Duke University
2020–2023
Imaging the spine under load
ML Research Scientist, Medical Imaging (PhD, Biomedical Engineering)
The problem
Low back pain is common and costly, yet we have few ways to measure how a living spinal disc actually responds to everyday loading. Tracking subtle changes in disc shape and composition from MRI by hand is slow and inconsistent.
My role
PhD researcher in the Musculoskeletal Bioengineering Lab (advisor: Louis DeFrate). I built the ML and signal-processing pipeline and ran the studies. I also helped manage NIH-funded grants.
Approach
- Deep-learning models that automatically segment intervertebral discs in MR images.
- An unsupervised, diffeomorphic image-registration method that estimates voxel-wise disc strain in vivo, without anything invasive.
- Controlled loading studies (e.g., treadmill walking) imaged before and after exercise.
Outcome
- Dissertation: Development of Imaging-Based Models for Analyzing the Spatiotemporal Function of Intervertebral Discs (Duke, 2023).
- Finding: higher BMI means greater lumbar disc deformation after a walking stress test (Journal of Biomechanics, 2021).
- See Research for papers and the patent.
What I’d tell a new PhD student
Explore your interests. Try something hard, fail at it, learn from it, and build something better. And pay attention to the bleeding edge of the science, because the method you need may have been published last month.