[home] [Personal Program] [Help]

---

Session: Musculoskeletal System
Session starts: Thursday 24 January, 10:40
Presentation starts: 10:55
Room: Lecture room 558

---

René Fluit (University of Twente)
Marjolein van der Krogt (University of Twente)
Nico Verdonschot (University of Twente)
Bart Koopman (University of Twente)

Abstract:
To predict the functional outcome of orthopaedic interventions on the musculo-skeletal model, we developed an inverted inverse dynamic model in order to predict whether and how patients walk postoperatively. In an inverse-inverse model, both the inverse dynamics input and the problem of unknown muscle forces are solved by formulating a minimization function[1]. Hence, the motion and muscle forces are optimized simultaneously. Gait lab measurements of ten healthy subjects were performed at the Radboud University Medical Centre. Three gait trials at comfortable speed of each subject, 30 in total, were analysed using a musculoskeletal analysis tool (AnyBody 4.2.1, AnyBody Technology A/S). Then, following Schwartz and Rozumalski (2008)[2], for each gait trial, all inverse-dynamic input over a complete gait cycle, i.e. all joint angles, pelvis position, ground reaction forces and center of pressure, were arranged in a single gait vector g. 30 singular vectors fk, referred to as gait features, were obtained by computing the singular value decomposition of the matrix G containing all concatenated gait vectors g. The inverse dynamics input was defined as an mth order approximation by multiplying the m main gait features fk with the feature components ck. In an inverse-inverse dynamic simulation, a complete gait cycle was predicted by optimizing only m feature components ck, based on an energy-cost function and subject to the constraint that the muscle activity should stay below 100 percent activation. The first gait feature described the average gait pattern, the following gait features described the variability present in the measured gait patterns. By optimizing only the first 5 or 10 gait features, already 79.5 and 90.3 percent of the variability was explained, respectively. The method does not guarantee that the optimized gait pattern is dynamically consistent, a situation in which unrealistically large residual forces are applied to the model. To prevent this, the ground reaction forces could be estimated in a separate optimization or the residual forces, together with the energy, could be minimized in a multi-objective optimization. Within the TLEMsafe project, software is developed to quickly generate subject-specific musculo-skeletal models. Combining this software with the inverse-inverse method may be an important step towards predicting the outcomes of orthopaedic surgeries.