain text

Kermany DS et al., 'Identifying Retinal Features Using a Self-Configuring CNN for Clinical Intervention', Investigative Ophthalmology & Visual Science 66(6):55 (2025). doi:10.1167/iovs.66.6.55 — Zenodo: https://zenodo.org/records/14580071

Source-stated terms

• Raw source string: CC BY-SA 4.0
• Normalized category: cc-by-sa
• Apparent scope: dataset_files
• Descriptive screening label: Standard label without an explicit NC clause; not a permission finding

⚠️ Source-stated terms, scope, and normalized labels are curation metadata, not legal advice or a permission finding. Review the current official source before transfer or reuse.

Related datasets with shared modalities

• syn_oct: SYN-OCT Synthetic Glaucoma OCT Dataset (200,000 records, cc-by)
• eyecare_100k: Eyecare-100K: Multimodal Ophthalmology VQA Corpus (102,000 records, unknown)
• kermany_oct: Kermany OCT 2018: Retinal OCT Image Classification (84,484 records, cc-by)
• multieye: MultiEYE: OCT-Enhanced Fundus Multi-Disease Benchmark (58,036 records, mit)
• lmod_plus: LMOD+ Multimodal Ophthalmology Benchmark (32,633 records, unknown)
• harvard_fairvision: Harvard-FairVision (AMD + DR + Glaucoma, paired SLO + OCT) (30,000 records, cc-by-nc-nd)
• mario: MARIO: AMD-Progression Longitudinal OCT (MICCAI 2024) (30,000 records, cc-by)
• mmrdr: MMRDR: Multi-Modal Retinal Diabetic Retinopathy Dataset (24,460 records, cc-by)