Symmetry and Reproducibility of Kinematic Parameters during Various Running Techniques

Abstract

KARAMANIDIS, K., A. ARAMPATZIS, and G.-P. BRÜGGEMANN. Symmetry and Reproducibility of Kinematic Parameters during Various Running Techniques. Med. Sci. Sports Exerc., Vol. 35, No. 6, pp. 1009–1016, 2003.

Purpose 

This study examined the validity of the assumption of lower-extremity kinematic parameter reproducibility and symmetry during running with different velocities and stride frequencies.

Methods 

Each of 12 female long-distance runners ran on a treadmill in combinations of three different velocities (2.5, 3.0, and 3.5 m·s⁻¹) and three different stride frequencies (preferred and ± 10% from preferred). The left and right sides of the athletes were filmed using video cameras placed orthogonally to the sagittal plane. A total number of three step cycles for each running condition were recorded (250 Hz). For each side of the body, 19 sagittal kinematic parameters from the lower extremity were evaluated.

Results 

Intraclass correlation coefficients (ICC) for both legs were high (generally > 0.80). Only the angular velocity parameters demonstrated correlation values below 0.70. The symmetry index for the linear and angular displacement parameters and the contact times during all running techniques were less than 8%, whereas those for the angular velocity parameters and flight times were higher than 15%.

Conclusion 

The present results suggested that the degree of reproducibility and symmetry of kinematic data do not vary with a deliberate change in running technique but rather depend on the parameter itself. With respect to the economy of data analysis, the present findings indicate that recording a single monolateral trial would provide reproducible and symmetric values for most kinematic parameters.

A large assortment of kinematic parameters have been examined for the purpose of analyzing characteristic human running (^3,27). These parameters may be temporal, linear, and angular displacement as well as linear and angular velocity parameters. In several studies, it has been reported that sagittal plane kinematic parameters may be influenced by varying the running velocity (^1,9,14) or stride frequency (^5,6,10). Although the purposes of the studies have been varied, most of them assumed that kinematic parameters are reproducible for running techniques associated with changes in running velocity or stride frequency. A review of relevant literature reveals that the validity of this assumption has not been directly addressed. Changes in the running velocity or stride frequency may be associated with parameter variability. Recently it was suggested that the variability of ground reaction force during walking is speed dependent (¹⁸) and that changes within a gait task could influence the reproducibility of kinematic parameters (²³). It is reasonable to assume that the reproducibility may vary with activity (⁷). Therefore, it is possible that the variation of running velocity or stride frequency might influence the reproducibility of some of the parameters within a single exercise and that reported differences may also result from a performance variability and not only from the experimental manipulation. The direct advantage of assuming reproducibility is a reduced data collection and analysis. Therefore, analyzing the reproducibility of kinematic parameters during various running techniques is of great interest with regard to the economy of data acquisition and treatment (^8,23).

Several studies have analyzed the symmetry between homologous body segments with respect to forces and movements developed during normal human running, and the reported degree of symmetry was controversial (^{2,4,13,15,21,24,25}). This is possibly due to differences in the definition of symmetry, to the applied methodology, and to the parameter itself. In studies examining the symmetry of ground reaction forces (^2,15), the symmetry index (SI) proposed by Robinson et al. (²²) has been widely used. However, only one prior investigation (¹⁶) has used the symmetry index to evaluate kinematic gait patterns. As it is of practical and clinical interest to quantify the kinematic symmetry that occurs in normal human running (¹⁵), the symmetry index may provide a useful tool for comparing kinematic symmetry across various experimental groups.

The kinematic analysis of running with deliberate change in velocity or stride frequency has been a subject of numerous investigations (^{1,5,6,9,10,14}). Most studies assumed kinematic symmetry between homologous body segments and simplified the experimental procedures by collecting and analyzing data from only one side of the body. The belief was that the contralateral limb behaves in a similar manner. However, the validity of this assumption was not studied.

The aims of this study, therefore, were to examine the reproducibility and the symmetry of a wide number of parameters obtained from the kinematic analysis of running at different velocities and deliberate changes in stride frequencies.

METHODS

Twelve female long-distance runners (regional level, age: 25 ± 3 yr, height: 167 ± 7 cm, weight: 58.58 ± 6.01 kg) participated in this study. Written consent was obtained from all subjects. The study was conducted according to the declaration of Berlin and was approved by the ethical committee of the University of Cologne. All chosen test subjects had no history of neuromuscular or muscular-skeletal impairments that might affect their running patterns.

The athletes ran on a treadmill in their own running shoes at three different running velocities (2.5, 3.0, and 3.5 m·s⁻¹) and three stride frequencies (preferred and ± 10% from preferred). Relatively low speeds were chosen such that subjects could maintain ± 10% of stride frequency for at least 5 min. On separate days before the main experiment, the participants had the chance to practice for accommodation to the treadmill. When the subjects were fully accommodated to the treadmill (subjective criteria), the preferred stride frequency was analyzed over a period of 30 s. The stride duration was defined as the time between two consecutive footfalls of the same leg. Ten percent slower and 10% faster stride frequencies were calculated for each velocity and practiced on the same day as the treadmill accommodation.

Athletes who had difficulty running at ± 10% of stride frequency for a test speed had a second day of practice. The appropriate stride frequency, including the preferred stride frequency, was indicated by an audio metronome. The order of velocities was 2.5 m·s⁻¹, 3.0 m·s⁻¹, and 3.5 m·s⁻¹. At each of these velocities, the athletes started with the preferred stride frequency. The 10% slower and 10% faster stride frequencies were randomized. The left and right sides of the athletes were recorded using two genlocked, high-speed video cameras (250 Hz, Redlake MotionScope 250 C (San Diego, CA)) placed orthogonally to the sagittal plane. To improve the quality of the video analysis, reflective markers (radius 10 mm) were used to identify the joints. All athletes wore tight-fitting shorts. Marker position determination and placement are described in Table 1. To ensure precise marker placement between right and left sides of the body, marker location was checked by visual inspection. Additionally segment lengths were computed at touch down during 2.5 m·s⁻¹ and preferred stride frequency (step cycle 1) to check the degree of symmetry in marker placement.

TABLE 1: Marker set: position determination and placement.

Left and right cameras were calibrated using a square frame (length 1 m, height 1 m) placed in the middle of the plane of motion of the corresponding leg. Light-emitting diodes (LED) were used to synchronize the video sequences. A whole running cycle, from heel strike to ipsilateral heel strike, was analyzed. The left and right running cycles always corresponded to consecutive footfalls. For each running condition, three step cycles were recorded over a period of 5 min: cycle 1 (between 0.5 and 1.5 min), cycle 2 (between 2.5 and 3.5 min), and cycle 3 (between 4 and 5 min). Fatigue effects were assumed not be present, as the running velocities were relatively slow for trained long-distance runners. Furthermore, the athletes had a 2–3 min of rest between the different test conditions.

The video sequences were digitized using the automatic tracking feature of the Peak-Motus analyzing software (Motus Workplace, Peak Performance Technologies, Inc., Denver, CO). The two-dimensional coordinates of the digitized markers were smoothed using a fourth-order low-pass Butterworth filter with a cut-off frequency of 15 Hz (²⁸). For the right leg, the instants of touch down and toe-off were determined using customized pressure measuring insoles (Paromed–System, Paromed Medizintechnik GmbH, Neubeuern, Germany, 1000 Hz) to achieve a higher accuracy for the events. Due to technical limitations, the events from the left leg were determined from the video sequences (250 Hz). The touch-down angle corresponded to the angle at the first frame after ground contact, and the toe-off angle corresponded to the first frame after toe-off. Nineteen parameters per body side were calculated and analyzed. The parameters chosen are typical sagittal plane parameters used to analyze running gait (^3,27). The definitions for the angles of the ankle (α), knee (β), and hip (γ) are illustrated in Figure 1.

FIGURE 1: Directions of joint angles at ankle (α), knee (β), and hip (γ).

Reproducibility test.

For each velocity, frequency, and body side, the analyzed cycles were divided into three groups: cycle 1, cycle 2, and cycle 3. The reproducibility of the measured parameters was analyzed using the statistical methods described below. Differences among groups were checked using a nonparametrical test for several dependent samples (Friedman test). For parameters showing differences, a nonparametrical test for two dependent samples (Wilcoxon test) was applied to determine where these differences occurred. Nonparametrical tests were chosen as the sample was relatively small and a normal distribution not assured. The level of significance was set at α < 0.05 (¹⁷). Intraclass correlation coefficients (ICC) were calculated to examine the relationship between groups (²).

Symmetry test.

Robinson et al. (²²) proposed the following equation for analyzing the symmetry in running:EQUATION

where X_r is the parameter recorded from the right leg and X_l is the corresponding parameter from the left leg. Although this equation might be beneficial when examining the actual bilateral dominance trends that characterize the subject’s sample, it might not demonstrate the real symmetry of the respective individuals, e.g., if half of the subjects have a symmetry index of −10% and the other half of + 10%, the mean index value for the group would be 0%. Therefore, the equation proposed by Robinson et al. (²²) was modified and then called the absolute symmetry index (ASI):EQUATION

A value of zero for the ASI indicates the variables X_r and X_l are identical and, therefore, that there is perfect gait symmetry as measured by that particular gait parameter. In addition, the presence of differences between the mean values of the right and left leg was checked using a nonparametrical test for two dependent samples (Wilcoxon test). The level of significance was set at α < 0.05 (¹⁷).

RESULTS

Reproducibility of kinematic parameters.

Figure 2 shows the angles of the ankle, knee, and hip at 3.0 m·s⁻¹ (preferred stride frequency) for the three measured cycles of the right and the left leg. For the preferred frequency, the kinematic and temporal parameters obtained in this study (Tables 2–4) agree with previously published results (^1,21). Table 5 contains the ICC for all parameters from both body sides during all velocities and stride frequencies. Generally, the left and right body side exhibited similar ICC for all parameters. Only the vertical displacement parameter at the hip showed different ICC values between body sides (generally ICC left < 0.80 and ICC right > 0.80). The ICC values of the temporal, linear, and angular displacement parameters for all running techniques and both legs were high (> 0.80). The lowest ICC values appeared for velocity parameters, especially at the hip and knee (generally < 0.75). Only the minimum ankle angle velocity at ground contact showed ICC values above 0.80. For the majority of parameters, there were no significant differences between step cycles. When significant (P < 0.05) differences between running step cycles were detected, these appeared predominantly between step cycles 1 and 2 or step cycles 1 and 3 but seldom between step cycles 2 and 3 (Tables 2–4). This is especially true at the instant of touch down and toe-off, for the angular and linear displacement parameters and the contact times of the right leg (Tables 2 and 4). There was no significant difference in the stride frequencies between the three analyzed step cycles, and the ICC values were high (Table 5).

FIGURE 2: Angles of the ankle, knee, and hip at 3.0 m·s⁻¹ (preferred stride frequency) for the three measured cycles (1, 2, and 3) of the right and the left leg (N = 12). The x-axis was normalized as follows: −200% to −100%: ground contact of the right leg; −100% to 0%: flight-phase; 0–100%: ground contact of the left leg; 100–200%: flight-phase.

TABLE 2: Analysis of various kinematic parameters from the right leg during different stride frequencies at 2.5 m·s⁻¹ for three step cycles; [mean (SD), N = 12].

TABLE 3: Analysis of various kinematic parameters from the left leg during different stride frequencies at 3.0 m·s⁻¹ for three step cycles; [mean (SD), N = 12].

TABLE 4: Analysis of various kinematic parameters from the right leg during different stride frequencies at 3.5 m·s⁻¹ for three step cycles; [mean (SD), N = 12].

TABLE 5: Intraclass correlation coefficients for all analyzed kinematic parameters for three velocities with three different stride frequencies from the left (L) and right (R) leg (N = 12).

Symmetry of kinematic parameters.

The ASI for the computed segment lengths were 4.47% for the foot segment, 0.44% for the shank segment, 2.13% for the thigh segment, and 2.99% for the trunk segment. Table 6 contains the mean ASI and SD of all analyzed parameters for three velocities and all stride frequencies. The lowest ASI was found for the linear displacement parameters at the hip (ASI generally < 5%), the angular displacement parameters at the knee (< 6%) and ankle (< 10%), and the contact times (< 8%). Only the linear and angular velocity parameters (for all joints) and the flight times showed values generally above 15%. Between the average values of the left and right body side, significant (P < 0.05) differences were detected for the angular displacement and temporal parameters, but no differences were found for the linear displacement parameters at the hip (Tables 2–4).

TABLE 6: The mean absolute symmetry index for all analyzed kinematic parameters during different stride frequencies at three velocities for step cycle 3 [mean (SD), N = 12].

DISCUSSION

Reproducibility of kinematic parameters.

Several authors have suggested that the type of activity or degrees of automation of the studied task are factors that might noticeably influence the reproducibility of parameters (^7,11,18,23). Systematic variation in stride frequency during different running velocities was certainly not a typical running task. The present study revealed that the reproducibility of kinematic parameters was high during all running techniques for both legs. Most of the data exhibit ICC values above 0.80 throughout the selected stride frequencies and velocities. Concerning the economy of data analysis, these findings indicate that a single trial would provide reproducible kinematic parameters.

In general, the linear and angular velocity parameters provided the lowest ICC values (most of data < 0.80). Velocity parameters probably contained bigger errors than position parameters because they are derived. The differentiation values may exaggerate higher frequency components such as noise (^20,23). In addition, parallax errors may have influenced the variation in hip vertical velocity. It is suggested that more than one trial is required if velocity parameters are of interest.

Several of the applied kinematic parameters have already been suggested to be reproducible during normal human running (^8,12,19). However, whether or not the reproducibility altered when running at various velocities or stride frequency has not been demonstrated previously. This study showed that most kinematic parameters were reproducible (ICC > 0.80) for all running velocities and stride frequencies. Significant (P < 0.05) differences for both legs detected between running step cycles were predominantly found between step cycles 1 and 2 or step cycles 1 and 3 but seldom between step cycles 2 and 3. It is suggested that data collection after 2–3 min of treadmill running could improve reproducibility. However, it cannot be excluded that the stride-to-stride variability within each of the three studied time intervals may play a role in observed differences. Therefore, further study is needed to investigate this possibility.

Differences in the degree of reproducibility of several parameters between the left and right side of the body were apparent. Generally, the ICC values of the vertical position of the hip at touch down and toe-off and the contact times of the right body side were higher than those for the left side. Significant (P < 0.05) differences between step cycles were regularly found for the parameters at touch down and toe-off and for the contact times for the right leg. This could be attributed to the methods used in assessing the left and right events. The visual determination might lead to errors ± 1 frame, whereas using pressure data allows identification of the precise frame. In addition, the pressure measuring insoles (1000 Hz) with a temporal resolution of 1 ms allowed a more reliable assessment of the events and the contact times of the right leg than those for the left leg (video camera 250 Hz). This, in turn, affected the calculated angles. This illustrated that the choice of parameters together with the type of equipment are important considerations to achieve acceptable levels of reproducibility.

Symmetry of kinematic parameters.

It was determined for all running techniques that the left-to-right symmetry differences for the linear and angular displacement parameters and the contact times were small (ASI for most data < 8%), whereas the angular and linear velocity parameters and the flight times were high (ASI > 15%). Therefore, the symmetry of the kinematic parameters obtained was not influenced by deliberate change in running technique but depends rather on the parameter itself.

In the present study, running symmetry was quantified by the absolute sum of the symmetry index. This equation might not be meaningful for examining the actual bilateral dominance trends that characterized the subject’s sample but might demonstrate the real symmetry of the individual, which was the actual purpose of the study. Although running represents a cyclic motor skill, most of the studies found in the literature used strides from different trials when analyzing symmetry (^2,13,15,16). The possibility of kinetic variability transfer between successive footfalls should not be excluded; using consecutive strides is surely a more valid method (²⁶).

The angular displacement and velocity as well as the temporal parameters often showed significant (P < 0.05) differences between homologous body segments. Vagenas and Hoshizaki (²⁴), who analyzed kinematic parameters of the rear foot motion, also found significant asymmetries during normal human running. They suggested that when different groups of runners are compared, average bilateral values should be used. Primary factors that could contribute to the differences in bilateral measurements were asymmetries involving anatomical body segment length, flexibility, or marker placement (^13,24,25). In the present study, we checked the degree of symmetry in marker placement by computing segment lengths at the same relative instants of the stride. The results revealed highly symmetric values for the computed segment lengths with an average ASI of 2.57%. However, whereas this method provides a crude insight into errors in relative proximal/distal placement of the markers, it would not identify errors in relative anteroposterior placement of the hip, knee, and ankle markers. In addition, different event identification methods for the left and right leg were employed, which might be a further factor for the differences in symmetry exhibited. However asymmetry in marker placement would only affect absolute angle values but not angular velocities.

The present findings revealed that the percentile differences, indicated by ASI, were generally smaller than 8% for the angular displacement parameters and the contact times for all running techniques. For the angular velocity parameters and the flight times, the ASI values were in general above 15%. It is suggested that each leg experienced different flight times and angular velocity parameters at the hip, knee, and ankle angle. The asymmetry could be partly attributed to the generally low reproducibility of flight times and velocity parameters.

Based upon the findings of the present study, the following conclusions can be drawn:

1) The reproducibility and symmetry of the kinematic data does not depend primarily on the running velocity or deliberate change in stride frequency but rather on the parameter itself.

2) The reproducibility of the kinematic data is generally high. Using data collected after 2–3 min in treadmill running reproducibility may be improved.

3) The left-right symmetry was lowest for the angular and linear velocity parameters, and highest for linear and angular displacement parameters. Those findings may imply that parameters assigned for analysis should be carefully chosen if only mono-lateral data are to be recorded.

4) With respect to the economy of data analysis, the present findings indicate that recording a single mono-lateral trial would provide reproducible and symmetric values for most kinematic para-meters.