Table 3.

Quantitative performance comparison of different models on both the validation and the hold-out test sets.

	Validation			Test			# Para. $\downarrow$	p-value
	Dice $\uparrow$	TC $\uparrow$	ASD $\downarrow$	Dice $\uparrow$	TC $\uparrow$	ASD $\downarrow$
3D U-Net	0.71 ± 0.06	0.78 ± 0.14	1.65 ± 0.28	0.69 ± 0.02	0.75 ± 0.12	2.34 ± 0.31	$\sim$6 M	< 0.0001
Unception	0.72 ± 0.13	0.77 ± 0.11	1.63 ± 0.21	0.69 ± 0.01	0.71 ± 0.13	2.31 ± 0.35	$\sim$9 M	< 0.0001
VC-Net	$\mathbf{0.78}\pm \mathbf{0.11}$	0.84 ± 0.13	$\mathbf{1.10}\pm \mathbf{0.19}$	0.72 ± 0.02	0.76 ± 0.11	1.69 ± 0.45	$\sim$24 M	< 0.0001
BRAVE-NET	0.77 ± 0.08	0.82 ± 0.12	1.12 ± 0.21	0.71 ± 0.01	0.80 ± 0.09	1.72 ± 0.23	$\sim$10 M	< 0.001
nnU-Net	0.76 ± 0.11	0.81 ± 0.15	1.15 ± 0.23	0.71 ± 0.02	0.77 ± 0.10	1.42 ± 0.25	$\sim$6 M	< 0.0001
DeepVessel-Net	0.68 ± 0.12	0.75 ± 0.09	2.52 ± 0.35	0.65 ± 0.04	0.69 ± 0.18	2.39 ± 0.41	$\sim \mathbf{0.06}\mathbf{M}$	< 0.00001
Proposed	0.77 ± 0.10	$\mathbf{0.90}\pm \mathbf{0.11}$	$\mathbf{1.10}\pm \mathbf{0.15}$	$\mathbf{0.74}\pm \mathbf{0.01}$	$\mathbf{0.87}\pm \mathbf{0.01}$	$\mathbf{1.11}\pm \mathbf{0.26}$	$\sim$8 M	–