Diffusion-weighted MRI

In diffusion-weighted MRI, we collect a set of images that are weighted by the displacements of water molecules in a set of different directions. We use these images to determine the preferential direction of diffusion at each brain location. Following these diffusion vectors around the brain allows us to infer the shape of white-matter axon bundles.

Algorithms for reconstructing white-matter pathways

Parsing diffusion-weighted MRI scans to tease apart different brain pathways is time consuming and requires extensive neuroanatomical expertise. We develop algorithms to extract that information automatically from the images. This allows researchers who study brain pathways in different disease populations to process large amounts of data robustly. Our algorithms rely on the observation that anatomists define brain pathways based on the structures that each pathway goes through or next to. We incorporate this knowledge into our algorithms to automate pathway reconstruction. We have tools for cross-sectional and longitudinal studies, and for processing brain scans from infancy through the entire human lifespan. Our tools are open-source and can be downloaded as part the FreeSurfer package. These tools include our atlas of 42+ white-matter tracts.

Ex vivo diffusion-weighted MRI

For in vivo neuroimaging, whole-brain scans must be acquired during the short period of time that a human subject can lie still inside a scanner. This places constraints on the resolution and overall quality of the images that we can collect. This means that, in some areas of the brain, the wiring is too complex to be reconstructed accurately from an in vivo scan. By scanning ex vivo brains, we can circumvent some of these issues. Ex vivo imaging allows us to perform much longer scans (lasting many hours or even days), to place receiver elements much closer to the brain area that we want to image, and to use scanners with much higher magnetic field strengths. [Image: Ex vivo diffusion-weighted MRI collected at 9.4 Tesla.]

Optical coherence tomography

Diffusion-weighted MRI gives us indirect measurements of the shape of white-matter axon bundles, by measuring the diffusion of water molecules along these bundles. Polarization-sensitive optical coherence tomography allows us to estimate the position and orientation of axons by detecting light that is backscattered from them. This can only be done in small brain samples, but it can serve as an independent source of measurements for the problem areas where axon configurations are too complex to resolve at the resolution that can be achieved with diffusion-weighted MRI. Our goal is to use these data not only to assess the accuracy of existing methods for inferring fiber geometries from diffusion-weighted MRI, but also to engineer the next generation of methods. Polarization-sensitive OCT

Anatomic tracing

Anatomic tracing Anatomic tracing has been the method of choice for neuroanatomists studying the wiring of the brain, and it is still the main source of gold standard data on long-range brain connections. Anatomic tracing data are very challenging to acquire and to annotate, and thus extensive collections of such data are only available in a handful of laboratories. We work with the laboratory of Dr. Suzanne Haber. Our goal is to provide a context for side-by-side comparisons of anatomic tracing and high-resolution, ex vivo diffusion-weighted MRI, by collecting both types of data in the same brains. Some of these data have been used for the IronTract Challenge.

Brain connections in depression and anxiety

We have been responsible for MRI data acquisition, as well as diffusion and structural MRI data analyses, for the Boston Adolescent Neuroimaging of Depression and Anxiety (BANDA) project. This a consortium led by MIT in collaboration with 3 clinical sites (BU, McLean Hospital, MGH) and funded by a phase II Human Connectome Project (HCP) award. The data collection includes detailed clinical characterization, as well as neuroimaging with HCP protocols, of a large number of adolescents with depression and anxiety disorders. The goal is to investigate if measures of brain structure and function obtained from MRI can be used to predict disease progression.