Table 2.

Segmentation performance on internal 2D ultrasound and CEUS domain-shift tests.

Model	Evaluation set	N images	Dice	IoU	Boundary F1	HD95	Pixel ECE
Residual U-Net	Internal 2D ultrasound test	382	0.750 (0.191)	0.632 (0.216)	0.167 (0.123)	97.986 (50.627)	0.039 (0.044)
Residual U-Net	CEUS domain-shift test	170	0.468 (0.190)	0.326 (0.167)	0.037 (0.035)	199.683 (65.159)	0.314 (0.143)
Combined augmentation	Internal 2D ultrasound test	382	0.745 (0.196)	0.628 (0.221)	0.178 (0.133)	93.448 (50.716)	0.037 (0.042)
Combined augmentation	CEUS domain-shift test	170	0.476 (0.192)	0.334 (0.171)	0.041 (0.032)	197.557 (64.357)	0.297 (0.138)
Blur augmentation	Internal 2D ultrasound test	382	0.748 (0.196)	0.631 (0.220)	0.168 (0.126)	99.917 (52.583)	0.037 (0.042)
Blur augmentation	CEUS domain-shift test	170	0.477 (0.212)	0.338 (0.187)	0.050 (0.037)	203.073 (64.607)	0.275 (0.132)
Photometric augmentation	Internal 2D ultrasound test	382	0.752 (0.190)	0.635 (0.217)	0.173 (0.127)	96.156 (49.826)	0.038 (0.043)
Photometric augmentation	CEUS domain-shift test	170	0.465 (0.190)	0.323 (0.165)	0.043 (0.033)	201.750 (64.956)	0.302 (0.124)
Noise augmentation	Internal 2D ultrasound test	382	0.746 (0.193)	0.627 (0.217)	0.167 (0.118)	98.379 (51.834)	0.039 (0.045)
Noise augmentation	CEUS domain-shift test	170	0.429 (0.158)	0.287 (0.136)	0.017 (0.019)	208.041 (68.450)	0.439 (0.126)
Attention U-Net	Internal 2D ultrasound test	382	0.749 (0.190)	0.631 (0.217)	0.168 (0.129)	98.487 (53.943)	0.038 (0.042)
Attention U-Net	CEUS domain-shift test	170	0.429 (0.175)	0.289 (0.146)	0.035 (0.028)	204.505 (65.856)	0.328 (0.132)
U-Net	Internal 2D ultrasound test	382	0.752 (0.189)	0.635 (0.214)	0.166 (0.123)	98.914 (51.314)	0.039 (0.044)
U-Net	CEUS domain-shift test	170	0.439 (0.162)	0.296 (0.141)	0.024 (0.028)	202.188 (66.009)	0.374 (0.114)

Values are mean (SD).