Table 9 Experimental results of multiple loss functions.

From: Hybrid transformer and convolution iteratively optimized pyramid network for brain large deformation image registration

Datasets

Methods

\(\:\text{D}\text{S}\text{C}\uparrow\:\)

\(\:\text{A}\text{S}\text{S}\text{D}\downarrow\:\)

\(\:\left|{J}_{\varnothing\:}\right|<0\downarrow\:\)

LPBA40

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{1}\mathcal{L}}_{diffusion}\left({{\varnothing}}_{1}\right)\)(Ours)

0.737±0.013

1.404±0.066

< 0.017%

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{2}\mathcal{L}}_{bending}\left({{\varnothing}}_{1}\right)\)

0.724±0.016

1.668±0.099

< 0.019%

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{1}\mathcal{L}}_{diffusion}\left({{\varnothing}}_{1}\right)+{{\gamma\:}_{2}\mathcal{L}}_{bending}\left({{\varnothing}}_{1}\right)\)

0.738±0.013

1.408±0.062

< 0.011%

Mindboggle101

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{1}\mathcal{L}}_{diffusion}\left({{\varnothing}}_{1}\right)\)(Ours)

0.637±0.013

0.878±0.023

< 0.050%

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{2}\mathcal{L}}_{bending}\left({{\varnothing}}_{1}\right)\)

0.633±0.016

0.856±0.025

< 0.051%

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{1}\mathcal{L}}_{diffusion}\left({{\varnothing}}_{1}\right)+{{\gamma\:}_{2}\mathcal{L}}_{bending}\left({{\varnothing}}_{1}\right)\)

0.635±0.013

0.983±0.035

< 0.042%

OASIS

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{1}\mathcal{L}}_{diffusion}\left({{\varnothing}}_{1}\right)\)(Ours)

0.830±0.020

0.786±0.046

< 0.030%

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{2}\mathcal{L}}_{bending}\left({{\varnothing}}_{1}\right)\)

0.810± 0.024

0.828±0.061

< 0.047%

\(\:{\mathcal{L}}_{sim}+{{\gamma\:}_{1}\mathcal{L}}_{diffusion}\left({{\varnothing}}_{1}\right)+{{\gamma\:}_{2}\mathcal{L}}_{bending}\left({{\varnothing}}_{1}\right)\)

0.821±0.024

0.821±0.056

< 0.035%

Back to article page