A Transfer Deep Generative Adversarial Network Model to Synthetic Brain CT Generation from MR Images

Table 2

Experimental environment.

Experimental environment	Parameter	Experimental environment	Parameter
Development system	Windows, Linux (Ubuntu 16.04)	CPU	Intel i7-6850K 3.60 GHz
Development environment	Matlab2018b, Pycharm	GPU	Ubuntu 16.04 (64 bit), 128 GB of RAM, NVIDIA TITAN XP(12 GB) GPU card
Development language	Python, matlab	Training time	About 31.6 hours
Model initial learning rate	1E-5	Training set	8 samples for each of the two types
Compilation framework	Tensorflow, Keras	Test set	1 sample for each of the two types
Training times	800 times	Network model	Dual3D&PatchGAN
Model dimensions	3D