Automated hippocampal segmentation by regional fluid registration of serial MRI: Validation and application in Alzheimer's disease

The application of voxel-level three-dimensional registration to serial mag netic resonance imaging (MRI) is described. This fluid registration determi nes deformation fields modeling brain change, which are consistent with a m odel describing a viscous fluid. The objective was to validate the measurem ent of hippocampal volumetric change by fluid registration in Alzheimer's d isease (AD) against current methodologies. The hippocampus was chosen for t his study because it is difficult to measure reproducibly by manual segment ation and is widely studied; however, the technique is applicable to any st ructure which can be delineated on a scan. First, suitable values for the v iscosity-body-force-ratio, alpha (0.01), and the number of iterations (300) , were established and the convergence, repeatability, linearity, and accur acy investigated and compared with expert manual segmentation. A simple mod el of hippocampal atrophy was used to compare simulated volumetric change a gainst that obtained by fluid registration. Finally the serial segmentation was compared with the current gold standard technique-expert human labelin g with a volume repeatability of similar to4%-in 27 subjects (15 normal con trols, 12 clinically diagnosed with Alzheimer's disease). The scan-rescan v olumetric consistency of serial segmentation by fluid-registration was show n to be superior to human serial segmentors (similar to2%). The mean absolu te volume difference between fluid and manual segmentation was 0.7%. Fluid registration has potential importance for tracking longitudinal structural changes in brain particularly in the context of the clinical trial where la rge numbers of subjects may have multiple MR scans. (C) 2001 Academic Press .