2023 Human Biology Project Fellow

Written by: Dr. Tam Maiuri 

Edited by: Dr. Kaitlyn Deschamps 

Seeing Beneath the Surface: How Artificial Intelligence Could Reveal Hidden Clues in the HD Brain 

Most brain scans using MRI look like a blur of grey and white, making it difficult to see the details. But Dr. Xiangrui (“Taylor”) Zeng, a postdoctoral researcher at Massachusetts General Hospital and Harvard Medical School, is teaching computers to sharpen that picture, revealing fine layers of the brain that may hold key insights into how Huntington’s disease (HD) begins and progresses. 

Taylor’s work is supported through the HDSA Human Biology Project, a research program that funds studies focused directly on HD in people. His project, Integrating clinical MRI and ultra-high-resolution ex vivo MRI to model laminar anomaly progression in Huntington’s disease, uses artificial intelligence to map which layers of the brain’s cortex are most vulnerable to HD and at what stage they become affected. 

From Curiosity to Collaboration 

Taylor’s interest in HD began to grow from a broader fascination with how the brain develops and deteriorates. “I was deeply interested in neurodevelopmental and neurodegenerative diseases since college,” he explained. “I’m curious about how neurons develop and how they deteriorate, and how that affects cognition and behaviour.” 

He first explored psychiatric conditions like schizophrenia and depression during his neuroscience studies, and then earned a PhD in computational biology at Carnegie Mellon University. Taylor became increasingly interested in diseases that connect small, molecular changes to real, measurable changes in brain structure. 

This curiosity led him to the HD field—a space defined not only by scientific challenges but by a unique sense of community. “Alzheimer’s disease is widely studied, but Huntington’s is rarer,” he said. Because of this, every HD dataset is especially valuable. Taylor explained that researchers today are fortunate to have access to shared brain scans, donated tissue, and collaborative studies built through years of dedicated partnership from the HD community. 

Now at Mass General, Taylor is collaborating with Dr. Diana Rosas, a longtime HDSA partner whose work has shown that the brain’s cortex begins thinning even before HD symptoms appear. Her research, supported by HD families through participation in clinical studies and brain donation, laid the foundation for Taylor’s project. As he put it, this field “offers very important biological insights into the disease—but we need ways to apply that knowledge in living patients.” 

Turning Blurry Brain Scans into Layer-by-Layer Maps 

The thin outer shell of the brain, called the cortex, is made up of multiple layers, each packed with specialized neurons. In HD, those neurons begin to die, leading to cortical thinning. But this loss doesn’t happen evenly. Some layers shrink first, while others remain intact for years. Understanding which layers go first could reveal how HD progresses throughout the brain. 

The challenge is that traditional clinical MRI scans can’t show these fine structural details. To tackle this problem, Taylor is using AI trained on ultra-high-resolution images of donated HD brains. These scans capture details that can’t be measured in living patients. By teaching AI to recognize the subtle differences between cortical layers in these high-resolution images, Taylor can then apply the same pattern recognition to standard hospital MRIs, effectively “boosting” their resolution. 

He compared it to enhancing an old photo: “Fifty years ago, we had low-resolution grey-scale images; now we can recolor and enhance them. We’re doing something similar with the cortex, but in 3D.”