Model-based Roentgen Stereophotogrammetric Analysis of Orthopaedic Implants

Dutch Technology Foundation, STW LPG.5678

September 2001 - September 2005 
B.L. Kaptein, P.W. de Bruin,  E.R. Valstar, B.C. Stoel 


Figure 3dRoentgen Stereophotogrammetric Analysis (RSA) is a highly accurate method for measuring the micro motion of endoprostheses with respect to the surrounding bone. To achieve the high accuracy, the following steps are carried out. Small roentgen opaque markers have to be introduced in the bone and attached to the prosthesis to serve as well-defined artificial landmarks. Two synchronized roentgen foci are used to obtain a stereo image of the bone and the prosthesis (Figure 1). Using a calibration object that holds tantalum markers at accurately known positions, the positions of the roentgen foci are assessed. The coordinates of the bone and prosthesis markers are accurately measured with RSA-CMS (Medis, Leiden, The Netherlands), a software package that automatically carries out RSA measurements. After the coordinates have been measured, the three-dimensional positions of the markers are reconstructed and the changes in the positions of the prosthesis markers relative to the bone markers is determined. Subsequently, the translation and rotations of the prosthesis can be calculated. The reported accuracy of RSA ranges between 0.05 and 0.5 mm for translations and between 0.15° and 1.15° for rotations (95%-confidence interval).Due to this high accuracy, RSA studies on the fixation of joint prostheses (e.g. total hip, total knee) can be performed in small study groups with a short follow-up period, while maintaining the sensitivity to accurately assess the fixation of these implants. Therefore, the exposure of large groups of patients to new, not fully validated, implants and fixation techniques (e.g. coatings, new cements) can be prevented. Former studies, in which less accurate measuring methods were used, failed to detect poor fixation at an early stage. As in these studies large numbers of patients were included, many implants were revised in order to restore the function and stability of the joint. The costs of treatment of a patient with a revision total hip prosthesis are three times higher than the costs of a primary total hip and the clinical success of the revision prosthesis is much poorer. So, the prevention of poorly designed implants and poor fixation techniques to be marketed will have a large influence on reducing the costs in total joint arthroplasty.
For conventional RSA, the bone is marked with tantalum beads with diameters of 0.5, 0.8, or 1.0 mm (Figure 2) since bony landmarks cannot be detected in a reproducible manner. Until now, in more than 3,000 patients about 30,000 tantalum beads have been implanted and no adverse effects have been reported.

Figure 1
Figure 1b
Figure 1: a) The RSA set-up consists of two synchronized roentgen tubes and a calibration box. The roentgen films are positioned underneath this box. b) The joint of interest is positioned at the cross section of the X-ray bundles, so that a stereo image is created.


Figure 2Figure 2b
Figure 2: For RSA, the bone is marked with tantalum beats and markers are attached to 
the prosthesis: a) an example of a total knee prosthesis and b) an example of a total hip prosthesis.

Most prostheses do not have distinct landmarks by themselves that can be detected in a reproducible manner. Therefore, prostheses have to be marked with at least three non-collinear markers. In some knee prostheses and in certain polyethylene cups, markers are inserted into the polyethylene. In other knee prostheses and hip stems, beads are attached to the metal surface of the implant. Preferably, the manufacturer of the implant will attach or insert these markers. As the regulatory bodies consider every type of prosthesis with additional markers as a new type of implant, a CE-mark has to be obtained for the implants for every RSA-study. This expensive and time-consuming procedure takes about one year. Therefore, this procedure holds back several implant manufacturers from starting-up collaborative RSA studies with research groups for using this technique for evaluation of their new prostheses.


The goal of the proposed project is to develop an RSA method that can be applied to any standard prosthesis without any modification, making the RSA method more accessible for clinical trials. This new method will be integrated in the current RSA software package in use at our institute to carry out the necessary validation studies. Furthermore, it is our aim to extensively validate and evaluate this new method, both theoretically, and in a clinical setting. .


The purpose of this project is the development, implementation and validation of a new approach for an accurate determination of the position and orientation of prostheses, without the demand for specific manufacturing 

adjustments to these prostheses. By this technique, the costs and time needed to start-up RSA-studies will be reduced dramatically and the RSA-technique can be applied more widely. Note that implanting tantalum beads into the surrounding bone remains necessary to obtain the position and orientation of the bone.
The technique is based on matching virtually projected boundaries (contours) of a 3D model of the prosthesis onto the actually detected contours of the prosthesis in an RSA radiograph (Figure 3). The position and orientation of the prosthesis are calculated by changing the orientation and position of this model such that the virtually projected contours form an optimal match with the actual contours. The parameters for the virtual projection, such as focus position and the position and orientation of the projection plane, are obtained by the same calibration procedure as used in conventional RSA using the calibration box as mentioned in Figure 1.

The 3-D model in model-based RSA is used to create virtually projected contours. To reach the high accuracy that is required in RSA, it is important that these virtually projected contours form a close match to the actually projected contours. For this, the model must be an accurate representation of the actual object.
A model type that is very suitable for these demands is a surface model that consists of triangular patches. The virtually projected contours of these models can be calculated relatively easy. Furthermore, the accuracy and the level of detail of the model can be improved by increasing the number of triangular patches.
Such a model can be generated from the Computer Aided Design (CAD) model that may be provided by the manufacturer of the prosthesis by using standard CAD software. A disadvantage of this method is that the accuracy of a model, that is generated this way, may not be high enough for model-based RSA. Another method to obtain surface models of orthopaedic implants is by means of reversed engineering techniques in which the actual model is scanned using a 3-D laser scanner.

Figure 3: The optimization process with an Interax tibial component illustrated by the projected contour in the left half of the RSA image:

a) After a region of interest has been specified by the user, the contour of the prosthesis is automatically detected by means of the Canny operator.   Figure 3a
b) The first estimation of the position and orientation of the prosthesis model is projected in the radiograph.   Figure 3b
c) An intermediate result of the optimization procedure: the overlap of both contours is increasing.   Figure 3c
d) The final result: of the optimization procedure: an optimum overlap of both contours has been obtained.   Figure 3d

In our pilot study, it was found that rather large dimensional tolerances exist in the implants that were used. These dimensional tolerances are not included in the three-dimensional model of the implant. As a result of this mismatch between the actual implant and the model, errors are introduced in the assessment of the position and orientation of the implant. During the start-up of this project, research will focus on finding a solution to this problem. We would like to focus on two approaches:

  1. Elimination of the mismatch between the actual prosthesis and the model. In this approach we will use reversed engineering techniques to define an accurate model of an actual prosthesis. Reversed engineering uses laser scanning techniques and commercially available software to generate a 3D prosthesis model with an accuracy of about 0.01 mm. With this technique, CAD models will not be necessary any more. In collaboration with TNO industries (Eindhoven, The Netherlands), a first trial with reversed engineering for a total knee prosthesis has been carried out and the results are promising. By using this technique we will get a much better insight in the propagation of the dimensional deviations into errors in the pose reconstruction of the actual implant.
  2. Implementation of matching algorithms that can handle incomplete contours. By this approach one is allowed to delete parts of the prosthesis contour that belong to badly scaled parts of the prosthesis. The Non Overlapping Area method that has already been implemented is principally not able to cope with incomplete contours. In the literature, several alternative matching algorithms have been described. Since the algorithms have not been used in RSA before and RSA has some specific requirements, several of these algorithms will need to be implemented and validated.




For further information, please contact:
B.L.Kaptein, PhD.
Department of Orthopaedics, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
Tel. +31 (0)71 526 4542
Fax. +31 (0)71 526 6801