Groupwise registration of infant brain diffusion tensor images using intermediate subgroup templates

Abstract

Registering infant brain images is challenging, as the infant brain undergoes rapid changes in size, shape and tissue contrast in the first months of life. Diffusion tensor images (DTI) have relatively consistent tissue properties over the course of infancy compared to commonly used T1 or T2-weighted images, presenting great potential for infant brain registration. Moreover, groupwise registration using intermediate templates can reduce deformation and bias introduced by predefined atlases, but most methods use scalar (e.g., fractional anisotropy) images, which lack the microstructural orientation information in tensor images that can help differentiate brain structures and further improve infant image registration accuracy. Here, we propose an intermediate subgroup tensor template-based groupwise (IST-G tensor) registration approach to align infant tensor images to a sample-specific common space. First, tensor images are clustered into more homogenous subgroups using Louvain clustering based on image similarity. Within each subgroup, tensor images are aligned using DTI-toolkit to generate subgroup tensor templates, which are subsequently aligned to a sample-specific common space. Results show that our approach significantly improved registration accuracy both globally and locally compared to standard tensor-based and fractional anisotropy-based approaches. Clustering based on image similarity yielded significantly higher registration accuracy than no clustering and performed comparably to clustering by chronological age. By leveraging the consistency of features in tensor maps across early infancy and reducing deformation through intermediate subgroup tensor templates, our IST-G tensor registration framework facilitates more accurate alignment of longitudinal infant brain tensor images.