It has been an exciting and eventful year for the ML team at DATA. The team has almost doubled in size compared to a year ago, with more international and graduate students joining the team. With the larger team, also comes new projects!

Some of our projects are still ongoing, such as our work with the Surgery team at UHN to create a real-time model that highlights ares that are safe vs not-safe to cut during laparascopic surgeries. On top of additional refinements being made to the model, extensive research and work has been done to further optimize the web platform to run the model in real-time. In its current capacity, the platform is able to perform semantic segmentation on a live video feed at 60fps with only a 100ms delay. This has severely enhanced the user experience, and has enabled the platform to be used more frequently during surgeries.

One of the exciting projects that we have been working on helps speed up patient intervention and care at the hospital. Working together with the Radiology department at UHN, we have developed a model that can predict the likelihood of a Pneumothorax being present based on an x-ray scan. With the prediction, we can flag the radiologist to review and confirm the diagnosis quicker, and as such the patient can receive the appropriate care they require faster. Preliminary tests show that our model can save upto 2 hours compared to the current practice, which can severely help patients in dire need of clinical attention. This, however, is only the first step. Our hope is to build a general infrastructure that can alert radiologists of other time-sensitive illnesses as well, thus improving patient care for all.

Another new project involves work with the Transplant team at UHN, where a machine-learning model is being developed to predict the likelihood of mortality after a liver transplant. The model allows for dynamic configuration, meaning staff members can update target values and see (in real-time) the effect it has on risk of mortality, thus allowing them to curate personalized treatment plans for their patients to maximize survivability. Although the project is in its infancy, our data shows great promise for a positive outcome.

We are honored to be working at UHN, alongside some of the most forward-thinking members of the healthcare world. With their clinical expertise, we continue our work on advancing healthcare, and creating real clinical impact for both patients and staff.