Issues with the Estimation of Nonlinear Hemodynamic Response Functions for BOLD fMRI Data

• Activity Number: 177
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Xiaowen Hu*+ and Richard Gunst
• Companies: Southern Methodist University and Southern Methodist University
• Keywords: HRF ; two-gamma model ; negative scale estimates ; constrained nonlinear optimization

Abstract

Functional magnetic resonance imaging (fMRI) is a procedure that measures brain activities by detecting associated changes in blood flow. Blood oxygen level dependent (BOLD) contrast is used as the primary measurement of fMRI brain activity, and it is often modeled by a nonlinear hemodynamic response function (HRF). Most of the common neuroimaging HRF model-fitting programs use canonical HRFs, a mixture of two gamma functions, each containing three parameters. As part of the analysis and modeling of data from a recent study on Gulf Wall illness, fitted models for a large number of brain locations resulted in negative estimates of scale parameters that should be strictly positive. Subsequent investigations uncovered several possible reasons for the negative parameter estimates. These investigations and possible solutions to the problem of negative scale estimates that are currently being studied will be reported.