Abstracts / Gait & Posture 30S (2009) S1–S153
models. We would, however, be tentative in using this system for faster movement speeds, at present. References  Taylor PN, et al. Arch Phys Med Rehabil 1999;80:1577–83.  Salarian A, et al. IEEE Trans Biomed Eng 2004;51(8):1434–43.
doi:10.1016/j.gaitpost.2009.08.039 O36 Human body acceleration: How to summarize the output from an inertial accelerometer? Vincent van Hees 1,∗ , Marcelo Pias 2 , Salman Taherian 2 , Soren Brage 1
Table 1 Comparison between metrics and the IMU reference values (based on optimized metric parameters). Metric
HPF VM HPF AV MAB VM MAB AV MAC VM MAC AV MAF VM MAF AV VM AV VM-1 AVE – mean
Agreement (g): mean (difference between 95% conﬁdence intervals)
0.86 0.87 0.89 0.90 0.42 0.46 0.63 0.62 0.58 0.32 0.58 0.33
0.95 0.90 0.93 0.90 0.94 0.91 0.90 0.83 0.95 0.92 0.95 0.77
−0.08 (0.23) −0.23 (0.46) −0.07 (0.21) −0.22 (0.44) −0.14 (0.50) −0.21 (0.45) −0.03 (0.43) −0.13 (0.38) 0.69 (1.39) 0.31 (0.62) 0.59 (1.19) −0.18 (0.53)
−0.05 (0.25) −1.03 (2.06) −0.08 (0.28) −1.03 (2.06) 0.22 (0.43) −0.26 (0.51) 0.40 (0.81) −0.11 (0.51) −0.10 (0.27) −0.62 (1.25) −0.21 (0.41) −1.14 (2.27)
MRC Epidemiology Unit, Institute of Metabolic Science, Cambridge, United Kingdom 2 Computer Laboratory, Cambridge University, Cambridge, United Kingdom Summary Twelve ways to summarize acceleration signals were evaluated in walking, jogging, and running. Results show that transparency in data processing is important; it enables a better understanding for and comparison between different summery measures and may help to reduce errors in the succeeding stages of data processing. Conclusions A high-pass ﬁlter is the best way to remove the gravitational component. Next the vector magnitude is the best way to combine the signals. Introduction Most accelerometers as applied in the ﬁeld of physical activity monitoring produce manufacturer dependent output values complicating the comparison and interpretation of results (e.g. Actigraph). Recently accelerometer devices, based on inertial acceleration sensors (MEMS), which allow for data storage at high sampling frequency and applicable in an epidemiological setting have become commercially available. The output of these devices is not summarized by the manufacturer allowing increased control over data processing by the customer. The current study evaluates 12 ways (metrics) of summarizing acceleration signals on their ability to remove the gravitational component. Patients, materials and methods Summary metrics were developed to (a) remove gravitational acceleration and (b) combine the different components of the acceleration signal. The gravitational component is removed by one of the following six methods: high-pass ﬁltering (HPF); minus one; minus the mean; and three subtractions of the signal by the backward moving average (MAB), the centred moving average (MAC), and the forward moving average (MAF). Signals are thereafter combined by taking either the vector magnitude (VM) or the average of the rectiﬁed sensor signals (AV). All metrics were evaluated in walking on level ground, and in a graded maximal treadmill test. One healthy participant wore an inertial measurement unit (MTx, Xsens Technologies B.V., Enschede, The Netherlands) on the wrist. The inertial measurement unit integrates three accelerometers, three gyroscopes, and three magnetometers. It allows for the estimation of segment orientation which was then used to rotate acceleration signals into a global coordinate system. This accurately separates the gravitational component from the linear acceleration component which is then used as a reference to evaluate the discussed metrics.
Results Following empirical tuning of the ﬁlter parameters, the metrics were applied to the acceleration signals. Metrics based on high-pass ﬁltering resulted in the best agreement and correlation with the reference values and the subtraction of the backward moving average serves as a good alternative (see Table 1). Combining the signals by taking the average instead of the vector magnitude resulted in scaling and lower agreement with the reference values (see Table 1). Discussion The current study shows that the removal of gravitational component is worth the effort as it increases the explained variation of up to 30%. Bouten et al.  concluded the opposite for an accelerometer worn on the waist. In the near future we will expand our analysis to other body locations. Reference  Bouten, et al. Med Biol Eng Comput 1997;35(1):50–6.
doi:10.1016/j.gaitpost.2009.08.040 O37 Feasibility of a functional procedure, operator independent, for wearable inertial measurement units in gait analysis Paolo Cappa 1 , Eduardo Palermo 1 , Fabrizio Rossi 2 , Maurizio Petrarca 2,∗ , Enrico Castelli 2
Patanè 1 , Stefano
Mechanics and Aeronautics Department, Sapienza University of Rome, Rome, Italy 2 Neurorehabilitation Department, Children’s Hospital Bambino Gesù, Rome, Italy Summary The aim of this study was to develop and test a protocol for the evaluation of the lower limb kinematics in gait analysis, using inertial measurement units (IMU). For this purpose a set of reference frames for each body segment (BRF), and a functional calibration procedure were introduced. The calibration was designed to be performed easily, independently by operator capability. The test consisted of four gait analysis trials on seven healthy subjects, performed with IMUs and optoelectronic system (OS). Conclusions Compared with OS, IMUs based protocol showed a mean absolute error of 5.4% and 8.2% on knee and hip Flex-Ext angles, respectively.
Abstracts / Gait & Posture 30S (2009) S1–S153
O38 Development and test of a protocol based on an Inertial and Magnetic Measurement System to measure the 3D kinematics of gait in real-life environment Andrea Giovanni Cutti 1,∗ , Pietro Garofalo 2 , Alberto Ferrari 2 , Michele Raggi 1 , Angelo Cappello 2 1 2
Fig. 1. Fl/Ex angles for knee (a) and hip (b) obtained with IMUs and OS.
Introduction The use of IMUs in Biomechanics represents an important alternative to magnetic and optoelectronic systems due to the low cost and the availability of open environment applications. An IMU measures the kinematics of the body segment which it is ﬁxed on, expressed in the same IMU reference frame (IRF). Therefore, in order to assess the absolute rotation of a BRF, one need to evaluate at ﬁrst the relative ﬁxed rotation between IRF and BRF. That relative rotation can be estimated by means of a proper calibration procedure, solutions previously proposed in scientiﬁc literature appear not feasible in clinical practice  and more speciﬁcally in the evaluation of lower limb kinematics . Patients/materials and methods Seven healthy subjects were asked to perform four gait analysis sessions. The Inertial Measurement System was composed by ﬁve IMUs placed on pelvis, thighs and shanks, a wearable Bluetooth unit for the synchronization of IMUs, and a PDA for data logging and displaying. For each body segment, BRFs were deﬁned relatively to the upright standing of the subject: z-axis parallel to the gravity vector pointing upside, y-axis belonging to sagittal plane pointing backward and x-axis automatically deﬁned. Therefore, the calibration procedure required the measurement of gravity vector in two subject conﬁgurations: standing upright and sitting with back and legs slightly inclined. The z-axis of BRF was evaluated by means of the gravity vector measured from the former subject position; the x-axis was evaluated by cross product of z and gravity vector obtained from the latter; the y-axis is calculated by cross product of z and x. The hip and knee Fl/Ex angles measured with IMUs and OS were analysed (Fig. 1). Results The mean of absolute difference between the Fl/Ex angle obtained from IMUs and OS was evaluated for hip and knee joints. The average and the standard deviation of this value over all trials were 4.1 ± 1.1◦ for the hip and 3.5 ± 0.9◦ for the knee, with a percentage on the maximum range of 8.2% and 5.4%, respectively. Discussion The comparison between IMU procedure and OS showed unnoticeable difference, then, the here proposed functional calibration for IMUs can be considered a good alternative to the OS in gait analysis. References  Picerno, et al. Gait Posture 2008;28(4):588–95.  Luinge, et al. J Biomech 2007;40(1):78–85.
INAIL Prostheses Centre, Vigorso di Budrio, Italy University of Bologna, Bologna, Italy
Summary A new gait analysis protocol (named Outwalk) was developed based on an Inertial and Magnetic Measurement System (IMMS) to collect 3D kinematic data during hundreds of gait cycles in real-life environments. Outwalk was speciﬁcally designed for amputees and children with cerebral palsy, and it requires only two simple calibration steps from “sensor mounting” to “gait measurement”. Outwalk kinematics was measured on four healthy subjects and compared with the kinematics of the CAST protocol. Conclusions Current results support the use of Outwalk for clinical measurements. Introduction The use of 3D gait analysis is currently limited to few medical centres, and its ability to monitor a patient’s typical walking capacity has still to be fully explored. IMMSs might overcome these limitations, being low-cost and fully wearable. Moreover, since the 3D orientation of their sensing units (SU) is known in a global earth-based coordinate system (CS) which is ubiquitous, long measurements are possible “out-of-the-lab”, in real-life environment. The aims of this work were to: (1) deﬁne a gait protocol (Outwalk) to be used in combination with the Xsens IMMS (Xsens Technologies, NL), suitable for amputees and cerebral palsy children; (2) compare the kinematics of Outwalk with a standard clinical protocol (CAST) in vivo ; (3) develop a software to use Outwalk in clinical setting (aim addressed in ). Patients/materials and methods To measure the pelvis-trunk, hips, knees, and ankle 3D kinematics, Outwalk requires 8 SUs, positioned on thorax, pelvis, thighs, shanks and feet. The anatomical CSs required to measure the trunkpelvis, hips and ankles kinematics are deﬁned and linked to the SUs on the segments through a static trial, with the subject either assuming a predeﬁned upright or supine posture. For the knees, the anatomical CSs are based on a functional estimation of the mean axis of rotation, executed either actively of passively. Before entering into clinical trial, Outwalk was tested on 4 healthy subjects (26–31 years old), each over 14 gait cycles. In particular, by synchronous measurements, we assessed the differences between CAST and Outwalk kinematics in terms of correlation, offset, range of motion, and overall kinematic patterns’ similarity (through the coefﬁcient of multiple correlation, CMC). Moreover, for one of the subject who also participated in , we compared the differences in terms of correlation and mean-absolute-variability between Outwalk and other four popular protocols in clinical gait analysis, including the Plug-in Gait. Results Over subjects, the CMC showed that Outwalk and CAST kinematics can be interchanged, offset included, for hip, knee and ankle ﬂexion-extension (CMC > 0.95) and hip ab-adduction (CMC > 0.93). The other joint angles can be interchanged offset excluded (CMC > 0.85). Outwalk kinematics showed the same