A deep learning-based algorithm for 2-D cell segmentation in microscopy images

BMC Bioinformatics

Table 1 The used U-net architecture

L#	Type	Size	Output	L#	Type	Size	Output
1	Input		1,160,160	17	Concatenate		256,20,20
2	Convolution	32 filters	32,160,160	18	Dropout	50%	256,20,20
3	Max pool	2 stride 2x2	32,80,80	19	Convolution	128 filters	128,20,20
4	Convolution	64 filters	64,80,80	20	Deconvolution	2 stride, 128x2x2	128,40,40
5	Max pool	2 stride 2x2	64,40,40	21	Convolution	128 filters	128,40,40
6	Convolution	128 filters	128,40,40	22	Concatenate		192,40,40
7	Max pool	2 stride 2x2	128,20,20	23	Dropout	50%	192,40,40
8	Convolution	128 filters	128,20,20	24	Convolution	128 filters	128,40,40
9	Max pool	2 stride 2x2	128,10,10	25	Deconvolution	2 stride, 128x2x2	128,80,80
10	Convolution	256 filters	256,10,10	26	Convolution	128 filters	128,80,80
11	Max pool	2 stride 2x2	256,5,5	27	Concatenate		160,80,80
12	Dropout	50%	256,5,5	28	Concatenate		160,80,80
13	Deconvolution	2 stride, 256x2x2	256,10,10	29	Convolution	64 filters	64,80,80
14	Convolution	128 filters	128,10,10	30	Deconvolution	2 stride, 128x2x2	64,160,160
15	Deconvolution	2 stride, 128x2x2	128,20,20	31	Convolution	64 filters	64,160,160
16	Convolution	128 filters	128,20,20	32	Output		3,160,160

ISSN: 1471-2105