Reproducibility of Loading Measurements With Skin-Mounted Accelerometers During Walking

Article Outline

• Abstract
• Methods
  • Participants
  • Acceleration Measurements
  • Gait Analysis
  • Data Reduction and Analysis
  • Statistical Analysis
• Results
  • Gait Characteristics
  • Repeatability of SMAs
  • Correlations Between Ground Reaction Force and SMA Parameters
  • Gait Symmetry
• Discussion
  • Study Limitations
• Conclusions
• Acknowledgment
• References
• Copyright

Abstract 

Liikavainio T, Bragge T, Hakkarainen M, Jurvelin JS, Karjalainen PA, Arokoski JP. Reproducibility of loading measurements with skin-mounted accelerometers during walking.

Objective

To examine reproducibility of load measurements with skin-mounted accelerometers (SMAs) during walking.

Design

Reliability study.

Setting

A motion analysis laboratory.

Participants

Ten healthy young men.

Interventions

Not applicable.

Main Outcome Measures

Two triaxial accelerometers were fixed to the subjects’ skin above and below the knee joint. The subjects walked barefoot at their preferred speed and at a constant speed (1.3m/s, ±5%) in a gait laboratory and along a corridor. The same protocol was repeated over 2 days. Initial peak acceleration (IPA), peak-to-peak (PP) acceleration, and maximal and average acceleration transient rates (ATRs) were calculated. The coefficient of variation (CV) and Pearson linear correlation coefficient were calculated to measure reproducibility of SMA load measurements.

Results

IPA and PP acceleration had good interday repeatability (CV <15%). The repeatability of average ATR and maximal ATR parameters was generally not acceptable. The loading variables obtained from ground reaction forces and SMA measurements during gait revealed high linear correlations, indicating that with SMA measurements it is possible to predict certain ground reaction force loading parameters.

Conclusions

SMAs are practical for use in clinical environments to collect acceleration data that may be used to estimate joint loads.

Key Words: Acceleration, Biomechanics, Gait, Reproducibility of results, Rehabilitation

ALTHOUGH GAIT LABORATORIES have become more common, there is a need to develop more simple methods for clinical gait measurements, for example, in order to estimate joint loading. Load measurements are believed to be clinically important because of claims that repetitive impulsive forces at heel strike during walking result in joint degeneration or low back pain.¹, ², ³ Excessive impact loading has also been linked to prosthetic joint loosening and to headaches.⁴

Skin-mounted accelerometers (SMAs) are noninvasive, small and inexpensive devices that appear to be well suited for use in gait analyses. SMAs fixed on the lumbar spine have been widely used in studies of gait kinematics.⁵, ⁶, ⁷, ⁸ Furthermore, there have been studies⁴, ⁹, ¹⁰, ¹¹, ¹², ¹³ of impulsive loading and shock absorption by the skeleton with an SMA attached to the lower limb during human locomotion. This method has also been used to measure heel pad¹⁴, ¹⁵ and footwear properties,¹², ¹⁶ as well as soft-tissue resonance.¹⁷, ¹⁸

Gait measurements must be accurate and reproducible if they are to obtain data that can be used for clinical or scientific purposes. Several authors have concluded that SMAs fixed on the lower limb provide a valid method for quantitating the magnitude of bone or joint acceleration,¹⁹, ²⁰, ²¹ provided that the sensor is properly preloaded (ie, fixed) and its mass is low. Moe-Nilssen²² and Henriksen et al²³ reported that the day-to-day reliability of the trunk accelerometer was good when gait kinematics and mean accelerations were measured during walking. However, reproducibility of the load measurements, that is, peak accelerations and loading rates, with SMAs attached to the lower limb has not been examined when the subjects are walking.

Various methods have been used to determine the magnitude and rate of impulsive loading from ground reaction force or accelerometer data during walking. Some researchers have used different formulas to measure the heel strike transient¹, ²⁴ or loading rate²⁵, ²⁶, ²⁷ from the vertical ground reaction force signal. Others have concentrated on initial peak accelerations (IPAs)¹⁰, ¹², ¹⁶ or the acceleration transient rate.¹⁶ There is no consensus in the literature about how to calculate impulsive loading using accelerometers, or even force platforms. Furthermore, the repeatability of previously used methods is unknown.

Our main purpose in this study was to investigate the intra- and interday repeatability of SMAs in measuring impulsive loading in level walking in combination with simultaneous ground reaction force measurements with force platforms. Gait trials were performed in a laboratory and along a corridor to find possible differences in “constructed” and “natural” circumstances. Our second purpose was to test different methods of calculating impulsive loading and their symmetry between limbs. We hypothesized that using SMAs would be a repeatable method for investigating impulsive loading during gait, and intraday repeatability would be better than interday variability. We also anticipated that certain vertical ground reaction force loading variables could be predicted with SMA measurements, that is, that they are linearly correlated.

Back to Article Outline

Methods 

Participants 

Ten healthy young men (mean age, 29±4.7y; height, 1.77±0.06m; body mass, 79.2±8.4kg) participated voluntarily in the study. They provided their written consent after receiving detailed information about the study design and methods and their rights as study participants. The Ethics Committee of the Kuopio University Hospital approved the study design. Subjects had no history of musculoskeletal trauma, surgery, or pain in the lower limbs.

Acceleration Measurements 

The same researcher (TL) attached 2 lightweight triaxial (range, ±10g; sensitivity, 100mV/g) Meac-x accelerometers,^a fixed onto an aluminum plate, tightly to the skin above and below the right knee joint with an adhesive bandage.^b The accelerometers were calibrated to give comparable outputs. The position of the lower sensor was in the medial surface of the proximal tibia at 20% of the distance between the medial malleolus and the medial knee joint space (tibial plateau). The upper sensor was fixed on the thigh at 20% of the distance from the lateral knee joint space to the trochanter major so that the medial border of the accelerometer was in the same vertical line as the lateral margin of the patella. The positive z axis (a_z, or axial acceleration) of the sensor was aligned downward parallel to the straight limb.

A portable biosignal Biomonitor ME6000 T16 data-acquisition unit^a (mass, 0.34kg) was fixed with a belt onto each subject’s back. The overall weight of the whole system was 1.6kg. Accelerometer data were collected with a sampling frequency of 2000Hz and saved on an exchangeable memory card in the corridor. The data were further transferred to Megawin software (version 2.3).^a During laboratory measurements the data were sampled on-line at 1000Hz and transferred telemetrically (WLAN) to the Megawin software.

Gait Analysis 

The gait analysis was done in our laboratory and in a corridor. Two force platforms^c were mounted in the middle of a walkway in the laboratory to measure the 3-dimensional ground reaction forces. Walking speed was measured with a pair of photocells placed 2.5m apart on either side of the force platforms. Data collection began when the subject passed the first photocell. With this method, we collected ground reaction forces and accelerometer data simultaneously during consecutive steps. The 10-m long walkway, covered with a thin rubber mat, permitted the measurement of 1 gait cycle with 2 consecutive steps on the force platforms, which meant that subjects could take 3 acceleration and braking steps. Ground reaction forces were collected and stored with AMTI software^c at 2000Hz. The measurement area in the long corridor was covered with the same rubber mat. The photocells were positioned 6m apart in the middle of the walkway. Two to 3 consecutive gait cycles were measured for every subject.

All subjects performed warm-up trials in both locations before the measurements to become familiar with the experiment’s procedure. During the measurements, the subjects walked barefoot 6 times at both their preferred speed and at a predetermined constant speed (1.3m/s, ±5%) along the walkway in the laboratory. The walking trials were then repeated 4 times at both speeds in the corridor. Subjects were instructed to walk naturally at a steady speed. Trials in which subjects did not walk at the required speed, or contacted the force platforms improperly, were rejected. The trial order was randomized. The test protocol was repeated over 2 days in order to calculate the repeatability of the SMA measurements.

Data Reduction and Analysis 

We omitted the first and last steps from the analysis; only the steps between the photocells were accepted for processing. All data were further analyzed with the Matlab-based software^d developed in the Department of Physics, University of Kuopio. Force parameters were expressed in proportion to a subject’s weight. Contact times were determined on the basis of vertical ground reaction force, with an adaptive threshold level of .001 of body weight (BW), and also indirectly from a_z acceleration based on the vertical ground reaction force and signal shape analysis of the axial acceleration. Stride frequency and length were calculated from a_z acceleration data as the known time period of gait cycle and gait speed. Local maximum and minimum forces and their timing were measured in both the vertical (F_z) (fig 1) and horizontal directions (F_x, F_y) (fig 2). Loading parameters were defined from the beginning of the contact phase. Peak forces (F_z1, ie, heel strike transient, and F_z2), as well as maximal (LR_max) and average loading rates (LR_ave), were measured from the vertical ground reaction force data. Average loading rates were calculated according to Stacoff et al²⁷:

where Δt_F is the time between beginning of contact and 0.8F_z2max.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 1. 

  Gait parameters defined from contact phase of vertical (F_z) ground reaction force (upper figure) and of a_z acceleration signals (lower). The initial parts of the signals are on an expanded scale on the right demonstrating the definitions of the rate of loading parameters. The ground reaction force of a left limb is demonstrated as a grey line in the upper left figure. Abbreviation: BW, body weight.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 2. 

  Gait parameters defined from contact phase of AP (F_x) and ML (F_y) ground reaction forces.

IPA and peak-to-peak (PP) acceleration, as well as maximal and average acceleration transient rates (ATRs), were determined from a_z and the resultant (a_r) directions (see fig 1). Average ATR was calculated according to Lafortune and Hennig.¹⁶ The gravitational acceleration (g) was reduced from all measured acceleration values of the axial and resultant directions. The IPA and PP accelerations were also measured in the resultant horizontal direction (a_rxy).

where Δt_a is the time between 0.2 IPA and 0.9 IPA. where a_y is mediolateral (ML) and a_x is anteroposterior (AP) acceleration, respectively.

Because of the SMA instrumentation on the right leg, the symmetry indices of the ground reaction force and kinematic parameters were determined using the absolute symmetry index (ASI) according to Giakas and Baltzopoulos²⁸ and White et al.²⁹ This equation (equation 6) is a modified version of the symmetry index equation used by Herzog et al.³⁰ The ground reaction force was broken down to its components so that we could calculate symmetry in the F_z, F_x, and F_y directions. The symmetry was considered acceptable if the ASI was less than 10%; perfect symmetry requires an ASI of 0%.²⁸, ²⁹

where X_r is the variable for the right limb and X_l is the variable for the left limb.

Statistical Analysis 

The mean and standard deviation (SD) were calculated from the measured parameters. Generally, there was no asymmetry between the values of the left and right limb and therefore only the results of the right side are presented here. The coefficient of variation (CV) was calculated with equation 7 to measure interday variability for the 10 subjects. The intraday variability was measured between single trials performed during the first day as the ratio of the SD divided by the mean and multiplied by 100 to yield a percentage. The repeatability was considered to be good if the CV was less than 15%.²² The Pearson correlation coefficient was used to quantify linear correlation between the ground reaction force and acceleration variables to see if it is possible to predict the loading parameters of ground reaction force with SMA measurements. We used the 2-tailed Student t test for dependent samples (paired) to compare walking in the laboratory and along the corridor. Results were considered significant at P less than .05. Data were analyzed with SPSS sofware^e for Windows.

where N is the number of subjects and the SD (between days 1 and 2) is where is the mean of parameter x from all trials of ith subject on day d_k.

Back to Article Outline

Results 

Gait Characteristics 

Preferred gait speed was 1.36±0.06m/s in the laboratory and 1.41±0.15m/s (P>.05) in the corridor. The basic kinematic gait parameters are presented in table 1. The stride length and stride frequency, but not the contact time, differed significantly at the constant speed in the laboratory compared with the respective values in the corridor. Stride frequency was slower in the laboratory during normal speed walking (P<.05). The measured acceleration variables in the a_z direction were significantly lower in the laboratory than in the corridor, even at constant gait speed (table 2). The finding was the same in the a_r direction (data not shown).

Table 1. Descriptive Gait Variables of All Subjects From Day 1

NOTE. Values are mean ± SD.

Abbreviations: GRF, ground reaction force; NA, not available.

⁎P<.05;

†P<.01 (laboratory compared with corridor at constant and normal speeds).

Table 2. Selected Vertical Ground Reaction Force and Acceleration Parameters in the Direction of the Lower Limb

NOTE. Values are mean ± SD for all subjects from multiple trials measured in the laboratory (Lab) and the corridor on day 1.

Abbreviations: a1z, acceleration in the direction of the lower limb below the knee; az2, acceleration in the direction of the lower limb above the knee; F_{z1 max}, first vertical peak force; F_{z2 max}, second vertical peak force; F_{z2 min}, local minimum after F_{z2 max}; F_{z3 max}, third vertical peak force.

⁎P<.05;

†P<.01;

‡P<.001 (laboratory compared with corridor at constant and normal speeds).

IPAs in the a_z direction were highest during normal speed walking along the corridor (above knee, 1.83±0.59g; below knee, 2.59±0.88g). These values were 25.1% and 21.1% higher than the corresponding values in the laboratory at normal speed. The mean IPA and maximal ATR values in the resultant (a_r) direction varied between 1.61 and 3.14g and 153 and 396g/s, respectively, depending on walking circumstance and SMA placement. The lowest values were measured above the knee in the laboratory at a constant speed similar to axial direction measurements. The horizontal resultant IPA and PP acceleration mean values were 1.73 and 2.65g and 1.48 and 2.45g, respectively. Average loading rate was 8.08±1.10 BW/s and 8.69±1.75 BW/s at constant and normal speeds, respectively. The maximal loading rates showed almost 10 times higher values (see table 2).

Repeatability of SMAs 

The kinematic gait parameters showed excellent intra- and interday repeatability when the subjects walked in the laboratory (table 3). The interday repeatability of the local maximum and minimum vertical ground reaction force parameters was also high (CV range, 1.8%−8.2%). The CVs of vertical maximal and average loading rates were 14.0% or less and 8.4% or less, respectively, indicating good interday repeatability. Intraday repeatability between single trials seemed to be in the same range as interday repeatability between the averaged trials. IPA and PP acceleration parameters in a_z and a_r, as well as a_rxy (a_rxy data not shown), exhibited good (CV <15%) interday repeatability. The CV of maximal ATR below the knee in the direction of the lower limb was 13.9% to 14.4% in the laboratory and 14.8% to 16.0% in the corridor. The corresponding values above the knee were estimated as poor (CV range, 17.2%−18.5%). The repeatability of average ATR in the a_z direction and ATR_a1r parameters showed generally unacceptable repeatability. The gait speed tended to have only a minor effect on CV values.

Table 3. CV Within a Day (intraday) and Between Days (interday) of Selected Vertical Ground Reaction Force and Acceleration Parameters at Constant and Normal Walking Speeds in the Laboratory and Corridor

NOTE. All values are percent.

Abbreviations: a1r, acceleration in resultant direction below the knee; a2r, acceleration in resultant direction above the knee; a1z, acceleration in the direction of the lower limb below the knee; az2, acceleration in the direction of the lower limb above the knee; CT acceleration, contact time from acceleration; F_{z1 max}, first vertical peak force; F_{z2 max}, second vertical peak force; F_{z2 min}, local minimum after F_{z2 max}; F_{z3 max}, third vertical peak force; NA, not available.

Repeatability was generally similar in the corridor and laboratory (see table 3). Average ATR in the a_z direction as well as ATR parameters in the resultant direction showed better interday repeatability in the corridor than in the laboratory for both gait speeds. The CV values of axial average ATR and ATR in resultant direction, however, were over 15% except for the resultant maximal ATR in corridor. The CV values seemed to be inferior in the corridor at constant speed.

Correlations Between Ground Reaction Force and SMA Parameters 

The linear correlation coefficients at normal speed between maximal loading rates and axial maximal ATR were .919 (P<.01) and .659 (P<.05) below and above the knee, respectively (fig 3). The corresponding values in the a_r direction were .882 (P<.01) and .711 (P<.05). At a constant walking speed, maximal loading rates also showed a significant linear correlation with axial and resultant maximal ATR (P range, <.01 to .001). Average loading rates correlated neither with average ATR nor with maximal loading rates in any circumstance; the maximal loading rate, however, was significantly related to average ATR (P range, <.05 to .01), except for the pair maximal loading rate versus average ATR_a2r at normal walking speed (r=.594, P=.07). F_z1 showed a linear correlation with axial and resultant IPA measured below and above the knee (P range, <.01 to .001) at both speeds. The best correlation coefficients between F_z1 and IPAs were found below the knee in the a_r direction (constant speed, r=.909; normal speed, r=.942). The pair maximal loading rate and IPA in the resultant (a_r) direction measured below knee had the very highest correlation. The IPA_1r accounted for 89.9% (R²) and 91.8% (both P<.001) of the variation of maximal loading rates at constant and normal speeds, respectively (see fig 3).

• 
  • View Large Image
  • Download to PowerPoint

• Fig 3. 

  Linear correlations between loading parameters of vertical (F_z) ground reaction force and a_z and a_r accelerations during walking with normal speed. (A) F_{z1 max} and IPA_a1z; (B) maximal loading rate (LR_max) and ATR_{max a1z}; (C) average loading rate (LR_ave) and ATR_{ave a1z}; (D) maximal loading rate and ATR_{ave a1z}; (E) maximal loading rate and IPA_a1z; and (F) LR_max and IPA_a1r. Acceleration was measured with the SMAs attached below the knee.

Gait Symmetry 

Vertical ground reaction force parameters (contact time, F_z2max, F_z2min, F_z3max) generally showed high symmetry (ASI <10%) at both gait velocities (ASI range, 0.2%−4.8%), but the ASI of the vertical loading parameters (ie, F_z1max) maximal loading rates and average loading rates were 15.1%, 26.5%, and 14.0% at constant speed walking, respectively (table 4). The corresponding values in normal speed were 15.1%, 27.8%, and 12.7%, indicating unacceptable symmetry. The ASI values were low in the AP direction (range, 0.7%−8.4%), except for the F_x1max parameter at normal speed walking (10.8%). The asymmetry increased in the ML direction. The ASI of F_y2max and its timing were 14.3% and 15.6% at constant walking speed, respectively. The corresponding values were 16.9% and 16.2% in normal speed. The F_y1min and F_y3max and their timing parameters showed good symmetry (range, 1.5%−8.8%), except for F_y3max at constant speed (ASI 10.2%).

Table 4. ASI Values of Selected Vertical Ground Reaction Force and Acceleration Parameters at Constant and Normal Walking Speeds From the Laboratory

NOTE: Values are percent. The acceptable symmetry requires ASI less than 10%. See Fig 1, Fig 2 for definitions of parameters.

Abbreviations: CT, GRF; contact time from ground reaction force; time, timing of ground reaction force parameter from the beginning of contact phase; x, AP direction; y, ML direction; z, vertical direction.

Back to Article Outline

Discussion 

In this study, we found that IPA and PP acceleration in the resultant axial direction as well as in the resultant horizontal direction had good interday repeatability (CV <15%). The maximal ATR and average ATR parameters, however, showed unacceptable interday repeatability because only half of the maximal ATR and 13% of the average ATR parameters showed acceptable reproducibility. To our knowledge, this is the first study that has evaluated the repeatability of loading measurement with SMAs.

Moe-Nilssen²² reported the test-retest reliability of trunk accelerometry during walking in 19 subjects who walked barefoot on an uneven mat and on a flat, wooden floor with the accelerometer fixed over the lumbar spine. The root mean square value of accelerations in 3 different directions and the resultant direction were calculated. The relative variabilities measured via the CV were 3.1% to 6.8% and 2.2% to 5.7% in the floor and mat, respectively. Henriksen et al²³ later conducted a similar study with 20 subjects. Their results showed excellent intraclass correlation coefficients (ICC range, .77–.96) values for mean acceleration, stride length, and rate, which is indicative of high reliability. We did not calculate ICC values, because several study variables had restricted range and therefore the correlation coefficients would have been difficult to interpret.

Recently, Kavanagh et al³¹ studied the inter- and intraexaminer reliability of acceleration data during walking that was taken from the head, neck, lower trunk, and shank. In their study, the reliability of the 3-dimensional accelerations was determined by using a waveform reliability statistic that considered the time-series evolution of the signal rather than the mean signal amplitude computed over several gait cycles. They showed that the SMAs at different locations are highly repeatable in inter- and intratester settings over a range of walking speeds. Because the placement of SMAs and data analysis was different, no direct comparison can be made with our results.

In our study, the accelerometer measurements indicated that the variability between single trials within 1 day were in the same range as the variability of averaged trials between days. This supports the belief that the interday variation in loading measurements of SMAs during walking is not caused primarily by subtle differences in the attachment of sensors on different days. The variation is possibly primarily caused by the sum effect of sensor placement, random and/or systematic changes in gait and the measured parameters. The CV of loading measurements below the knee on the tibial plateau was lower than the corresponding SMA located above the knee. The soft tissue beneath the SMA above the knee is obviously thicker. High impact loading can produce a minor vibration of the sensor, resulting in poor repeatability in measurement of certain loading parameters. In making clinical decisions, the higher variation means that the change in that parameter between 2 consecutive gait measurements should be highly enhanced so that one can draw conclusions about improvement in rehabilitation.

The kinematic and ground reaction force variables, derived from gait tests of healthy people at their preferred speed, are believed to provide excellent repeatability in test-retest settings within the same day or between days.³², ³³ The CV values of kinematic parameters derived from axial acceleration signal between days in our study were 1.0% to 2.9%, which is in agreement with those earlier results. The basic kinetic ground reaction force parameters (ie, F_z2max, F_z3max, F_z2min) also showed excellent CV values (range, 1.8%−3.0%). The ground reaction force parameters measuring impulsive loading demonstrated larger variability, both within day (range, 8.0%−19.4%) and between days (range, 7.1%−14.0%), highlighting the random variation between gait trials within a single day and the significance of the variables used to assess the interday repeatability of gait. It should be noted, however, that in this study inter- and intraday CV values are not directly comparable because the SMAs were not detached between separate trials in the same day and because of a slightly different methodology in averaging the trials.

The repeatability of any method is not worth investigating if the validity of measurements is poor. Several researchers¹⁹, ²⁰, ²¹ have used either cadavers or in vivo to examine the validity of SMAs in measuring bone loading by comparing concurrently the output of SMA to that of a bone-mounted accelerometer (BMA). These authors have concluded that SMA is a valid method if the sensor is properly fixed to the skin and its mass is low. On the contrary, Lafortune et al¹¹ observed that SMA measurements would cause signal distortion in some subjects during running. They also found, however, that frequency corrected SMA signals mimicked the signals registered with the BMAs. Lafortune stated that a frequency transformation of SMAs recordings could provide a sufficient estimate of the shock experienced by the tibia. Although bone-mounted sensors offer a more precise method for measuring shock waves that impact on bone, BMAs are not practical for use in clinical experiments.

We used the forceplate measurements as the reference method in our analysis of acceleration. The high linear correlations that we found between vertical ground reaction forces and SMA variables, both in the axial and resultant directions, confirm that it is possible to predict certain vertical ground reaction force parameters with SMA measurements. Especially, maximal ATR explained 43.4% to 85.3% and IPAs explained 49.1% to 91.8% of the variation of maximal loading rates. Also, IPAs predicted 58.8% to 88.7% of the variation of F_z1. Although SMAs cannot measure the real forces on joint surfaces, they could be practical for use in clinical gait analysis because they provide reliable estimates of joint loading in a noninvasive manner. In our study, the mean peak axial loading (IPA_a1z) was 1.87 to 2.59g, as measured just below the knee. The corresponding values in resultant horizontal direction (range, 1.74−2.65g) demonstrated that the vertical component of the transmitted stress wave is accompanied by transverse components of the same magnitude. Thus, the traveling wave is not purely compressive but contains significant shear components that may, in addition to vertical load, be harmful for a joint and an articular cartilage.³⁴

Accelerometers placed at the level of the knee joint could estimate more precisely actual joint loading than could a trunk accelerometer or ground reaction force measurements. There is no established method for parameterization of the vertical ground reaction force or acceleration curve for loading measurements, which makes it difficult to compare different studies. One purpose of this study was to test the different methods of calculating impulsive loading. We found that average-loading rates did not correlate with average ATR or maximal loading rate, but maximal loading rate was associated with average ATR. This could perhaps mean that average-loading rate is not a valid parameter for measuring impact loading because it does not take into account the possible heel strike transient and therefore it underestimates the true loading rate.

Greig et al³⁵ reported that in healthy elderly people walking on treadmill differed from walking in a corridor despite the fact that they were asked to maintain a standardized gait speed. The subjects in that study had lower heart rates and higher step rates in conjunction with a shorter step length when walking along the corridor. According to these authors, treadmill walking by younger adults was more representative than corridor walking, that is, more of a real-life situation. Our study showed that laboratory walking produced lower loading values as measured with SMAs, as well as slight differences in gait kinematics compared with walking along the corridor (ie, a natural situation). Gait speed or dissimilar walking surfaces did not explain the finding because both were kept constant. Instead, subjects may have been more familiar with corridor walking despite participating in several warm-up trials in both locations. The shorter walkway in the laboratory might also have caused slight differences in gait kinematics, possibly resulting in diminished loading when walking.

Asymmetry between the right and left legs in normal gait appears to be low in vertical and AP directions of ground reaction forces. The asymmetry, however, increases in the ML direction.²⁸, ²⁹, ³⁰ Our study confirms earlier findings that symmetry is generally high (ASI <10%) in the vertical and AP directions. The vertical loading parameters, however, had unacceptable ASI values. The asymmetry in ML direction (F_y) was less pronounced when compared with previously reported values. This might be explained by the more specifically determined parameters in the ML direction in recent experiments,²⁸, ³⁰ or the different age group of the subjects.²⁹ The conventional symmetry between limbs does indicate that having an SMA tightly fixed onto the right leg did not affect the natural gait of subjects.

Study Limitations 

We investigated the repeatability of SMAs during level walking with 10 healthy subjects. It is not possible to identify which of the confounding factors had the greatest effect on the CV. One source of systematic error in clinical studies is called “the learning effect,” because subjects are likely to improve their performance when repeating a task. This effect could be diminished, but not necessarily totally abolished, by giving proper instructions to the subjects and allowing them to adequately familiarize themselves with the test procedures.²², ³⁶ Stair walking is a daily activity in which impact loading can be more forceful.¹² Our results cannot be directly generalized to different activities or different gait speeds. Intertester repeatability of SMAs for loading measurements is an important factor in clinical gait analysis and will need to be investigated in the future.

Back to Article Outline

Conclusions 

This study demonstrated that IPA and PP acceleration exhibited good interday repeatability, but ATR parameters showed unacceptable repeatability. The high linear correlations between vertical ground reaction forces and acceleration in both axial and resultant directions are evidence of the possibility of predicting certain ground reaction force loading parameters with SMA measurements. Average loading rate is probably not a valid parameter for measuring impact loading because it does not take the possible heel strike transient into account and therefore underestimates the true loading rate. Walking in the laboratory produced lower loading values as measured with the SMAs, as well as slight differences in gait kinematics compared with the corresponding values estimated while walking along the corridor; this indicates that these 2 situations are not comparable. In conclusion, SMAs are practical for use in a clinical environment to provide acceleration data that may be used to estimate joint loads.

Suppliers

Back to Article Outline

Acknowledgment 

We thank Vesa Kiviniemi, PhL, for his valuable help in the statistical methods.

Back to Article Outline

References 

1. Radin EL, Whittle MW, Yang KH, et al. The heelstrike transient, its relationship with the angular velocity of the shank, and the effect of quadriceps paralysis. In:  Lantz SH,  King AI editor. Advances in bioengineering. New York: American Society of Mechanical Engineers; 1986;p. 121–123
   • View In Article
2. Voloshin A, Wosk J. An in vivo study of low back pain and shock absorption in the human locomotor system. J Biomech. 1982;15:21–27
   • View In Article
   • MEDLINE
   • CrossRef
3. Wosk J, Voloshin AS. Low back pain: conservative treatment with artificial shock absorbers. Arch Phys Med Rehabil. 1985;66:145–148
   • View In Article
   • MEDLINE
4. Light LH, McLellan GE, Klenerman L. Skeletal transients on heel strike in normal walking with different footwear. J Biomech. 1980;13:477–480
   • View In Article
   • MEDLINE
   • CrossRef
5. Auvinet B, Berrut G, Touzard C, et al. Reference data for normal subjects obtained with an accelerometric device. Gait Posture. 2002;16:124–134
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (347 KB)
   • CrossRef
6. Currie G, Rafferty D, Duncan G, Bell F, Evans AL. Measurement of gait by accelerometer and walkway: a comparison study. Med Biol Eng Comput. 1992;30:669–670
   • View In Article
   • MEDLINE
   • CrossRef
7. Mayagoitia RE, Nene AV, Veltink PH. Accelerometer and rate gyroscope measurement of kinematics: an inexpensive alternative to optical motion analysis systems. J Biomech. 2002;35:537–542
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (211 KB)
   • CrossRef
8. Moe-Nilssen R, Helbostad JL. Estimation of gait cycle characteristics by trunk accelerometry. J Biomech. 2004;37:121–126
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (234 KB)
   • CrossRef
9. Folman Y, Wosk J, Voloshin A, Liberty S. Cyclic impacts on heel strike: a possible biomechanical factor in etiology of degenerative disease of the human locomotor system. Arch Orthop Trauma Surg. 1986;104:363–365
   • View In Article
   • MEDLINE
   • CrossRef
10. Gill HS, O’Connor JJ. Heelstrike and the pathomechanics of osteoarthrosis: a pilot gait study. J Biomech. 2003;36:1625–1631
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (601 KB)
    • CrossRef
11. Lafortune MA, Henning E, Valiant GA. Tibial shock measured with bone and skin mounted transducers. J Biomech. 1995;28:989–993
    • View In Article
    • Abstract
    • Full-Text PDF (407 KB)
    • CrossRef
12. Loy DJ, Voloshin AS. Biomechanics of stair walking and jumping. J Sports Sci. 1991;9:137–149
    • View In Article
    • MEDLINE
    • CrossRef
13. Wosk J, Voloshin A. Wave attenuation in skeletons of young healthy persons. J Biomech. 1981;14:261–267
    • View In Article
    • MEDLINE
    • CrossRef
14. Kinoshita H, Ogawa T, Kuzuhara K, Ikuta K. In vivo examination of the dynamic properties of the human heel pad. Int J Sports Med. 1993;14:312–319
    • View In Article
    • MEDLINE
    • CrossRef
15. Kinoshita H, Francis PR, Murase T, Kawai S, Ogawa T. The mechanical properties of the heel pad in elderly adults. Eur J Appl Physiol Occup Physiol. 1996;73:404–409
    • View In Article
    • MEDLINE
    • CrossRef
16. Lafortune MA, Hennig EM. Cushioning properties of footwear during walking: accelerometer and force platform measurement. Clin Biomech. 1992;7:181–184
    • View In Article
17. Boyer KA, Nigg BM. Muscle activity in the leg is tuned in response to impact force characteristics. J Biomech. 2004;37:1583–1588
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (341 KB)
    • CrossRef
18. Wakeling JM, Liphardt AM, Nigg BM. Muscle activity reduces soft-tissue resonance at heel-strike during walking. J Biomech. 2003;36:1761–1769
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (373 KB)
    • CrossRef
19. Nokes L, Fairclough JA, Mintowt-Czyz WJ, Mackie I, Williams J. Vibration analysis of human tibia: the effect of soft tissue on the output from skin-mounted accelerometers. J Biomed Eng. 1984;6:223–226
    • View In Article
    • MEDLINE
    • CrossRef
20. Saha S, Lakes RS. The effect of soft tissue on wave-propagation and vibration tests for determining the in vivo properties of bone. J Biomech. 1977;10:393–401
    • View In Article
    • MEDLINE
    • CrossRef
21. Ziegert JC, Lewis JL. The effect of soft tissue on measurements of vibrational bone motion by skin-mounted accelerometers. J Biomech Eng. 1979;101:218–220
    • View In Article
    • CrossRef
22. Moe-Nilssen R. Test-retest reliability of trunk accelerometry during standing and walking. Arch Phys Med Rehabil. 1998;79:1377–1385
    • View In Article
    • Abstract
    • Full-Text PDF (800 KB)
    • CrossRef
23. Henriksen M, Lund H, Moe-Nilssen R, Bliddal H, Danneskiod-Samsoe B. Test-retest reliability of trunk accelerometric gait analysis. Gait Posture. 2004;19:288–297
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (208 KB)
    • CrossRef
24. Gillespie KA, Dickey JP. Determination of the effectiveness of materials in attenuating high frequency shock during gait using filterbank analysis. Clin Biomech (Bristol, Avon). 2003;18:50–59
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (210 KB)
    • CrossRef
25. Chen WL, O’Connor JJ, Radin EL. A comparison of the gaits of Chinese and Caucasian women with particular reference to their heelstrike transients. Clin Biomech (Bristol, Avon). 2003;18:207–213
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (114 KB)
    • CrossRef
26. McCrory JL, White SC, Lifeso RM. Vertical ground reaction forces: objective measures of gait following hip arthroplasty. Gait Posture. 2001;14:104–109
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (78 KB)
    • CrossRef
27. Stacoff A, Diezi C, Luder G, Stussi E, Kramers-de Quervain IA. Ground reaction forces on stairs: effects of stair inclination and age. Gait Posture. 2005;21:24–38
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (225 KB)
    • CrossRef
28. Giakas G, Baltzopoulos V. Time and frequency domain analysis of ground reaction forces during walking: an investigation of symmetry and variability. Gait Posture. 1997;5:189–197
    • View In Article
    • Abstract
    • Full-Text PDF (1502 KB)
    • CrossRef
29. White R, Agouris I, Selbie RD, Kirkpatrick M. The variability of force platform data in normal and cerebral palsy gait. Clin Biomech (Bristol, Avon). 1999;14:185–192
    • View In Article
    • Abstract
    • Full-Text PDF (3302 KB)
    • CrossRef
30. Herzog W, Nigg BM, Read LJ, Olsson E. Asymmetries in ground reaction force patterns in normal human gait. Med Sci Sports Exerc. 1989;21:110–114
    • View In Article
    • MEDLINE
    • CrossRef
31. Kavanagh JJ, Morrison S, James DA, Barrett R. Reliability of segmental accelerations measured using a new wireless gait analysis system. J Biomech. 2006;39:2863–2872
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (322 KB)
    • CrossRef
32. Kadaba MP, Ramakrishnan HK, Wootten ME, Gainey J, Gorton G, Cochran GV. Repeatability of kinematic, kinetic, and electromyographic data in normal adult gait. J Orthop Res. 1989;7:849–860
    • View In Article
    • MEDLINE
    • CrossRef
33. Winter DA. Kinematic and patterns in human gait: variability and compensating effects. Hum Mov Sci. 1984;3:51–56
    • View In Article
    • CrossRef
34. O’Connor JJ, Johnston I. Transmission of rapidly applied load through articular cartilage: the mechanics of osteoarthrosis. In:  Turner-Smith AR editors. Micromovement in orthopaedics. Oxford: Clarendon Pr; 1993;p. 244–268
    • View In Article
35. Greig C, Butler F, Skelton D, Mahmud S, Young A. Treadmill walking in old age may not reproduce the real life situation. J Am Geriatr Soc. 1993;41:15–18
    • View In Article
    • MEDLINE
36. Schmidt RA. Motor control and learning: a behavioral emphasis. 2nd ed. Champaign: Human Kinetics; 1997;
    • View In Article

• a Mega Electronics Ltd, Microkatu 1, Kuopio, 70211, Finland.
• b Acrylastic; BSN Medical SAS, Rue Du Millénaire, Vibraye, 72320, France.
• c Model OR6-7MA; Advanced Mechanical Technology Inc, 176 Waltham St, Watertown, MA 02472.
• d Version 7.0.4; The MathWorks Inc, 2 Apple Hill Dr, Natick, MA 01760.
• e Version 11.5; SPSS Inc, 233 S Wacker Dr, 11th Fl, Chicago, IL 60606.