Dataset Viewer
Auto-converted to Parquet Duplicate
filename
stringlengths
14
14
age
float64
4.34
104
gender
stringclasses
2 values
race
stringclasses
3 values
ethnicity
stringclasses
2 values
language
stringclasses
3 values
maritalstatus
stringclasses
6 values
amd
stringclasses
4 values
use
stringclasses
3 values
data_00001.npz
71.93
female
white
non-hispanic
english
married or partnered
late amd
training
data_00002.npz
62.65
male
white
hispanic
english
single
early amd
training
data_00003.npz
62.59
female
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00004.npz
90.08
female
white
non-hispanic
english
widowed
late amd
training
data_00005.npz
71.89
female
white
non-hispanic
english
unknown
early amd
training
data_00006.npz
92.03
female
white
non-hispanic
english
unknown
late amd
training
data_00007.npz
73.2
male
white
non-hispanic
english
unknown
normal
training
data_00008.npz
31.45
female
white
non-hispanic
english
married or partnered
normal
training
data_00009.npz
16.08
male
white
non-hispanic
english
single
normal
training
data_00010.npz
51.75
female
white
non-hispanic
english
married or partnered
normal
training
data_00011.npz
69.13
female
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00012.npz
84.95
female
white
non-hispanic
english
widowed
late amd
training
data_00013.npz
92.18
female
white
non-hispanic
english
widowed
late amd
training
data_00014.npz
52.81
female
white
non-hispanic
english
married or partnered
normal
training
data_00015.npz
80.29
male
white
non-hispanic
english
married or partnered
normal
training
data_00016.npz
51.55
male
white
non-hispanic
english
married or partnered
normal
training
data_00017.npz
78.56
male
white
non-hispanic
english
married or partnered
normal
training
data_00018.npz
54.65
female
white
non-hispanic
english
married or partnered
normal
training
data_00019.npz
89.12
female
white
non-hispanic
other languages
widowed
normal
training
data_00020.npz
57.56
male
black
non-hispanic
english
single
normal
training
data_00021.npz
35.84
male
white
non-hispanic
english
married or partnered
normal
training
data_00022.npz
41.41
female
white
non-hispanic
english
married or partnered
normal
training
data_00023.npz
71.03
female
white
non-hispanic
english
single
normal
training
data_00024.npz
70.29
female
white
non-hispanic
english
divorced
normal
training
data_00025.npz
85.62
female
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00026.npz
58.16
female
white
non-hispanic
english
single
normal
training
data_00027.npz
23.96
male
white
non-hispanic
english
single
normal
training
data_00028.npz
74.18
male
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00029.npz
48.52
female
white
non-hispanic
english
married or partnered
normal
training
data_00030.npz
83.52
female
white
non-hispanic
english
married or partnered
late amd
training
data_00031.npz
76.63
female
white
non-hispanic
english
unknown
late amd
training
data_00032.npz
76.82
male
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00033.npz
81.2
female
asian
non-hispanic
other languages
single
late amd
training
data_00034.npz
50.85
male
white
non-hispanic
english
married or partnered
normal
training
data_00035.npz
90.9
female
white
non-hispanic
english
widowed
normal
training
data_00036.npz
78.58
male
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00037.npz
70.16
female
white
non-hispanic
english
married or partnered
early amd
training
data_00038.npz
72.79
male
asian
non-hispanic
english
married or partnered
normal
training
data_00039.npz
69.07
female
white
non-hispanic
english
divorced
normal
training
data_00040.npz
70.96
male
white
non-hispanic
english
married or partnered
normal
training
data_00041.npz
29.72
female
asian
non-hispanic
english
married or partnered
normal
training
data_00042.npz
72.13
female
black
non-hispanic
english
widowed
normal
training
data_00043.npz
83.85
male
white
non-hispanic
english
married or partnered
normal
training
data_00044.npz
73.85
female
white
non-hispanic
english
widowed
late amd
training
data_00045.npz
75.18
female
asian
non-hispanic
other languages
divorced
early amd
training
data_00046.npz
81.45
male
white
non-hispanic
english
single
normal
training
data_00047.npz
65.26
male
black
non-hispanic
english
married or partnered
normal
training
data_00048.npz
61.35
female
white
non-hispanic
english
married or partnered
normal
training
data_00049.npz
66.72
female
white
non-hispanic
english
married or partnered
normal
training
data_00050.npz
61.92
male
white
non-hispanic
english
married or partnered
normal
training
data_00051.npz
70.52
male
white
non-hispanic
english
divorced
normal
training
data_00052.npz
45.39
male
white
non-hispanic
english
married or partnered
normal
training
data_00053.npz
56.3
female
white
non-hispanic
english
divorced
early amd
training
data_00054.npz
81.02
female
white
non-hispanic
other languages
married or partnered
normal
training
data_00055.npz
66.18
male
white
non-hispanic
english
single
normal
training
data_00056.npz
56.84
male
white
non-hispanic
english
married or partnered
normal
training
data_00057.npz
81.96
female
white
non-hispanic
english
married or partnered
normal
training
data_00058.npz
80.55
male
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00059.npz
60.47
female
white
non-hispanic
english
married or partnered
normal
training
data_00060.npz
74.79
male
white
non-hispanic
english
widowed
intermediate amd
training
data_00061.npz
86.45
female
white
non-hispanic
english
widowed
early amd
training
data_00062.npz
83.69
female
white
non-hispanic
english
widowed
normal
training
data_00063.npz
24.49
female
asian
non-hispanic
english
single
normal
training
data_00064.npz
76.96
male
white
non-hispanic
english
married or partnered
early amd
training
data_00065.npz
76.66
female
white
non-hispanic
english
unknown
late amd
training
data_00066.npz
84.08
male
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00067.npz
93.24
male
asian
non-hispanic
other languages
widowed
early amd
training
data_00068.npz
43.48
female
asian
non-hispanic
english
married or partnered
normal
training
data_00069.npz
72
female
white
non-hispanic
english
unknown
normal
training
data_00070.npz
51.16
female
white
non-hispanic
english
single
normal
training
data_00071.npz
55.08
male
white
non-hispanic
english
single
normal
training
data_00072.npz
58.02
male
asian
non-hispanic
english
married or partnered
normal
training
data_00073.npz
79.05
female
white
non-hispanic
english
widowed
late amd
training
data_00074.npz
69.47
male
white
non-hispanic
english
married or partnered
normal
training
data_00075.npz
42.38
male
white
non-hispanic
english
single
normal
training
data_00076.npz
82.66
male
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00077.npz
68.72
male
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00078.npz
80.08
male
white
non-hispanic
english
married or partnered
normal
training
data_00079.npz
80.23
female
white
non-hispanic
english
single
normal
training
data_00080.npz
49.77
female
black
non-hispanic
english
married or partnered
normal
training
data_00081.npz
93.21
male
white
non-hispanic
english
married or partnered
normal
training
data_00082.npz
69.24
female
black
non-hispanic
english
married or partnered
normal
training
data_00083.npz
68.63
female
white
non-hispanic
english
married or partnered
normal
training
data_00084.npz
67.48
female
white
non-hispanic
english
married or partnered
intermediate amd
training
data_00085.npz
78.17
female
white
non-hispanic
english
married or partnered
early amd
training
data_00086.npz
69.29
male
white
non-hispanic
english
married or partnered
late amd
training
data_00087.npz
62.15
female
white
non-hispanic
english
married or partnered
normal
training
data_00088.npz
80.11
female
white
non-hispanic
english
married or partnered
late amd
training
data_00089.npz
37.82
female
white
non-hispanic
english
single
normal
training
data_00090.npz
80.87
female
white
non-hispanic
english
married or partnered
late amd
training
data_00091.npz
88.02
female
white
non-hispanic
english
unknown
normal
training
data_00092.npz
49.92
male
white
non-hispanic
english
divorced
normal
training
data_00093.npz
54.65
female
white
non-hispanic
english
divorced
normal
training
data_00094.npz
70.73
male
white
non-hispanic
english
married or partnered
normal
training
data_00095.npz
74.56
male
white
non-hispanic
english
divorced
late amd
training
data_00096.npz
87.33
male
white
non-hispanic
other languages
married or partnered
intermediate amd
training
data_00097.npz
70.98
female
black
non-hispanic
other languages
single
normal
training
data_00098.npz
55.6
male
white
non-hispanic
english
married or partnered
normal
training
data_00099.npz
86.47
male
white
non-hispanic
english
single
late amd
training
data_00100.npz
67.91
male
white
non-hispanic
english
married or partnered
normal
training
End of preview. Expand in Data Studio

Dataset Card: Harvard-FairVision

Dataset Summary

Harvard-FairVision is the first large-scale medical fairness dataset with both 2D and 3D imaging data, covering three major eye diseases affecting approximately 380 million people worldwide. It contains 30,000 subjects (10,000 per disease) across Age-Related Macular Degeneration (AMD), Diabetic Retinopathy (DR), and glaucoma, each with paired SLO fundus photos and 3D OCT B-scans and six demographic identity attributes.

This dataset was introduced in the paper: FairVision: Equitable Deep Learning for Eye Disease Screening via Fair Identity Scaling.

Dataset Details

Dataset Description

Field Value
Institution Department of Ophthalmology, Harvard Medical School
Tasks AMD detection, diabetic retinopathy detection, glaucoma detection
Modalities Scanning Laser Ophthalmoscopy (SLO) fundus images, 3D OCT B-scans
Scale 30,000 subjects (10,000 per disease)
OCT size 200 Γ— 200 Γ— 200 (glaucoma), 128 Γ— 200 Γ— 200 (AMD, DR)
SLO size 512 Γ— 664 (folders), 200 Γ— 200 (NPZ, normalized to [0, 255])
Total size ~600 GB
License CC BY-NC-ND 4.0

Dataset Structure

FairVision
β”œβ”€β”€ AMD
β”‚   β”œβ”€β”€ Training
β”‚   β”œβ”€β”€ Validation
β”‚   └── Test
β”œβ”€β”€ data_summary_amd.csv
β”œβ”€β”€ DR
β”‚   β”œβ”€β”€ Training
β”‚   β”œβ”€β”€ Validation
β”‚   └── Test
β”œβ”€β”€ data_summary_dr.csv
β”œβ”€β”€ Glaucoma
β”‚   β”œβ”€β”€ Training
β”‚   β”œβ”€β”€ Validation
β”‚   └── Test
└── data_summary_glaucoma.csv

Each split folder contains SLO fundus photos (slo_xxxxx.jpg) and NPZ files (data_xxxxx.npz). Per-disease metadata CSVs (data_summary_*.csv) provide race, gender, ethnicity, marital status, age, and preferred language for all subjects.

Data Fields

All NPZ files share the following demographic and imaging fields:

Field Description
slo_fundus SLO fundus image, 200 Γ— 200 (normalized)
oct_bscans 3D OCT B-scans (200 Γ— 200 Γ— 200 for glaucoma; 128 Γ— 200 Γ— 200 for AMD/DR)
race 0 = Asian, 1 = Black, 2 = White
male 0 = Female, 1 = Male
hispanic 0 = Non-Hispanic, 1 = Hispanic
maritalstatus 0 = Married, 1 = Single, 2 = Divorced, 3 = Widowed, 4 = Legally Separated
language 0 = English, 1 = Spanish, 2 = Other

Disease-specific label fields:

Disease Field Values
Glaucoma glaucoma 0 = non-glaucoma, 1 = glaucoma
AMD amd_condition 9-class condition string, mapped to 0 = no AMD, 1 = early dry, 2 = intermediate dry, 3 = advanced
DR dr_subtype 6-class condition string, mapped to 0 = non-vision-threatening, 1 = vision-threatening (severe NPDR or PDR)

Uses

Direct Use

  • Fairness benchmarking for 2D and 3D ophthalmic disease classification across race, gender, and ethnicity
  • Multi-disease fairness analysis (AMD, DR, glaucoma) under a unified framework
  • Development and evaluation of fairness learning methods for medical imaging
  • Comparative study of 2D vs. 3D model fairness in clinical AI

Out-of-Scope Use

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

Access

The "Harvard" designation indicates the dataset originates from the Department of Ophthalmology at Harvard Medical School. It does not imply endorsement, sponsorship, or legal responsibility by Harvard University or Harvard Medical School.

Citation

BibTeX:

@misc{luo2024fairvisionequitabledeeplearning,
  title={FairVision: Equitable Deep Learning for Eye Disease Screening via Fair Identity Scaling},
  author={Yan Luo and Muhammad Osama Khan and Yu Tian and Min Shi and Zehao Dou and Tobias Elze and Yi Fang and Mengyu Wang},
  year={2024},
  eprint={2310.02492},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2310.02492}
}

APA:

Luo, Y., Khan, M. O., Tian, Y., Shi, M., Dou, Z., Elze, T., Fang, Y., & Wang, M. (2024). FairVision: Equitable Deep Learning for Eye Disease Screening via Fair Identity Scaling. arXiv preprint arXiv:2310.02492.

Downloads last month
42

Paper for harvardairobotics/FairVision