13:30   Motion and Neuro muscular The dynamics of non-upright castor wheels and their application for pushing wheelchairs from the side Matto Leeuwis, Lucy Bennett, Bram Sterke, Heike Vallery Abstract: Some wheelchair users, such as the elderly or children with Profound Intellectual and Multiple Disabilities (PIMD) require eye contact to communicate and observe safety. Conventional wheelchairs place the caretaker behind the patient when pushing the wheelchair, but this position obstructs communication and makes it harder to assess the user’s health. To facilitate face-to-face communication while walking, we proposed a steering compensation method at BME2021 using a banked castor wheel to allow the caretaker to walk next to the wheelchair. We have developed a mathematical model to determine how the design variables and external parameters such as rolling resistance affect the steering compensation. This model can be used to optimize the castor wheel for different users and terrains and provide insight into the steering mechanism’s capabilities and limitations. Both the Lagrange and virtual power method are used to find the equations of motion of the castor wheel. In addition to the wheel radius and castor trail, the three-dimensional orientation of the castor swivel axis and wheel spin axis are considered. The steering effect was quantified as the resultant moment around the castor wheel's up axis exerted on the wheelchair. The simulated results show that in steady-state, the steering effect on the wheelchair is a linear function of the vertical load on the castor wheel. The slope of this relation is determined by the vertical height change for an infinitesimal change in the steering angle. Since rolling resistance can also be approximated as a linear function of vertical load, the wheelchair can be pushed in steady-state from the same position regardless of the weight of the wheelchair user. Variation of the rolling resistance coefficient does result in a different pushing position. In the future, the developed mathematical model will be applied to calculate the influence of design variables to build a better prototype and eventually advance the quality of life of manual wheelchair users. Scaling of 2D gait skeleton data for quantitative assessment of movement disorders using freehand single camera video recordings Wei Tang, Peter M.A. van Ooijen, Deborah A. Sival, Natasha M. Maurits Abstract: Gait can clinically be assessed semi-quantitatively using rating scales. However, such clinical assessment requires extensive clinical knowledge, and is subjective and time-consuming. Therefore, wearable inertial sensors have been considered for objective and more accurate assessment. Yet, this approach still requires special hardware, preparation time and can by itself not be used to derive important distance-based features, such as step width. In this work, we propose a 2D skeleton-based method for quantitative gait analysis using video images taken in the coronal plane, assessing five different scaling methods to preprocess 2D skeleton keypoint data as derived from pose estimation deep neural networks. Fifteen children (5 ataxia, 5 developmental coordination disorder and 5 controls) walked in a straight line, with a single 2D camera placed in front of them. A deep neural network model trained on the MSCOCO dataset based on Alphapose was first used to extract 17 skeleton keypoints. Then a PoseFlow framework was used to match the skeleton to the same participant in a recording. Four of the assessed scaling methods used width and height of the bounding boxes of: 1) all keypoints (Box scale or BS method), 2) the two shoulder keypoints (S method), 3) the left shoulder and right hip keypoints (LS/RH method) and 4) the two hip keypoints (H method). The fifth method used the distance between left shoulder and right hip for both width and height (LS/RH-d method). The mean likelihood for all keypoints from the pose estimation model was 0.89 (std=0.02) after cleaning the data, indicating high keypoint reliability. Based on mean absolute angle error and mean variance of distance between certain keypoints, the LS/RH method performed best because it maintained expected low variability in shoulder and hip distance, while allowing for expected variability in hands and ankle distance. This study is a first step towards quantitative analysis of clinically observed gait using single camera freehand video recording, that may be used to differentiate movement disorders with machine learning in the future. Does ankle push-off correct for errors in sagittal plane foot placement relative to center-of mass state? Jian Jin, Jaap van Dieën, Dinant Kistemaker, Andreas Daffertshofer, Sjoerd Bruijn Abstract: Understanding the mechanisms of stable walking is important for unravelling the causes of falls. Modelling studies suggested two basic principles to stabilize walking that can subsequently be tested in humans: foot placement control and push-off control. Foot placement control found support in two way: the center-of-mass (CoM) state from mid-stance to heel-strike can serve to predict sagittal plane foot placement (Wang & Srinivasan, 2014); and, the covariance between CoM state and foot placement seems to subserve a stable gait pattern (Hof, 2008). Sagittal plane push-off control, by contrast, has only been tested in perturbed walking (van Mierlo et al., 2021) and not during steady-state human gait. To fill this lacuna, we investigated whether a push-off control can be identified in steady-state walking. Because of the covariance between CoM state and anterior-posterior foot place, we tested whether foot placement error, i.e. the difference between actual and predicted foot placement based on the CoM state at heel-strike, may serve as feedback signal for push-off control. Using kinematics and kinetics from steady-state walking in 30 healthy participants (van Leeuwen et al., 2020), we compared Pearson’s correlations between foot placement error and ankle moment during the subsequent double-stance phase across participants. We found that foot placement errors were corrected by modulations of sagittal plane ankle push-off moments, with mean correlations up to 0.45. Positive foot placement errors were corrected by larger push-off moments to enhance CoM accelerations and vice versa for negative errors. Our results indicate that humans use a push-off strategy to control their sagittal plane CoM state. They correct foot placement errors not only in perturbed walking but also in steady-state walking. Hof, A. L. (2008). The “extrapolated center of mass” concept suggests a simple control of balance in walking. Human Movement Science, 27(1). van Leeuwen, M., Bruijn, S. M., van Dieën, J. H., & Daffertshofer, A. (2020). Dataset and analyses for: Active foot placement control ensures stable gait: Effect of constraints on foot placement and ankle moments. Zenodo. van Mierlo, M., Vlutters, M., van Asseldonk, E. H. F., & van der Kooij, H. (2021). Centre of pressure modulations in double support effectively counteract anteroposterior perturbations during gait. Journal of Biomechanics, 126. Wang, Y., & Srinivasan, M. (2014). Stepping in the direction of the fall: The next foot placement can be predicted from current upper body state in steady-state walking. Biology Letters, 10(9). Investigation of the effect of different stiffnesses of foot’s plantar soft tissue on the contact pressure Zeinab Kamal, Edsko E.G. Hekman, G.J. Bart Verkerke Abstract: Introduction: Changes in the material properties of the foot’s soft tissue have been reported as a possible cause of ulceration development[1]. The effect of different stiffnesses of the foot’s soft tissue on the foot stresses has not been investigated yet. In this work, it was hypothesized that increased stiffness of the plantar soft tissue can increase the plantar contact pressure. For this purpose, a 3D ankle-foot finite element model was constructed and analyzed in the late-midstance phase. The results showed a significant correlation between forefoot stiffness and contact pressure. Method: A foot model was created based on one individual’s foot CT images(49yrs, height of 160cm, and weight of 67kg) in the neutral posture. The bones and ligaments were considered as rigid bodies and linear elastic springs (E=260MPa), respectively[2]. Soft tissue was added, having an elastic modulus E=1.15MPa and Poisson's ratio=0.49)[2]. For the simulation of stiffened tissue, the E-modulus was increased by 50%. The upper surface of the soft tissue and Tibia was fully constrained. The interaction between the foot plantar surface and the ground was defined with a frictional coefficient of 0.6[2]. The muscle forces and ground reaction forces (GRF) were estimated using OpenSim. After this, the model was loaded with these forces using Abaqus. Results: Fig.1 shows the results of the contact pressure of the foot model with different stiffness. Conclusion: The results showed a 47% increase in the contact pressure after 50% amplification in the soft tissue’s stiffness. This finding the greater pressure at the interface of plantar tissue with the ground due to stiffening is compatible with the reported association between the soft tissue stiffening and ulceration in diabetic feet [1] and can be deemed when preventive strategies in the diabetic foot are designed. References:[1] Naemi et al. J. Diabetes-Complicat(2016)30(7):1293-1299. [2] Akrami et al.Biomech-Model-Mechanobiol(2018)17:559–576. Comparing finger tapping tasks and MDS-UPDRS-III features for treatment classification in Parkinson’s disease patients Willem O. Elzinga, Soma Makai-Bölöni, Eva Thijssen, Ingrid Koopmans, Geert J. Groeneveld, Robert-Jan Doll Abstract: The Movement Disorder Society - Unified Parkinson Disease Rating Scale part III (UPDRS-III) is the gold standard to assess motor symptom severity in patients with Parkinson’s Disease (PD). As rating requires a trained rater, such a scale may be considered time-consuming, expensive, and subject to inter-rater variability. Recently, we developed three different tapping tasks that may complement the UPDRS-III assessment. Here, we used Machine Learning techniques to study the performance of (the combination of) multi-modal finger tapping tasks and UPDRS-III scores for classification of treatment classes in PD patients. Data were collected during a double-blind, randomized, placebo-controlled, cross-over design study with 20 PD patients with a median Hoehn-Yahr stage of 2 (range 1-3). During all visits, three tapping tasks and MDS-UPDRS-III scoring were performed repeatedly before and after treatment with either placebo or levodopa/carbidopa. Two finger tapping tasks were performed on a touchscreen laptop, whereas the third tapping task consisted of a goniometer recording the angle between the index finger metacarpal and proximal phalanx (wrist). Machine Learning techniques were employed to build treatment classifiers using either the tapping task features, or the MDS-UPDRS-III score features. The best performing classifier using the MDS-UPDRS-III score features reached an accuracy of 58% (± 4% std). By contrast, the best performing classifier for the tapping features had an average accuracy of 75% (± 21% std). This treatment classifier identified the tapping features related to fatigue (e.g., change in tapping velocity) from a single tapping task as most predictive. Here we identified important tapping features to discriminate placebo from active treatment and contrasted this performance to a model using the MDS-UPDRS-III scores. We demonstrated that finger tapping features are better than MDS-UPDRS-III scores for classifying the treatment. This suggests that a finger tapping task is more informative than the gold standard in demonstrating dopaminergic treatment effects in PD patients. Effect of dynamic arm support on upper extremity muscle coordination in people with facioscapulohumeral dystrophy performing functional tasks Hans Essers, Kenneth Meijer, Anneliek Peters, Alessio Murgia Abstract: Facioscapulohumeral Dystrophy (FSHD) is a neuromuscular disorder characterized by muscles strength loss in the upper extremity and compensatory muscle activations. Dynamic arm supports compensate for gravity thus reducing muscle efforts and enhancing task performance. It is postulated that the use of support in persons with FSHD leads to larger individual-specific alterations of muscle coordination than in healthy controls. Therefore, we investigate the effect of muscular weakness and support on muscle coordination in persons with FSHD, when performing functional tasks. We hypothesized firstly, that the FSHD population presents less consistent muscle coordination than healthy controls and secondly, that the support greatly improves consistency of muscle coordination within the FSHD population. Electromyograms of eight upper extremity muscles were recorded for twelve FSHD (56.0±14.5yrs, 1.76±0.10m, 75±20kg) and twelve matched control participants (55.5±13.4yrs, 1.76±0.08m, 72±14kg) while they performed ipsi-/contralateral reaching, eating/drinking, and a push/pull task with/without support. Muscle synergies were extracted by non-negative matrix factorization and clustered based on the Pearson’s correlation coefficients (r) between the participants’ synergy weights, respectively for population, w/o support, and task. The r-values within (consistency) and between clusters (similarity) were investigated to quantify the effect of FSHD and support. Synergy consistency was tested with a non-parametric analysis of variance. Synergy similarity was interpreted as low (r<0.3), medium (r: 0.3-0.5), high (r: 0.5-0.8), or very high (r>0.8). A group and group*support effect revealed that muscle synergies in FSHDs were less consistent than in controls (r: -0.41) and were not affected by support; for controls the consistency increased significantly (r: +0.38) when using support. Furthermore, synergy similarity between FSHDs and controls was higher when FSHDs used the support and controls did not (r: +0.52). Similarity between the two groups also increased (r: +0.61) when both groups used the support. These findings support our first but not our second hypothesis and indicate that muscle coordination in FSHDs is heterogeneous, i.e. it is not affected by the arm support to similar levels as in the controls. However, the support does increase the similarity between populations, which seems a generalizable effect in FSHDs leading towards a muscle coordination that resembles that of healthy controls.

end %-->