Multi-manifold diffeomorphic metric mapping for aligning cortical hemispheric surfaces.

Area of Science:

• Neuroimaging
• Computational anatomy
• Medical image analysis

Background:

• Cortical surface-based analysis is crucial for anatomical and functional brain studies.
• Aligning cortical hemispheric surfaces across individuals presents a significant challenge.
• Existing methods struggle with simultaneous surface and curve alignment.

Purpose of the Study:

• To introduce a novel multi-manifold large deformation diffeomorphic metric mapping (MM-LDDMM) algorithm.
• To enable simultaneous mapping of cortical surfaces and their sulcal curves.
• To improve the accuracy of anatomical variation analysis and template generation.

Main Methods:

• Developed an algorithm based on momentum conservation for diffeomorphic flow geodesics.
• Utilized initial momentum as a space for shape analysis via geodesic flow.
• Employed a gradient descent scheme to optimize initial momenta for anatomical variation.
• Applied the MM-LDDMM algorithm to construct templates for cortical surfaces and sulcal curves from 40 subjects.

Main Results:

• The MM-LDDMM algorithm optimizes initial momenta, encoding anatomical variations.
• Template generation for cortical surfaces and sulcal curves was successfully performed.
• The algorithm effectively captures highly variable regions across subjects.
• MM-LDDMM demonstrated superior surface-to-surface distance results compared to single-manifold LDDMM algorithms.

Conclusions:

• The MM-LDDMM algorithm offers a robust framework for cortical surface and sulcal curve alignment.
• It facilitates more accurate shape deformation averaging and template generation.
• This method enhances comparative analyses in neuroimaging studies.
• MM-LDDMM provides improved accuracy over existing LDDMM approaches for cortical mapping.