Monitoring Structural Change in the Brain: Application to Neurodegeneration

Magnetic resonance imaging (MRI) is used in clinical routine to map the brain’s morphology. Structural changes due to brain growth, aging, surgical intervention or pathological processes may be detected by image registration of time-series imaging data. To monitor structural change a three step approach is pursued here: (1) rigid registration and intensity matching of an initial (reference) and follow-up MRI scans, (2) a non-rigid registration of the scans, and (3) the segmentation of the resulting displacement field. Cros-correlation is used as a similarity measure for rigid registration. Non-rigid registration is based on a fluid dynamical model. The resulting displacement fields are usually large and, therefore, hard to interpret. For a simplified but sufficient description of such vector fields, contraction mapping is proposed to detect vector field singularities. This enables the detection and analysis of singularities of any order as critical points which reflect the topology of the vector field. An application demonstrates how this method helps to increase the understanding of pathological processes in the brain. Keywords— non-rigid registration, critical points, neurodegeneration, Alzheimer dementia