Table 3:

Quantitative evaluation results of the inter-patient (i.e., the JHU dataset) and the atlas-to-patient (i.e., the IXI dataset) brain MRI registration. Dice score and percentage of voxels with a non-positive Jacobian determinant (i.e., folded voxels) are evaluated for different methods. The bolded numbers denote the highest scores, while the italicized ones indicate the second highest.

	Inter-patient MRI		Atlas-to-patient MRI
Model	DSC	% of $\mid J_{\varphi }\mid \le 0$	DSC	% of $\mid J_{\varphi }\mid \le 0$
Affine	0.572±0.166	-	0.386±0.195	-
SyN	0.729±0.127	<0.0001	0.645±0.152	<0.0001
NiftyReg	0.723±0.131	0.061±0.093	0.645±0.167	0.020±0.046
LDDMM	0.716±0.131	<0.0001	0.680±0.135	<0.0001
deedsBCV	0.719±0.130	0.253±0.110	0.733±0.126	0.147±0.050
VoxelMorph-1	0.718±0.134	0.426±0.231	0.729±0.129	1.590±0.339
VoxelMorph-2	0.723±0.132	0.389±0.222	0.732±0.123	1.522±0.336
VoxelMorph-diff	0.715±0.137	<0.0001	0.580±0.165	<0.0001
CycleMorph	0.719±0.134	0.231±0.168	0.737±0.123	1.719±0.382
MIDIR	0.710±0.132	<0.0001	0.742±0.128	<0.0001
ViT-V-Net	0.729±0.128	0.402±0.249	0.734±0.124	1.609±0.319
PVT	0.729±0.130	0.427±0.254	0.727±0.128	1.858±0.314
CoTr	0.725±0.131	0.415±0.258	0.735±0.135	1.292±0.342
nnFormer	0.729±0.128	0.399±0.234	0.747±0.135	1.595±0.358
TransMorph-Bayes	0.744±0.125	0.389±0.241	0.753±0.123	1.560±0.333
TransMorph-diff	0.730±0.129	<0.0001	0.594±0.163	<0.0001
TransMorph-bspl	0.740±0.123	<0.0001	0.761±0.122	<0.0001
TransMorph	0.745±0.125	0.396±0.240	0.754±0.124	1.579±0.328