Differentiation of the Follicular Neoplasm on the Gray-Scale US by Image Selection Subsampling along with the Marginal Outline Using Convolutional Neural Network

Table 5

Training structure of the convolutional neural net (5-conv, 2-pool, 2-fully-conn structure).

Layer		Activation
Conv.	(3 × 3) × 16	ReLu
Conv.	(3 × 3) × 256	ReLu
Max-Pooling	kernel size: (2 × 2)		Strides: 2
Conv.	(3 × 3) × 512	ReLu
Conv.	(3 × 3) × 2048	ReLu
Conv.	(3 × 3) × 4096	ReLu
Max-Pooling	kernel size: (2 × 2)		Strides: 2
Fully-Conn.	512	ReLu
Fully-Conn.	256	ReLu	(Dropout rate: 50%)
Fully-Conn.		Softmax	Output units: 2