Fig. 5.

Clavicle segmentation results of real and synthetic data-trained segmentation models. Three subject images where the patient is adversarial-positioned ($\alpha =-{20}^{\circ }$ and ${20}^{\circ }$) and well-positioned ($\alpha =0^{\circ }$). The segmentation contours color for ground truth (GT) are red, U-Net(Real) are blue and U-Net($\text{Real}+{\text{Syn}}_{\mathit{cen}}$) are cyan, and U-Net($\text{Real}+{\text{Syn}}_{\mathit{all}}$) are yellow. The Dice score analysis with respect to GT for each patient and angle are shown below each image, left is U-Net(Real), middle is U-Net($\text{Real}+{\text{Syn}}_{\mathit{cen}}$), and right is U-Net($\text{Real}+{\text{Syn}}_{\mathit{all}}$). Bold indicates the highest value among the triplets, and underline indicates the second-highest value.