Projects

Navigated Orthopaedic Fixation using Ultrasound System. Pelvic fractures are usually fixed with iliosacral screws inserted under fluoroscopy, which exposes patients and staff to radiation. NOFUSS replaces fluoroscopy with a conventional ultrasound probe and an optical tracking camera: it segments bone surfaces automatically, registers them to the patient’s CT scan and guides the surgeon with targeting displays validated in a user study.

In surgeries on eight human cadavers, NOFUSS matched the screw-entry accuracy of fluoroscopy, needed no intraoperative radiation, and reduced median insertion time from 12 min 36 s to 7 min 37 s per screw.

Retinographer is a project dedicated to making retinal disease screening more accessible to rural and remote parts of Canada. We are aiming to build a cheap device, powered by artificial intelligence, that can be used by frontline healthcare workers to screen patients’ retinas. This can help identify patients who require specialized care, which is often located in large cities, and those patients who do not.

The deep-learning models behind it were published in the British Journal of Ophthalmology: an ensemble of CNNs that classified fundus photographs more accurately than board-certified ophthalmologists, and latent diffusion models that generate realistic retinal images to augment training data. I also developed hardware and algorithms to automate screening with smartphone cameras. The project received a Joule Innovation Grant from the Canadian Medical Association.