This work presents an approach towards reconstructing ligament and tendon attachment sites from 3D medical image data. We apply statistical shape models with an additional free form step to reconstruct the anatomical shape of the distal femur from CT data. After the shape fitting process the locations of ligament attachment sites, which are incorporated in the geometric model, are extracted and their positions optimized via an image based approach. To show the potential of our method we manually extracted the surface of 11 distal femora and their corresponding ligament attachment sites in clinical CT data and compared them to the results of our method.