A Novel U-Net Based Deep Learning Method for 3D Cardiovascular MRI Segmentation

Table 1

Our 3D convolutional model.

Input image				Output				Layer (type)	Stride	Kernel	Parameters
64	64	64	1	32	32	32	16	Conv_1 (convolution)	2	3	448
32	32	32	16	16	16	16	16	Conv_2 (convolution)	2	3	6928
16	16	16	16	16	16	16	16	Spatial attention	2	1	816
16	16	16	16	16	16	16	100	Sparse Block_1 (sparse block)	1	3	43300
16	16	16	100	16	16	16	100	Conv_3 (convolution)	1	1	10100
16	16	16	100	16	16	16	184	Sparse Block_2 (sparse block)	1	3	496984
16	16	16	184	16	16	16	64	Conv_4 (convolution)	1	1	11840
16	16	16	64	32	32	32	64	Deconv_1 (deconvolution)	2	4	262208
32	32	32	64	64	64	64	64	Deconv_2 (deconvolution)	2	4	262208
16	16	16	100	64	64	64	64	Skip connection	1	1	6464