Assessment of Bias in MRI Diffusion Tensor Imaging Parameters Using SIMEX

• Activity Number: 621
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Carolyn Lauzon*+ and Brian Caffo and Bennett Landman
• Companies: Vanderbilt University and The Johns Hopkins University and Vanderbilt University
• Keywords: DTI ; bias ; SIMEX ; parameter estimation ; bias correction ; diffusion tensor imaging

Abstract

Diffusion Tensor Imaging (DTI) is a Magnetic Resonance Imaging methodology used to measure water diffusion in-vivo. Water diffusion is sensitive to anatomical and physiological properties of the surrounding local brain tissue and measurement of water diffusion provides powerful information, enabling techniques such as neuronal fiber tracking. As such, DTI metrics are clinically relevant, but the quantitative value of these metrics is attenuated by the well established presence of bias. The bias in DTI data is difficult to quantify because it is a multi-variable problem including, measurement error, six diffusion tensor parameters, and the specific DTI analysis method used. SIMEX is a Monte-Carlo based modern statistical technique that estimates bias by tracking measurement error as a function of noise added to the data, then extrapolates the results to zero noise. Here we show that DTI metrics respond smoothly to SIMEX and the method provides, i) procedural constancy regardless of the specific DTI method, ii) graphical representation of bias that is data set specific and accessible to non-statisticians/clinicians, and iv) a first time possibility for bias correction of DTI metrics.