The MLOps team is the newest (and one of the smallest) team within DATA. However, with the continued support from the broader team, we have already been able to make exciting progress to the advancement of healthcare delivery. Our mission is to fully leverage and adopt machine learning in the clinical setting such that we can create a real clinical impact to both institution staff as well as patients. 

One of our projects includes working directly with a Surgeon at the University Health Network (UHN) to create a real-time model that highlights areas that are safe vs. not-safe to cut during laparoscopic surgeries. The primary purpose of such a tool will be to help train new residents and to improve the overall safety of such surgeries. The team is also determined to minimize both the hardware and software requirements of running such a model to allow it to be deployed to anywhere in the world, where perhaps surgical resources and expertise might be lacking. This has been done through a light web app that leverages multiple GPUs in order to allow real-time inference anywhere in the world.

Another high-impact project in the pipeline is a collaboration with a Psychiatrist at Princess Margaret Cancer Centre to create a Complexity-Alert model for UHN oncology patients. The goal of this project is to screen the patient-reported outcomes (i.e., PROs, or questionnaires) of upcoming appointments, and identify medically relevant adverse patients and health system outcomes. These patients can then receive more appropriate care based on the model’s output, in an effort to help optimize their individual treatment plans. This project will help set the pace of precision-health medicine solutions being developed and deployed at UHN.

We are excited about the potential for these projects both inside the walls of UHN and beyond. We also look forward to future work that will continue to advance the healthcare landscape through digital approaches to care delivery. For any questions about the projects mentioned please feel free to contact data.team@uhn.ca.