Hello, I'm Ziyang Long

Bioengineering Ph.D. Candidate & MRI Research Specialist

Welcome to my corner of the web! 🧲 When I’m not wrestling with magnetic fields or persuading protons to cooperate, I’m probably explaining to friends why their MRI scans take forever. My mission: Bridging physics and AI to create MRI magic.

Ziyang Long

Research Highlights

Teaching AI to Fix Bent MR Images

What if distortions in MRI scans weren't just annoying artifacts, but actually useful clues? I built a physics-informed diffusion model that learns from distortions to correct them β€” basically using the problem as part of the solution. Radiologists rated the results a full point higher on average.

Learn more

An AI Brain for Long MRI Workflows

MRI analysis pipelines are ridiculously long: reconstruct β†’ denoise β†’ segment β†’ quantify β†’ write a report. I built BCER, an AI agent that breaks this into brain / cerebellum / extremities (yes, named after brain parts) and handles the chaos. It took workflow success from 22% to 93% on cardiac tasks and hit 99% across 1,000+ cases.

Learn more

About Me

I'm a Ph.D. candidate in Bioengineering at UCLA with a passion for making MRIs better for everyone. πŸŽ“ My research focuses on advanced shimming techniques and foundation models that make MRI scans clearer, faster, and more accessible.
When I'm not in the lab, I'm exploring all that Los Angeles has to offer! 🌴 After three years in LA, I've come to love the city's vibrant culture, diverse food scene, and year-round sunshine. You might find me playing tennis on weekends 🎾 or escaping to the mountains for skiing during winter breaks ⛷️.
I earned my Bachelor's in Electronic Information Engineering through a joint program between UESTC and the University of Glasgow (2018-2022), which gave me a global perspective I bring to all my work and life. Turns out magnetic fields are just as fascinating on different continents! 🌍

Recent Publications

View all publications β†’

Let Distortion Guide Restoration (DGR): A Physics-informed Learning Framework for Prostate Diffusion MRI

Z. Long et al., arXiv preprint, 2025

Ever tried to read text through a funhouse mirror? That's what geometric distortion does to prostate MRI scans. In this work, I turned that around β€” the distortion itself teaches a diffusion model how to undo it. The approach outperformed existing methods by a solid margin, both in math-y metrics (NMSE) and in what actually matters: radiologists thought the images looked better.

Improving Fat-Saturation Robustness with Local Bβ‚€ Shimming

Z. Long et al., Magnetic Resonance in Medicine, 2025

If you've ever had a knee MRI, there's a decent chance the fat suppression wasn't perfectly uniform (sorry!). I designed a custom shimming coil that sits right where the problem is β€” near the joint β€” and cleans up the magnetic field locally. Think of it as giving the MRI scanner a pair of reading glasses for your extremities.