Multivariate and repeated measures (MRM): A new toolbox for dependent and multimodal group-level neuroimaging data

Abstract

Repeated measurements and multimodal data are common in neuroimaging research. Despite this, conventional approaches to group level analysis ignore these repeated measurements in favour of multiple between-subject models using contrasts of interest. This approach has a number of drawbacks as certain designs and comparisons of interest are either not possible or complex to implement. Unfortunately, even when attempting to analyse group level data within a repeated-measures framework, the methods implemented in popular software packages make potentially unrealistic assumptions about the covariance structure across the brain. In this paper, we describe how this issue can be addressed in a simple and efficient manner using the multivariate form of the familiar general linear model (GLM), as implemented in a new MATLAB toolbox. This multivariate framework is discussed, paying particular attention to methods of inference by permutation. Comparisons with existing approaches and software packages for dependent group-level neuroimaging data are made. We also demonstrate how this method is easily adapted for dependency at the group level when multiple modalities of imaging are collected from the same individuals. Follow-up of these multimodal models using linear discriminant functions (LDA) is also discussed, with applications to future studies wishing to integrate multiple scanning techniques into investigating populations of interest.

Extending Inferential Group Analysis in Type 2 Diabetic Patients with Multivariate GLM Implemented in SPM8

“…On the other hand, the multivariate GLM can handle dependent neuroimaging data (such as repeated measurement and multimodal imaging data at the group level) [7], which cannot be easily addressed with univariate approaches [6, 7]. Therefore, we recommend the use of both approaches to have a fuller picture of the underlying pathology.…”

“…The Roy’s Largest Root has been reported as a test that should be avoided because has much poorer Type I error given that it provides a lower-bound on the p-value. The others have been reported as similar [7]. Though any of these approaches could have been used, the Wilks’ Lambda was chosen because it presents the most balanced behavior, whereas the Hotelling-Lawley Trace is more liberal and the Pillai’s Trace is the most conservative (7):…”

A Hitchhiker's Guide to Functional Magnetic Resonance Imaging

“…Inducing small direct currents in the brain using transcranial magnetic stimulation (TMS) or transcranial direct current stimulation (tDCS), make possible relatively focal excitation or inhibition and, when performed concurrently with fMRI, allows the study of functional interactions (Ruff et al, 2009; Peters et al, 2013; Weber et al, 2014; Leitão et al, 2015). The rapid growth of multimodal neuroimaging techniques has triggered the parallel development of computing methods and workflows capable of analyzing the resultant complex data sets (for review Liu et al, 2015b), and has led to the development of several tools dedicated to this type of study (Casanova et al, 2007; McFarquhar et al, 2016). …”

Main Effects of Energy Drinks on Mood, Reaction Time and Brain Regions

“…As was described by McFarquhar et al [15], the multivariate form of the univariate General Linear Model could be expressed as:…”

“…Hypothesis testing in the multivariate GLM is based on the contrast (2) The hypotheses testing approach is conceptualized as combining hypotheses about groups using A, and hypotheses about repeated measures using C. The test statistics from the multivariate GLM was carried out using the MRM toolbox [15] considering the Wilks'lambda statistics, and a voxel-level permutation test to provide a FEW correction, thresholding the images at p FWE < 0.05 with 5000 permutations. Table 1 show the main significant interactions from POMS questionnaire analysis.…”