Spherical demons: fast diffeomorphic landmark-free surface registration

Area of Science:

• Medical image analysis
• Computational anatomy
• Neuroimaging

Background:

• Accurate registration of spherical images and cortical surfaces is crucial for comparative neuroimaging studies.
• Existing registration algorithms may lack speed, accuracy, or the ability to ensure diffeomorphic transformations.

Purpose of the Study:

• To introduce the Spherical Demons algorithm for efficient and accurate registration of spherical images and cortical surfaces.
• To demonstrate its utility in transferring segmentation labels for neuroimaging applications.

Main Methods:

• The Spherical Demons algorithm utilizes spherical vector spline interpolation and iterative smoothing for regularization.
• It is based on one-parameter subgroups of diffeomorphisms, ensuring a diffeomorphic and fast registration process.
• The algorithm can be adapted to register images to probabilistic atlases, with variants for warping either the atlas or the subject.

Main Results:

• Registration of large cortical surface meshes (>160k nodes) is achieved in under 5 minutes (warping atlas) or 3 minutes (warping subject).
• The method demonstrates comparable speed to non-diffeomorphic registration algorithms and favorable accuracy against FreeSurfer.
• Validated in transferring segmentation labels for in vivo cortical surface parcellation and ex vivo Brodmann area localization.

Conclusions:

• Spherical Demons provides a fast, accurate, and diffeomorphic approach for spherical image and cortical surface registration.
• Its efficiency and accuracy make it suitable for transferring labels in neuroimaging, aiding in cortical parcellation and localization studies.