Dynamic Modeling of Knee Mechanics, 18-R8167Printer Friendly Version
Inclusive Dates: 07/01/10 Current
Background - Osteoarthritis (OA) is the most common form of arthritis and, as the major cause of activity limitation and physical disability in older people, is a tremendous public health concern. Arthritis causes pain, swelling, and reduced motion in joints caused by the breakdown or degradation of the articular cartilage covering the joint surfaces. While it is generally accepted that differences in knee mechanics, or alterations in knee mechanics due to certain risk factors, lead to knee OA, the precise dynamic mechanical environment of the knee and its anatomical structures during routine physical movements is largely unknown. Thus, a remaining unmet, technically difficult challenge in musculoskeletal research, and the focus of this project, is determining the detailed dynamic mechanical environment of the healthy knee joint and understanding alterations in knee mechanics caused by injury, aging, or disease.
Approach - The primary objectives of this project are to:
Accomplishments - A first-generation dynamic finite element model of the lower body has been developed using individual anatomic surface models from a digital human anatomy collection. Contact conditions were defined at the knee joint between the femoral cartilage, tibial cartilage, menisci, and patella cartilage to constrain articular joint motion based on anatomic geometry; all other joints are fixed in this version of the model. The remaining joints of the lower limb will be implemented in the coming quarter. All the knee ligaments (anterior cruciate, posterior cruciate, medial, and lateral ligaments) were modeled as non-linear, one-dimensional springs. The femoral cartilage, tibial cartilage and menisci were modeled as linear elastic materials using published material properties. Fifty-nine muscles were implemented using one-dimensional active contraction Hill-type muscle springs. Anatomic insertion points and Hill muscle model parameters were derived from data obtained from an open source neuromuscular biomechanics website. Simple movements were simulated using manually determined muscle activation patterns to test and debug the high fidelity finite element model. The model was reconfigured to simulate a seated leg extension exercise and the quadriceps muscles (vastus lateralis, vastus medialis, rectus femoris, and vastus intermedius) were activated to perform the leg extension while the activations for the remaining muscles were kept at zero or a minimal level for stability. This simulation required approximately three hours to run using four CPUs. To facilitate the evaluation of the model and its gradient at each desired point, a program was written to construct the input decks for all n+1 model runs and then invoke a shell script that uses MPI to "push" each analysis to a different compute node. By performing these analyses in parallel, all n+1 models needed at each iteration during the optimization can be evaluated in approximately the same amount of time as a single analysis. This code was tested by optimizing the cross-sectional area of a loaded cantilever beam subject to a constraint on the beam's deflection, which was defined via a finite element model. In addition, the ability to control and parallelize the high-fidelity leg extension finite element model for a single iteration of the intended optimization problem has been demonstrated.