Table 2.

Comparison of the performance of the U-Net and cGAN for different metrics. The values are from the model trained on the largest training dataset used in this study (100 patients). Statistical analyses showed no significant differences between U-Net and cGAN.

Organ	Metric	DSC	Precision	Sensitivity	Hausdorff distance
Rectum	U-Net	0.84 ± 0.08	0.85 ± 0.11	0.84 ± 0.09	7.77 ± 6.62
	GAN	0.84 ± 0.08	0.85 ± 0.11	0.85 ± 0.09	6.79 ± 5.08
Bladder	U-Net	0.88 ± 0.09	0.90 ± 0.10	0.88 ± 0.11	6.90 ± 13.18
	GAN	0.89 ± 0.08	0.90 ± 0.09	0.88 ± 0.11	6.02 ± 6.95
Femoral head left	U-Net	0.93 ± 0.06	0.93 ± 0.08	0.93 ± 0.05	4.06 ± 8.41
	GAN	0.93 ± 0.06	0.93 ± 0.08	0.93 ± 0.05	3.53 ± 4.35
Femoral head right	U-Net	0.93 ± 0.06	0.93 ± 0.08	0.93 ± 0.05	3.66 ± 4.48
	GAN	0.93 ± 0.06	0.93 ± 0.08	0.93 ± 0.05	3.59 ± 4.44