DiffeoRaptor

DiffeoRaptor performs diffeomorphic inter-modal and contrast image registration by integrating the RaPTOR (Robust PaTch-based cOrrelation Ratio) metric with bandlimited geodesic shooting to align medical images while preserving topology.

Key Features:

• Inter-Modal/Contrast Registration: Enables alignment across different imaging modalities and contrasts for tasks that challenge intra-contrast algorithms.
• Diffeomorphic Transformations: Produces smooth, continuous deformations that preserve anatomical topology during registration.
• RaPTOR Metric Integration: Uses the Robust PaTch-based cOrrelation Ratio (RaPTOR) metric to improve robustness in inter-modal similarity estimation.
• Bandlimited Geodesic Shooting: Implements bandlimited geodesic shooting as in the FLASH (Fourier-Approximated Lie Algebras) framework for computational efficiency.
• Validation on Public Databases: Validated on three public databases for brain and abdominal image registration tasks.
• Comparative Performance: Demonstrated comparable or superior registration accuracy relative to FLASH, NiftyReg, and Symmetric Normalization (SyN), with smoother deformations than SyN in inter-modal settings.

Scientific Applications:

• Neuroimaging: Facilitates inter-modal registration of brain images for studies requiring cross-contrast alignment.
• Abdominal Imaging: Applied to abdominal image registration validated in the reported public datasets.
• Oncology Imaging: Supports multi-contrast alignment relevant to tumor localization and treatment planning.
• Multi-modal Diagnostics: Integrates information across modalities and contrasts for diagnostic procedures combining complementary imaging sources.

Methodology:

Combines the RaPTOR metric for robust inter-modal similarity estimation with bandlimited geodesic shooting (as implemented in FLASH/Fourier-Approximated Lie Algebras) to compute diffeomorphic transformations.