Table 2.

Detail of the architecture of U-Net.

Layers	Architectures	Output
Input	Image (96 × 96)	96 × 96 × 1
conv1	2@Conv (3 × 3)/Relu padding = “same”	96 × 96 × 64
	Max pooling stride = 2
conv2	2@Conv (3 × 3)/Relu padding = “same”	48 × 48 × 128
	Max pooling stride = 2
conv3	2@Conv (3 × 3)/Relu padding = “same”	24 × 24 × 256
	Max pooling stride = 2
conv4	2@Conv (3 × 3)/Relu padding = “same”	12 × 12 × 512
drop4	Dropout (p = 0.5)
	Max pooling stride = 2
conv5	2@Conv (3 × 3)/Relu padding = “same”	6 × 6 × 1024
drop5	Dropout (p = 0.5)
up6	Upsampling conv (2 × 2)/Relu	12 × 12 × 512
Concatination	[drop4, up6]
conv6	2@Conv (3 × 3)/Relu padding = “same”	12 × 12 × 512
up7	Upsampling conv (2 × 2)/Relu	24 × 24 × 256
Concatination	[conv3, up7]
conv7	2@Conv (3 × 3)/Relu padding = “same”	24 × 24 × 256
up8	Up-sampling conv (2 × 2)/Relu	48 × 48 × 128
Concatination	[conv2, up8]
conv8	2@Conv (3 × 3)/Relu padding = “same”	48 × 48 × 128
up9	Upsampling conv (2 × 2)/Relu	96 × 96 × 64
Concatination	[conv1, up9]
conv9	2@Conv (3 × 3)/Relu padding = “same”	96 × 96 × 64
conv10	Conv (1 × 1) Sigmoid	96 × 96 × 1
Output	Segmentation map	96 × 96 × 1