MVComp toolbox: MultiVariate Comparisons of brain MRI features accounting for common information across measures

Abstract

Introduction Multivariate approaches have recently gained in popularity to address the physiological unspecificity of neuroimaging measures and to better characterize the complexity of biological processes underlying behavior. However, commonly used approaches are biased by the intrinsic associations between variables, or they are computationally expensive and may be more complicated to implement than standard univariate approaches. Here, we propose using the Mahalanobis distance (D2), an individual-level measure of deviation relative to a reference distribution that accounts for covariance between measures. To facilitate its use, we introduce an open-source python-based tool for computing D2 relative to a reference group or within a single individual: the MultiVariate Comparison (MVComp) toolbox (https://github.com/neuralabc/mvcomp (https://github.com/neuralabc/mvcomp)). The toolbox allows different levels of analysis (i.e., group- or subject-level), resolutions (e.g., voxel-wise, ROI-wise) and dimensions considered (e.g., combining MRI measures or WM tracts). Several example cases are presented to showcase the wide range of possible applications of MVComp and to demonstrate the functionality of the toolbox. Methods & Results The D2 framework was applied to the assessment of white matter (WM) microstructure at 1) the group-level, where D2 can be computed between a subject and a reference group to yield an individualized measure of deviation. We observed that clustering applied to D2 in the corpus callosum yields parcellations that highly resemble known topography based on neuroanatomy, suggesting that D2 provides an integrative index that meaningfully reflects the underlying microstructure. 2) At the subject level, D2 was computed between voxels to obtain a measure of (dis)similarity. The loadings of each MRI measure (i.e., its relative contribution to D2) were then extracted in voxels of interest to showcase a useful option of the MVComp toolbox. These relative contributions can provide important insights into the physiological underpinnings of differences observed. Discussion Integrative multivariate models are crucial to expand our understanding of the complex brain-behavior relationships and the multiple factors underlying disease development and progression. Our toolbox facilitates the implementation of a useful multivariate method, making it more widely accessible.

White Matter Microstructural Correlates of Cognitive and Motor Functioning Revealed via Multimodal Multivariate Analysis

“…While the metrics derived from DTI, FOD, and NODDI have been unquestionably useful in human neuroscience and often interpreted as specific physiological parameters, there is no clear one-to-one metric-parameter mapping. Further, we and others have shown that there is a high degree of correlation between metrics within and across models (Carter et al, 2022;Figley et al, 2022;Tremblay et al, 2024;Uddin et al, 2019). This suggests that strong conclusions based on theoretical interpretation of a single metric are often difficult to justify.…”

“…The distance between each subject's microstructural features (i.e., the 10 features) and the average of those features at a given voxel is calculated, resulting in a subject distance vector of the shape 10x1. Then, the distance vector is divided by the covariance between features (in our mvComp approach, this is achieved by multiplication with the pseudoinverse of the covariance matrix; see Tremblay et al, 2024). The same procedure is conducted for each voxel in every subject, resulting in voxelwise D2 maps for each of them.…”

“…Previous research has largely relied on assessing individual contrasts with univariate approaches (most commonly assessing fractional anisotropy [FA] and mean diffusivity [MD]), which problematically assumes independence when MRI metrics have overlapping characteristics that cannot be directly tied to a single microstructural or physiological source (Raghavan et al, 2021;Tardif et al, 2016). In contrast, each metric can be considered within a multivariate framework as indirect measures that provide overlapping complementary information for a more comprehensive characterization of WM microstructure (Tremblay et al, 2024). dMRI-derived metrics provide measurements of WM microstructure such as myelination and fibre characteristics (Raffelt et al, 2017;Raghavan et al, 2021;Tardif et al, 2016).…”

“…The non-specific nature of the metrics makes it difficult to disentangle the contribution of tissue properties to the imaging modalities, and give rise to shared and overlapping changes in the MR signal. To help overcome this problem, we recently developed an integrative quantitative assessment of WM microstructure using multimodal neuroimaging (Tremblay et al, 2024). We proposed that combining neuroimaging metrics that are highly correlated and represent overlapping physiological tissue properties tie intimately to the underlying microstructure.…”

“…It also explicitly incorporates the covariance between features so as to not bias the distance with collinearity. D2 has already been used in a number of human MRI studies to assess microstructural deviation as a severity marker in traumatic brain injury (Taylor et al, 2020), relate epilepsy duration to alterations in WM tract FA (Owen et al, 2021), and quantify individual microstructural variation to drive clustering in the corpus callosum (Tremblay et al, 2024).…”