Table 4.

Evaluation of two 2D nnUNet models trained on the SAROS dataset.

Body Regions	Dice	NSD	Body Parts	Dice	NSD
Abdominal Cavity	0.988	0.997	Arm Left	0.856	0.889
Bones	0.974	0.996	Arm Right	0.868	0.905
Brain	0.991	0.997	Head	0.739	0.764
Breast Implant	0.585	0.599	Leg Left	0.942	0.945
Mediastinum	0.906	0.957	Leg Right	0.954	0.957
Muscle	0.971	0.994	Torso	0.992	0.995
Parotid Glands	0.762	0.848
Pericardium	0.966	0.985
Spinal Cord	0.955	0.986
Subcutaneous Tissue	0.98	0.995
Submandibular Glands	0.56	0.651
Thoracic Cavity	0.988	0.997
Thyroid Glands	0.825	0.931
Average	0.881	0.918		0.892	0.909

The models were evaluated using the Dice and the Normalized Surface Dice (NSD) scores. The scores were computed as the average of each label over the CT scans. Both scores range from 0 to 1, where 0 is the worst and 1 is the best possible score.