A Novel Deep Learning Network and Its Application for Pulmonary Nodule Segmentation

Table 1

Parameters in each layer.

Layers	Size of feature map	Size/step of convolution
Input	64 × 64	—
Dense connection	64 × 64	[3 × 3 Conv-64] × 2
Max pooling	32 × 32	2 × 2/2
Dense connection	32 × 32	[3 × 3 Conv-96] × 2
Max pooling	16 × 16	2×2/2
Dense connection	16 × 16	[3 × 3 Conv-128] × 2
Max pooling	8 × 8	2 × 2/2
Dense connection	8 × 8	[3 × 3 Conv-256] × 2
Max pooling	4 × 4	2 × 2/2
Dense connection	4 × 4	[3 × 3 Conv-512] × 2
Max pooling	8 × 8	2 × 2/2
Dense connection	8 × 8	[3 × 3 Conv-256] × 2
Max pooling	16 × 16	2 × 2/2
Dense connection	16 × 16	[3 × 3 Conv-128] × 2
Max pooling	32 × 32	2 × 2/2
Dense connection	32 × 32	[3 × 3 Conv-96] × 2
Max pooling	64 × 64	2 × 2/2
Dense connection	64 × 64	[3 × 3 Conv-64] × 2
Max pooling	64 × 64	1 × 1 Conv