Table A2.

p-values using Kruskal–Wallis for the dice scores obtained in the cross-validation of the transition segmentation task. On the top area we compare the performance of the different architectures on the full data, and on the bottom on the cropped data. On the diagonal we present the comparison of the same architecture on both data types. Significant values ($p<0.05$) are highlighted in green.

TZ	unet	unet++	d2unet	d2aunet	runet	aunet	daunet	r2unet	r2aunet	segresnet	highresnet	vnet	nnunet
unet	0.009	0.7540	0.7540	0.4647	0.4647	0.2506	0.7540	0.7540	0.7540	0.1172	0.9168	0.7540	0.0163
unet++	0.7540	0.009	0.6015	0.6015	0.9168	0.7540	0.9168	0.4647	0.9168	0.1172	0.9168	0.4647	0.009
d2unet	0.7540	0.6015	0.009	0.3472	0.9168	0.6015	0.7540	0.6015	0.7540	0.0472	0.9168	0.3472	0.009
d2aunet	0.4647	0.6015	0.3472	0.0163	0.2506	0.2506	0.6015	0.9168	0.6015	0.4647	0.6015	0.7540	0.0163
runet	0.4647	0.9168	0.9168	0.2506	0.0472	0.3472	0.7540	0.6015	0.7540	0.0472	0.7540	0.1745	0.0163
aunet	0.2506	0.7540	0.6015	0.2506	0.3472	0.0163	0.7540	0.3472	0.6015	0.0472	0.7540	0.1745	0.0163
daunet	0.7540	0.9168	0.7540	0.6015	0.7540	0.7540	0.009	0.3472	0.4647	0.1172	0.7540	0.4647	0.009
r2unet	0.7540	0.4647	0.6015	0.9168	0.6015	0.3472	0.3472	0.0283	0.6015	0.4647	0.3472	0.9168	0.009	Full
r2aunet	0.7540	0.9168	0.7540	0.6015	0.7540	0.6015	0.4647	0.6015	0.009	0.0758	0.7540	0.3472	0.009
segresnet	0.1172	0.1172	0.0472	0.4647	0.0472	0.0472	0.1172	0.4647	0.0758	0.0163	0.1172	0.1745	0.009
highresnet	0.9168	0.9168	0.9168	0.6015	0.7540	0.7540	0.7540	0.3472	0.7540	0.1172	0.0283	0.7540	0.009
vnet	0.7540	0.4647	0.3472	0.7540	0.1745	0.1745	0.4647	0.9168	0.3472	0.1745	0.7540	0.0472	0.009
nnunet	0.0163	0.009	0.009	0.0163	0.0163	0.0163	0.009	0.009	0.009	0.009	0.009	0.009	0.0283
	Cropped													full vs. cropped