A multi-resolution scheme for distortion-minimizing mapping between human subcortical structures based on geodesic...

Area of Science:

• Neuroimaging
• Computational Anatomy
• Medical Image Analysis

Background:

• Subcortical surface registration is challenging due to limited salient features.
• Existing methods struggle with consistent landmark selection for subcortical structures.
• Accurate registration is crucial for understanding subcortical neuroanatomy.

Purpose of the Study:

• To develop a landmark-free approach for subcortical surface registration.
• To enable accurate alignment of subcortical structures like hippocampi and caudates.
• To provide a robust method independent of surface modeling tools.

Main Methods:

• Developed a registration method based on measuring mesh distortion.
• Represented surfaces as points on a Riemannian manifold with a novel distance metric.
• Utilized a multi-resolution framework to minimize distortion and preserve shape.
• The method iteratively displaces source mesh vertices to match the target mesh.

Main Results:

• The proposed landmark-free method effectively registers subcortical surfaces.
• Validation demonstrated robustness to mesh variations and consistency with anatomical landmarks.
• The approach aligns well with clinical findings and outperforms a comparable surface-based method.
• The distance metric accurately quantifies mesh distortion.

Conclusions:

• The novel landmark-free approach provides accurate and robust subcortical surface registration.
• This method overcomes limitations of feature-based and landmark-dependent techniques.
• It offers a valuable tool for neuroimaging research and clinical applications involving subcortical structures.