Joint T1 and brain fiber log-demons registration using currents to model geometry

Area of Science:

• Medical image analysis
• Computational anatomy
• Neuroimaging

Background:

• Diffeomorphic registration is crucial for medical image analysis.
• Existing methods may not fully capture complex geometric structures.
• Accurate registration of anatomical structures and white matter tracts is essential.

Purpose of the Study:

• To present an extension of the diffeomorphic Geometric Demons algorithm.
• To incorporate geometric constraints into the registration process.
• To improve the accuracy of medical image registration, particularly for fiber bundles.

Main Methods:

• Developed a Log-domain Geometric Demons algorithm combining iconic and geometric constraints.
• Utilized the space of currents for shape representation, sensitive to location and structure.
• Computed deformation fields efficiently in the log-domain space.
• Registered T1 images and 65 fiber bundles simultaneously across 12 subjects.

Main Results:

• The Log-domain Geometric Demons algorithm demonstrated superior performance.
• Achieved more accurate registration compared to non-linear T1, tensor, and multi-modal T1 + Fractional Anisotropy (FA) registration.
• The framework effectively registered both anatomical images and complex white matter tracts.

Conclusions:

• The proposed Log-domain Geometric Demons algorithm offers enhanced accuracy in medical image registration.
• This method provides a robust framework for simultaneous registration of anatomical images and fiber bundles.
• The use of currents in log-domain space is effective for capturing geometric information without explicit correspondences.