Recognition of Thyroid Ultrasound Standard Plane Images Based on Residual Network

Table 1

Architecture of ResNet-18.

Block	Layers	Output size
Input	Input layer	500 × 500 × 1
Conv 1	7 × 7 conv	250 × 250 × 64
Pooling	2 × 2 max pooling	125 × 125 × 64
Residual block 1	3 × 3 conv	125 × 125 × 64
	3 × 3 conv
Residual block 2	3 × 3 conv	125 × 125 × 64
	3 × 3 conv
Residual block 3	3 × 3 conv	63 × 63 × 128
	3 × 3 conv
Residual block 4	3 × 3 conv	63 × 63 × 128
	3 × 3 conv
Residual block 5	3 × 3 conv	32 × 32 × 256
	3 × 3 conv
Residual block 6	3 × 3 conv	32 × 32 × 256
	3 × 3 conv
Residual block 7	3 × 3 conv	16 × 16 × 512
	3 × 3 conv
Residual block 8	3 × 3 conv	16 × 16 × 512
	3 × 3 conv
Avg pooling	16 × 16 avg pooling	1 × 1 × 512
FC layer	FC softmax	1 × 1 × 8