Detecting microstructural deviations in individuals with deep diffusion MRI tractometry

Abstract

Most diffusion MRI (dMRI) studies of disease rely on statistical comparisons between large groups of patients and healthy controls to infer altered tissue state. Such studies often require data from a significant number of patients before robust inferences can be made, and clinical heterogeneity can greatly challenge their discriminative power. Moreover, for clinicians and researchers studying small datasets, rare cases, or individual patients, this approach is clearly inappropriate. There is a clear and unmet need to shift away from the current standard approach of group-wise comparisons to methods with the sensitivity for detection of altered tissue states at the individual level. This would ultimately enable the early detection and interpretation of microstructural abnormalities in individual patients, an important step towards personalised-medicine in translational imaging. To this end, Detect was developed to advance dMRI-based Tractometry towards single-subject analysis. By: 1) operating on the manifold of white matter pathways; and 2) learning normative microstructural features to better discriminate patients from controls, our framework captures idiosyncrasies in patterns along brain white matter pathways in the individual. This novel approach paves the way from traditional group-based comparisons to true personalised radiology, taking microstructural imaging from the bench to the bedside.

Prevalence of white matter pathways coming into a single diffusion MRI voxel orientation: the bottleneck issue in tractography

“…This bottleneck problem has further implications on quantitative tractography, or tractometry, where the alongtract profile of measures along multiple tracts allows a comprehensive characterization of white matter. Most methods map voxel-wise values along the tract (such as FA) [43], which we know are affected by crossing fibers, while recent studies map fixel-wise values (such as apparent fiber density), along the tract [44,45]. Yet, we show here that these measures are still not yet truly specific to one bundle.…”

Astrogliosis mapping in individual brains using multidimensional MRI

“…Acquiring such data in a clinical setting has become feasible following recent developments of multidimensional MRI clinical protocols by multiple groups [28, 46, 47]. As is the case with any other single-patient analysis methods, substantial amounts of normative data will be required to establish a reference ‘atlas’ [48, 49]. Additional limitations and confounds specific to our study include the effects of postmortem decay, fixation and resulting dehydration.…”

Through-plane super-resolution with autoencoders in diffusion magnetic resonance imaging of the developing human brain

“…Although unsupervised learning via autoencoders has been recently used in DW-MRI to cluster individuals based on their microstructural properties (Chamberland et al, 2021;Rokem, 2021)), this is to the best of our knowledge, the first unsupervised learning study for super-resolution enhancement in DW-MRI using autoencoders.…”