Table 1.

Comparative evaluation of the proposed method segmentation performance with the existing SOTA algorithms was reported on the JSRT dataset.

Methods	Clavicles				Lungs				Heart
	IoU	IoU-FR	Dice	Dice-FR	IoU	IoU-FR	Dice	Dice-FR	IoU	IoU-FR	Dice	Dice-FR
Human8	–	0.896	–	0.945	–	0.946	–	0.972	–	0.878	–	0.935
X-Net+54	0.848	0.874	–	0.933	0.951	0.956	–	0.978	0.881	0.884		0.938
RX-Net+54	0.838	0.859	–	0.924	0.947	0.948	–	0.973	0.876	0.876	–	0.934
TVC55	–	–	–	–	–	0.951	–	0.975	–	0.893	–	0.943
SCIA56	0.863	–	0.926	–	0.959	–	0.979	–	0.899	–	0.947	–
TMI16	0.833	–	0.929	–	0.95	-	0.974	–	0.882	–	0.937	–
DED-CNN	0.86	–	0.909	–	0.954	–	0.974	–	0.906	–	0.949	–
DED-CNN*	0.868	–	0.91	–	0.955	–	0.976	–	0.907	–	0.95	–

The performance of the models was trained using full resolution (FR) (i.e., 2048 $\times$ 2048) and reported with IoU-FR and Dice-FR. The proposed method (*) represents the segmentation results of the model trained with the augmented dataset.

Significant values are in bold.