AI Algorithm Maps Brainstem White Matter Pathways in MRI Scans

A collaborative team from MIT, Harvard University, and Massachusetts General Hospital has introduced an artificial intelligence (AI) tool capable of automatically mapping crucial white matter pathways in the human brainstem. This advancement, reported in the 6 February issue of Proceedings of the National Academy of Sciences, addresses long-standing challenges in visualizing and analyzing these neural fibers using existing imaging technologies.

The software, named the BrainStem Bundle Tool (BSBT), leverages convolutional neural networks—a type of AI that excels at identifying patterns in visual data—to segment eight distinct axonal bundles from diffusion MRI scans. Diffusion MRI is a specialist brain imaging method that highlights how water diffuses along nerve fibers, but technical limitations have previously hindered fine-grained analysis of the brainstem’s white matter.

"The brainstem is a region of the brain that is essentially not explored because it is tough to image," said Mark Olchanyi, MIT doctoral candidate and lead researcher of the project. "People don't really understand its makeup from an imaging perspective. We need to understand what the organization of the white matter is in humans and how this organization breaks down in certain disorders."

To create BSBT, the research team trained the AI using annotated MRI data sets from the Human Connectome Project. The convolutional neural network learned to differentiate between the tightly-packed fiber tracts by analysing existing scans as well as post-mortem microscopic dissections to establish anatomical ground truths. Rigorous testing was performed to ensure stability and reproducibility of results across multiple scans and data sources.

Once developed, BSBT was tested on MRI data from patients with various neurological conditions, including Alzheimer's disease, Parkinson's disease, multiple sclerosis (MS), and traumatic brain injury (TBI). The tool measured changes in bundle volume and the directionality of water diffusion (fractional anisotropy)—key indicators of white matter integrity and potential biomarkers for disease progression.

Results indicated that, in each clinical condition, BSBT detected characteristic patterns. For example, in Parkinson’s disease, three of the eight bundles showed reduced structural integrity, while Alzheimer’s disease affected mainly one bundle. In MS patients, the greatest changes emerged in four bundles. In cases of traumatic brain injury, although there was little volume loss, most bundles showed reduced integrity.

Importantly, BSBT also provided insights into patient recovery. In one case involving a 29-year-old man in a coma following severe TBI, BSBT revealed the slow recovery and repositioning of the white matter bundles over seven months, potentially serving as a prognostic marker for outcome.

According to senior co-author Professor Emery N. Brown of MIT, “Mark’s algorithms are a significant contribution to imaging research and to our ability to understand regulation of fundamental physiology. By enhancing our capacity to image the brainstem, he offers us new access to vital physiological functions such as control of the respiratory and cardiovascular systems, temperature regulation, how we stay awake during the day, and how we sleep at night.”

BSBT is now available as an open-access tool, supporting further research and potentially improving clinical assessments of brain disorders involving the brainstem. The study was funded by numerous U.S.-based foundations and research agencies.

For more, see news.mit.edu.