Grant: Doctoral Award - Frederick Banting and Charles Best Canada Graduate Scholarships

The spine is the most common site of secondary tumour growth for patients with breast, prostate, lung and renal cancers. These tumours can cause osteolytic and/or osteoblastic diseases and in severe cases lead to vertebral compression fracture (VCF). An effective way to treat these tumours is with Stereotactic Body Radiation Therapy (SBRT). However, a complication with SBRT is induced VCF ranging from 10% to 40% occurrence rate. Current evaluation of the risk of SBRT induced VCF is done based on a scoring of six criteria, but this score only provides a qualitative metric to determine risk of fracture. Previous work has shown that the size of the osteolytic tumour compared to overall size of the vertebral body is highly linked to VCF occurring. However, that work did not provide a quantifiable probability of fracture, only a threshold for when it is more likely to occur. The proposed work here is to use deep learning to evaluate both Computed Tomography (CT) images and Magnetic Resonance Imaging (MRI) to provide a quantitative probability of fracture occurring. This will be done by developing a network that can first isolate a vertebral body from the different images and extract information of the vertebral body including size of the tumour and vertebral body and shape of the spine. This information will then be used to determine the probability of the vertebra fracturing.