Comparison of resection-intersection algorithms and projection geometries in radiostereometry

Three resection-intersection algorithms were applied to simulated projections and clinical data from radiostereometric patients. On simulated data, the more advanced bundle-adjustment-based algorithms outperformed the classical Selvik algorithm, even if the error reductions were small for some parameters. On clinical data, the results were inconclusive. The two different projection geometries had a much larger influence on the error size and distribution. For the biplanar configuration, the position and motion errors were small and almost isotropic. For the uniplanar configuration, the position errors were comparably high and anisotropic, but still resulted in a high accuracy for some motion parameters at the expense of others. The simplified resection-intersection algorithm by Selvik may still be considered a good and robust algorithm for radiostereometry. More studies will have to be performed to find out how the theoretical advantages of the bundle methods can be utilized in clinical radiostereometry. (C) 2002 Elsevier Science B.V. All rights reserved.