Patientenspezifisch optimierte Frakturversorgung unter Berücksichtigung der biomechanischen Knochenstrukturparameter sowie der zu erwartenden Belastungen ----------------------------------------------------------------- Patient specific optimization of fracture treatment considering the inhomogeneous material properties of bone tissue and the expected load situation M. Schimmelpfennig A,C. Wittkowske B,S. Raith B,A. Nolte C, B. König D,S. Döbele D,J. Bauer E,E. Grande Gracia E,L. Kovacs B E D A B C Dynardo GmbH, Weimar,Germany Research Group CAPS, TU München, Germany CADFEM GmbH, Grafing, Germany Berufsgenossenschaftliche Unfallklinik Tübingen, Germany Institut für diagnostische und interventionelle Radiologie der TU München 2
Outline Introduction and Objective Methods Results Limitations and Outlook Summary 3
Outline Introduction and Objective Methods Results Limitations and Outlook Summary 4
Introduction and Objective Young men: Traffic and sports accidents Elderly women: Low-impact trauma, osteoporotic bone Adequate treatment of osteoporotic bone Optimizing the screw layout and the plate position Fracture of femoral shaft 5
Introduction and Objective Millions of different screw layouts are possible Selection of best screw layout in terms of optimal healing depends on: Bone quality Bone shape Fracture type Difficult decision for surgeons Monocortical Bicortical 6
Introduction and Objective Development of atool which supports surgeon in screw and plate placement decision Requirements of tool: Integration of patient-specific data Ease of use Automatization and short computing time Parallel computing Robust data handling Optimization is necessary Objective: treatment of the fracture with the lowest amount of screws Don t violate different constraints 7
Outline Introduction and Objective Methods Results Limitations and Outlook Summary 8
Methods -Workflow MANUAL ONCE FOR EVERY PATIENT AUTOMATED Segmentation Position Bone/Plate Screw Generation Finite Element Mesh Hounsfield Units Correlation HU Optimization = Find the best design out of several other designs FE Simulation Optimization 9
Methods -Workflow MANUAL ONCE FOR EVERY PATIENT AUTOMATED Segmentation Position Bone/Plate Screw Generation Finite Element Mesh Hounsfield Units Correlation HU Optimization = Find the best design out of several other designs FE Simulation Optimization 10
Methods -Segmentation CT Data converted to CAD geometry (stl-file) Half automated process Requires 20 to 40 min Layer thickness 0.6 mm Smoothing 3x3 11
Methods -Workflow MANUAL ONCE FOR EVERY PATIENT AUTOMATED Segmentation Position Bone/Plate Screw Generation Finite Element Mesh Hounsfield Units Correlation HU Optimization = Find the best design out of several other designs FE Simulation Optimization 12
Methods -Repositioning of Bone Fragments Aim: Virtual repositioning of bone fragments Using an artificial fracture as afirst step Matching plate with shape of bone Requires 20 to 40 min Artificial fracture 13
Methods -Workflow MANUAL ONCE FOR EVERY PATIENT AUTOMATED Segmentation Position Bone/Plate Screw Generation Finite Element Mesh Hounsfield Units Correlation HU Optimization = Find the best design out of several other designs FE Simulation Optimization 14
Methods Screw setting and plate position Simplified screw and plate geometry Variable length depending on input parameters 0: No screw 1: Monocortical screw 2: Bicortical screw Plate position (x, y, z) Predefined position can be variated Requires1min 15
Methods -Workflow MANUAL ONCE FOR EVERY PATIENT AUTOMATED Segmentation Position Bone/Plate Screw Generation Finite Element Mesh Hounsfield Units Correlation HU Optimization = Find the best design out of several other designs FE Simulation Optimization 16
Methods -Generation of Finite Element Mesh ICEMCFD as abatch call Unstructured mesh (Tetrahedrons with middle nodes) ~500k Elements Requires 5min Laplace Smoothing 17
Methods -Workflow MANUAL ONCE FOR EVERY PATIENT AUTOMATED Segmentation Position Bone/Plate Screw Generation Finite Element Mesh Hounsfield Units Correlation HU Optimization = Find the best design out of several other designs FE Simulation Optimization 18
Methods -Assignment of Hounsfield Units CT Scan ANSYS Simulation Values of voxels in CT scans are Hounsfield Units Each node of the FE Mesh gets assigned a Hounsfield Unit Requires 1min 19
Methods -Workflow MANUAL ONCE FOR EVERY PATIENT AUTOMATED Segmentation Position Bone/Plate Screw Generation Finite Element Mesh Hounsfield Units Correlation HU Optimization = Find the best design out of several other designs FE Simulation Optimization 20
Methods -Finite Element Simulation Pre-processing: fixed support Assignment of material properties (titanium alloy, stainless steel) Boundary conditions ( 150N max. force on injured leg after surgery - 50N Weight force of lower leg and foot 100N max. force from distal) Simulation: 2x parallel requires ~10 min 21
Methods -Finite Element Simulation Post-Processing: Display of results (e.g. displacement, stress and strain) Interfragmentary Movement: Optimal healing: 0.5 to 1.0mm Different interfragmentary movement on near cortex and far cortex 22
Methods -Workflow MANUAL ONCE FOR EVERY PATIENT AUTOMATED Segmentation Position Bone/Plate Screw Generation Finite Element Mesh Hounsfield Units Correlation HU Optimization = Find the best design out of several other designs FE Simulation Optimization 23
Methods -Optimization optislang controls the workflow (one design ~15 min) By defining the screw layout and the plate position Analyzing the output parameters (responses) Objective: Minimize total screw number Constraints: Minimum number of screws per fragment: 2 Maximum interfragmentary movement: 1.6 mm Maximum stress in titanium implant: 800 MPa (overload) Maximum stress in bone: 100 MPa (overload) Overload = 3x load situation 24
Methods -Optimization optislang controls the workflow By defining the screw layout and the plate position Analyzing the output parameters (responses) Sensitivity Analysis to get abetter understanding of the process Evolutionary Algorithm to create new designs and find the best 25
Outline Introduction and Objective Methods Results Limitations and Outlook Summary 26
Results -Bridging Length and Movement Bridging Length influences interfragmentary movement Movement on near cortex smaller compared to far cortex Bridging length of 3to 6unoccupied screw holes achieves good results 27
Results Optimizing using EA Best design: Number of screws: 4 Bridging Length: 4unoccupied screw holes No screws at distal end needed Deformation [mm] 28
Results Robustness of best design Layout of the best design of the EA is transferred to Robustness analysis Robustness parameters: Screw layout is fixed Plate position movement Position in X-Axis: ± 1mm (distance: plate bone) Position in Y-Axis: ± 1mm Position in Z-Axis: ± 0.5mm (axial direction to the bone) 29
Results Robustness of best design Robustness evaluation: Constraints in movement are fulfilled Constraints of stress inside the plate and bone are fulfilled 30
Outline Introduction and Objective Methods Results Limitations and Outlook Summary 31
Limitations - Monocortical vs. Bicortical Screws No difference between mono- or bicortical screws in term of interfragmentary movement Only small changes in stress distribution FE simulation is linear elastic and geometrical linear Healthy bone Osteoporotic bone 32
Limitations -Stress singularities Screw holes with sharp edges Maximum stress values Screwhead with sharp edges Maximum stress values at sharp edges of screw holes Less than 0.01% of all elements contributing towards the top 20% of stress Possible solution: 10 th / 20 th /100 th highest stress element 33
Limitations -Physioligical Boundary Conditions Forces from distal fixed support Forces from proximal (Taylor et al. 1996) fixed support Both models: 100N force at distal end of femur Differences in interfragmentary movement and stress distributions Careful selection of model is important Integration of full musculoskeletal model (AnyBody software) 34
Outlook Human Model Repository: Muscles Bones Joints Motion Rigid Bodies, Inverse Dynamic & Optimizationcriterion Output: Muscle work Muscle forces Activity Joint reaction forces Motion data External load Muscle and joint reaction forces as boundary conditions for FE Simulation 35
Outlook Influence of different loading situations Evaluation of more data, in particular osteoporotic bone Improve the FE Model (linear non-linear) Extension of the robustness parameters Tests with other types of bones and plates Transfer the workflow inside ANSYS Workbench to use the HPC Parametric Pack Licenses 36
Outline Introduction and Objective Methods Results Limitations and Outlook Summary 37
Summary Development of automatic workflow Validation of workflow Selection of best design for first patient Best design 38
Project partners Thanks to the involved partners and the Federal Ministry of Economics and Technology for funding. 39