Patient-specific unicompartmental knee replacements could restore osteoarthritic knees closer to their non-pathological states than off-the-shelf products. However, current state-of-the-art still relies on the skill and interpretation of technicians to reproduce healthy geometries. Therefore, this study focused on developing a reliable, semi-automated implant design method. The variation in a populations' knees was captured by statistical shape models in order to estimate new individuals’ non-pathological situations. These estimates formed the base of the implants' femoral components, and matching tibial components were created through motion guidance. All components were generated via automated B-spline parametrisation. The design is ready for pre-clinical testing and evaluation. ​