Semiparametric Wavelet Estimation of the Hemodynamic Response Function (HRF) and Its Application in a Gustatory Functional Magnetic Resonance Imaging (fMRI) Study

• Activity Number: 116
• Type: Topic Contributed
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Imaging
• Author(s): Maria Aleksandra Kudela*+ and Mario Dzemidzic and Brandon Oberlin and David Kareken and Jaroslaw Harezlak
• Companies: Department of Biostatistics, Indiana University Fairbanks School of Public Health and Departments of Neurology and Radiology, Indiana University School Medicine and Department of Neurology, Indiana University School Medicine and Departments of Neurology, Psychiatry and Radiology, Indiana University School Medicine and Indiana Univ Fairbanks School of Public Health
• Keywords: brain imaging ; brain imaging ; wavelets ; HRF estimation ; semiparametric penalized regression ; mixed models

Abstract

Analyses of blood oxygen level dependent (BOLD) responses in fMRI studies usually involve modeling an observed signal using a canonical HRF. The response is assumed to be independent of brain location, stimulus type and onset latency. We propose a combined wavelet and penalized regression approach to: (1) estimate a subject- and session-specific BOLD response in a pre-specified region of interest (ROI) and (2) use the derived HRF to improve the stimulus/response association studies. Specifically, we use a discrete wavelet transform with thresholding to extract the signal at each voxel in a pre-defined ROI. Semiparametric penalized regression is used next to obtain stable HRF estimates while accounting for within-subject correlations. Our analysis shows that there is notable between-subject HRF variation. Using our method we can detect increased activity in the orbitofrontal cortex area. Comparing our method to the reference basis function HRF models, we find that differentially activated clusters contain more voxels in the left caudate and posterior cingulate areas. In summary, the method proposed increases sensitivity to activation in associative regions.