Automatic Semantic Segmentation of Brain Gliomas from MRI Images Using a Deep Cascaded Neural Network

Table 3

Parameters used in the subnet TLN. In each convolutional layer, the feature maps had been padded by 1 prior to the convolution so that all intermediate feature maps do not change their sizes before and after the convolution.

Number	Layer name	Filter size	Stride	Number of Filters	Output
1	Conv 1_1 + ReLU	33	1	64	43843864
2	Conv 1_2 + ReLU	33	1	64	43843864
3	Max pooling 1	22	2	—	21921964
4	Conv 2_1 + ReLU	33	1	128	219219128
5	Conv 2_2 + ReLU	33	1	128	219219128
6	Max pooling 2	22	2	—	110110128
7	Conv 3_1 + ReLU	33	1	256	110110256
8	Conv 3_2 + ReLU	33	1	256	110110256
9	Conv 3_3 + ReLU	33	1	256	110110256
10	Max pooling 3	22	2	—	5555256
11	Conv 4_1 + ReLU	33	1	512	5555512
12	Conv 4_2 + ReLU	33	1	512	5555512
13	Conv 4_3 + ReLU	33	1	512	5555512
14	Max pooling 4	22	2	—	2828512
15	Conv 5_1 + ReLU	33	1	512	2828512
16	Conv 5_2 + ReLU	33	1	512	2828512
17	Conv 5_3 + ReLU	33	1	512	2828512
18	Max pooling 5	22	2	—	1414512
19	Conv 6 + ReLU	77	1	4096	884096
20	Conv 7 + ReLU	11	1	4096	884096