• shape modeling
• statistical shape model
• total knee replacement
• osteophytes
• patient-specific prosthesis

In the last decades the number of knee replacement surgeries has increased considerably, also thanks to the ever-better outcomes that new technologies can achieve. Among the most popular innovations, there are patient-specific prostheses. The Rejoint company, with which this project is developed, deals with the creation of customized knee prostheses designed on patient’s anatomy and fabricated with Additive Manufacturing techniques. The workflow starts with 3D modelling the patient's bone by segmenting his CT scan and continues with the model analysis to identify the most suitable prosthesis. Bones that require knee replacement are often osteoarthritic ones that present osteophytes on their surfaces. Osteophytes must not be reproduced in the bone model as they alter its shape and volume, so an operator must perform a manual segmentation to remove them since an automatic thresholding segmentation method can’t do it. This step increases the overall workflow time, and for this reason the goal of this thesis is to develop an algorithm that can automatically remove osteophytes from osteoarthritic knee bone models generated with thresholding. A statistical shape model (SSM) that can assume all the shapes that can be reproduced by a combination of Gaussian distributed deformations learned from a dataset of example shapes is developed, both for the femur and the tibia. Since the provided example shapes are healthy bones, when the model is fitted to an osteoarthritic bone it assumes the healthy shape that best resembles it, without reproducing osteophytes. The results are consistent with the state of the art. Future developments will investigate how to modify the dataset of example shapes and the model parameters to improve the final result.