Automatic Reconstruction of Cerebellar Cortex from Standard MRI Using Diffeomorphic Registration of a High-Resolution Template (ARCUS)

Abstract

As cumulating evidence points to a wider range of functional tasks and neurological conditions that involve the cerebellum than previously known, the interest for examining the cerebellum with non-invasive neuroimaging techniques is growing. However, the standard methods of computational neuroanatomy for segmenting and reconstructing the cerebral cortex work poorly for the cerebellar cortex at the resolutions attainable with contemporary MRI technology because of its extremely intricate folding, making detailed and topologically correct reconstructions of the geometry of the cerebellar cortical surface unfeasible. Recently, a detailed surface reconstruction of the human cerebellar cortex was achieved from an ex-vivo specimen. These novel anatomical data enable a new reconstruction technique where this detailed surface reconstruction is morphed to subject space based on standard in-vivo MRI data. The result is an approximate reconstruction of the cerebellar cortex that requires only standard-resolution MRI data and can be used e.g., in functional neuroimaging, for integrating topographic population data or for visualizing topographic data on flattened surface patches.