Multichannel Retinal Blood Vessel Segmentation Based on the Combination of Matched Filter and U-Net Network

Table 3

Comparison of the proposed method with other methods on the DRIVE dataset.

Method	Se	Sp	ACC	AUC
Khan et al. (2016) [9]	0.7373	0.9670	0.9501	0.8522
Khan et al. (2016) [29]	0.780	0.972	0.952	0.876
Soomro et al. (2017) [11]	0.746	0.917	0.946	0.8315
Ngo and Han (2017) [12]	0.7464	0.9836	0.9533	0.8650
Biswal et al. (2017) [30]	0.71	0.97	0.95	0.84
Yan et al. (2018) [13]	0.7653	0.9818	0.9542	0.8736
Oliveira et al. (2018) [15]	0.8039	0.9804	0.9576	0.8922
Wang et al. (2019) [25]	0.7648	0.9817	0.9541	0.8733
Guo et al. (2019) [31]	0.7800	0.9806	0.9551	0.8803
Feng et al. (2019) [14]	0.7625	0.9809	0.9528	0.8717
Ribeiro et al. (2019) [32]	0.7880	0.9819	0.9569	0.8850
Dharmawan et al. (2019) [18]	0.8314	0.9726	—	0.902
Saroj et al. (2020) [5]	0.7307	0.9761	0.9544	0.8534
Dash and Senapati (2020) [33]	0.7403	0.9905	0.9661	0.8654
Biswas et al. (2020) [16]	0.7823	0.9814	0.9561	0.8819
Budak et al. (2020) [34]	0.7439	0.9900	0.9685	0.8670
2^nd human observer	0.7760	0.9724	0.9472	0.8742
Proposed method	0.8745	0.9624	0.9546	0.9185