In total knee replacement, the clinical outcome is significantly influenced by the implant design. Different concepts of gender-, race- or patient-specific designs have been developed. This is based on reported anatomical differences in the knee shape (morphology), as reported in literature. However, little is known whether these differences originate from phenotypic differences between men and women or different ethnicities, or e.g. overall body size, and thus can be eliminated by a simple scaling. Therefore, we develop parametric models to quantify the knee morphology. This includes fully automatic methods for landmark recognition and feature extraction. Subsequently, we use statistical, correlation, and cluster analysis to investigate the knee morphology for patient specific implant design.