Table 7.

Key characteristics of the winning algorithms. Colored circles in Awards represent first (gold), second (silver) and third (bronze) prizes a team has won in the challenge. Abbreviations: resolution (res), binary (bin) or multiclass (mul) segmentation (seg), localization and extraction of custom ROI (ROI crop), Cross-Entropy (CE), intersection over union (IoU), centerline-boundary Dice (cb-Dice), binary topological interaction (BTI), skeleton recall (SkelRecall).

Team	Awards	Input requiring paired modalities	Training using both modalities	Additional data used	3D data used	Data augmentation	CoW ROI used	# Stages	Multi-stage	Architectures	Loss function	Ensemble	Topological optimizations
NexToU					✓	✓		2	low-res bin seg + full-res seg	nnUNet; NexToU	Dice + CE + cb-Dice + BTI	✓	cb-Dice, BTI
NIC-VICOROB-1					✓	✓		1 – 2	bin seg + mul seg	nnUNet	Dice + CE	✓
Organizers			✓		✓	✓	custom	2	ROI crop + seg	nnDetection; nnUNet	CE + IoU; Dice + CE	✓
refrain					✓	✓	custom	2	ROI crop + seg	nnUNet	Dice
sjtu_eiee_2–426lab					✓	✓	custom	2	ROI crop + seg	nnUNet	Dice + CE
UW					✓	✓		1		nnUNet	Dice + CE + TopK	✓
WilliWillsWissen			✓		✓	✓		1		nnUNet	Dice + CE + clDice + SkelRecall	✓	clDice, skeleton recall