Landmark Segmentation and Image Registration of Abdominal Point-of-Care Ultrasound Images using Deep Learning

According to the Canadian Liver Foundation, non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease in Canada, affecting 20% of the Canadian population. Over decades, NAFLD can progress to liver cirrhosis, with deceased liver function and decompensation associated with mortality and requiring consideration of liver transplantation. Thus, early diagnosis is crucial to implement therapeutic strategies that prevent further progression of the disease. Current early detection and diagnosis of NAFLD/NASH include invasive biopsies, elastography, and magnetic resonance imaging/computed tomography scans, all of which are expensive, requiring vast medical expertise and complex workflows. The focus of this research is to develop deep learning algorithms for image segmentation, registration and volume estimation of the liver using two-dimensional point-of-care ultrasound images. Oncoustics can leverage these potential algorithms to hardware agnostic software. Such technology would assist clinicians in observing and detecting possible liver disease to improve workflow speed.

Faculty Supervisor:

Eranga Ukwatta

Student:

Partner:

Oncoustics

Discipline:

Engineering

Sector:

Manufacturing; Professional, scientific and technical services

University:

University of Guelph

Program: