Automated Pulmonary Nodule Classification in Computed Tomography Images Using a Deep Convolutional Neural Network Trained by Generative Adversarial Networks

Table 4

Classification result by difference in network models.

Model	Pretraining method	Classification accuracy [%]
		Benign	Malignant	Overall
Proposed method; (AlexNet)	None	51.9	84.9	70.0
	ImageNet	40.7	93.9	70.0
	WGAN (60000)	66.7	93.9	81.7
GoogLeNet	None	40.7	84.9	65.0
	ImageNet	48.2	87.9	70.0
	WGAN (60000)	48.2	97.0	75.0
VGG16	None	29.6	90.9	63.3
	ImageNet	33.3	87.9	63.3
	WGAN (60000)	48.2	84.9	68.3