Project summary: The overarching goals of this project are to image human white-matter circuitry ex vivo with
cutting-edge technologies, including ultra-high gradient-strength diffusion MRI (dMRI) and direct
measurements of axonal orientations with polarization-sensitive optical coherence tomography (PS-OCT); and
use the microscopic-scale PS-OCT data to inform the development of more accurate algorithms for
reconstructing brain circuitry from mesoscopic-scale dMRI. In the previous funding cycle, we constructed a
unique, 48-channel receive array coil for imaging whole human brains on the 3T Connectome MRI scanner
with high sensitivity at sub-mm resolution; we blocked human brains and imaged small samples with even
higher-resolution dMRI in a pre-clinical scanner; and we compared the mesoscopic diffusion orientation
estimates obtained from different q-space sampling schemes and reconstruction methods to microscopic
measurements of in-plane axonal orientations from PS-OCT. This work provided us with several insights on
the fiber configurations that are most challenging to dMRI techniques, as well as the relative contribution of
acquisition parameters such as spatial and angular resolution to the accuracy of dMRI. In the next funding
cycle, we propose to extend this work in ways that will allow us to use the optical microscopy data not only to
assess the accuracy of existing dMRI analysis methods, but also to engineer the next generation of methods.
First, we will use a modified PS-OCT setup that will allow us to measure not only in-plane but 3D orientations.
Second, we will extend our ex vivo dMRI data acquisition to allow microstructural modeling, and we will use the
ground-truth axonal orientations from PS-OCT to investigate whether MR-based microstructural indices are
continuous along axon bundles and different between bundles, i.e., to test the basic premise of microstructure-
guided tractography. Third, we will use the dMRI and PS-OCT data to train models for classifying different fiber
configurations (crossing, branching, turning, etc) directly from dMRI. The ultimate goal of the proposed data
acquisition and algorithmic development is to explore alternative (model-based or model-free) approaches to
tackling the ambiguities that lead to errors in conventional dMRI tractography methods. This project will take
advantage of the cutting-edge imaging technologies available at the MGH Martinos Center, and an
investigative team with complementary expertise in tractography, microstructural analysis, and optics. The PI
has a track record of developing open-source software tools and, during the previous cycle, spearheaded the
IronTract Challenge, with the participation of tractography developers from 20 institutions worldwide. We will
continue these practices of sharing software and data, as well as engaging the research community, during the
next cycle. The proposed work will produce (i) a unique set of post mortem MRI and optical imaging data that
can provide new insights on the relationship between fiber architectures and dMRI signals, and (ii) novel
algorithms that can take advantage of these data to improve the reconstruction of brain circuitry from dMRI.