Anastasia Yendiki

Diffusion MRI tractography has received attention both for its value, as the only tool for imaging the wiring of the brain in vivo, and for its limitations, as a tool that cannot reach the precision of invasive anatomic techniques. In this talk, I will discuss the lessons learned from our experience developing tractography tools for in vivo studies, as well as comparing tractography to anatomy in ex vivo studies. What works? Diffusion MRI is remarkably good at reconstructing known anatomy. In its supervised form, where the tractography algorithm can incorporate prior anatomical knowledge, it can reconstruct the highways of the brain accurately, even from diffusion MRI data collected with a relatively modest acquisition protocol. The latest advances in hardware and sequences allow us to go well beyond these highways, and reconstruct the detailed organization of small fiber bundles, replicating the results of invasive anatomic studies. What doesn't work? Diffusion tractography cannot be used in lieu of anatomic studies, i.e., it cannot be used as the sole source of evidence that two brain regions are or are not connected to each other. That is because, in its purely unsupervised form, tractography will produce not only anatomically correct solutions, but also erroneous ones. Post mortem studies that compare diffusion MRI and ground-truth fiber trajectories from microscopy in the same samples, while limited in throughput, provide important clues on where and why tractography goes wrong. What next? These studies suggest that improvements in diffusion MRI data quality will not be sufficient to solve these errors of tractography, as they stem from a limitation of the current analytic paradigm. The path forward will involve increasing the throughput of post mortem studies to collect paired diffusion MRI and ground truth axonal orientations at a larger scale, and taking advantage of the rapid advances in machine learning to engineer the next generation of tractography algorithms.

Anastasia Yendiki is Associate Professor in Radiology at Harvard Medical School and Associate Investigator at Massachusetts General Hospital, Martinos Center for Biomedical Imaging. She is the lead PI of the center for Large-scale Imaging of Neural Circuits (LINC), a multi-institutional consortium funded by the NIH BRAIN Initiative CONNECTS program, with the aim of imaging human and non-human primate brain circuitry across scales. Anastasia received her Ph.D in Electrical Engineering: Systems from the University of Michigan at Ann Arbor, where she worked on inverse problems in tomographic image reconstruction. She then moved to the Martinos Center, first as a postdoc and then as a faculty member. There she developed TRACULA, the diffusion tractography toolbox in the FreeSurfer software package. She has served as MGH site PI in the Connectomes Related to Human Disease, and spearheaded the IronTract Challenge, an initiative bringing together tractography developers from around the world to compare and optimize the accuracy of their methods using gold standard post mortem data. Her current interests are in obtaining accurate models of white-matter fiber bundles from microscopy techniques, such as anatomic tracing and optical imaging, and developing methods that can take advantage of these post mortem models to infer connectional anatomy from in vivo diffusion MRI.

* To register for remote access, follow this link : Zoom Link