Table 8.

Selected learning paradigm configuration per architecture.

Model	Loss	Optimiser	Include background
nnFormer	DiceCE	SGD	✗
nnUNet	DiceCE	Adam	✗
SegmentationNet	DiceCE	Adam	✓
Swin-UNetr	DiceCE	AdamW	✓
Transbts	Dice	SGD	✗
UNetr	Dice	Adam	✗
VT-UNet	DiceCE	AdamW	✗