Kermany OCT 2018: Retinal OCT Image Classification

~84,000 retinal OCT B-scan images across 4 classes: CNV, DME, DRUSEN, NORMAL. Train: ~83,484 / Test: 1000.

At a glance

Field	Value
Short name	kermany_oct
Full name	Kermany OCT 2018: Retinal OCT Image Classification
First published	2017-12-25
Publication date precision	day
Publication date evidence	data.mendeley.com/datasets
Publication date source field	Mendeley Data version 1 page: Published
Publication date reviewed	2026-09-11
Primary category	oct
Resource role	current_dataset
Dataset family	kermany_oct
Contained modalities	oct
Tasks	classification
Primary reported quantity	84,484 images
Classes	4 (CNV, DME, DRUSEN, NORMAL)
Splits	train, test, val
Size	6.0 GB
Source-stated terms	CC BY 4.0
Normalized terms	cc-by
Descriptive screening label	Standard label without an explicit NC clause; not a permission finding
Terms scope	dataset_files
Access friction	self_service_authenticated
Route backend	Kaggle
Availability	available (checked 2026-07-21)
Acquisition support	standard_platform_supported
Legacy sample-loader status	Standard loader included

Reported quantities

Role	Count	Unit	Scope	Basis	Evidence
Primary	84,484	images	Primary quantity reported in the reviewed catalog source	legacy_catalog_field	kaggle.com/datasets

Counts retain their source-reported units. Additional rows can describe components, paired items, or derivative copies and are not automatically added to the primary quantity.

Documented relationships

These links record source-supported lineage or overlap, not merely similar modality tags.

• intraretinal_cystoid_fluid is derived from this record: The source states that 1,000 training images were selected from the Kermany Retinal OCT Images DME class; 200 test images were collected separately. (evidence)
• mm_retinal_reason is derived from this record: The version-pinned official dataset card lists this record among the CFP or OCT sources used to construct MM-Retinal-Reason. (evidence)
• multieye is derived from this record: The MultiEYE paper names this record as one of the public fundus or OCT sources assembled for the benchmark. (evidence)
• synthetic_retinal_oct_biomarkers is derived from this record: The official dataset card documents the Kermany retinal OCT collection as the source for the four synthetic diagnostic classes. (evidence)
• x_pcr is derived from this record: Source labels in the version-pinned public X-PCR deposit identify this catalog record as upstream material. (evidence)

Access information and download

CLI

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

# Download, only when preflight reports supported behavior
eyehub download kermany_oct --data-dir ./data

Python

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

ds = REGISTRY.get_dataset('kermany_oct')
print(preflight_dataset(ds, './data'))  # no download

Upstream page: kaggle.com/datasets

Source-term evidence: kaggle.com/datasets

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('kermany_oct')
samples = ds.load(data_dir, split='test')
for s in samples[:5]:
    print(s.sample_id, s.label, s.image_path)

Citation

BibTeX

@misc{kermany_oct,
  title  = { Kermany OCT 2018: Retinal OCT Image Classification },
  note   = { Kermany et al., 'Identifying medical diagnoses and treatable diseases by image-based deep learning', Cell 2018 },
  year   = { 2018 },
  url    = { https://www.kaggle.com/datasets/paultimothymooney/kermany2018 },
}

Plain text

Kermany et al., 'Identifying medical diagnoses and treatable diseases by image-based deep learning', Cell 2018.

Source-stated terms

• Raw source string: CC BY 4.0
• Normalized category: cc-by
• 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.

Similar resources by shared modality

• syn_oct: SYN-OCT Synthetic Glaucoma OCT Dataset (200,000 images, cc-by)
• multieye: MultiEYE: OCT-Enhanced Fundus Multi-Disease Benchmark (103,959 images, mit)
• eyecare_100k: Eyecare-100K: Multimodal Ophthalmology VQA Corpus (102,000 question answer pairs, unknown)
• lmod_plus: LMOD+ Multimodal Ophthalmology Benchmark (32,633 annotated instances, unknown)
• harvard_fairvision: Harvard-FairVision (AMD + DR + Glaucoma, paired SLO + OCT) (30,000 participants, cc-by-nc-nd)
• mario: MARIO: AMD-Progression Longitudinal OCT (MICCAI 2024) (30,000 images, cc-by)
• mmrdr: MMRDR: Multi-Modal Retinal Diabetic Retinopathy Dataset (24,460 images, cc-by)
• oct_c8: Retinal OCT-C8: 8-Class OCT Classification (24,000 images, unknown)