Joint surface modeling with thin-plate splines.

Mathematical joint surface models based on experimentally determined data points can be used to investigate joint characteristics such as curvature, congruency, cartilage thickness, joint contact areas, as well as to provide geometric information well suited for finite element analysis. Commonly, surface modeling methods are based on B-splines, which involve tensor products. These methods have had success; however, they are limited due to the complex organizational aspect of working with surface patches, and modeling unordered, scattered experimental data points. An alternative method for mathematical joint surface modeling is presented based on the thin-plate spline (TPS). It has the advantage that it does not involve surface patches, and can model scattered data points without experimental data preparation. An analytical surface was developed and modeled with the TPS to quantify its interpolating and smoothing characteristics. Some limitations of the TPS include discontinuity of curvature at exactly the experimental surface data points, and numerical problems dealing with data sets in excess of 2000 points. However, suggestions for overcoming these limitations are presented. Testing the TPS with real experimental data, the patellofemoral joint of a cat was measured with multistation digital photogrammetry and modeled using the TPS to determine cartilage thicknesses and surface curvature. The cartilage thickness distribution ranged between 100 to 550 microns on the patella, and 100 to 300 microns on the femur. It was found that the TPS was an effective tool for modeling joint surfaces because no preparation of the experimental data points was necessary, and the resulting unique function representing the entire surface does not involve surface patches. A detailed algorithm is presented for implementation of the TPS.

Precise measurement of cat patellofemoral joint surface geometry with multistation digital photogrammetry.

Three-dimensional joint models are important tools for investigating mechanisms related to normal and pathological joints. Often these models necessitate accurate three-dimensional joint surface geometric data so that reliable model results can be obtained; however, in models based on small joints, this is often problematic due to limitations of the present techniques. These limitations include insufficient measurement precision the requirement of contact for the measurement process, and lack of entire joint description. This study presents a new non-contact method for precise determination of entire joint surfaces using multistation digital photogrammetry (MDPG) and is demonstrated by determining the cartilage and subchondral bone surfaces of the cat patellofemoral (PF) joint. The digital camera-lens setup was precisely calibrated using 16 photographs arranged to achieve highly convergent geometry to estimate interior and distortion parameters of the camera-lens setup. Subsequently, six photographs of each joint surface were then acquired for surface measurement. The digital images were directly imported to a computer and newly introduced semi-automatic computer algorithms were used to precisely determine the image coordinates. Finally, a rigorous mathematical procedure named the bundle adjustment was used to determine the three-dimensional coordinates of the joint surfaces and to estimate the precision of the coordinates. These estimations were validated by comparing the MDPG measurements of a cylinder and plane to an analytical model. The joint surfaces were successfully measured using the MDPG method with mean precision estimates in the least favorable coordinate direction being 10.3 microns for subchondral bone and 17.9 microns for cartilage. The difference in measurement precision for bone and cartilage primarily reflects differences in the translucent properties of the surfaces.