Purpose: Knee osteoarthritis (OA) is diagnosed and graded with a radiographic evaluation of the joint. However, there is no clear correlation between this grade and patient functional outcome scores (Sanghi et al. 2011; Zifchock et al. 2011, Alkan et al. 2013). This leads to the need of a quantitative and precise measure to represent the knee function. In the recent years, to evaluate healthy and OA knee function, the trend was to track skin markers displacements, known to be affected by soft tissue artefacts (Leardini et al., 2005). On the other hand, with imagery-based methods, 3D model acquired either from CT scan or Magnetic resonance imaging are matched on 2D fluoroscopic images (kinematics) or several radiographs (pseudo-kinematics) (Moro-Oka et al., 2006). With objects of known geometry in the radiographic scene (bone-embedded tantalum beads or prostheses), this matching allows attaining kinematic measurement accuracy <1degree and <1mm. Without tantalum beads or prostheses, the process lacks reliability and accuracy. Thus, our goal is to present an intrinsic computation method - without any object added in the scene - and to validate its reliability on patients' images as well as its accuracy on simulated radiographs in a pseudokinematic context. Methods: The low radiation dose biplanar EOSTM system was used to acquire two orthogonal radiographs at 0degree, 15degree, 30degree, 45degree and 70degree knee flexion of the full limb during static squat. Personalised 3D models of the femur, tibia and fibula were reconstructed from the radiographs at 0degree of flexion . The knee 3D pseudo-kinematic was evaluated matching the 3D models on the orthogonal radiographs, computing the joint displacement as 3D rotations and 3D displacements as follows - for each pair of images at the different knee flexion angles: 1-Manual positioning of the 3D model in the radiographic scene by the user (Fig.a, b) 2-Extraction of each bone contour in the radiographs (Fig.b) (Chav et al., 2009) 3-Automatic estimation of the 3D model pose with a matching on the bones' contours (Fig.c) Coordinate systems were then defined on the 3D models of the femur and tibia according to Sudhoff (2007). For each position, the pseudokinematics was expressed in terms of abduction/adduction (AA), internal/external tibial rotation (IER), antero-posterior (AP), mediolateral (ML) and proximo-distal (PD) displacements with respect to knee flexion (Fig.c). 3D pseudo-kinematic measurement repeatability was assessed for 5 severe OA patients and 3 asymptomatic (AS) subjects. Intra-observer repeatability was assessed by one observer 3 times, for the 8 subjects. Similarly, 2 observers computed the 3D pseudo-kinematics for the same 8 subjects for inter-observers repeatability assessment. To evaluate the accuracy of the method, we generated synthetic radiographs in 5 known positions for the femur, tibia and fibula to create a simulated 3D pseudo-kinematic. Then, the bone contours were extracted in the synthetic radiographs, the 3D models' pose estimated and the 3D pseudo-kinematic computed. The method accuracy was evaluated as the mean absolute difference of the estimated kinematic and the simulated one. Results: The inter and intra-observer repeatability of the 3D pseudokinematic measurement were <1and <1mm, even for IER, that rely on the intricate registration of the femur and tibia tubular shapes (Table 1). Mean measurement accuracy was 0.32degree (RMS 0.28, max 0.92o) for abduction and rotation and 0.30mm (RMS 0.24mm, max 0.71mm) for the three Displacements. These results were similar to Sharma's (2012) who obtained repeatable (<1degree/<1mm) and accurate (0.5degree/0.84mm) measurements with patient images, matching 3D CT models with prostheses and biplanar fluoroscopy. It is worth noticing that we were able to compute 3D pseudo-kinematics for AS aswell as for OA subjects: the cartilage degeneration seemed to not affect adversely our results. Conclusions: We presented an intrinsic method that allows computing 3D pseudo-kinematics of the knee. For the 5 flexion positions, the acquisition time remained short (<30 minutes) and the bone matching and pseudo-kinematic calculation took approximately one hour per subject. Moreover, the results on OA patients show the feasibility of the method in a clinical context: this method could potentially offer a measure representing the knee function, and be used for follow-up of OA patients, to understand how OA affects the knee kinematics. (Figure Presented).
Kanhonou, M., Cresson, T., Clément, J., Lavoie, F., Hagemeister, N., & de Guise, J. A. (2015). A method to evaluate the 3D pseudo-kinematic of the osteoarthritic knee. Osteoarthritis and Cartilage, 23, A117–A118. https://doi.org/10.1016/j.joca.2015.02.839