12/16/2022
By Irma Silva
The Kennedy College of Science, Department of Biological Sciences, invites you to attend a M.S. thesis proposal defense by Samuel Delap entitled “Moving Morphology - Using AI for High-throughput Reconstruction of Musculoskeletal Motion.”
Candidate: Samuel Delap
Date: Tuesday, Dec. 20, 2022
Time: 2:30-4:30 p.m.
Location: Olsen 235
Committee members:
- Nicolai Konow, Assistant Professor, Biological Sciences, University of Massachusetts Lowell
- Wenjin Zhou, Assistant Professor, Computer Science, University of Massachusetts Lowell
- L Fahn-Lai, Research Associate, Design Laboratory, Harvard T. H. Chan School of Public Health
Title: Moving Morphology - Using AI for High-throughput Reconstruction of Musculoskeletal Motion
Brief abstract:
Biomechanists seek to determine how bones and joints rotate and translate and how muscles contract to coordinate musculoskeletal movements, which requires precise and accurate data. These data are often collected using the X-Ray Reconstruction of Moving Morphology (XROMM) pipeline, a set of experimental and computational techniques designed to generate precise 3D models of musculoskeletal systems. One critical step of the XROMM pipeline is digitizing bio-inert markers implanted into the target bones and muscles. Although this process yields precise and accurate data (± 0.1mm), it is also time-consuming, taking (student) researchers 25.1 ± 28.0 hours per trial to complete. To reduce the data analysis burden for biomechanics researchers, we built upon existing (published and unpublished) artificial intelligence (AI) image recognition algorithms for XROMM. First, we made the existing algorithm more accessible by reducing its complexity and moving more functionality to systems that researchers are familiar with (uploading files and selecting a folder). This approach allowed students with minimal coding experience to successfully track a trial after 3 hours of work (an 88% time reduction). However, this time reduction comes at the cost of increased tracking error (1.48 ± 2.36 pixels, n=3). Moving forward, we will continue to improve the AI pipeline’s accessibility, allowing biomechanical researchers to focus more on understanding how complex movements are coordinated, and less on how to extract their data.