A Cross-Sectional, Multi-Site Study of T1 Weighted Magnetic Resonance Image Properties

[摘要] Cross sectional Magnetic Resonance Imaging (MRI) studies of the human brain have shown age related changes in morphometric measures as well as signal properties like tissue contrast. Greater access to open data sets and an emphasis on multicenter studies make repeatability of measures important. This study takes a novel dataset and MRI from the IXI open dataset to examine age related changes in MRI signal properties across sites. We processed T1 weighted MRIs of 1131 subjects from 3 sites using established tissue segmentation tools and region of interest sampling techniques to get regional and global estimates of signal intensity and volume. Signal intensity estimates were used to calculate gray-white contrast values. We created multiple linear regression models of the age related changes in gray-white contrast, within tissue coefficient of variation (CV), and tissue volume while controlling for the effects of gender and site. In addition we compared the contrast results computed using our method to the recent methods provided with FreeSurfer.Signal intensity decreased with age in two sites while increasing with age in the third. Gray-white contrast showed a decline with age in most regions across all sites. CV increased with age across all regions, sites and tissue types. GM and WM volume decreases with age globally and regionally while CSF volume increased with age. The three methods show strong proportional bias and poor agreement.It is problematic to make inferences about MRI signal intensity without first normalizing the result. Measures of contrast and CV are both normalized views of signal intensity and both show similar if inverse relationships with age. Volume changes seen here are in agreement with previously published results, but the rates differ by site. Normalizing volume measurements by inter-cranial volume removes the site-specific rate differences. While the gray-white contrast calculation methods show strong correlation, the estimates are not comparable.Age related MRI signal property changes exist and may affect morphometric measurements. The public health relevance of this research relates to the possibility that MRI contrast may be a biomarker for use in diagnosing neurodegenerative diseases like Alzheimer’s that affect elderly populations allowing earlier detection and treatment.

[发布机构] the University of Pittsburgh