Journal of Biomechanics 47 (2014) 1704–1711

Contents lists available at ScienceDirect

Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com

A wearable system for multi-segment foot kinetics measurement H. Rouhani a,n, J. Favre a, X. Crevoisier b, K. Aminian a a b

Ecole Polytechnique Fédérale de Lausanne (EPFL), Laboratory of Movement Analysis and Measurement, ELH 137/ Station 11, CH-1015 Lausanne, Switzerland Centre Hospitalier Universitaire Vaudois and University of Lausanne (CHUV), Department of Orthopaedic Surgery and Traumatology, Lausanne, Switzerland

art ic l e i nf o

a b s t r a c t

Article history: Accepted 20 February 2014

This study aims to design a wearable system for kinetics measurement of multi-segment foot joints in long-distance walking and to investigate its suitability for clinical evaluations. The wearable system consisted of inertial sensors (3D gyroscopes and 3D accelerometers) on toes, forefoot, hindfoot, and shank, and a plantar pressure insole. After calibration in a laboratory, 10 healthy elderly subjects and 12 patients with ankle osteoarthritis walked 50 m twice wearing this system. Using inverse dynamics, 3D forces, moments, and power were calculated in the joint sections among toes, forefoot, hindfoot, and shank. Compared to those we previously estimated for a one-segment foot model, the sagittal and transverse moments and power in the ankle joint, as measured via multi-segment foot model, showed a normalized RMS difference of less than 11%, 14%, and 13%, respectively, for healthy subjects, and 13%, 15%, and 14%, for patients. Similar to our previous study, the coronal moments were not analyzed. Maxima–minima values of anteriorposterior and vertical force, sagittal moment, and power in shank-hindfoot and hindfoot-forefoot joints were significantly different between patients and healthy subjects. Except for power, the inter-subject repeatability of these parameters was CMC40.90 for healthy subjects and CMC40.70 for patients. Repeatability of these parameters was lower for the forefoot-toes joint. The proposed measurement system estimated multi-segment foot joints kinetics with acceptable repeatability but showed difference, compared to those previously estimated for the one-segment foot model. These parameters also could distinguish patients from healthy subjects. Thus, this system is suggested for outcome evaluations of foot treatments. & 2014 Elsevier Ltd. All rights reserved.

Keywords: Multi-segment foot kinetics Pressure insole Inertial sensors Gait Inverse dynamics

1. Introduction Assessment of ankle-joint kinetics has been applied in clinical evaluations (Valderrabano et al., 2007; Ingrosso et al., 2009). However, few studies have assessed the multi-segment foot kinetics (MacWilliams et al., 2003; Buczek et al., 2006; Bruening et al., 2012; Dixon et al., 2012). Particularly, the application of multi-segment foot kinetics for clinical evaluation is rare, and it is currently unclear whether these models are sensitive enough to detect kinetic differences between healthy and pathological feet. Furthermore, in all these prior works, stationary devices (e.g., force-plate and cameras) were used for measurement of ground reaction force (GRF) and foot kinematics. Only a few consecutive gait cycles can be recorded in the constrained space of a laboratory, which confines a natural gait (Bussmann et al., 1995). Therefore, field measurements using body-worn systems might be preferable to lab measurements. Recently, we suggested estimation of three-dimensional (3D) GRF using pressure insoles and verified its accuracy for pathological gaits n

Corresponding author. Tel.: þ 41 21 693 5675; fax: þ41 21 693 6915. E-mail addresses: hossein.rouhani@epfl.ch (H. Rouhani), julien.favre@epfl.ch (J. Favre), [email protected] (X. Crevoisier), kamiar.aminian@epfl.ch (K. Aminian). http://dx.doi.org/10.1016/j.jbiomech.2014.02.027 0021-9290 & 2014 Elsevier Ltd. All rights reserved.

(Rouhani et al., 2010). Then, we suggested inertial measurement units (IMU) for multi-segment foot kinematics measurement (Rouhani et al., 2012b). Based on these two systems, we developed a method for ambulatory assessment of ankle kinetics using a onesegment foot model and demonstrated its suitability for clinical evaluations (Rouhani et al., 2011b). The present study aimed to extend the mentioned method for one-segment foot kinetics measurement, to ambulatory measurement of multi-segment foot kinetics during long-distance walking. Specifically, our first objective was to assess the influence of multisegment foot modeling on the ankle kinetics estimation compared to one-segment foot modeling. Our second objective was to assess the sensitivity of the measured foot kinetics using multi-segment foot model to the difference between gaits of healthy subjects and patients with ankle osteoarthritis. Thereby, we investigated the suitability of these measurements for clinical evaluations. 2. Materials and methods 2.1. Wearable measurement system We considered the foot-ankle complex as four segments: toes (TO), forefoot (FF), hindfoot (HF), and shank (SH), according to Rouhani et al. (2011c, 2012c) and

H. Rouhani et al. / Journal of Biomechanics 47 (2014) 1704–1711 Bruening et al. (2012). According to Rouhani et al. (2012a, 2012b), one IMU was placed on each segment where the soft tissue between skin and bone had its minimum thickness (Fig. 1). Each IMU consisted of a 3D accelerometer and a 3D gyroscope, and the four IMUs were connected to two portable data-loggers (Physilog, BioAGM, CH) embedded in a belt. Subjects also wore custom-made shoes embedding pressure insoles (Pedar, Novel, DE). The insoles were glued on the shoes. The foot was fixed in the shoe using medical tape around the foot and shoe to minimize sliding. These devices recorded synchronously at 200 Hz. These efforts aimed to calculate the forces, moments, and power in the FF  TO, HF  FF, and SH  HF joints based on the inverse dynamics approach. Calculation of these parameters requires assessment of 3D GRF and kinematics for each foot segment using these body-worn sensors. The following sections describe these required steps. 2.2. Three-dimensional segments orientation Prior to the measurement of the gait trials, an auxiliary system (Vicon, UK) was used for anatomical calibration of the IMUs in order to obtain repeatable results that can be compared among subjects (Cappozzo et al., 2005). Using this procedure, each IMU's output was expressed relative to the bone-embedded anatomical frame (BAF) of the segment instead of the technical frame of the IMU, following Rouhani et al. (2012b). During the gait trials, first, the stance phase of the gait cycle was detected based on the shank IMU measurements following Salarian et al. (2004). Second, for each stance phase, a floor-fixed reference frame (SRF), aligned with the direction of the foot's headway, was determined following Rouhani et al. (2011b). Third, the instantaneous rotation matrices of segments' BAF with respect to SRF (RSRF BAF SEG , SEG: TO, FF, HF, SH) were calculated using IMUs according to Rouhani et al. (2012b) SRF and Favre et al. (2006). For this purpose, RBAF SEG was estimated using the accelerometer output at a typical instant during foot-flat period, when the acceleration of foot segment was small. Then, this matrix was estimated in all other instants of the stance phase using forward and backward strap-down integration of the gyroscope output. 2.3. Three-dimensional GRF for multi-segment foot Similar to Rouhani et al. (2011b), prior to the walking trials, as a subject stood still, an auxiliary system (Vicon, UK) was used to calibrate the pressure insoles so that the 3D GRF, the COP, and the coordinates of the insole's pressure sensors were expressed relative to SRF instead of the insole's frame. Also, the foot joints location was measured using cameras based on foot anatomical landmarks. The joint locations were assumed to be the midpoints of the palpated maleolli for SH HF joint, navicular and cuboid for HF FF joint, and the heads of the 1st, 2nd, and 5th metatarsals (average 3D coordinate of these three recorded locations) for FF TO joint. Then, each pressure insole's sensor element was allocated to one of the three foot segments based on its position with respect to the HF FF and FF TO joints location. Next, during each stance phase of the walking trials, the 3D GRF components, frictional torque (T), and center of pressure (COP) for the one-segment foot in each stance phase were estimated. According to Rouhani et al. (2010), it was assumed that these parameters for whole foot can be estimated based on the vertical pressure distribution measured by the insoles. Prior to gait trials, each subject, wearing the insoles-embedding shoes, walked a few times over a force-plate. Then,

1705

mappings between recorded vertical pressure distribution and 3D GRF components (and T) were estimated. These mappings were later used to estimate 3D GRF components (and T) only based on insole measurements during long-term gait outdoor. Then, for each time sample of each stance phase, the GRF, T, and COP calculated for the one-segment foot were subdivided among the three foot segments to calculate the GRFSEG, TSEG, and COPSEG. To this end, the pressure insole elements were divided into three regions, based on the location of the HF  FF and FF  TO joints, measured previously. The COP of each segment (COPSEG) was calculated based on the pressure values measured by the pressure sensors allocated to the segment: 2

J

3T

J

6∑ FV j xj ∑ FV j yj 7 7 6j j COP SEG ¼ 6 07 7 6 J J 5 4 ∑ FV j ∑ FV j j

ð1Þ

j

where FVj is the vertical force measured by the insole's jth sensor, xj and yj are the coordinates of the jth sensor, and J includes all sensors allocated to the segment. According to Scott and Winter (1993) and MacWilliams et al. (2003), it was assumed that shear GRF components are distributed among foot segments in ! proportion to their vertical GRF. Therefore, the 3D GRF for each segment (G R F SEG ) was calculated as follows: ! FV SEG ! G R F SEG ¼ G R F FV

ð2Þ

where FVSEG is the sum of the vertical force measured by all insole sensors allocated ! to the segment, G R Fis the 3D ground reaction force for the one-segment foot estimated according to Rouhani et al. (2010), and FV is the sum of the vertical forces measured by all insole sensors. ! The distribution of T among the segments was more complex. In fact, in a three–segment foot model, the vectorial resultant of the anterior-posterior and ! ! medial-lateral components of each G R F SEG , hereafter referenced as shear G R F SEG ! (SG R F SEG ), is applied at the segment's COPSEG. In contrast, in a one-segment foot ! ! model, the whole foot's shear G R F(SG R F) is applied at the whole foot's COP. Therefore, equivalence of torque between one-segment and three-segment foot models is as follows (Fig. 2): ! ! ! ! ! T SUM ¼ T  ðC O P TO  C O PÞ  SG R F TO ! ! ! ! ! !  ðC O P FF  C O PÞ  SG R F FF  ðC O P HF  C O PÞ  SG R F HF ð3Þ ! where T SUM is the summation of frictional torques applied on the three segments. ! Then, similar to Eq. (2), the frictional torque of each segment ( T SEG , SEG: TO, FF, HF) was derived as follows: ! ! FV SEG T SEG ¼ T SUM FV

ð4Þ

2.4. Multi-segment foot kinetics To calculate the forces, moments, and power in the three joints, the equations described by Rouhani et al. (2011b) for the one-segment foot model were extended to the three-segment foot model where the mass and inertia of foot segments were neglected. Therefore, according to the equilibrium of forces and moments applied

Fig. 1. The wearable measurement system composed of IMUs, pressure insoles, cables, batteries, and synchronized data-loggers, altogether weighed 1.5 kg.

1706

H. Rouhani et al. / Journal of Biomechanics 47 (2014) 1704–1711

COPTO

TO

SH

TTO

SGRFTO TFF

COPFF SGRFFF

SH-HF MSH-HF

COP

FSH-HF

HF GRFHF

MHF-FF T HF

FF

FF HF-FF

GRFFF FF-TO

FHF-FF

COPHF

GRFTO

COPHF

T SGRF

THF

TFF

TTO TO MFF-TO COPFF FFF-TO COP TO

SGRFHF HF

Fig. 2. (a) Multi-segment foot i.e.—shank (SH), hindfoot (HF), forefoot (FF), and toes (TO) – and forces and moments applied on its joint sections i.e. – SH  HF, HF  FF, and ! ! FF  TO—during stance time. COPSEG, G R F SEG , and T SEG for each segment also are shown (in gray). The depicted joint forces and moments were calculated via Newton–Euler ! ! ! equations for each segment. (b) Partitioning of the foot sole into three segments. COP, shear G R F(SG R F), and T for whole foot (gray) were measured by pressure insole, and ! ! were estimated for each segment (COPSEG, SG R F SEG , and T SEG : black) based on pressure distribution. on the TO, FF, and HF segments:

Section 2.3:



ð5Þ

! ! BAF SEG ! BAF SEG P osJoint 1  Joint 2 ¼ RSRF BAF SEG ð P osJoint 1  P osJoint 2 Þ

ð6Þ

2.5. Measurement protocol



F FF  TO ¼ −GRF TO →







M FF  TO ¼ −T TO −P osCOPTO −FF  TO  GR F TO →





ð7Þ

F HF  FF ¼ −GR F FF þ F FF  TO

ð8Þ

P FF  TO ¼ M FF  TO : ðωTO −ωFF Þ →















M HF  FF ¼ −T FF −P osCOPFF −HF  FF  GRF FF þ P osFF  TO−HF  FF →



F FF  TO þ M FF  TO →



ð9Þ



P HF  FF ¼ M HF  FF : ðωFF −ωHF Þ →



ð10Þ



F SH  HF ¼ −GR F HF þ F HF  FF

ð15Þ

The walking trials of 10 healthy subjects (7 females; age 61 713 years; height 1667 9 cm; weight 677 10 kg) and 12 age-matched patients with unilateral endstage ankle osteoarthritis (4 females; age 58 7 13 years; height 1697 7 cm; weight 817 19 kg) collected during our previous study (Rouhani et al., 2011b) were used to evaluate this new methodology to measure the multi-segment foot kinetics. After calibration of the IMUs and pressure insoles using the auxiliary systems (camera and force-plate) as described above, the individuals walked two 50-m trials in a hospital corridor, during which only IMUs and pressure insoles were used. The patients reported a Foot Function Index of 48 716 and American Orthopaedic Foot and Ankle Society score for ankle-hindfoot of 46 712. The local ethics committee approved the experimental protocol.

ð11Þ 2.6. Data analysis











M SH  HF ¼ −T HF −P osCOP HF −SH  HF  GR F HF þ P osHF  FF−SH  HF →



F HF  FF þ M HF  FF →





P SH  HF ¼ M SH  HF : ðωHF −ωSH Þ

ð12Þ ð13Þ

where F, M, and P correspond to force, moment, and power in the FF  TO, HF  FF, ! and SH  HF joint sections, indicated as subscript. P osA  B indicates the relative position of point A with respect to point B. All parameters are expressed in SRF. Segments' angular velocity (ωSEG; SEG: TO, FF, HF, and SH), was expressed in SRF using RSRF BAF SEG . The position of each COPSEG relative to the proximal joint expressed in SRF, i.e., ! ! ! P osCOPTO −FF  TO , P osCOPFF −HF  FF , and P osCOPHF −SH  HF , was calculated following the method suggested by Rouhani et al. (2011b). To this end, first, a vector from the instantaneous COPSEG (measured by insole) to the initial position of the joint in a stand-still situation (as determined in the calibration procedure in Section 2.3) was ! SEG calculated and expressed in BAFSEG ( P osBAF Joint  COP SEG ). Then, COPSEG was assumed to be the instantaneous center of rotation of the segment. Therefore, the instanta! BAF SEG ! neous vector of P osSRF Joint  COP SEG was expressed as the rotation of P osJoint  COP SEG around the instantaneous COPSEG: ! BAF SEG ! SRF P osCOPSEG  Jo int ¼  RSRF BAF SEG P osJo int  COP SEG

ð14Þ

! ! Finally, the relative position of two joints ( P osJoint 1  Joint 2 ), i.e., P osFF  TO !  HF  FF and P osFF  TO  HF  FF , expressed in SRF, was obtained based on their relative position expressed in BAFSEG, measured in the calibration procedure in

The patterns (time series) of anterior-posterior, medial-lateral, and vertical forces (FAnt–Post, FMed–Lat, and FVertical), sagittal and transverse moments (MSagittal and MTransverse), and power (P) were calculated in the three joint sections during every stance phase of each subject's gait using the method described above. The three first and three last gait cycles of each trial were discarded in data analysis to minimize the start-up and end-up effects of gait. All time series were temporally normalized in stance time (1–100%) and averaged over all gait cycles (Favre et al., 2010). According to Rouhani et al. (2011b) and Dixon et al. (2012), the coronal moment (MCoronal) in foot joints is very sensitive to measurement errors, inconsistent among the subjects, and less suitable for clinical evaluations. Therefore, MCoronal components were not considered in data analysis. Forces were normalized to BW (body weight) and moments and powers were normalized to BW  BH (body height). For each gait cycle, the MSagittal, MTransverse, and P in the SH  HF joint, obtained using the multi-segment foot model, were compared to the ankle kinetics obtained using a one-segment foot model (Rouhani et al., 2011b). The calculated force in each joint section expressed the 3D GRF borne by the joint and applied to its distal segments, and thus the forces in the ankle joint is identical for both models. Multisegment and one-segment patterns were compared using the RMS difference normalized to the range (NRSMD) and correlation coefficients (R). The inter-subject repeatability of the kinetic parameters was calculated for each population using the coefficients of multiple correlations (CMC). Additionally, maxima–minima values of force, moment, and power in the three joints (Figs. 4 and 5) were compared between the healthy and patient populations using Wilcoxon rank-sum test. All statistical analyses were done using Matlab and the level of significance was set to α ¼ 0.05.

H. Rouhani et al. / Journal of Biomechanics 47 (2014) 1704–1711

1707

3. Results The median (over a population) NRSMD comparing the MSagittal, MTransverse, and P in the ankle joint, between the multi-segment and the one-segment models, was 10.5%, 13.7%, and 12.7%, for the healthy subjects. The median correlation coefficient between these patterns was 0.99, 0.98, and 0.94, and the median normalized difference between peaks of these patterns was 9.9%, 33.7%, and 49.5% for the healthy subjects (Table 1 and Fig. 3). The inter-subject repeatability of the parameters in the sagittal plane (FAnt–Post, FVertical, MSagittal) was high (CMC 40.90) for all joints of the healthy subjects and lower for the patients' joints (CMC between 0.74 and 0.90 in SH  HF and HF FF) (Table 2). Maxima–minima values of the FAnt-Post, FVertical, MSagittal, and P were significantly different in the SH–HF and HF–FF joints between the two populations (Table 3 and Figs. 4 and 5).

4. Discussions The present study introduced a wearable system to measure the multi-segment foot kinetics during long-distance walking. This study showed that the ankle kinetics estimated by this system based on multi-segment foot model and one-segment foot model was different. Moreover, the estimated multi-segment foot kinetics was sensitive enough to detect kinetic differences between healthy subjects and patients with ankle osteoarthritis. 4.1. Comparison with one-segment foot model The MSagittal and MTransverse of the ankle joint obtained with the multi-segment foot model and one-segment foot model showed NRMSDs lower than 15% and correlation coefficients higher than 0.97 (Table 2). These observed differences were natural and due to the different estimated force lever arms in the two models. With the one-segment model, the foot was assumed as a rigid segment whose orientation was estimated by the forefoot IMU. However, with the multi-segment model, the foot was assumed as three rigid segments in a chain with different 3D orientations, which accommodates the flexibility of the foot. This flexibility affected the calculation of the lever arms used to calculate the ankle moments. The one-segment foot model markedly overestimates the ankle power. This difference is mainly due to the joint used for the angular velocity calculation. For the multi-segment foot model, the SH  HF rotation was considered in Eq. (13), which is smaller than the SH  FF rotation (considered for the one-segment foot model), and thus resulted in smaller and more realistic ankle power values. Similar overestimation of ankle power in a Table 1 Comparison between ankle moments (Sagittal and Transverse) and power obtained with one-segment and multi-segment models using the proposed wearable system. Correlation coefficient (R), normalized RMS difference (NRMSD%), and normalized difference between peaks (%), between two time series are presented for each population. Results are expressed as median (inter-quartile range) over the population.

Healthy subjects Patients

Correlation Coefficient (R) MSagittal MTransverse 0.99(0.01) 0.98(0.05) 1.00(0.01) 0.97(0.06)

Healthy subjects Patients

Normalized RMS Difference (NRMSD%) 10.5(6.0) 13.7(4.2) 12.7(5.9) 13.0(6.5) 15.0(9.1) 13.7(3.9)

Healthy subjects Patients

Normalized Difference Between Peaks (%) 9.9(20.0) 33.7(13.9) 49.5(27.4) 15.0(34.2) 21.8(35.1) 43.3(26.6)

Power 0.94(0.06) 0.90(0.17)

Fig. 3. Moments (Sagittal and Transverse) and power time series during stance time in the SH  HF joint for healthy subjects. Mean curves (solid black) and mean7 std (dashed black) over subjects obtained with the multi-segment foot model are presented. Results for the SH  Foot joint obtained via the one-segment foot model also are presented (red). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

one-segment foot model was observed by MacWilliams et al. (2003) and Dixon et al. (2012). In general, the multi-segment foot model considered the flexibility among foot segments and not only resulted in more realistic estimation of the ankle joint kinetics but also estimated other foot joints kinetics.

1708

H. Rouhani et al. / Journal of Biomechanics 47 (2014) 1704–1711

Table 2 Repeatability of patterns (time series) over subjects of each population for kinetic parameters in the three joints, presented as coefficient of multiple correlations (CMC). FAnt–Post

FMed–Lat

FVertical

MSagittal

MTransverse

Power

4.3. Suitability for clinical evaluations

Healthy subjects Patients

SH  HF 0.93 0.84

0.74 0.74

0.90 0.9

0.93 0.81

0.86 0.77

0.70 0.53

Healthy subjects Patients

HF  FF 0.92 0.74

0.84 0.79

0.95 0.86

0.95 0.82

0.85 0.78

0.77 0.7

Healthy subjects Patients

FF  TO 0.93 0.64

0.80 0.66

0.92 0.62

0.91 0.68

0.77 0.62

0.87 0.69

Table 3 Forces (FAnt–Post, FMed–Lat and FVertical), moments (MSagittal and MTranscerse), and power (P) features in the SH  HF, HF FF, and FF  TO joints for healthy and patient populations presented as median (inter-quartile range). In the case of significant difference between populations (p-value o 0.05), p-value is shown in bold. Force maxima–minima values of the SH  HF joint were presented previously by Rouhani et al., (2011b). Healthy subjects

Patients

P-value

FAnt-Post (Max) FMed-Lat (Max1) FVertical (Min1) FAnt-Post (Min) FMed-Lat (Max2) FVertical (Max) FVertical (Min2) MSagittal (Max) MTransverse (Max) P (Max)

16.91(9.28) 4.96(2.26)  115.46(16.72)  18.83(9.03) 8.32(4.38)  72.83(14.39)  114.97(18.27) 7.44(1.27) 0.87(0.58) 22.55(17.01)

8.82(4.44) 7.08(2.08)  101.03(7.38)  10.81(3.6) 9.59(2.42)  90.36(7.22)  104.03(9.58) 5.98(0.58) 0.73(0.48) 6.41(10.86)

p¼ 0.003 p¼ 0.081 p¼ 0.003 p¼ 0.001 p¼ 0.717 p¼ 0.002 p¼ 0.006 p¼ 0.001 p¼ 0.223 p¼ 0.023

HF  FF

FAnt-Post (Min) FMed–Lat (Max) FVertical (Min) MSagittal (Max MTransverse (Max) P Max

 18.83(9.17) 8.3(4.38)  114.97(18.25) 5.22(0.68) 0.75(0.53) 9.83(8.46)

 10.81(3.62) 9.58(2.06)  103.97(9.7) 3.89(1) 0.53(0.44) 3.44(2.91)

p¼ 0.001 p¼ 0.717 p¼ 0.006 p¼ 0.001 p¼ 0.199 p¼ 0.005

FF  TO

FAnt–Post (Min) FMed–Lat (Max) FVertical (Min) MSagittal (Max) MTransverse (Max) P (Max)

 10.99(3.81) 4.41(2.87)  58.49(10.83) 0.77(0.22) 0.23(0.14)  3.84(0.79)

SH  HF

who measured it using IMUs and 3D force sensors beneath the shoe's outsole. Instead of thick shoe outsoles, we used pressure insoles, which resulted in more natural foot–ground interface during gaits (Liedtke et al., 2007).

 5.16(4.75) 4.39(3.18)  48.35(22.26) 0.62(0.3) 0.12(0.16)  1.54(1.24)

p¼ 0.001 p¼ 0.489 p¼ 0.052 p¼ 0.07 p¼ 0.106 p¼ 0.008

4.2. Comparison with previous studies Few previous studies calculated joint kinetics in a multisegment foot and there is no comprehensive agreement on the range of these parameters, even when measured by stationary systems (force-plates and cameras). Nevertheless, the MSagittal, MTransverse, and power obtained in the three joints in our study demonstrated solid agreement with the results of MacWilliams et al. (2003), although they used different foot segmentation. Our results also agreed with those of Bruening et al. (2012), particularly for MSagittal and Power in the foot joints, although they used two adjacent force-plates for GRF assessment where the gait can be different with long-distance gait. Additionally, both these studies assessed gaits of children, with natural differences from our elderly subjects. Our results for the ankle joint kinetics obtained with the multi-segment and one-segment foot models agreed with those of Dixon et al. (2012). Also, the inter-subject variability of our results was similar to these studies that used stationary motion capture systems. Finally, our results for the ankle joint moment agreed with those of Schepers et al. (2007)

Despite its relevance for clinical evaluations, it is unclear whether the SH HF, HF FF, and FF TO joints kinetics estimated using multi-segment models can detect kinetic differences within the foot between different populations. We compared patients with ankle osteoarthritis to healthy subjects, since ankle osteoarthritis has been shown to alter multi-segment foot kinematics (Rouhani et al., 2012a). Estimated force and moment components showed higher inter-subject repeatability in the sagittal plane compared to the other planes and in the SH  HF and HF  FF joints compared to the FF TO joint (Table 2), and thus were more suitable for clinical evaluations. Maxima–minima values of FAnt– Post, FVertical, and MSagittal in the SH  HF and HF  FF joints of patients were significantly different compared to healthy subjects, indicating significant alteration of multi-segment foot kinetics due to ankle osteoarthritis (Table 3 and Figs. 4 and 5). The sensibility of the estimated FF  TO kinetics was not found high enough for this clinical evaluation, potentially because of its lower inter-subject repeatability. In contrast, the power maxima–minima values had large and significant differences between populations in all foot joints, despite their lower repeatability. This first application of a multi-segment model showed great potential for future clinical evaluations, mainly for FAnt-Post, FVertical, MSagittal, and P in the SH HF and HF  FF joints. 4.4. Modeling limitations Standard measurement of the multi-segment foot joints kinetics, besides the ankle, in gait lab requires a system to measure the 3D GRF separately for each foot segment, ideally 3D GRF distribution measurement insoles (Dobrzynska and Gijs, 2012). Such a reference system does not currently exist as a gold standard. Therefore, similar to previous publications (MacWilliams et al., 2003; Dixon et al., 2012), currently we do not have the possibility of validating our results for multi-segment foot joints with a gold-standard system. Asking the subject to step with a foot joint exactly astride two forceplates (Bruening et al., 2012) might be suggested as an alternative but this approach raises practical issues, changes the natural gait patterns, can subdivide the foot to only two segments, and particularly does not allow long-distance gait assessments. In the absence of a reference system, our algorithm for the first time could divide the 3D GRF among multiple segments using a 1D (vertical) pressure insole, and assess the multi-segment foot kinetics. We had previously validated estimations of the foot segments kinematics (Rouhani et al., 2012b), 3D GRF for whole foot (Rouhani et al., 2010), COP, and force level arms used in inverse dynamics approach to calculate the ankle kinetics using our wearable system (Rouhani et al., 2011b). The key point for this methodology extension to the multi-segment foot model was the distribution of 3D GRF among the HF, FF, and TO segments. Following Scott and Winter (1993) and MacWilliams et al. (2003), we assumed that shear GRF components are distributed among foot segments in proportion to vertical GRF. Bruening et al. (2010) showed that although this assumption may result in exacerbated errors for more dynamic activities, it results in smaller errors and can be suitable for applications such as clinical gait analysis, which was targeted in our study. Moreover, any assumptions for the shear GRF and frictional torque distribution among the foot segments was mainly applied during the foot-flat period when all shear force components and frictional torque had small amplitudes with no maxima–minima

H. Rouhani et al. / Journal of Biomechanics 47 (2014) 1704–1711

1709

Fig. 4. Moments (Sagittal and Transverse) and power time series during stance time in the SH  HF, HF  FF, and FF  TO joints. Mean curves (solid) and mean7 std (dashed) over subjects are presented. Results of healthy subjects (black) and patients (red) are presented. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

values. Consequently, these assumptions minimally affected the calculated maxima–minima values of joints moments and power, used for clinical evaluations. Therefore, although we cannot validate our results for the multi-segment foot joints, we expect that the differences between the multi-segment foot joint kinetics measured by our ambulatory system and stationary systems would be similar to those differences we previously assessed for the ankle joint kinetics in (Rouhani et al., 2011b). In our study, the 3D GRF was estimated for all toes as one segment but the IMU had to be placed on one rigid bone of this segment (on the first toe) to estimate the orientation of all toes. However, this limitation is expected to minimally affect the results, since the relative motions among toes minimally affected the lever arms calculation. Besides, the GRF applied on the first toes is larger than GRF applied on any other toe (Rouhani et al., 2011a), and thus it was the most relevant choice for IMU placement.

5. Conclusions We proposed a wearable system to measure the multi-segment foot kinetics during long-distance walking. The measured ankle kinetics was different and more realistic compared to our previous results obtained with a one-segment foot model. The measured force, moment, and power, particularly in the sagittal plane and in the SH  HF and HF  FF joints showed acceptable inter-subject repeatability. Moreover, their maxima–minima values distinguished patients with gait disorder from healthy subjects during 50-meter walks. The proposed system can be accessible for most clinics (unlike prototypes), rapidly installed, used without need for skillful engineers, and integrated in instrumented shoes. In its current form, this system required auxiliary devices (e.g. cameras and force-plate) to calibrate the IMUs' and the pressure insole's measured quantities prior to clinical measurements. However,

1710

H. Rouhani et al. / Journal of Biomechanics 47 (2014) 1704–1711

Fig. 5. Forces in joint sections (Anterior–Posterior, Medial–Lateral, and Vertical) time series during stance time in the SH HF, HF FF, and FF TO joints. Mean curves (solid) and mean7std (dashed) over subjects are presented. Results of healthy subjects (black) and patients (red) are presented. The force patterns (time series) of the SH HF joint were presented previously by Rouhani et al. (2011b). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

calibration without auxiliary devices is expected in the future (Favre et al., 2009; Rouhani et al., 2010). The present study provided the multi-segment foot kinetics measurement for clinical evaluations. Nevertheless, as we have previously shown, the pressure insoles and inertial sensors used in this system can be used separately to measure the plantar pressure distribution (Rouhani et al., 2011a) and multi-segment foot kinematics (Rouhani et al., 2012a) for outcome evaluation of foot and ankle treatments during long-distance gait.

authors would like to thank Prof. Brigitte Jolles for her advice and Mr. Pascal Morel and Mr. Jean Gramiger for their assistance in measurements.

Conflict of interest

References

None. Acknowledgments The work was supported by the Fonds National Suisse de la Recherche Scientifique (SNSF), Grant no. 3200B0-120422/1. The

Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.jbiomech.2014.02.027.

Bruening, D.A., Cooney, K.M., Buczek, F.L., Richards, J.G., 2010. Measured and estimated ground reaction forces for multi-segment foot models. J. Biomech. 43, 3222–3226. Bruening, D.A., Cooney, K.M., Buczek, F.L., 2012. Analysis of a kinetic multi-segment foot model part II: kinetics and clinical implications. Gait Posture 35, 535–540. Buczek, F.L., Walker, M.R., Rainbow, M.J., Cooney, K.M., Sanders, J.O., 2006. Impact of mediolateral segmentation on a multi-segment foot model. Gait Posture 23, 519–522.

H. Rouhani et al. / Journal of Biomechanics 47 (2014) 1704–1711

Bussmann, J.B.J., Veltink, P.H., Koelma, F., Van Lummel, R.C., Stam, H.J., 1995. Ambulatory monitoring of mobility-related activities: The initial phase of the development of an activity monitor. Eur J Phys Med Rehabil. 5, 2–7. Cappozzo, A., Della Croce, U., Leardini, A., Chiari, L., 2005. Human movement analysis using stereophotogrammetry: part 1: theoretical background. Gait Posture 21, 186–196. Dixon, P.C., Bohm, H., Doderlein, L., 2012. Ankle and midfoot kinetics during normal gait: A multi-segment approach. J. Biomech. 45, 1011–1016. Dobrzynska, J.A., Gijs, M.A.M., 2012. Flexible polyimide-based force sensor. Sens. Actuators A 173, 127–135. Favre, J., Jolles, B.M., Siegrist, O., Aminian, K., 2006. Quaternion-based fusion of gyroscopes and accelerometers to improve 3D angle measurement. Electron. Lett. 42, 612–614. Favre, J., Aissaoui, R., Jolles, B.M., de Guise, J.A., Aminian, K., 2009. Functional calibration procedure for 3D knee joint angle description using inertial sensors. J. Biomech. 42, 2330–2335. Favre, J., Crevoisier, X., Jolles, B.M., Aminian, K., 2010. Evaluation of a mixed approach combining stationary and wearable systems to monitor gait over long distance. J. Biomech. 43, 2196–2202. Ingrosso, S., Benedetti, M.G., Leardini, A., Casanelli, S., Sforza, T., Giannini, S., 2009. Gait analysis in patients operated with a novel total ankle prosthesis. Gait Posture 30, 132–137. Liedtke, C., Fokkenrood, S.A.W., Menger, J.T., van der Kooij, H., Veltink, P.H., 2007. Evaluation of instrumented shoes for ambulatory assessment of ground reaction forces. Gait Posture 26, 39–47. MacWilliams, B.A., Cowley, M., Nicholson, D.E., 2003. Foot kinematics and kinetics during adolescent gait. Gait Posture 17, 214–224. Rouhani, H., Favre, J., Crevoisier, X., Aminian, K., 2010. Ambulatory assessment of 3D ground reaction force using plantar pressure distribution. Gait Posture 32, 311–316.

1711

Rouhani, H., Crevoisier, X., Favre, J., Aminian, K., 2011a. Outcome evaluation of ankle osteoarthritis treatments: Plantar pressure analysis during relatively longdistance walking. Clin. Biomech. 26, 397–404. Rouhani, H., Favre, J., Crevoisier, X., Aminian, K., 2011b. Ambulatory measurement of ankle kinetics for clinical applications. J. Biomech. 44, 2712–2718. Rouhani, H., Favre, J., Crevoisier, X., Jolles, B.M., Aminian, K., 2011c. Segmentation of foot and ankle complex based on kinematic criteria. Comput. Methods Biomech. Biomed. Eng. 14, 773–781. Rouhani, H., Favre, J., Aminian, K., Crevoisier, X., 2012a. Multi-segment foot kinematics after total ankle replacement and ankle arthrodesis during relatively long-distance gait. Gait Posture 36, 561–566. Rouhani, H., Favre, J., Crevoisier, X., Aminian, K., 2012b. Measurement of multisegment foot joint angles during gait using a wearable system. J. Biomech. Eng. 134 (article no. 061006). Rouhani, H., Favre, J., Crevoisier, X., Jolles, B.M., Aminian, K., 2012. A comparison between joint coordinate system and attitude vector for multi-segment foot kinematics. J. Biomech. 45, 2041–2045. Salarian, A., Russmann, H., Vingerhoets, F.J., Dehollain, C., Blanc, Y., Burkhard, P.R., Aminian, K., 2004. Gait assessment in Parkinson's disease: toward an ambulatory system for long-term monitoring. IEEE Trans Biomed Eng 51, 1434–1443. Schepers, H.M., Koopman, H.F.J.M., Veltink, P.H., 2007. Ambulatory assessment of ankle and foot dynamics. IEEE Trans. Biomed. Eng. 54, 895–902. Scott, S.H., Winter, D.A., 1993. Biomechanical model of the human foot: kinematics and kinetics during the stance phase of walking. J. Biomech. 26, 1091–1104. Valderrabano, V., Nigg, B.M., von Tscharner, V., Stefanyshyn, D.J., Goepfert, B., Hintermann, B., 2007. Gait analysis in ankle osteoarthritis and total ankle replacement. Clin. Biomech. 22, 894–904.

A wearable system for multi-segment foot kinetics measurement.

This study aims to design a wearable system for kinetics measurement of multi-segment foot joints in long-distance walking and to investigate its suit...
934KB Sizes 3 Downloads 3 Views