The signals that power most of the brain and body's most vital functions – consciousness, sleep, respiration, heart rate and movement – travel through bundles of “white matter” fibers within the brainstem, but imaging systems haven’t been capable of accurately resolve these crucial nerve cables. This has left researchers and doctors with little ability to evaluate how they’re affected by trauma or neurodegeneration.
In a brand new study, a team of researchers from MIT, Harvard University and Massachusetts General Hospital present AI-powered software able to mechanically segmenting eight different bundles in each diffusion MRI sequence.
In the open access study published on February sixth within the The research team, led by MIT graduate student Mark Olchanyi, reports that they developed their BrainStem Bundle Tool (BSBT). publicly accessiblerevealed distinct patterns of structural changes in patients with Parkinson's disease, multiple sclerosis and traumatic brain injury, and in addition make clear Alzheimer's disease. Furthermore, the study shows that BSBT retrospectively enabled tracking of bundle healing in a coma patient, reflecting the patient's seven-month journey to recovery.
“The brainstem is a region of the brain that is actually unstudied since it is difficult to image,” says Olchanyi, a doctoral candidate in MIT’s Medical Engineering and Medical Physics Program. “From an imaging perspective, people don't really understand its structure. We need to know how white matter is organized in humans and the way that organization breaks down in certain diseases.”
Adds Professor Emery N. BrownOlchanyi's doctoral advisor and co-senior creator of the study, “the brainstem is one in every of the body's most vital control centers. Mark's algorithms are a big contribution to imaging research and our ability to know the regulation of basic physiology. By improving our ability to image the brainstem, he offers us recent access to vital physiological functions equivalent to control of the respiratory and cardiovascular systems, temperature regulation, how we not sleep throughout the day, and the way we sleep at night.”
Brown is the Edward Hood Taplin Professor of Computational Neuroscience and Medical Engineering on the Picower Institute for Learning and Memory, the Institute for Medical Engineering and Science, and the Division of Brain and Cognitive Sciences at MIT. He can also be an anesthesiologist at MGH and a professor at Harvard Medical School.
Creating the algorithm
Diffusion MRI helps trace the long branches, or “axons,” that neurons extend to speak with one another. Axons are typically covered in a fatty sheath called myelin, and water diffuses along the axons throughout the myelin, also called the brain's “white matter.” Diffusion MRI can reveal this very directional water displacement. However, segmenting different axon bundles within the brainstem has proven difficult because they’re small and obscured by the flow of brain fluids and the movements produced by respiration and heartbeat.
As a part of his graduate work to higher understand the neural mechanisms underlying consciousness, Olchanyi desired to develop an AI algorithm to beat these obstacles. BSBT tracks the fiber bundles entering the brainstem from neighboring areas higher up within the brain, equivalent to the thalamus and cerebellum, to create a “probable fiber map.” An artificial intelligence module called a convolutional neural network then combines the map with multiple channels of image information from the brainstem to differentiate eight individual bundles.
To train the neural network to segment the bundles, Olchanyi “showed” it 30 live diffusion MRI scans from volunteers within the Human Connectome Project (HCP). The scans were manually annotated to show the neural network to discover the bundles. He then validated BSBT by testing his results on “ground truth” sections of postmortem human brains, where the bundles might be well delineated by microscopic examination or very slow but ultra-high-resolution imaging. After training, BSBT mastered the automated identification of the eight different fiber bundles in recent scans.
In an experiment to check consistency and reliability, Olchanyi tasked BSBT with finding the bundles in 40 volunteers who underwent separate scans two months apart. In each case, the tool was capable of find the identical bundles in the identical patients in each of their two scans. Olchanyi also tested BSBT on multiple data sets (not only the HCP) and even examined how each component of the neural network contributed to BSBT's evaluation by hindering them one after the other.
“We put the neural network to the test,” says Olchanyi. “We desired to be certain that that these plausible segmentations are literally being performed and that every individual component is getting used in a way that improves accuracy.”
Possible recent biomarkers
Once the algorithm was properly trained and validated, the research team tested whether the flexibility to segment different fiber bundles in diffusion MRI scans could make it possible to trace how the amount and structure of every bundle modified depending on disease or injury, thereby making a novel style of biomarker. Although it has been difficult to check the brainstem intimately, many studies show that neurodegenerative diseases affect the brainstem, often early of their progression.
Olchanyi, Brown, and their co-authors applied BSBT to quite a few datasets of diffusion MRI scans from patients with Alzheimer's disease, Parkinson's disease, MS, and traumatic brain injury (TBI). Patients were compared over time with controls and sometimes with themselves. In the scans, the tool measured bundle volume and “fractional anisotropy” (FA), which tracks how much water flows along myelinated axons and the way much diffuses in other directions, an indicator of the structural integrity of white matter.
In each condition, the tool found consistent patterns of change across the bundles. While just one bundle showed a big decrease in Alzheimer's, the tool showed a discount in FA in three of the eight bundles in Parkinson's. Additionally, volume loss in one other bundle was noted in patients between a baseline scan and a two-year follow-up. Patients with MS showed the best FA reductions in 4 bundles and the best volume loss in three. Meanwhile, TBI patients didn’t show significant volume loss in any of the bundles, but FA reductions were evident in most bundles.
Tests within the study showed that BSBT was more accurate than other classification methods in distinguishing between patients with health problems and controls.
BSBT can subsequently be “a crucial tool that supports current diagnostic imaging methods by providing a fine-grained assessment of white matter structure within the brainstem and, in some cases, longitudinal information,” the authors write.
Finally, within the case of a 29-year-old man who suffered a severe traumatic brain injury, Olchanyi applied BSBT to scans taken throughout the man's seven-month coma. The device showed that the person's brainstem bundles had been displaced but not severed, and showed that the amount of lesions on the nerve bundles decreased by an element of three while in a coma. As they healed, the bundles also moved back into place.
The authors wrote that BSBT “has significant prognostic potential by identifying preserved brainstem bundles which will facilitate recovery from coma.”
The study's other senior authors are Juan Eugenio Iglesias and Brian Edlow. Additional co-authors include David Schreier, Jian Li, Chiara Maffei, Annabel Sorby-Adams, Hannah Kinney, Brian Healy, Holly Freeman, Jared Shless, Christophe Destrieux and Hendry Tregidgo.
Funding for the study was provided by the National Institutes of Health, the U.S. Department of Defense, the James S. McDonnell Foundation, the Rappaport Foundation, the American SidS Institute, the American Brain Foundation, the American Academy of Neurology, the Center for Integration of Medicine and Innovative Technology, Blueprint for Neuroscience Research, and the Massachusetts Life Sciences Center.
