Effect of different walking speeds on joint and muscle force estimation using AnyBody and OpenSim

• Published in: Gait & posture Vol. 90; pp. 197 - 203
• Main Authors: Alexander, Nathalie, Schwameder, Hermann, Baker, Richard, Trinler, Ursula
• Format: Journal Article
• Language: English
• Published: England Elsevier B.V 01.10.2021
• Subjects: Biomechanical Phenomena; Correlation coefficient; Gait; Generic model; Humans; Knee Joint; Level walking; Mean absolute error; Muscle, Skeletal; Muscluloskeletal model; Trend analysis; Walking; Walking Speed; Correlation coefficient; Generic model; Muscluloskeletal model; Level walking; Mean absolute error; Trend analysis
• ISSN: 0966-6362; 1879-2219
• DOI: 10.1016/j.gaitpost.2021.08.026

•Trend analyses showed no difference between AnyBody and OpenSim.•MAEs increased sign. from slow to fast walking for knee joint compression forces.•Some hip extensors, glut. med. and vast. int. MAEs also increased sign.•Lower correlations for some knee extensors and hip extensors during slow walking. To be able to use muscluloskeletal models in clinical settings, it is important to understand the effect of walking speed on joint and muscle force estimations in different generic musculoskeletal models. The aim of the current study is to compare estimated joint and muscle forces as a function of walking speed between two standard approaches offered in two different modelling environments (AnyBody and OpenSim). Experimental data of 10 healthy participants were recorded at three different walking speeds (self-selected, 25 % slower, 25 % faster) using a ten-camera motion capture system together with four force plates embedded into a ten-meter walkway. Joint compression forces and muscle forces were calculated with a generic model in AnyBody and OpenSim. Trend analyses, mean absolute error (MAE) and correlation coefficients were used to compare joint compression forces and muscle forces between the two approaches. A one-way and two-way ANOVA with repeated measures were used to compare MAE and trend analysis changes, respectively (α = 0.05, Bonferroni corrected post-hoc tests). Trend analyses showed the same speed effect for AnyBody and OpenSim. MAEs increased significantly from slow to fast walking for knee joint compression forces, biceps femoris long head, gluteus maximus, gluteus medius and vastus intermedius. Lower correlation coefficients during slower walking were found for quadriceps muscles, gluteus maximus and biceps femoris compared to normal and faster walking. Lower correlation coefficients during slower walking are assumed to be due to a higher amount of solutions solving the muscle recruitment in musculoskeletal models. This indicates that decreasing walking speed is more prone to speed dependent differences regarding variability, while the absolute error increased with increasing walking speed. To conclude, different modelling environments can react differently to changes in walking speed, but overall results are promising regarding the generalization across different generic musculoskeletal models.