Modeling pelvis contact during lateral falls:​ multi-criteria evaluation of model performance and error source identification

Levine, I. C.; Pretty, S. P.; Laing, A. C.

Affiliations

¹Injury Biomechanics and Aging Laboratory, University of Waterloo, Waterloo, ON

²Home & Community Research Team, Toronto Rehabilitation Institute-UHN, Toronto, ON

Abstract

Hip fractures are a substantive public health issue. Simple mass-spring or mass-spring-damper systems have been utilized to model lateral fall on the hip. However, the biofidelity of these models is questionable as the femur/pelvis system is comprised of complex interactions between biological soft and skeletal tissues. This study investigated how increasing the complexity of contact models (from geometric and damping perspectives) influenced the accuracy of impact dynamics predictions during sideways falls, and the biomechanical sources of errors for each model. Fortysix participants (<35 years) underwent simulated sideways falls which involved their pelvis impacting a force plate with a low (but clinically relevant) velocity of 1 m/s. Simulations implementing five contact models (mass-spring(MS), Voigt(VG), Hertzian (HZ), HuntCrossley(HC), and volumetric(VO)) estimated normal force during impact. Subject-specific input parameters (mass, stiffness, and damping) were incorporated using previously- derived regression equations. Model predictions were evaluated against subject-specific experimental data to determine five error metrics including: peak force magnitude (Err_max), loading duration (ErrTTP), RMSE error over the impact period (Err_RMSE), impulse (Err_imp), and prediction within an experimental corridor (Err_corr)). Peak force estimates were substantively over-predictive for MS and VO, substantively under-predictive for VG, and best for HZ and HC. Time- to- peak force and impulse predictions were best for models with damping components (i.e. VG and HC, VO) but significantly over-predictive for MS and HZ. Err_corr and Err_RMSE were substantially improved for HC compared to all other models. Model errors were primarily linked to body composition, particularly body fat, overall body size, and floor-pelvis contact profile. Overall model performance was best for HC compared to all other models. Future model iterations should focus on characterizing the influence of body fat and adjusting contact geometry assumptions approximate shape and size of the floor-pelvis contact profile.

Links

URL: http://ibrc.osu.edu/wp-content/uploads/2018/05/Levine_manuscript.pdf

Cited Works (7)

Year	Entry
1991	Hirokawa S. Three-dimensional mathematical model analysis of the patellofemoral joint. J Biomech. 1991;24(8):659-671.
2017	Levine IC, Laing AC. Peak pressure and contact profile during sideways falls on the hip: links with individual characteristics and falling configuration. In: Proceedings of the 13th Ohio State University Injury Biomechanics Symposium. May 21-23, 2017; Columbus, OH.
2017	Levine IC. The Effects of Body Composition and Body Configuration on Impact Dynamics During Lateral Falls: Insights from in-Vivo, in-Vitro, and in-Silico Approaches [PhD thesis]. University of Waterloo; 2017.
1997	Robinovitch SN, Hayes WC, McMahon TA. Distribution of contact force during impact to the hip. Ann Biomed Eng. May 1997;25(3):499-508.
2007	Bouxsein ML, Szulc P, Munoz F, Thrall E, Sornay‐Rendu E, Delmas PD. Contribution of trochanteric soft tissues to fall force estimates, the factor of risk, and prediction of hip fracture risk. J Bone Miner Res*. June 2007;22(6):825-831.
1991	Eberhardt AW, Lewis JL, Keer LM. Normal contact of elastic spheres with two elastic layers as a model of joint articulation. J Biomech Eng. November 1991;113(4):410-417.
1991	Robinovitch SN, Hayes WC, McMahon TA. Prediction of femoral impact forces in falls on the hip. J Biomech Eng. November 1991;113(4):366-374.