FuseMorph: accurate and time-efficient MRI 3D T1 Image deformable registration with iterative search and deep learning

Motivation: Deformable MRI brain registration is critical in various research areas. However, it is generally time-consuming. Advanced deep learning methods can enhance both efficiency and accuracy. Goal(s): This study aims to develop an optimized, iterative approach for deformable MRI image registration, targeting both improved alignment accuracy and reduced computation time. Approach: We introduce FuseMorph, which integrates a VoxelMorph-based model with iterative optimization and grid search. We benchmark its performance against ANTs' SyNCC. Results: FuseMorph improves MRI registration accuracy compared to SyNCC and significantly reduces overall processing time. Impact: This method improves MRI alignment accuracy and accelerates processing, offering a reliable tool for both research and clinical applications. It also enhances downstream tasks, such as VBM analysis, allowing them to be performed with greater speed and precision.