Table 3.

Mean (and SD) per cent agreement between confidence and accuracy of the deep convolutional ensemble and ophthalmologists

	Deep convolutional ensemble		Ophthalmologists
	Correct	Incorrect	Correct	Incorrect
Confident	77.4% (2.5%)	16.6% (2.5%)	58.7% (4.3%)	15.7% (8.6%)
Not confident	1.8% (0.8%)	4.2% (0.8%)	14% (6.0%)	11.6% (3.7%)

Green cells indicate agreement between the ground-truth labels and predictions by the deep convolutional ensemble or ophthalmologists, and red cells similarly indicate disagreement