A team of researchers from MIT, Harvard University, and Massachusetts General Hospital has developed an AI-powered algorithm capable of automatically identifying and tracking eight key white matter fiber bundles in the brainstem using standard diffusion MRI scans. The tool, called the BrainStem Bundle Tool (BSBT), overcomes longstanding challenges in imaging the brainstem—a region critical for regulating consciousness, breathing, heart rate, sleep, and movement—due to its small size and susceptibility to motion artifacts from breathing and heartbeat. Published in the Proceedings of the National Academy of Sciences on February 6, the open-access study demonstrates that BSBT can produce detailed, accurate maps of these neural pathways, enabling new insights into how they are affected by neurological disorders. The algorithm uses a convolutional neural network trained on 30 diffusion MRI scans from the Human Connectome Project, where experts manually labeled the bundles. After training, BSBT was validated against post-mortem brain dissections and high-resolution imaging, confirming its ability to reliably identify the same bundles across multiple scans and datasets. The researchers tested BSBT’s consistency by analyzing scans from 40 volunteers taken two months apart—finding identical bundle patterns in each case. They also evaluated the contribution of each component of the AI system, ensuring that the tool’s performance was based on meaningful, accurate processing. In clinical applications, BSBT revealed distinct structural changes in patients with Parkinson’s disease, multiple sclerosis, Alzheimer’s disease, and traumatic brain injury. For example, Parkinson’s patients showed reduced fractional anisotropy—a measure of white matter integrity—in three bundles, while those with multiple sclerosis exhibited the greatest FA reductions in four bundles and volume loss in three. Alzheimer’s patients showed changes in only one bundle, but the tool detected subtle, early signs of degeneration. In traumatic brain injury, while no significant volume loss was observed, widespread FA reductions were found across most bundles. Most strikingly, BSBT was used to track the recovery of a 29-year-old man who spent seven months in a coma after a severe brain injury. The algorithm detected that his brainstem bundles were displaced but not severed, and that the size of the lesions decreased by a factor of three over time. As the bundles healed and repositioned, the patient regained consciousness, suggesting BSBT could serve as a powerful prognostic tool for coma recovery. The study’s senior authors include Mark Olchanyi, a PhD candidate at MIT; Emery N. Brown, a professor at MIT and MGH; Juan Eugenio Iglesias; and Brian Edlow. The work was supported by the National Institutes of Health, U.S. Department of Defense, and several private foundations. BSBT is now publicly available, offering researchers and clinicians a new way to study the brainstem and develop early biomarkers for neurodegenerative and traumatic brain conditions.