The current practice of diagnosing loosening of total knee arthroplasty (TKA) is based on radiological and nuclear imaging for indirect clues and can only be confirmed after surgery. Therefore, this project aims at providing a new method that automatically analyses the CT scans and calculates the displacement in terms of translation and rotation of the TKA component relative to the bone. The method should automatically detect and segment the TKA components and the bone in CT scans and calculate the induced displacement of the arthroplasty component. Furthermore, we aim to create a model that is able to calculate the probability of loosening by combining the prosthesis displacements with geometric characteristics of the arthroplasty components, local bone geometry, patients’ characteristics, and clinical outcome.