Table 4.

Post-hoc analysis of the performance of single-modal and multi-modal T1 vibe Dixon network inputs for automated liver parenchyma, portal veins, and hepatic veins segmentation.

			Liver parenchyma	Portal veins	Hepatic veins
In-phase	vs	Opposed-phase	0.129	< 0.001	< 0.001
In-phase	vs	Water	1	0.331	1
In-phase	vs	In-phase, water	1	1	1
In-phase	vs	In-phase, opposed-phase	1	1	1
In-phase	vs	In-phase, opposed-phase, water	1	1	1
In-phase	vs	In-phase, opposed-phase, water, fat	1	1	1
Opposed-phase	vs	Water	1	0.365	0.112
Opposed-phase	vs	In-phase, water	0.310	< 0.001	0.002
Opposed-phase	vs	In-phase, opposed-phase	1	0.001	0.002
Opposed-phase	vs	In-phase, opposed-phase, water	0.256	< 0.001	0.003
Opposed-phase	vs	In-phase, opposed-phase, water, fat	0.302	< 0.001	0.004
Water	vs	In-phase, water	1	0.403	1
Water	vs	In-phase, opposed-phase	1	0.921	1
Water	vs	In-phase, opposed-phase, water	1	0.641	1
Water	vs	In-phase, opposed-phase, water, fat	1	0.629	1
In-phase, water	vs	In-phase, opposed-phase	1	1	1
In-phase, water	vs	In-phase, opposed-phase, water	1	1	1
In-phase, water	vs	In-phase, opposed-phase, water, fat	1	1	1
In-phase, opposed-phase	vs	In-phase, opposed-phase, water	0.444	1	1
In-phase, opposed-phase	vs	In-phase, opposed-phase, water, fat	1	1	1
In-phase, opposed-phase, water	vs	In-phase, opposed-phase, water, fat	1	1	1

The p-values were calculated with the Kruskal–Wallis test with Dunn’s multiple comparison post-hoc test using the Dice similarity coefficients reported in Table 3. Any values below the significance level α = 0.05 are highlighted in bold. The single-modal neural network inputs are in-phase; water; and opposed phase. The multi-modal neural network inputs are in-phase, water; in-phase, opposed-phase; in-phase, opposed-phase, water; and in-phase, opposed-phase, fat, water.