Table 1 Baseline characteristics of the development and clinical validation datasets.

From: Predicting optical coherence tomography-derived diabetic macular edema grades from fundus photographs using deep learning

Characteristics

Development set

Primary clinical validation set

Secondary clinical validation set

Dataset

Thailand dataset

Thailand dataset

EyePACS-DME dataset

Number of patients

4035

697

554

Number of fundus images

6039

1033

990

Camera used for fundus images

Kowa VX-10

Kowa VX-10

Canon CR-DGi

OCT device used for determining ci-DME

Heidelberg Spectralis

Heidelberg Spectralis

Optovue iVue

Age: mean, years (SD)

55.6 (10.8)

n = 6038

55.8 (10.8)

n = 1033

62.0 (9.8)

n = 990

Gender (% male)

60.8%

n = 6036

62.4%

n = 1031

50.1%

n = 990

Central retinal thickness: mean, μm (SD)

263.8 (146.5)

n = 6039

258.4 (132.8)

n = 1033

254.4 (56.3)

n = 990

ci-DME, Center Point Thickness ≥ 250 μm in Thailand dataset. Central Subfield Thickness ≥ 300 μm in the Eyepacs-DME dataset

28.3%

n = 6039

27.2%

n = 1033

7.8%

n = 990

Subretinal fluid presence

15.7%

n = 6039

15.1%

n = 1033

NA

Intraretinal fluid presence

45.5%

n = 6039

46.3%

n = 1033

NA

  1. It is noteworthy that the difference between total n and subcategories is missing data (e.g., not all images had age or sex)