Ground Reaction Force Prediction During Weighted Leg Press and Weighted Squat in a Flywheel Exercise Device
Total Page:16
File Type:pdf, Size:1020Kb
DEGREE PROJECT IN MEDICAL ENGINEERING, SECOND CYCLE, 30 CREDITS STOCKHOLM, SWEDEN 2017 Ground Reaction Force Prediction during Weighted Leg Press and Weighted Squat in a Flywheel Exercise Device Estimering av markreaktionskraften vid viktad benpress och viktad knäböj i ett svänghjulsbaserat träningsredskap TOBIAS MUNKHAMMAR KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF TECHNOLOGY AND HEALTH Acknowledgement First of I would like to thank my supervisor, Maria J¨onsson,for guidance and encouragement during the whole project and Lena Norrbrand, who, together with Maria collected all experi- mental data used in this study. Furthermore, my thanks goes to Rodrigo Moreno, Jan H¨ornfeldt, Jonathan Munkhammar and Ola Eiken for proof-reading and general feedback on the report and Elena Gutierrez Farewik, for being a link towards the musculoskeletal software company whenever the licence struggled. Lastly, I would like to thank all people at the Department of Environmental Physiology, for making me feel welcome and showing interest in my work. Abstract When performing a biomechanical analysis of human movement, knowledge about the ground reaction force (GRF) is necessary to compute forces and moments within joints. This is important when analysing a movement and its effect on the human body. To obtain knowledge about the GRF, the gold standard is to use force plates which directly measure all three components of the GRF (mediolateral, anteroposterior and normal). However, force plates are heavy, clunky and expensive, setting constraints on possible experimental setups, which make it desirable to exclude them and instead use a predictive method to obtain the full GRF. Several predictive methods exist. The node model is a GRF predictive method included in a musculoskeletal modeling software. The tool use motion capture and virtual actuators to predict all three GRF components. However, this model has not yet been validated during weighted leg press and weighted squat. Furthermore, the normal component of the GRF can be measured continuously during the activity with pressure sensitive insoles (PSIs), which might provide better accuracy of the GRF prediction. The objectives of this thesis were to investigate whether force plates can be exluded during weighted leg press and weighted squat and to investigate whether PSIs can improve the GRF prediction. To investigate this, the node model and a developed shear model was validated. The shear model computes the two shear GRF components based on data from PSIs, an external load acting upon the body and data from a motion capture system. Both the node model and the shear model were analysed with two test subjects performing two successive repetitions of both weighted squat and weighted leg press in a flywheel exercise device. During the leg press exercise, the node model had a mean coefficient of correlation (Pearson's) ranging from 0.70 to 0.98 for all three directions with a mean root mean square error ranging between 8 % to 20 % of the test person's body weight. The developed shear model had a coefficient of correlation (Pearson's) between 0.64 to 0.99 and a mean root mean square error between 3 % and 21 % of the test person's body weight. This indicates that it is possible to exclude force plates and instead predict the GRF during weighted leg press. During squat, neither the node model nor the shear model provided accurate results regarding the mediolateral and anteroposterior components of the GRF, suggesting that force plates can not yet be excluded to obtain the full GRF during weighted squat. The results of the normal component during leg press was somewhat improved with the shear model compared to the node model, indicating that using PSIs can improve the results to some extent. Keywords: Ground reaction force prediction, musculoskeletal model, validation, leg press, squat, biomechanical analysis, pressure sensitive insoles, flywheel exercise device. Sammanfattning Biomekaniska analyser av m¨anniskans r¨orelsem¨onster inneb¨ar bland annat att analysera krafter och moment i kroppens leder. F¨or att g¨ora detta ¨ar information om reaktionskraften fr˚anmarken n¨odv¨andig. Markreaktionskraften ¨ar en tredimensionell vektor som kan komposantuppdelas i tre riktningar; normal-, anteroposterior- och mediolateralriktning. Vanligtvis m¨ats dessa med hj¨alp av kraftplattor. Kraftplattor ¨ar otympliga och dyra, vilket g¨or dem sv˚araatt anv¨anda under exempelvis f¨altstudier. D¨arf¨or ¨ar det ¨onskv¨art att ist¨allet estimera markreaktionskraften. Det finns flertalet metoder f¨or att estimera markreaktionskraften. Ingen av dessa metoder har anv¨ants f¨or att estimera markreaktionskraften vid viktad kn¨ab¨oj eller viktad benpress. Huvudsyftet med denna studie var att unders¨oka huruvida kraftplattor kan ers¨attas med en estimeringsmetod f¨or markreaktionskraften vid viktad kn¨ab¨oj och viktad benpress. F¨or att unders¨oka om kraftplattor kan exkluderas anv¨andes ett tr¨aningsredskap med sv¨anghjulsteknik och en estimeringsmetod som ¨ar inkluderad i en muskuloskeletal modelleringsmjukvara. Estime- ringsmetoden ("nodmodellen") fick approximera markreaktionskraften baserat p˚ar¨orelsedata och data fr˚anen kraftsensor som m¨atte kraften fr˚antr¨aningsredskapet. Nodmodellen, likt de flesta andra estimeringsmetoderna, approximerar alla tre komponenterna av markreaktionskraften. Ge- nom att anv¨anda tryckk¨ansliga skosulor som m¨ater normalkomponenten av markreaktionskraften skulle estimeringen av de andra tv˚akomponenterna kunna f¨orb¨attras d˚ade ofta ber¨aknas med hj¨alp av normalkomponenten, vilket i sig skulle kunna ge en mer korrekt biomekanisk analys. F¨or att unders¨oka detta utvecklades ytterligare en approximationsmetod ("skjuvmodellen"). Skjuv- modellen anv¨ande sig av normalkomponentsdata fr˚antryckk¨ansliga skosulor f¨or att estimera de tv˚askjuvkomponenterna av markreaktionskraften. I denna studie j¨amf¨ordes nodmodellen och skjuvmodellen mot data uppm¨att med hj¨alp av kraft- plattor. Nodmodellen hade h¨og Pearson's korrelationskoefficient (r = 0.79 till r = 0.98) j¨amf¨ort med data uppm¨att med kraftplattor vid viktad benpress medan skjuvmodellen hade m˚attligtill h¨og korrelation (r = 0.64 till r = 0.99). Resultaten indikerar att en estimeringsmetod kan ers¨atta kraftplattor vid viktad benpress. Varken nodmodellen eller skjuvmodellen har tillr¨ackligt h¨og korrelation vid viktad kn¨ab¨oj. Resultatet indikerar ¨aven att tryckk¨ansliga skosulor kan ge en mer korrekt normalkomponent j¨amf¨ort med om komponenten hade estimerats. Contents 1 Introduction 1 2 Method and materials 2 2.1 Data collection . 3 2.2 Node model . 4 2.3 Shear model . 6 2.3.1 Friction coefficient . 6 2.3.2 Shear components of the ground reaction force . 9 2.4 Data analysis . 11 3 Results 12 3.1 Squat . 12 3.2 Leg press . 16 4 Discussion 20 4.1 Potential sources of error . 21 4.2 Future work . 21 5 Conclusion 22 Appendices i A Background ii A.1 Introduction . ii A.2 The flywheel exercise device . iii A.3 Biomechanics . iv A.3.1 Anatomical planes, axes and terms . iv A.3.2 Ground reaction force . iv A.4 Biomechanical analysis . vii A.4.1 Motion capture . vii A.4.2 Inverse dynamics . ix A.5 Problems with today's ground reaction force measurements . xi A.6 AnyBody Modeling System . xii A.6.1 The ground reaction force prediction tool . xiii B Markerplacement during data collection xiv 1 Introduction Biomechanical analysis of human movement is an important tool when, for example, analysing the effects of an equipment on the human body, pathological gait or when trying to improve a certain movement, such as a baseball throw [1]. One of the important aspects of a biomechanical analysis of human movement is how forces and moments within joints act when performing a motion. This can be analysed in several ways, most often using inverse dynamics, see appendix chapter A.4.2. One of the most powerful ways is to use a musculoskeletal modeling software since it provides a way to not only determine the forces and moments but also muscle activation. No matter which approach is applied knowledge about the ground reaction force (GRF) is necessary [1]. The GRF is a three dimensional vector which is often decomposed into three anatomically im- portant directions; normal, anteroposterior and mediolateral, see appendix chapters A.3.1 and A.3.2. To determine the full GRF experimentally, the gold standard is to use force plates, which directly measure all three components [2]. However, force plates are bulky, expensive and sens- itive to environmental factors such as temperature [3, 4]. This set constraints on the possible ways to use the equipment, especially since the test person has to be in contact with the force plates in order to obtain a GRF signal [3]. To be able to perform studies without constraints on how a test person is allowed to move, several studies, using different methods, have been made on GRF prediction. Most of them focus on gait, which is one of the most common movements to analyse biomechanically [5]. Some of these methods, see appendix chapter A.5, have shown promising results during gait [6, 7, 8]. A recent, more general method showed promising result during both gait and some sports related movements. The method was used with a motion capture system as input to a musculoskeletal modeling software, which computed the full GRF [3, 9, 10, 11]. This method has however not been validated with an external load present and not during squat or leg press. Futhermore, the normal component of the GRF can be experimentally obtained using pressure sensitive insoles (PSIs) [12], see appendix chapter A.3.2. These insoles have only been used in a few GRF prediction models [6, 7]. If the PSIs are more accurate in the normal direction compared to a GRF estimation model, this could increase the accuracy of the two shear components as well. In turn, this could increase the accuracy of a biomechanical analysis.