WP11: Modelling and simulation for NND Thomas Geijtenbeek, Frans van der Helm Delft University of Technology Amsterdam – February 2015
Status of the WP T11.1 Construction of a scalable mass distribution model – Literature study on sensitivity of scaling methods – Process results from joint center calibration Amsterdam – February 2015
Status of the WP T11.2 Development of a personalized disease specific skeletal model – Functional calibration Find joint rotation centers and axes Scale segment lengths Scale segment weights – MRI recordings Muscle volume – Physiological Cross-Sectional Area – Muscle optimum length from cadaver studies & scaled Via points – muscle attachments – Finally through Statistical State Model Amsterdam – February 2015
Scaled musculoskeletal models Constructing a scaled model Brussels – 6-7 May 2014 Functional joint center calibration
Status of the WP T11.3 Construction of a disease specific muscle model – Plan to include spasticity model from VUMC (Van der Krogt et al.) – Muscle constraints: » Contractures: Passive constraints » Aberrant reflexes: low threshold on muscle contraction velocity reflex Amsterdam – February 2015
Spastic controller: increased stretch reflex (vdKrogt et al., 2015) If: Stretch velocity of muscle fibers > Threshold Then: Spastic excitation ( t + Delay ) = Gain * stretch velocity … lead to efferent impulses causing contraction Afferent impulses from spinal cord… Threshold: extracted from data Delay: fixed to 30 ms Gain: individually tuned (0-4) METHODS
Force-length curves - with optimal stiffness parameters Cerebral Palsy (CP) vs Typically Developing (TD) vdKrogt et al. (2015) Relative fiber length Relative fiber force Active Default passive TD hamstrings CP hamstrings TD vasti CP vasti RESULTS
Status of the WP T11.4 Design of models driven by the dynamics of gait perturbations – OpenSim model (available) Gait2392 model – 23 DOF – 92 musculotendon actuators representing 76 muscles To be adapted by available gait and morphological data – Optimization toolbox (connected) Covariance matrix adaptation (CMA) – Neural control signal, feedback parameters – Unknown model parameters – Predictive simulations (in progress) Optimization – Simulate optimal neural control model in pathological state » Mimick pathological gait – Predict optimal neural control after therapeutic intervention » Prediction of outcome after intervention and rehabilitation practice – Use gait data + EMG for validation Amsterdam – February 2015
Predictive Simulations Find optimal gait pattern for any given musculoskeletal model Use high-level optimization criteria – Target speed – Metabolic energy expenditure Method does not require motion capture data – Can be used for validation Can incorporate neuromuscular deficiencies Amsterdam – February 2015
Adapt to Model Scaling Amsterdam – February 2015
The Optimization Process Amsterdam – February 2015
Adapt to Target Speed Amsterdam – February 2015
Find Optimized Muscle Attachments Amsterdam – February 2015
What we are going to present to the EC (Annual review) Significant results Use cases Amsterdam – February 2015
Current open issues Issues/criticalities Corrective actions Amsterdam – February 2015