Photo courtesy of Faculty of Applied Sciences and Engineering, University of Toronto (credit: Neil Ta)
By: Cormac Rea
A major challenge for the Canadian healthcare system involves creating biomedical implants such as knee and hip replacements that will not require extensive follow-up or revision surgery. The demand for expensive revision surgeries continues to grow as the population ages, so there is an urgent need to reduce the revision rates. When University of Toronto researcher, Yu Zou, learned of the problem – he wanted to help.
“I’m a material scientist and really want to make materials that are useful to people and society,” said Zou.
Zou also needed to understand how and why implants fail, as post-surgery complications are attributed to various failure modes of implant materials are also associated with patients’ identity factors, such as sex, age, physical disability, activity level, and body mass index, as well as the regions that patients live in. An interdisciplinary team was needed to employ sound data science methods to identify these variables from national health data sets.
“I had a chance to speak with some hospital doctors and they told me there can be problems with the materials, specifically the durability of implants. Millions of dollars from the healthcare system are spent on joint replacements, often leading to revision surgery if certain parts don’t work well,” he added.
Supported by a Data Sciences Institute catalyst seed grant, Professors Zou (Associate Professor, Faculty of Applied Science & Engineering, University of Toronto), Qiang Sun (Associate Professor Department of Statistical Sciences and Department of Computer Science, University of Toronto), and Adele Changoor (Staff Scientist Orthopaedic Surgery, Lunenfeld-Tanenbaum Research Institute and Assistant Professor, Department of Laboratory Medicine & Pathology, Temerty Faculty of Medicine, University of Toronto) came together to employ data science methodologies combined with AI tools to analyze massive datasets on joint replacement patients to help design complex microstructure materials.
In developing new implants, Zou’s team needed to work with expensive materials that were more common in aerospace or airplane engineering to come up with microstructures that could provide the necessary strength, durability but lower elastic modulus required of a human joint.
“In our lab we use data science tools and AI tools together to help us develop and manufacture new generation materials for extreme environments,” said Zou.
Using data science insights from the hip and knee replacement revision surgery data registries, the researchers created algorithms to help drive insights from machine learning tools, in turn expediting the development of new implant materials.
“It is just like ‘cooking’ meals,” said Zou. “We tried something and tested it, tried something different and tested again, and so on. So initially the efficiency was very low and there was a very high cost, both in terms of the funding required but also the time cost for those working on the project.”
“Statistics and AI can streamline the lengthy trial-and-error process, narrowing thousands of possibilities down to a select few best options,” added Sun.
“In this way, we only need to test about ten samples instead of thousands. This greatly shortens the research cycles and associated costs,” concluded Zou.
The researchers continue to develop the data sets and necessary microstructures with the intent of further developing partnerships with hospitals, with a vision to develop a product that can be used by frontline hospital clinicians. Given that patient-specific biology (e.g. bone density, activity levels) contributes to implant survivability, the long-term goal is to build open-source tools for clinicians to be able to easily use at hospitals.
“In the future, doctors could possibly visualize an accelerated simulation of the joint implant’s suitability, based on the patient, and see how the materials would change or degrade over five, ten or twenty years,” revealed Zou.
With preliminary results of their research in place, Zou’s team was successful in applying for external funding in 2024.
“The initial support funding from DSI was very helpful in securing external funding streams,” said Zou. “The New Frontiers in Research Fund from the federal government will support us in our work for another two years.”
“Statistics and data sciences, including AI, have the potential to transform fields that heavily rely on trial-and-error approaches,” said Sun. “Their impact will likely be seen across many disciplines.”