gms | German Medical Science

Questionnaire: Precise implantation in orthopaedic surgery often requires intra-operative imaging. We have developed a computer assisted surgery (CAS) system that detects the exact location of the femoral head from just two c-arm shots by utilizing a priori knowledge of the femoral head's 3D anatomy. Much research has been performed on the anatomy of the femur but earlier evaluations were done on a very limited number of bones. There is little current literature on anatomical definitions including the third dimension, although a 3D representation of bone dimensions is presumed to be essential. In this paper we describe a database of more than 1400 femurs and how we have evaluated the 3D anatomical models for use in the CAS system.

Methods: We developed a bone database of 3D models based on CT scans from patients with different metadata (ethnic group, age etc.). The database includes over 1400 models of the femur. Using the most up-to-date methods from point distribution models (PDM) and principal components analysis (PCA), key bone geometries can be analyzed automatically for a large amount of 3D models. Additionally, procedures were established to standardize and unify anatomical measurements and define landmarks on the 3D bone models. A software tool can then automatically transfer the landmarks to all bones included in the database. In the new CAS system, 3D information on the femoral head cortex is provided. This information guides the surgeon to position a screw as closely as possible to the femoral head cortex to assure a proper fixation of the fracture. The 3D reconstruction of the femoral head cortex is performed using only two distinct images, assuming that the cortex has a spherical surface. For application in the OR, estimating the system's reconstruction accuracy is essential. Therefore, we investigated the sphericity of the femoral head in order to determine the possible maximum expected error. We defined twelve points on the femoral head and measured the distances between the pre-defined head center and those twelve points. All measurements were done on 611 datasets from the database­. The 611 datasets were from adult persons of varying gender and age.

Results and conclusions: The maximum radius of a femoral head in the dataset was approximately 30.4 mm, whereas the minimum radius was about 19.8 mm. For each distance measure the difference to the particular mean radius was computed. From these differences, we obtained a standard deviation of 0.35 mm. Looking at an interval of three times the standard deviation, which is about 99.7% of the expected shapes, the error would compute to ±1.05 mm. This is acceptable for the CAS system, where an error criterion of ±2.0 mm is required. Our analysis verified that reconstructing the femoral head as a sphere is appropriate. A CAS system can compute the radius and the location of the femoral head from just two c-arm shots with sufficient accuracy.