S8 Abstracts / Gait & Posture 39S (2014) S1–S141 Fig. 1. Example of a strict knee flexion–extension simulation with combined 3D orientation errors of the femoral frame (Flexion 8 ◦ , Adduction 7 ◦ , Rotation 22 ◦ ). The crosstalk range does not exceed 10 ◦ in this very inaccurate configuration. knee joint compression forces and to a higher risk of developing knee osteoarthritis. Reference [1] Simonsen EB, et al. Clin Biomech (Bristol, Avon) 2012;27(6):573–7. [2] Remvig L, et al. J Rheumatol 2007;34(4):804–9. [3] Galli M, et al. Res Dev Disabil 2011;32(5):1663–8. http://dx.doi.org/10.1016/j.gaitpost.2014.04.013 009 Knee crosstalk is not a quality indicator of the hip rotation Eric Desailly Fondation Ellen Poidatz, Motion Analysis Unit, St Fargeau Ponthierry, France Introduction and aim: The knee crosstalk or varus-valgus arti- fact is usually used to access the potential error of transverse orientation of the thigh kinematic frame and therefore to provide an indicator of quality of the hip rotation measures. This phenomenon has been widely studied, and some authors even propose to intro- duce an offset of femoral rotation to minimize the crosstalk range [1,2]. However, only the impact of femoral rotation on the knee crosstalk was studied. The impact on the crosstalk of all the uncer- tainties related to the constitution of the femoral frame remains to be determined. Our hypothesis is that other uncertainties may affect the almost linear relationship between crosstalk and mea- surement error of the femoral rotation. Patients/materials and methods: Uncertainties location of the hip center (±0.02 m) [3], and location of the knee center (X: ±0.015 m, Z: 0.01 m) [4,5] have been used to consider potential errors in flexion and adduction of the femoral frame according to the size of the femur (0.25–0.45 m). A simulation of the kinematics of the knee in strict flexion-extension was used to test the effect of the combined 3D orientation errors of the femoral frame on the crosstalk values (Fig. 1). Results: The crosstalk effect is confirmed for isolated rotation errors. Respectively, for different femurs (0.25–0.45 m) adding flex- ion (±8 ◦ –4 ◦ ) and adduction errors (±7 ◦ –4 ◦ ) of the femur introduce coupling effects (Fig. 2). Crosstalk of 0 ◦ ,5 ◦ and 10 ◦ may then corre- spond to errors of 6 ◦ (0.25) to 3 ◦ (0.45), 14 ◦ (0.25) to 11 ◦ (0.45) and 22 ◦ (0.25) to 18 ◦ (0.45) of hip rotation. Fig. 2. Crosstalk (color map intensity) is not solely dependent of the femur rotation error but also of its adduction and flexion errors. Discussion and conclusions: The phenomenon of crosstalk is not questioned. However, it does not depend only on the hip rota- tion error. A moderate knee varus-valgus artifact is therefore not a guarantee of the hip rotation measurement quality. Reference [1] Baker R, et al. Hum Mov Sci 1999;18(5):655–67. [2] Rivest LP. J Biomech 2005;38(8):1604–11. [3] Harrington ME, et al. J Biomech 2007;40(3):595–602. [4] Della Croce U, et al. Med Biol Eng Comput 1999;37(2):155–61. [5] Della Croce U, et al. Med Eng Phys 2003;25(5):425–31. http://dx.doi.org/10.1016/j.gaitpost.2014.04.014 010 How good is your instrumented treadmill?—A comprehensive protocol L.H. Sloot 1,2,∗ , H. Houdijk 2 , J. Harlaar 1,2 1 Department of Rehabilitation Medicine, VU University Medical Center, Amsterdam, The Netherlands 2 Research Institute MOVE Amsterdam, The Netherlands Introduction and aim: With the introduction of instrumented treadmills in (clinical) gait analysis, it is imperative to evaluate the accuracy of force data, which has been shown to significantly affect joint torque calculations. The inaccuracy of force data is expected to be higher than measurements from floor-embedded force plates, due to the mounting of the load sensors in a large compliant tread- mill structure, the friction caused by the moving belt as well as variability in the belt speed. Therefore, we designed a standard and comprehensive protocol to check, improve and systematically report the performance of instrumented treadmills. To create a benchmark, the protocol was applied to two in-house instrumented treadmills. Materials and methods: The protocol was directed towards measurement of potential sources of systematic errors. First, the accuracy of force and centre of pressure (CoP) data was deter- mined using an instrumented stick [1]. In addition, factors affecting static CoP measurement were determined, i.e. crosstalk between and within belt plates using calibrated weights, while linearity and hysteresis of the force sensors were based on the instrumented stick data. Second, the effect of dynamic use of the treadmill was determined. The response to foot impact was estimated using an instrumented hammer and the variability in belt speed due to (un)loading was measured with a tachometer. Lastly, measurement variability over time was determined, including drift, warming