Multidimensional analysis and detection of informative features in human brain 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) to quantify tissue properties along the trajectories of these connections. Statistical inference from tractometry usually either averages these quantities along the length of each fiber bundle or computes regression models 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. We developed a method based on the sparse group lasso (SGL) that takes into account tissue properties along all of the bundles and selects informative features by enforcing both global and bundle-level sparsity. We demonstrate the performance of the method in two settings: i) in a classification setting, patients with amyotrophic lateral sclerosis (ALS) are accurately distinguished from matched controls. Furthermore, SGL identifies 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, SGL accurately predicts “brain age.” In this case, the weights are distributed throughout the white matter indicating that many different regions of the white matter change over the lifespan. Thus, SGL leverages the multivariate relationships between diffusion properties in multiple bundles to make accurate phenotypic predictions while simultaneously discovering the most relevant features of the white matter.

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

“…In a recent study, age has been accurately predicted by FA and ADC metrics. 105 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.…”

An open, analysis-ready, and quality controlled resource for pediatric brain white-matter research

“…Failure to do so may cause spurious associations or degrade predictive performance. To demonstrate this, we selected participant age as a representative phenotypic benchmark because (i) it operates on a natural scale with meaningful units and (ii) despite the unique methodological challenges it presents for biomarker identi cation ( Nelson et al, 2020 ), brain age prediction may be diagnostic of overall brain health ( Franke et al, 2010 ; Cole et al, 2019 ; Richie-Halford et al, 2021 ). We observed the effect of varying QC cutoff on the predictive performance of an age prediction model.…”

“…We retain only the mean diffusivity (MD) and the fractional anisotropy (FA) from a diffusion kurtosis imaging (DKI) model ( Jensen et al, 2005 ), implemented in DIPY ( Henriques et al, 2021 ), and impute missing bundles using median imputation as implemented by scikit-learn ’s . Because the HBN-POD2 bundle profiles exhibit strong site effects ( Richie-Halford et al, 2021 ), we used the ComBat harmonization method to robustly adjust for site effects in the tract profiles. Initially designed to correct for site effects in gene expression studies ( Johnson et al, 2007 ), ComBat employs a parametric empirical Bayes approach to adjust for batch effects and has since been applied to multi-site cortical thickness measurements ( Fortin et al, 2018 ), multi-site DTI studies ( Fortin et al, 2017 ), and functional MRI data in the Adolescent Brain Cognitive Development Study (ABCD) ( Nielson et al, 2018 ).…”

“…We further demonstrated the impact of QC in a benchmark age prediction task (Figure 9). In this case, the increase in model performance from imposing a QC cutoff is intuitive: we know from Figure 8 that participants with low QC scores have reduced MD, but MD also decreases as participants mature (Yeatman et al, 2014;Richie-Halford et al, 2021). Eliminating participants with low QC therefore removes the ones who may look artificially older from the analysis, improving overall performance.…”

Automated Multiclass Artifact Detection in Diffusion MRI Volumes via 3D Residual Squeeze-and-Excitation Convolutional Neural Networks

“…The HBN dataset was selected because the age range was appropriate for the desired age-based analysis, and because many of the MRI scans were known to be corrupted with artifacts based on our prior experience with this dataset ( Luna et al, 2021 ). Prior research has demonstrated a linear relationship between participant age and FA ( Rathee et al, 2016 ; Richie-Halford et al, 2021 ). Exploiting this fact, a proof-of-concept analysis examined the bivariate correlation between mean whole-brain FA and participant age.…”