Table 1

Key characteristics of the dataset

	Training set	Validation set
Number of patients	19 531	2133
Number of images	140 614	15 313
Age: mean, years (SD)	55.15 (10.68), n=19 495	55.19 (10.38), n=2130
Gender (% male)	0.37, n=19 102	0.37, n=2090
HbA1c: mean, % (SD) Disease duration: mean, years (SD) Ethnicity	7.98 (2.26), n=15 672 7.46 (6.44), n=18 803 62.1% Latin American, 11.4% ethnicity not specified, 9.1% African Descent, 7.4% Caucasian, 5.4% Asian, 2.8% Indian subcontinent origin, 1.1% Other, n=19 076	7.99 (2.75), n=1703 7.21 (6.20), n=2045 61.4% Latin American, 10.7% ethnicity not specified, 9.7% African Descent, 7.8% Caucasian, 5.6% Asian, 3.1% Indian subcontinent origin, 1.1% Other, n=2086