Full Bayesian Modeling for fMRI Group Analysis

TL;DR

This paper introduces a comprehensive Bayesian framework for analyzing fMRI data at individual and group levels, utilizing multivariate dynamic linear models and Gaussian process ANOVA to improve brain activation detection.

Contribution

It presents a novel Bayesian approach combining MDLM and Gaussian process models for fMRI group analysis, enabling online estimation and improved activation assessment.

Findings

Effective on-line estimation of brain activation curves.

Reduced false-positive rates in real resting-state data.

Flexible modeling of individual and group fMRI data.

Abstract

Functional magnetic resonance imaging or functional MRI (fMRI) is a non-invasive way to assess brain activity by detecting changes associated with blood flow. In this work, we propose a full Bayesian procedure to analyze fMRI data for individual and group stages. For the individual stage we use a multivariate dynamic linear model (MDLM), where the temporal dependence is modeled through the state parameters and the spatial dependence is modeled only locally, taking the nearest neighbors of each voxel location. For the group stage we take advantage of the posterior distribution of the state parameters obtained in the individual stage and create a new posterior distribution that represents the updated beliefs for the group analysis. Since the posterior distribution for the state parameters is indexed by the time $t$, we propose an algorithm that allows on-line estimated curves of the state parameters to be drawn and posterior probabilities computed in order to assess brain activation for both individual and group analysis. We propose an alternative analysis for the group stage using a Gaussian process ANOVA model, where the on-line estimated curves obtained in the individual stage are modeled as a functional response. Finally, we assess our proposed modeling procedure using real resting-state data and computing empirical false-positive brain activation rates.