Multidimensional analysis and detection of informative features in diffusion MRI measurements of human white matter

Abstract

The white matter contains long-range connections between different brain regions and the organization of these connections holds important implications for brain function in health and disease. Tractometry uses diffusion-weighted magnetic resonance imaging (dMRI) data to quantify tissue properties (e.g. fractional anisotropy (FA), mean diffusivity (MD), etc.), along the trajectories of these connections [1]. Statistical inference from tractometry usually either (a) averages these quantities along the length of each bundle in each individual, or (b) performs analysis point-by-point along each bundle, with group comparisons or regression models computed separately for each point along every one of the bundles. These approaches are limited in their sensitivity, in the former case, or in their statistical power, in the latter. In the present work, we developed a method based on the sparse group lasso (SGL) [2] that takes into account tissue properties measured along all of the bundles, and selects informative features by enforcing sparsity, not only at the level of individual bundles, but also across the entire set of bundles and all of the measured tissue properties. The sparsity penalties for each of these constraints is identified using a nested cross-validation scheme that guards against over-fitting and simultaneously identifies the correct level of sparsity. We demonstrate the accuracy of the method in two settings: i) In a classification setting, patients with amyotrophic lateral sclerosis (ALS) are accurately distinguished from matched controls [3]. Furthermore, SGL automatically identifies FA in the corticospinal tract as important for this classification -correctly finding the parts of the white matter known to be affected by the disease. ii) In a regression setting, dMRI is used to accurately predict "brain age" [4,5]. In this case, the weights are distributed throughout the white matter indicating that many different regions of the white matter change with development and contribute to the prediction of age. Thus, SGL makes it possible to leverage the multivariate relationship between diffusion properties measured along multiple bundles to make accurate predictions of subject characteristics while simultaneously discovering the most relevant features of the white matter for the characteristic of interest. Introduction 1 Diffusion-weighted Magnetic Resonance Imaging (dMRI) provides a unique view into 2 the physical properties of the connections that comprise the brain white matter. While 3 the measurements are usually conducted with voxels at the millimeter scale, water 4 molecules within each voxel diffuse with characteristic lengths at the micrometer scale, 5 providing aggregate information about the physical structure of the white matter, 6 including the density of axons and distribution of fiber orientations within each voxel [6]. 7 Even though metrics derived from diffusion measurements are ambiguous in terms of 8 their underlying biological interpretation [7], analyzing the variance in these properti...

Evaluating the reliability of human brain white matter tractometry

“…Instead, the outputs of pyAFQ are provided as "tidy" CSV tables (26). This means that it is compatible as inputs to the AFQ Insight tool for statistical analysis (19), but also amenable to many other statistical analysis approaches. This output should facilitate interdisciplinary use of dMRI data, as it is provided in a format that is widely used in statistics and machine learning.…”

“…This could be from QSIprep (25) or dMRIprep (https://github.com/nipreps/dmriprep). Bottom right: pyAFQ outputs can serve as inputs to AFQ Browser for further interaction and visualization (50) or AFQ Insight for statistical analysis (19). Bottom left: pyAFQ uses DIPY (27) for the implementation of dMRI algorithms.…”

Bundle analytics, a computational framework for investigating the shapes and profiles of brain pathways across populations

“…For statistical analysis of bundle profiles, BUAN uses linear mixed models. However, we are planning to include other methods such as functional data analysis 49 , 50 and predictive machine learning 37 . Bundle adjacency (BA) is used for shape similarity analysis of the same bundles across subjects.…”

“…Furthermore, tractometric analysis methods look at how bundles differ between specified groups. Recently, many studies [27][28][29][30][31][32][33][34]37 have applied statistical methods to the study of group differences along the length of tracts. These are often called bundle profiles 28 .…”

Machine learning-based prediction of motor status in glioma patients using diffusion MRI metrics along the corticospinal tract

“…This confirms previous findings [50][51][52] in which WM changes in relation to age and its variation as a function of age were investigated. In a recent study, age has been accurately predicted by FA and ADC metrics 48 . These results justify as well that if taking into account the microstructural measures, age is of critical importance in distinguishing between the two motor groups.…”