TRD 2: Improved spatio-temporal acquisitions for in vivo mesoscale functional/structural MR imaging

In vivo non-invasive human brain imaging with MRI has lately undergone tremendous advancement. This has enabled collection of rich functional and structural information at the macroscale across the brain in a timeframe of 20-30 minutes, amendable for use in recent exciting large-scale population studies. The advancements in MR technology have also aided in the push towards higher spatial resolution, where imaging at the mesoscopic scale is starting to become feasible, with critical barriers remaining to achieve adequate specificity and sensitivity.

This project will create a program for MR technology development to overcome current “encoding limits” in MRI to achieve in vivo imaging at the mesoscopic scale: diffusion, functional, and structural imaging of the human brain at the 400–600 µm isotropic voxel size with high sensitivity and high spatial accuracy. This will push in vivo MRI from the macro-scale toward the meso-scale of cerebral cortical columns and layers and subcortical nuclei to transfer new insights from invasive animal and post mortem micro-scale imaging to non-invasive human imaging. Because fundamental modules of brain organization can be observed in the meso-scale architecture, this project will allow for in vivo investigations in humans at relevant spatial scales and approaching whole-brain coverage.

We will use a synergistic ‘from-the-ground-up’ approach joining novel encoding and reconstruction strategies with newly available instrumentation to achieve high imaging fidelity and sensitivity at the target resolution. SNR-efficient volumetric and continuous acquisitions along with highly-accelerated spatio-temporal controlled-aliasing encoding will be developed. New approaches to image encoding will be created that utilize a recently introduced combined RF and B0 shim-array technology, not only for its original intended purpose of reducing B0 inhomogeneity, but also to complement conventional encoding schemes to increase acceleration performance, improve robustness, and achieve large artifacts mitigation, particularly for multi-shot EPI. Synergistic reconstruction schemes will also be developed using emerging concepts in low-rank and multi-dimensional sub-space modeling combined with powerful Machine Learning (ML) algorithms. The proposed time-resolved reconstruction of both functional and structural data will provide a new, rich imaging dataset with hundreds of TEs and TIs from a single scan. With this approach, the detrimental image blurring from relaxation effects and distortion from B0 inhomogeneity, will also be removed to create sharp, high-fidelity datasets at mesoscopic resolutions.