4D registration of serial brain’s MR images: a robust measure of changes applied to Alzheimer’s disease

The study of neurodegenerative pathologies like Alzheimer’s disease led to an increasing interest in the evaluation of the morphological changes in the brain over time. The recent availability of public longitudinal datasets requires new approaches to consistently measure the changes through sequences of MR images of a specific subject. Nonrigid registration represents an instrument to measure atrophy as geometric differences between pairs of scans. Among these methods, the Symmetric Log-Demons algorithm is a computationally efficient registration algorithm which defines the transformations as diffeomorphisms. In this work we propose a robust framework for the intra-subject nonrigid registration of serial MR images to evaluate the brain changes in time. The temporal consistency is obtained by integration of the structural changes at each time point into a 4-dimensional warping algorithm, to describe the subject-specific temporal trajectory. Moreover, we will show how to derive measurements of brain changes consistently along the spatial dimension, from the voxel to the regional level. Results on synthetic and real data show that, under this approach, the resulting deformations define smoother trajectories for the evolution of the changes. The accuracy of the measurements is also improved by reducing the influence of intrasubject variability and the biases affecting the data. The present method could represent the basis for the development of a robust and consistent model of longitudinal changes at the population level.