Table 2.

Comparison of the performance for each method

	Sensitivity	Specificity	Accuracy	Time cost (s)
Dual-input	1:000 (0:832; 1:000)	0.450 (0.231, 0.685)	0.725 (0.561, 0.854)	0:0366 ± 0:0001
Dual-input + RAGS	0.960 (0.751, 0.999)	0:910 (0:683; 0:988)	0:935 (0:796; 0:984)	0:0569 ± 0:0013
Rahmany et al. [15]	0.950 (0.751, 0.999)	0.300 (0.119, 0.543)	0.625 (0.458, 0.773)	62:5460 ± 23:222
YOLOv3 [22]	0.850 (0.621, 0.968)	0.700 (0.457, 0.881)	0.775 (0.616, 0.892)	0:0232 ± 0:0005
RetinaNet [31]	1:000 (0:832; 1:000)	0.420 (0.191, 0.639)	0.710 (0.535, 0.834)	0:0456 ± 0:0036
Physician	0.900 (0.683, 0.988)	0.900 (0.683, 0.988)	0.900 (0.763, 0.972)	–

Data shown in “italic” are the highest value of the column

Data in parentheses are 95% CI