PAMM · Proc. Appl. Math. Mech. 17, 217 – 218 (2017) / DOI 10.1002/pamm.201710078 Personalized simulation of a bone–implant–system during a step forward Michael Roland 1, * , Thorsten Tjardes 2 , Bertil Bouillon 2 , and Stefan Diebels 1 1 Chair of Applied Mechanics, Saarland University, Campus A4 2, D–66123 Saarbrücken, Germany 2 Department of Trauma and Orthopedic Surgery, University of Witten/Herdecke, Faculty of Health – School of Medicine, Cologne Merheim Medical Center, Ostmerheimerstr. 200, D–51109 Cologne, Germany In this study, a workflow is presented for a personalized simulation of a bone–implant–system of a fractured tibia during a step forward. The workflow is based on routinely acquired tomographic data, segmentation, material assignment, mesh generation, setup of realistic boundary conditions and a finite element simulation (FEM). In the absence of patient–specific monitoring data, a dataset from the OrthoLoad database is implemented as individualized boundary conditions. This allows a simulation of the bone–implant–system close to reality of patient’s step forward. c  2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 1 Introduction Fractures of the long bones, e.g. tibia or femur, are treated surgically. This means that implants are used that allow early post operative weight bearing and physiotherapy of the injured limb. To deduce a better understanding of the patient–specific mechanical processes appearing in such bone–implant–systems, numerical simulations are a key technology. Therefore, routinely acquired computed tomography data sets are subjected to an image processing workflow resulting in personalized computational models. These models can be either equipped with monitoring data of the patient itself [9] or with avaiable in vivo data [7]. This leads to patient–specific simulations allowing the analysis of the stresses and strains of bone–implant– systems and are the starting point for optimization algorithms in orthopedic trauma surgery, cf. [1] and [2]. 2 Methods and Results The personalized simulations were performed for a 32 years old male patient with a bodyweight of 82 kg. The patient has a distal tibia fracture and was treated with an intramedullary nail. The presented simulation workflow starts with an image processing step, illustrated in Figure 1. Figure 1 a) shows one slice of the original computed tomography image stack. In Figure 1 b), the result of the image segmentation is shown. Here, an edge–enhancing non–linear anisotropic diffusion (EED) filtering was used in combination with an adaptive thresholding based on histogram analysis [3]. In Figure 1 c), the grayscale values based on the CT image given in Hounsfield units are mapped to the segmented part. In the last step, the fracture area is marked red in a semi–automated procedure guided by a health professional, cf. 1 d). The result of this image processing is a volume mesh, storing the segmented image stack in a homogenous grid with an uniform resolution given by the tomographic data. This leads to a 1:1 relation between voxels from the CT data and cells in the volume mesh, illustrated in Figure 2. Besides the geometric properties, material properties needed to be passed to the simulation. Because of the variations in the grayscale distribution and the artefacts in the image stack produced by the metallic parts, e.g. nail and screws, a calibration phantom is used to calibrate and to correct the image information, cf. Figure 1 e). After that step, the image information can be mapped to material parameters stored at the center of each mesh cell. For the intramedullary nail and the screws, standard values for medical implants are used [4]. For the soft tissue in the fracture area and the cortical as well as the trabecular bone parts, material classes are used, cf. [5] and [6]. In order to execute the simulations of the individual bone–implant–system also with realistic boundary conditions, knee joint reaction forces given by the OrthoLoad database were mapped into the computational model and calibrated via the bodyweight of the considered patient. The knee joint reaction forces measured during a gait analysis setting are shown in Figure 3. The data is provided by the OrthoLoad database which includes in vivo measurements of orthopedic implants in different clinical settings [7]. The FEM simulations were performed in the deal.II environment, an open source C++ library [8]. The computing grid representing the full tomographic resolution has 6,157,537 mesh cells corresponding to 19,201,212 degrees of freedom for linear Lagrange finite elements and a linear elasticity material model. Figure 4 shows the maximum of the von Mises stress arising during the simulation of a step forward of the patient. The maximum of 218 MPa occurs as expected in the lower part of the nail located near to the fracture (shown in yellow) and is also in the range of stresses which are measured and reported for such loading scenarios. ∗ Corresponding author: e-mail m.roland@mx.uni-saarland.de, phone +49 681 302 3789, fax +49 681 302 3992 c  2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim