MRI segmentation of the rhesus monkey brain for volumetric analysis of aging and age-related disease

The goals of this study were to develop automated methods to segment and analyze MRI scans of the rhesus monkey brain and then to apply these methods to identify brain changes in normal aging and experimentally induced hypertension. To accomplish these goals, a reference atlas consisting of 15 specific neuroanatomical structures was manually segmented from dual-echo spin-echo (DE SE) images of one behaviorally normal, young, rhesus monkey (age 7). The resulting atlas provides an anatomical reference for the automated Template Driven Segmentation (TDS) algorithm (Warfield, 1995). TDS elastically matches (or warps) the atlas to images which have been statistically segmented for gray matter, white matter and cerebrospinal fluid (CSF), thus anatomically parcellating these regions into the structures contained in the atlas. The validity of TDS was determined by comparing manual segmentations of five normal, young (age 5–12) monkey brains by five expert raters. Intra-class coefficient analysis and mean volume comparison revealed good agreement, sensitivity and specificity for 11 of the 15 structures. The relative reliability of TDS compared to one manual rater was assessed by repeating segmentations of two monkey scans. The volumes segmented for all 15 structures by the manual rater varied between the two assessments, but there was no variation in TDS volumes. This automated method was then used to segment DE SE brain images of eight young (5–12 years), six middle-aged (16–19 years) and eight old (24–30 years) monkeys. Results demonstrated a statistically significant, age-related decrease in the volumes of forebrain parenchyma, gray matter and white matter, but an increase in the volume of the third ventricle. The cognitive status of these monkeys was assessed by a battery of learning and memory tests, but volume changes were not associated with any behavioral results. TDS was also applied to scans from 10 young monkeys with experimentally induced hypertension (12 months duration). Despite hypertension-induced cognitive impairments, TDS did not detect any statistically significant volume changes in brain structures analyzed. The effectiveness of the method developed here will facilitate efficient analysis of MRI data in the search for the relationship between brain volumes and changes in brain function.