Knee Osteoarthritis is the most common degenerative joint disease, which has no cure, and its ultimate treatment is total knee replacement (TKR) surgery. Complications from surgery and the length of recovery are difficult to predict in advance, and a variety of factors may affect them. Among the data used to study surgical outcomes, imaging is often overlooked or not fully utilised.
This project builds on the extensive experience of the research team in developing methods to automatically analyse skeletal structures in medical images. The student will develop and validate advanced Machine Learning techniques to study the ability to: (1) predict patient-reported outcomes from pre-surgery clinical data and images, (2) predict the incidence of revision surgery, and (3) generalise Machine Learning techniques across populations in Finland and the UK. The student will investigate both more classical approaches, which leverage explicit imaging-based biomarkers (e.g. joint space width), and Deep Learning techniques for end-to-end prediction. This may lead to the identification of novel musculoskeletal biomarkers for assessing TKR outcomes.
The project will leverage a unique dataset containing pre- and post-operative radiographic images, clinical data and patient-reported outcomes data of over 1800 patients who underwent TKR surgery in the UK. Furthermore, external validation will be performed using data from the Oulu University Hospital in Finland.
