VoxelMorph

Overview

VxmDense is the U-Net registration baseline of Balakrishnan et al., included for comparison and as a light-weight backbone (0.3 M parameters with the default feature configuration). It predicts a displacement field.

Usage

from MIR.models import VxmDense
import MIR.models.configs_VoxelMorph as configs_VoxelMorph

config = configs_VoxelMorph.get_VXM_default_config()     # also get_VXM_1_config, get_VXM_BJ_config
config.img_size = fix.shape[2:]
model = VxmDense(config, gen_output=False).cuda()
flow = model((mov, fix))

Key modules

• MIR.models.VoxelMorph, MIR.models.configs_VoxelMorph

Related tutorials

• LUMIR25 VoxelMorph

References

• Balakrishnan, Zhao, Sabuncu, Guttag, Dalca. VoxelMorph: a learning framework for deformable medical image registration. IEEE TMI 2019.