Innovative technology from University of Alberta computing scientists enables clinicians to get a more in-depth view of patients’ joints in motion.
“What we’re trying to do is create a complete dynamic model of the entire knee joint in motion, including bone, muscle and cartilage,” explains Pierre Boulanger, professor in the Department of Computing Science and Cisco Chair in Healthcare Solutions. “We’re trying to use advanced neural network techniques to segment these knee structures using standard MRI and then track them using a novel real-time open MRI machine to capture the motion.”
The technique uses a deep learning algorithm to segment the knee anatomy automatically from examples provided by clinicians allowing the dynamic movements of individual bones inside a patient’s knee to shine through. Additionally, it lowers the harmful effects of X-ray radiation due to its MRI imaging base.
While the researchers used the knee joint as a case study, they said the technology can be applied to any joint or any motion in the body.
Physiotherapists and orthopedic surgeons may also be able to use this technology to understand how their work affects the body over time, which will come in useful when a pronounced issue may be hard to diagnose.
“This is the beginning of a new generation of imaging tools that can create personalized models of individual patients’ joints,” adds Boulanger.
Boulanger, who worked on the study with Constance Lebrun and Fateme Esfandiarpour, U of A family physicians from the Glen Sather Sports Medicine Clinic, explains: “Ultimately, our goal is to develop patient-specific modelling, technology and tools for use in a clinical setting.”
This study, “A Structured Deep-Learning Based Approach for the Automated Segmentation of Human Leg Muscle from 3D MRI,” was presented at the 2017 Conference on Computer and Robot Vision.