Meeting Banner
Abstract #2004

A Framework for Analysis of Living Phantom Data in a Multicenter DTI Study

Lindsay Walker1, Nicholas Lange2, Lin-Ching Chang3, Carlo Pierpaoli1, Tthe Brain Development Cooperative Group4

1STBB, NICHD, National Institutes of Health, Bethesda, MD, United States; 2Departments of Psychiatry & Biostatistics, Harvard Schools of Medicine & Public Health, Boston, MA, United States; 3Department of Electronic Engineering & Computer Science, The Catholic University of America, Washington, DC, United States;

In Multicenter DTI studies it is important to assess the impact of site related variability to meaningfully pool quantitative imaging data across sites. We propose two tools for the assessment of living phantom data. 1) Outlier identification using the median of tensor derived metrics, and 2) analysis of inter-site and intra-site variability for assessment of sources of variance. High variability is seen at tissue interfaces, suggesting inconsistent morphology is a significant issue, even in a single subject. Thus, a single subject should be used as living phantom in multicenter studies in order to avoid the confound of inter-subject anatomical variability.