A Bayesian Variable Selection Approach Yields Improved Brain Activation from Complex-Valued fMRI

• Activity Number: 399 - ASA Statistics in Imaging Section Student Paper Competition
• Type: Topic Contributed
• Date/Time: Tuesday, August 1, 2017 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Cheng-Han Yu* and Raquel Prado and Hernando Ombao and Daniel B. Rowe
• Companies: UC Santa Cruz and UC Santa Cruz and KAUST and UC Irvine and Marquette University
• Keywords: fMRI ; complex-valued time series ; variable selection ; Bayesian modeling

Abstract

Voxel fMRI time courses are complex-valued signals giving rise to magnitude and phase data. Nevertheless, most studies use only the magnitude signals and discard half of the data that could potentially contain important information. Methods that make use of complex-valued fMRI (cfMRI) data have been shown to lead to superior power in detecting active voxels when compared to magnitude-only methods, particularly for small signal-to-noise ratios. We develop models with complex-valued spike-and-slab priors on the activation parameters that are able to combine the magnitude and phase information. We present a complex-valued EM variable selection algorithm that leads to fast detection at the voxel level in cfMRI slices and also considers full inference via MCMC. Model performance is illustrated through simulation studies, including the analysis of physically-based simulated cfMRI slices. Finally, we use the approaches to detect active voxels in human cfMRI data from a healthy individual who performed unilateral tapping in a designed experiment. The proposed approach leads to activation in the expected motor-related brain regions and produces fewer spurious results than other methods.