Convolutional neural network for automated mass segmentation in mammography

BMC Bioinformatics

Table 1 The performance of the proposed Vanilla U-Net model, original U-Net, Faster R-CNN, SegNet, Dilated-Net, FCN, and RG

Model	Mean accuracy	Mean DI	Mean IOU	Mean BFscore	Mean inference time (second)/image
-Proposed Vanilla U-Net, with Aug.	0.926	0.951	0.909	0.964	0.115
-Proposed Vanilla U-Net, without Aug.	0.910	0.922	0.856	0.940	0.115
-Original U-Net, with Aug.	0.842	0.818	0.693	0.800	0.118
-Original U-Net, without Aug.	0.821	0.801	0.668	0.776	0.118
-Faster R-CNN, with Aug.	0.702	-	0.601	-	0.454
-SegNet, with Aug.	0.853	0.824	0.701	0.822	0.098
-Dilated-Net, with Aug.	0.832	0.799	0.665	0.701	0.094
-FCN, with Aug.	0.843	0.802	0.669	0.752	0.091
-RG	0.801	0.602	0.401	0.603	0.320

ISSN: 1471-2105