- Email: [email protected]
A comparison of the ground reaction force frequency content during rearfoot and non-rearfoot running patterns
Accepted Manuscript Title: A comparison of the ground reaction force frequency content during rearfoot and non-rearfoot running patterns Authors: Allison H. Gruber, W. Brent Edwards, Joseph Hamill, Timothy R. Derrick, Katherine A. Boyer PII: DOI: Reference:
S0966-6362(17)30170-4 http://dx.doi.org/doi:10.1016/j.gaitpost.2017.04.037 GAIPOS 5410
To appear in:
Gait & Posture
Received date: Revised date: Accepted date:
10-10-2016 26-4-2017 27-4-2017
Please cite this article as: Gruber Allison H, Edwards W Brent, Hamill Joseph, Derrick Timothy R, Boyer Katherine A.A comparison of the ground reaction force frequency content during rearfoot and non-rearfoot running patterns.Gait and Posture http://dx.doi.org/10.1016/j.gaitpost.2017.04.037 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Original Article – Full Paper A comparison of the ground reaction force frequency content during rearfoot and nonrearfoot running patterns *Allison H. Grubera,b, W. Brent Edwardsc, Joseph Hamillb, Timothy R. Derrickd, Katherine A. Boyerb a Department of Kinesiology, Indiana University, SPH Building, 112, 1025 E. Seventh ST, Bloomington, IN, 47405-7109, United States b
c
Department of Kinesiology, University of Massachusetts, 110 Totman Building, 30 Eastman Lane, Amherst, MA, 01003-9258, United States
Human Performance Laboratory, University of Calgary, KNB 418, 2500 University Drive NW, Calgary, AB, T2N 1N4, Canada
d
Department of Kinesiology, Iowa State University, 249 Forker, 534 Wallace RD, Ames, IA, 50011-3191, United States
*Corresponding author: Allison H. Gruber Affiliation Address: Department of Kinesiology University of Massachusetts 110 Totman Building 30 Eastman Lane Amherst, MA 01003-9258 Permanent Address: Department of Kinesiology Indiana University SPH Building 112 1025 E. Seventh ST Bloomington, IN 47405-7109 Phone: +1(812) 856-2447 Fax: +1(812) 855-3193 Email: [email protected]
(Title) A comparison of the ground reaction force frequency content during rearfoot and nonrearfoot running patterns Highlights:
The presence of an impact transient during non-rearfoot running is proposed. Rearfoot and non-rearfoot running generate frequencies associated with an impact transient. Traditional methods for analyzing impact loading rate and magnitude may be inappropriate.
Abstract Running with a non-rearfoot pattern has been claimed to reduce injury risk because the impact peak in the vertical ground reaction force (GRF) is visually absent in the time-domain compared with a rearfoot pattern. However, running results in a rapid deceleration of the lower extremity segments immediately following initial contact with the ground, regardless of footfall pattern. Therefore, the frequency content of the GRF is expected to contain evidence of this collision. The purpose of the present study was to characterize the waveform components of the GRF generated during the impact phase by habitual rearfoot and habitual non-rearfoot pattern groups using the continuous wavelet transform. Twenty rearfoot and 20 non-rearfoot participants ran over-ground at a standardized speed with their habitual footfall pattern. The continuous wavelet transform was performed on the resultant GRF vector and the vertical GRF. GRF signals generated by the non-rearfoot pattern group during early stance had maximum signal power of 15.4±9.1 Hz occurring at 23.1±6.3% of stance, which is within the 10-20 Hz range previously associated with impact in rearfoot runners. Maximum signal power occurred earlier in the impact phase (11.5±1.5%) and with a higher frequency (27.2±3.9 Hz) in the rearfoot pattern group verses the non-rearfoot pattern group (P<0.05). While the impact force transient may not appear as a prominent feature within the time-domain GRF with a non-rearfoot pattern, the results indicate that both footfall patterns generate frequencies associated with the impact peak in the vertical GRF.
2
Keywords: foot strike pattern; ; ; ; , ground reaction force, running biomechanics, wavelet transform, impact peak
3
1.
Introduction Non-rearfoot running footfall patterns (NRFP, i.e. midfoot or forefoot) have been popularized
because of claims that they may reduce the risk of developing running related injuries verses running with a rearfoot footfall pattern (RFP) (e.g. [1, 2]). Given that characteristics of the impact peak within the vertical ground reaction force (GRF) component were related to running injury in some studies (e.g. [3-5]), it has been suggested that a NRFP will reduce the risk of injury because the vertical impact peak is visually blunted or absent in the time-domain with these patterns [6]. Regardless of footfall pattern, there is a rapid deceleration of the lower extremity segments immediately following initial contact with the ground [7]. Although a distinctive vertical impact peak may not be visible in the time-domain during NRFP running [8], the frequency content of the GRF likely provides evidence of this collision. An analysis of vertical tibial shock indicated that frequencies representative of a significant vertical impact (i.e. 10 – 20 Hz [7]) are present in both RFP and NRFP running [9]. Time-domain analyses have revealed evidence of an impact peak in the anteroposterior and mediolateral GRF during NRFP running [10-12]. However, time-domain analyses of the vertical GRF component may not be sufficient to identify a vertical impact force during NRFP running. The time-domain GRF profile generated during RFP running provides information about the acceleration of the whole-body center of mass (COM), which typically illustrates two distinct peaks in the vertical direction [13]. The first peak is associated with the impact that occurs as a result of the rapid deceleration of the lower extremity segments at initial contact. The second peak is associated with the active motion of the rest of the body when the COM reaches its minimum vertical position during the stance phase [7]. The timing and magnitude of the vertical impact peak depends on parameters such as running speed, stride length, and segment position at initial ground contact [14-16]. Footfall pattern has also been shown to alter vertical loading rate and impact peak magnitude [6, 10, 11]. The impact force may not appear visually as a prominent feature within the time-domain GRF
4
signal generated during NRFP running because the timing of the maximum lower extremity acceleration may be shifted relative to initial ground contact compared with RFP running. Therefore, the calculation of impact loading variables in the time-domain without the visual presence of an impact peak may be inappropriate. The frequencies present in the GRF and tibial acceleration signals during walking and running have typically been analyzed using the Fourier Transform (FT) (e.g. [17, 18]). The FT is well understood, commonly used, and relatively easy to perform with many available software packages. Although the FT can indicate the frequencies and amplitude of each frequency that exist within a signal, its primary disadvantage is the loss of the time resolution of a signal’s frequency components. For example, the FT of the vertical GRF during running will result in lower (4 – 8 Hz) and higher (10 – 20 Hz) frequency bands that represent the active and impact forces, respectively [16, 19]. However, the time course of these frequency components and when the maximum signal power occurs during running is lost. The continuous wavelet transform maintains both the time and frequency resolutions and is used to analyze the similarity of an original signal with different scales and time-shifted versions of a basis function. The wavelet transform has been used to analyze biological signals such as muscle activation [20], tissue vibration [21], and the GRF in walking of healthy individuals [22, 23] and clinical populations [24]. Examining the GRF signal in time-frequency space with a wavelet transform may provide a better understanding of the differences in loading conditions between RFP runners and NRFP runners. The purpose of the present study was to characterize the waveform components of the GRF generated during impact by habitual RFP runners and habitual NRFP runners using the continuous wavelet transform. We hypothesized that the resultant and vertical component of the GRF generated by NRFP runners will include frequency components representative of the impact force that occurs during the impact phase of stance. The vertical impact peak generated by NRFP runners may not be visible in the time-domain because there may be a smaller time delay between the peak accelerations of the lower
5
extremity verses the rest of the body mass and is thus obscured by the active force. Therefore, we also hypothesize that maximum signal power of impact within the resultant and vertical GRF generated by NRFP runners will occur later in stance compared with RFP runners. 2.
Methods
2.1.
Participants Twenty habitual RFP runners and 20 habitual NRFP runners participated in this study (Table 1).
Participants were recreational and/or competitive middle to long distance runners who were free of lower extremity or back injury in the past year and ran a minimum of 16 km/week. Participants were excluded if they had any neurological or cardiovascular pathology. The study was approved by the university Institutional Review Board. All runners read and signed an informed consent document prior to participation. [Insert Table 1 here] Habitual footfall pattern was determined from the data collected during five over-ground running trials performed at each participant’s preferred speed. These trials were collected separately from those used in the wavelet analysis. Footfall pattern was classified using the strike index [12] and sagittal plane foot segment angle at initial ground contact [25]. For the purpose of this study, participants were included in the NRFP group if at least three of the five trials resulted in a strike index greater than 34% and an initial sagittal plane foot segment angle of less than eight degrees of dorsiflexion. The NRFP group consisted of 15 midfoot runners and five forefoot runners classified by these methods. All other participants were placed in the RFP group. 2.2.
Protocol
6
Participants wore fitted shorts and neutral racing flats (RC 550, New Balance, Brighton, MA, USA) provided by the laboratory for all procedures. Three-dimensional motion, center of pressure, and GRFs were recorded for the footfall pattern classification trials and the experimental trials. Calibration and tracking retro-reflective markers were placed on the pelvis and right lower limb based on an adapted Cleveland Clinic Model [26]. Position data of the markers during the static calibration trial and the running trials were recorded by an eight-camera Qualisys Oqus 3-Series optical motion capture system sampling at 240 Hz (Qualisys, Inc., Gothenberg, Sweden). Raw GRF data were recorded by a strain-gauge force platform sampling at 1200 Hz (OR6-5, AMTI, Watertown, MA, USA) that was flush with the floor and in the center of a 30 m runway. Participants were asked to perform 10 over-ground running trials at speed of 3.5 m/s ± 5% while making a right foot contact on the force platform without adjusting their speed or stride. Running speed was monitored by timing gates placed six meters before and after the center of the force platform. Marker data were processed with Visual 3D software (CMotion, Bethesda, MD, USA) using a fourth-order, zero-lag, low-pass Butterworth filter with a cut-off frequency of 12 Hz. Strike index, stance time, and horizontal distance from heel to whole-body COM at initial contact were calculated. 2.3.
Continuous Wavelet Transform The resultant GRF vector and the vertical GRF component were processed with the continuous
wavelet transform (CWT) function in MATLAB (Mathworks, Inc., Natick, MA). The CWT calculates the inner product of a signal and the translated and scaled versions of an analyzing function called the mother wavelet. The mother wavelet is of a finite length and oscillates with specific time and frequency properties. With each iteration of the transform, the mother wavelet was scaled by stretching or compressing it in the time-domain, therefore altering its frequency. Each scaled wavelet represents a region of frequencies rather than a single frequency due to the uncertainty principle [27]. Lower scales
7
result in a more compressed wavelet with higher frequency content whereas higher scales result in a more stretched wavelet with a lower frequency content. Because of the linear but imprecise relationship between scale and frequency, the term pseudo-frequency is used to describe the frequency content results of the CWT. The output of the CWT is a wavelet coefficient for each scale (i.e. pseudofrequency) and time point of the original signal. The wavelet coefficients have a larger magnitude (i.e. signal power) when the scaled wavelet and the original signal are more similar. The CWT using the Mexican Hat mother wavelet was performed on the raw resultant GRF vector and the raw vertical GRF component from the entire stance phase of all trials and participants for scale values 1 – 200 (i.e. pseudo-frequency values 1.5 – 300 Hz). This mother wavelet was chosen because it has features that closely match the acceleration profiles seen during impact in running. Only positive wavelet coefficients were considered because they indicate a positive correlation between the scaled wavelet and the original signal. It was assumed that wavelet coefficients with a signal power intensity of less than 200 represented negligible power and were excluded from the analysis of each trial. The GRF data were time-normalized to 101 data points after the CWT was performed. We examined the time and frequency characteristics of the resultant and vertical GRF signal within an impact phase of each trial. The impact phase was defined in the time-domain as the time from initial contact to the instant of the first peak vertical acceleration of the leg segment COM during stance, multiplied by two. Although earlier studies have identified a frequency range of the vertical impact force generated during rearfoot running between 10 – 20 Hz, more recent evidence suggests impact frequencies of substantial power may include frequencies above 35 Hz [22, 28] and frequencies representing impact in forefoot running may be as low as 8 Hz [29]. Therefore, the frequency range defining the impact phase for both footfall patterns was 8 – 50 Hz (presented visually later in the text).
8
The mean pseudo-frequency (PFmean) weighted by signal power was calculated as the product of each wavelet coefficient and its associated pseudo-frequency divided by the total sum of all wavelet coefficients calculated within the impact phase. The sum of the signal power (Psum) and the maximum power (Pmax) were calculated as the sum of the wavelet coefficients and the maximum wavelet coefficient within the impact range, respectively. The pseudo-frequency (PFmax) and percent of stance (%Smax) in which the maximum signal power occurred was also identified.
2.4.
Statistical Analysis An independent sample student’s t-test was used to assess the differences between the RFP and
NRFP groups for participant characteristics and each of the following variables calculated from the impact phase: PFmean, Psum, Pmax, PFmax, and %Smax of the resultant and vertical GRF. Differences were considered significant if P < 0.05. Cohen’s d was calculated to determine the size of the effect of each variable between RFP and NRFP groups. A Cohen’s d of ≤0.4, 0.5–0.7, and ≥0.8 indicated a small, moderate, and large effect size, respectively. 3.
Results The RFP and NRFP groups ran with a mean±SD strike index of 13.0±5.7% and 62.1±5.7%,
respectively (P<0.001, d=8.6). At the same running speed, the RFP group had a 9.7% longer total stance time and a 30.6% greater horizontal distance from heel to whole-body COM at initial contact compared with the NRFP group (P<0.001, d>1.4). The RFP group generated an impact peak visible in the timedomain for both the vertical and resultant GRF, whereas the NRFP group generated a visible impact peak in the anteroposterior and mediolateral components (Figure 1). The impact phase ended at 18.2±3.2% and 24.4±3.6% of the stance phase in the RFP and NRFP groups, respectively (P<0.001, d=1.9; range: RFP=10.0–28.0%, NRFP=15.0–33.0%). Visual inspection of the plots for the group mean wavelet
9
coefficients in time-frequency space of the resultant (Figure 2) and vertical GRF (Figure 3) generated by the NRFP group revealed distinctive signal power occurring during the impact phase in both signals (bottom panels of Figure 2 and Figure 3). [Insert Figures 1-3 here] The statistical results of the CWT performed on the resultant GRF and the vertical GRF component were similar (Table 2). For both signals, PFmean was significantly lower in the NRFP group compared with the RFP group (P<0.001, d>1.0). The NRFP group had a greater Psum during the impact phase compared with the RFP group but the difference was not significant (P > 0.05, d=0.1). Pmax and PFmax were significantly greater in the RFP group than in the NRFP group (Pmax: P<0.001, d≤0.6; PFmax: P<0.001, d>1.0). %Smax was significantly greater in the NRFP group compared with the RFP group (P<0.001, d>1.0). [Insert Table 2 here] 4.
Discussion Non-rearfoot running has been advocated to reduce the risk of running related overuse injury
because these footfall patterns do not generate a visible impact within the time-domain GRF profile. The aim of this study was to characterize the waveform components of the GRF generated during impact by RFP runners and NRFP runners using the continuous wavelet transform. In support of our first hypothesis, the NRFP group generated a resultant GRF and vertical GRF component that contained signal power within the 10 – 20 Hz frequency range which is associated with impact in typical RFP runners [7]. Specifically, maximum signal power occurred at a pseudo-frequency of 15.4±9.1 Hz in the NRFP group, which indicates that non-rearfoot running generates an impact force that is not necessarily visible in the time-domain GRF.
10
The impact force may not appear as a prominent feature within the time-domain GRF signal during NRFP running because the timing of the maximum lower extremity segment deceleration occurs later after initial ground contact compared with RFP running. In support of our second hypothesis, the maximum signal power of the frequencies associated with impact in the NRFP group occurred ~12% later in the stance phase and with a lower frequency compared with the RFP group. These results indicate that the time course of the impact force may be delayed in NRFP running due to a reduced impact loading rate compared with the RFP group. The motion of the ankle joint and foot segment after initial contact by NRFP runners [10] may be responsible for the delay in the time course of maximum signal power of the impact energy compared with rearfoot running and may explain why a distinctive impact peak is visually absent from the resultant GRF and vertical GRF component in the time-domain. Traditional time-domain analyses have only been able to identify a visible impact peak in the anteroposterior and mediolateral GRF components generated during forefoot running [10-12]. The impact peak in these components generated during forefoot running may be a result of differences in foot COM velocity prior to foot contact and center of pressure velocity immediately after foot contact, as described by other authors [10, 12]. Differences in stride length will also contribute to time-domain differences in primarily the vertical GRF component between footfall patterns. Given the similarity in the results between the vertical GRF component and the resultant GRF in the NRFP group, it is likely that the vertical GRF component was the primary contributor to the impact frequencies observed in the resultant GRF rather than the anteroposterior and mediolateral components. However, the impact signal power in the mediolateral and anteroposterior directions should not be ignored given the potential role of shear loading on tissue health [30]. Calculating impact related variables such as vertical loading rate and impact magnitude from the time-domain representations of the vertical ground reaction force may be inappropriate because the active GRF visually obscures the impact force in non-rearfoot running. By isolating the frequencies
11
associated only with foot-ground collision, the present study found that the sum of the signal power of these frequencies within the impact phase was statistically similar between groups for both the resultant and vertical GRF. The similarity of the total impact energy observed in the present study cannot be detected from traditional time-domain analyses. The magnitude of the impact peak will be greater when it is calculated from the original time-domain GRF waveform than the peak magnitude calculated after the impact force is extracted. When a distinctive impact peak is not visually present, conventional time-domain methods for quantifying impact characteristics may not truly represent features of the signal resulting from the foot-ground collision specifically. Therefore, it may be misleading to suggest that one footfall pattern is more or less injurious than another due to differences in vertical impact parameters because these parameters are being calculated on different waveforms for each footfall pattern. Impact parameters of both footfall patterns may be more accurately assessed and compared if they are calculated independent of the active portion of the curve or in the frequency domain. It is currently unknown whether signal power, the frequency of waveform components, the duration that signal power occurs during stance, and/or some other frequency characteristic has the greatest effect on running injury development. Although the sum of the power of the impact energy was similar between groups in the present study, maximum power, the frequency that peak power occurred, and the weighted mean pseudo-frequency were greater in the RFP group than in the NRFP group. The effect of these frequency-domain characteristics on the body tissues and how that effect may differ between rearfoot and non-rearfoot runners should be examined. A limitation of the continuous wavelet transform and other frequency analyses is that individual frequency components cannot be fully isolated due to the uncertainty principle [27]. Each scaled wavelet represents a region of frequencies rather than a single frequency component, thus there is an imprecise relationship between scale and frequency. Therefore, the power and frequency of the
12
waveform components that result from the impact force overlap with those of the active force and cannot be completely isolated. This complication may cause some of the signal power resulting from the active force being included in the calculation of signal power during the impact phase and may explain why several participants had the greatest signal power occurring in the lower end of the frequency range defined for the impact phase (see Table 2). However, a benefit of the CWT analysis over the Fourier Transform is that the time-course of frequency information is maintained, which allowed for the analysis of the frequencies due specifically to impact occurring during an isolated timeperiod of the stance phase. 5.
Conclusions The results of the present study indicate that the foot-ground collision during running generates
frequencies associated with an impact transient, regardless of the portion of the foot that makes initial contact with the ground. The impact force may not appear as a prominent feature within the timedomain signal in non-rearfoot running because the timing of the maximum lower extremity segment acceleration is likely shifted relative to initial ground contact. Given that the vertical impact peak is visually absent in the time-domain with forefoot running, conventional time-domain methods for analyzing loading parameters may be inappropriate. Caution should be used when comparing timedomain impact parameters between footfall patterns without decomposing the impact and active forces into separate waveforms. Conflict of Interest None.
13
References
[1] Daoud AI, Geissler GJ, Wang F, Saretsky J, Daoud YA, Lieberman DE. Foot Strike and Injury Rates in Endurance Runners: a retrospective study. Med Sci Sports Exerc. 2012; 44:132534. [2] Willson JD, Ratcliff OM, Meardon SA, Willy RW. Influence of step length and landing pattern on patellofemoral joint kinetics during running. Scand J Med Sci Sports. 2015; 25:73643. [3] Milner CE, Ferber R, Pollard CD, Hamill J, Davis IS. Biomechanical factors associated with tibial stress fracture in female runners. Med Sci Sports Exerc. 2006; 38:323-8. [4] Pohl MB, Hamill J, Davis IS. Biomechanical and anatomic factors associated with a history of plantar fasciitis in female runners. Clin J Sport Med. 2009; 19:372-6. [5] Zadpoor AA, Nikooyan AA. The relationship between lower-extremity stress fractures and the ground reaction force: a systematic review. Clin Biomech (Bristol, Avon). 2011; 26:23-8. [6] Lieberman DE, Venkadesan M, Werbel WA, Daoud AI, D'Andrea S, Davis IS, et al. Foot strike patterns and collision forces in habitually barefoot versus shod runners. Nature. 2010; 463:531-5. [7] Bobbert MF, Schamhardt HC, Nigg BM. Calculation of vertical ground reaction force estimates during running from positional data. J Biomech. 1991; 24:1095-105. [8] Shorten M, Mientjes MI. The ‘heel impact’ force peak during running is neither ‘heel’ nor ‘impact’ and does not quantify shoe cushioning effects. Footwear Science. 2011; 3:41-58. [9] Gruber AH, Boyer KA, Derrick TR, Hamill J. Impact shock frequency components and attenuation in rearfoot and forefoot running. Journal of Sport and Health Science. 2014; 3:11321. [10] Boyer ER, Rooney BD, Derrick TR. Rearfoot and midfoot or forefoot impacts in habitually shod runners. Med Sci Sports Exerc. 2014; 46:1384-91. [11] Nordin AD, Dufek JS, Mercer JA. Three-dimensional impact kinetics with foot-strike manipulations during running. Journal of Sport and Health Science. 2015; ePub ahead of print. [12] Cavanagh PR, Lafortune MA. Ground reaction forces in distance running. J Biomech. 1980; 13:397-406. [13] Bobbert MF, Yeadon MR, Nigg BM. Mechanical analysis of the landing phase in heel-toe running. J Biomech. 1992; 25:223-34.
14
[14] Gerritsen KG, van den Bogert AJ, Nigg BM. Direct dynamics simulation of the impact phase in heel-toe running. J Biomech. 1995; 28:661-8. [15] Mercer JA, Vance J, Hreljac A, Hamill J. Relationship between shock attenuation and stride length during running at different velocities. Eur J Appl Physiol. 2002; 87:403-8. [16] Potthast W, Bruggemann GP, Lundberg A, Arndt A. The influences of impact interface, muscle activity, and knee angle on impact forces and tibial and femoral accelerations occurring after external impacts. J Appl Biomech. 2010; 26:1-9. [17] Shorten MR, Winslow DS. Spectral Analysis of Impact Shock During Running. Int J Sports Biomech. 1992; 8:288-304. [18] Stergiou N, Giakas G, Byrne JB, Pomeroy V. Frequency domain characteristics of ground reaction forces during walking of young and elderly females. Clin Biomech (Bristol, Avon). 2002; 17:615-7. [19] Nigg BM, Wakeling JM. Impact forces and muscle tuning: a new paradigm. Exerc Sport Sci Rev. 2001; 29:37-41. [20] von Tscharner V. Intensity analysis in time-frequency space of surface myoelectric signals by wavelets of specified resolution. J Electromyogr Kinesiol. 2000; 10:433-45. [21] Enders H, von Tscharner V, Nigg BM. Analysis of damped tissue vibrations in timefrequency space: a wavelet-based approach. J Biomech. 2012; 45:2855-9. [22] 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; 18:50-9. [23] Verdini F, Leo T, Fioretti S, Benedetti MG, Catani F, Giannini S. Analysis of ground reaction forces by means of wavelet transform. Clin Biomech (Bristol, Avon). 2000; 15:607-10. [24] Giakas G, Baltzopoulos V, Dangerfield PH, Dorgan JC, Dalmira S. Comparison of gait patterns between healthy and scoliotic patients using time and frequency domain analysis of ground reaction forces. Spine (Phila Pa 1976). 1996; 21:2235-42. [25] Altman AR, Davis IS. A kinematic method for footstrike pattern detection in barefoot and shod runners. Gait Posture. 2012; 35:298-300. [26] Hamill J, Selbie WS, Kepple TM. Three-Dimensional Kinematics. In: Robertson DGE, Caldwell GE, Hamill J, Kamen G, Whittlesey SN, editors. Research Methods in Biomechanics. 2 ed. Champaign, IL: Human Kinetics; 2014. p. 35-61.
15
[27] Mallat SG. A Theory for Multiresolution Signal Decomposition: The Wavelet Representation. IEEE Trans Pattern Anal Mach Intell. 1989; 11:674-93. [28] Edwards WB, Derrick TR, Hamill J. Musculoskeletal attenuation of impact shock in response to knee angle manipulation. J Appl Biomech. 2012; 28:502-10. [29] Gruber AH, Boyer KA, Derrick TR, Hamill J. Impact shock frequency components and attenuation in rearfoot and forefoot running. Journal of Sport and Health Science. 2014; 3:11321. [30] Turner CH, Wang T, Burr DB. Shear strength and fatigue properties of human cortical bone determined from pure shear tests. Calcif Tissue Int. 2001; 69:373-8. Figure Caption
16
Figr-1
17
Figr-2
18
Figr-3
19
Figr-4Figure 1: Mean stance phase ground reaction forces in the (A) vertical, (B) anteroposterior,
and (C) mediolateral directions generated by the rearfoot pattern group (black) and non-rearfoot pattern group (grey). Figure 2: Continuous wavelet transform results of the resultant ground reaction force. Values are wavelet coefficients for the rearfoot pattern group (RFP, right column) and the non-rearfoot pattern group (NRFP, left column). Coefficient magnitudes are indicated by color intensity. White space represents coefficient magnitudes less than 200 (arbitrary units). Results are plotted across the stance phase for pseudo-frequencies 0 – 100 Hz (top row) and for the impact phase within 0 – 25% of stance and for pseudo-frequencies 8 – 50 Hz (bottom row). The thick vertical black line in the plots of the bottom row represent the mean end of the impact phase in the time domain (RFP = 18.2±3.2% of stance; NRFP = 24.4±3.2% of stance). Figure 3: Continuous wavelet transform results of the vertical ground reaction force component. Values are wavelet coefficients for the rearfoot pattern group (RFP, right column) and the nonrearfoot pattern group (NRFP, left column). Coefficient magnitudes are indicated by color intensity. White space represents coefficient magnitudes less than 200 (arbitrary units). Results are plotted across the stance phase for pseudo-frequencies 0 – 100 Hz (top row) and for the impact phase within 0 – 25% of stance and for pseudo-frequencies 8 – 50 Hz (bottom row). The thick vertical black line in the plots of the bottom row represent the mean end of the impact phase in the time domain (RFP = 18.2±3.2% of stance; NRFP = 24.4±3.2% of stance).
20
Table 1: Participant characteristics of the rearfoot pattern group (RFP) and non-rearfoot pattern group (NRFP).
Group
Males/ Females (#)
Age (yrs)
Height (m)
Mass (kg)
Pref. Speed (m/s)
km/week (km)
RFP
13/7
26.3 ± 6.2
1.76 ± 0.09
69.9 ± 9.8
3.5 ± 0.9
46.3 ± 32.3
NRFP
15/5
25.6 ± 6.1
1.76 ± 0.10
70.3 ± 10.7
3.7 ± 0.3
59.5 ± 25.2
p-value
-
0.702
0.822
0.906
0.324
0.736
21
Table 2: Mean ± SD (range) frequency characteristics of the resultant ground reaction force (GRF) and the vertical GRF during the impact phase generated by the rearfoot pattern group (RFP) and nonrearfoot pattern group (NRFP).
Resultant GRF
RFP
NRFP
Vertical GRF % Difference
RFP
NRFP
(P, d) 26.6 ± 3.0 Hz
18.7 ± 5.8 Hz
(16.3–34.1 Hz)
(12.2–50.0 Hz)
93.4 ± 65.3
102.4 ± 92.0
(18.9– 460.6)
(0.2 – 426.0)
Max Power
1.5 ± 0.4
1.2 ± 0.7
(x103, arbitrary units)
(0.8–2.9)
(0.2 – 4.2)
27.4 ± 4.1 Hz
14.2 ± 7.0 Hz
(8.1–42.9 Hz)
(8.1 – 50.0 Hz)
11.6 ± 1.9%
24.0 ± 4.5%
(7.0– 28.0%)
(7.0 – 33.0%)
Weighted mean pseudofrequency
Power Sum (x103, arbitrary units)
Pseudo-frequency of Max Power
% Stance Max Power Occurred
22
35.0%
(P, d) 26.4 ± 2.8 18.1 ± 5.6 Hz Hz
(<0.001, (16.5–33.7 (12.0 – 46.4 1.8) Hz) Hz)
9.2% (0.278, 0.1)
22.8%
96.4 ± 64.3 (19.9 – 453.0)
1.5 ± 0.5
103.4 ± 92.6 (0.2 – 426.5) 1.2 ± 0.7
(<0.001, (0.8 – 2.9) (0.2 – 4.2) 0.5) 63.3%
% Differenc e
27.2 ± 3.9 15.4 ± 9.1 Hz Hz
37.5% (<0.001, 2.3)
7.0% (0.402, 0.1) 24.1% (<0.001, 0.6) 55.6%
(<0.001, (18.8–37.5 (8.1 – 50.0 2.4) Hz) Hz)
(<0.001, 1.8)
11.5 ± 1.5%
23.1 ± 6.3%
66.6%
(1.0 – 33.0%)
(<0.001, 2.9)
69.9% (<0.001, 3.8)
(8.0 – 16.0%)
S0966-6362(17)30170-4 http://dx.doi.org/doi:10.1016/j.gaitpost.2017.04.037 GAIPOS 5410
To appear in:
Gait & Posture
Received date: Revised date: Accepted date:
10-10-2016 26-4-2017 27-4-2017
Please cite this article as: Gruber Allison H, Edwards W Brent, Hamill Joseph, Derrick Timothy R, Boyer Katherine A.A comparison of the ground reaction force frequency content during rearfoot and non-rearfoot running patterns.Gait and Posture http://dx.doi.org/10.1016/j.gaitpost.2017.04.037 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Original Article – Full Paper A comparison of the ground reaction force frequency content during rearfoot and nonrearfoot running patterns *Allison H. Grubera,b, W. Brent Edwardsc, Joseph Hamillb, Timothy R. Derrickd, Katherine A. Boyerb a Department of Kinesiology, Indiana University, SPH Building, 112, 1025 E. Seventh ST, Bloomington, IN, 47405-7109, United States b
c
Department of Kinesiology, University of Massachusetts, 110 Totman Building, 30 Eastman Lane, Amherst, MA, 01003-9258, United States
Human Performance Laboratory, University of Calgary, KNB 418, 2500 University Drive NW, Calgary, AB, T2N 1N4, Canada
d
Department of Kinesiology, Iowa State University, 249 Forker, 534 Wallace RD, Ames, IA, 50011-3191, United States
*Corresponding author: Allison H. Gruber Affiliation Address: Department of Kinesiology University of Massachusetts 110 Totman Building 30 Eastman Lane Amherst, MA 01003-9258 Permanent Address: Department of Kinesiology Indiana University SPH Building 112 1025 E. Seventh ST Bloomington, IN 47405-7109 Phone: +1(812) 856-2447 Fax: +1(812) 855-3193 Email: [email protected]
(Title) A comparison of the ground reaction force frequency content during rearfoot and nonrearfoot running patterns Highlights:
The presence of an impact transient during non-rearfoot running is proposed. Rearfoot and non-rearfoot running generate frequencies associated with an impact transient. Traditional methods for analyzing impact loading rate and magnitude may be inappropriate.
Abstract Running with a non-rearfoot pattern has been claimed to reduce injury risk because the impact peak in the vertical ground reaction force (GRF) is visually absent in the time-domain compared with a rearfoot pattern. However, running results in a rapid deceleration of the lower extremity segments immediately following initial contact with the ground, regardless of footfall pattern. Therefore, the frequency content of the GRF is expected to contain evidence of this collision. The purpose of the present study was to characterize the waveform components of the GRF generated during the impact phase by habitual rearfoot and habitual non-rearfoot pattern groups using the continuous wavelet transform. Twenty rearfoot and 20 non-rearfoot participants ran over-ground at a standardized speed with their habitual footfall pattern. The continuous wavelet transform was performed on the resultant GRF vector and the vertical GRF. GRF signals generated by the non-rearfoot pattern group during early stance had maximum signal power of 15.4±9.1 Hz occurring at 23.1±6.3% of stance, which is within the 10-20 Hz range previously associated with impact in rearfoot runners. Maximum signal power occurred earlier in the impact phase (11.5±1.5%) and with a higher frequency (27.2±3.9 Hz) in the rearfoot pattern group verses the non-rearfoot pattern group (P<0.05). While the impact force transient may not appear as a prominent feature within the time-domain GRF with a non-rearfoot pattern, the results indicate that both footfall patterns generate frequencies associated with the impact peak in the vertical GRF.
2
Keywords: foot strike pattern; ; ; ; , ground reaction force, running biomechanics, wavelet transform, impact peak
3
1.
Introduction Non-rearfoot running footfall patterns (NRFP, i.e. midfoot or forefoot) have been popularized
because of claims that they may reduce the risk of developing running related injuries verses running with a rearfoot footfall pattern (RFP) (e.g. [1, 2]). Given that characteristics of the impact peak within the vertical ground reaction force (GRF) component were related to running injury in some studies (e.g. [3-5]), it has been suggested that a NRFP will reduce the risk of injury because the vertical impact peak is visually blunted or absent in the time-domain with these patterns [6]. Regardless of footfall pattern, there is a rapid deceleration of the lower extremity segments immediately following initial contact with the ground [7]. Although a distinctive vertical impact peak may not be visible in the time-domain during NRFP running [8], the frequency content of the GRF likely provides evidence of this collision. An analysis of vertical tibial shock indicated that frequencies representative of a significant vertical impact (i.e. 10 – 20 Hz [7]) are present in both RFP and NRFP running [9]. Time-domain analyses have revealed evidence of an impact peak in the anteroposterior and mediolateral GRF during NRFP running [10-12]. However, time-domain analyses of the vertical GRF component may not be sufficient to identify a vertical impact force during NRFP running. The time-domain GRF profile generated during RFP running provides information about the acceleration of the whole-body center of mass (COM), which typically illustrates two distinct peaks in the vertical direction [13]. The first peak is associated with the impact that occurs as a result of the rapid deceleration of the lower extremity segments at initial contact. The second peak is associated with the active motion of the rest of the body when the COM reaches its minimum vertical position during the stance phase [7]. The timing and magnitude of the vertical impact peak depends on parameters such as running speed, stride length, and segment position at initial ground contact [14-16]. Footfall pattern has also been shown to alter vertical loading rate and impact peak magnitude [6, 10, 11]. The impact force may not appear visually as a prominent feature within the time-domain GRF
4
signal generated during NRFP running because the timing of the maximum lower extremity acceleration may be shifted relative to initial ground contact compared with RFP running. Therefore, the calculation of impact loading variables in the time-domain without the visual presence of an impact peak may be inappropriate. The frequencies present in the GRF and tibial acceleration signals during walking and running have typically been analyzed using the Fourier Transform (FT) (e.g. [17, 18]). The FT is well understood, commonly used, and relatively easy to perform with many available software packages. Although the FT can indicate the frequencies and amplitude of each frequency that exist within a signal, its primary disadvantage is the loss of the time resolution of a signal’s frequency components. For example, the FT of the vertical GRF during running will result in lower (4 – 8 Hz) and higher (10 – 20 Hz) frequency bands that represent the active and impact forces, respectively [16, 19]. However, the time course of these frequency components and when the maximum signal power occurs during running is lost. The continuous wavelet transform maintains both the time and frequency resolutions and is used to analyze the similarity of an original signal with different scales and time-shifted versions of a basis function. The wavelet transform has been used to analyze biological signals such as muscle activation [20], tissue vibration [21], and the GRF in walking of healthy individuals [22, 23] and clinical populations [24]. Examining the GRF signal in time-frequency space with a wavelet transform may provide a better understanding of the differences in loading conditions between RFP runners and NRFP runners. The purpose of the present study was to characterize the waveform components of the GRF generated during impact by habitual RFP runners and habitual NRFP runners using the continuous wavelet transform. We hypothesized that the resultant and vertical component of the GRF generated by NRFP runners will include frequency components representative of the impact force that occurs during the impact phase of stance. The vertical impact peak generated by NRFP runners may not be visible in the time-domain because there may be a smaller time delay between the peak accelerations of the lower
5
extremity verses the rest of the body mass and is thus obscured by the active force. Therefore, we also hypothesize that maximum signal power of impact within the resultant and vertical GRF generated by NRFP runners will occur later in stance compared with RFP runners. 2.
Methods
2.1.
Participants Twenty habitual RFP runners and 20 habitual NRFP runners participated in this study (Table 1).
Participants were recreational and/or competitive middle to long distance runners who were free of lower extremity or back injury in the past year and ran a minimum of 16 km/week. Participants were excluded if they had any neurological or cardiovascular pathology. The study was approved by the university Institutional Review Board. All runners read and signed an informed consent document prior to participation. [Insert Table 1 here] Habitual footfall pattern was determined from the data collected during five over-ground running trials performed at each participant’s preferred speed. These trials were collected separately from those used in the wavelet analysis. Footfall pattern was classified using the strike index [12] and sagittal plane foot segment angle at initial ground contact [25]. For the purpose of this study, participants were included in the NRFP group if at least three of the five trials resulted in a strike index greater than 34% and an initial sagittal plane foot segment angle of less than eight degrees of dorsiflexion. The NRFP group consisted of 15 midfoot runners and five forefoot runners classified by these methods. All other participants were placed in the RFP group. 2.2.
Protocol
6
Participants wore fitted shorts and neutral racing flats (RC 550, New Balance, Brighton, MA, USA) provided by the laboratory for all procedures. Three-dimensional motion, center of pressure, and GRFs were recorded for the footfall pattern classification trials and the experimental trials. Calibration and tracking retro-reflective markers were placed on the pelvis and right lower limb based on an adapted Cleveland Clinic Model [26]. Position data of the markers during the static calibration trial and the running trials were recorded by an eight-camera Qualisys Oqus 3-Series optical motion capture system sampling at 240 Hz (Qualisys, Inc., Gothenberg, Sweden). Raw GRF data were recorded by a strain-gauge force platform sampling at 1200 Hz (OR6-5, AMTI, Watertown, MA, USA) that was flush with the floor and in the center of a 30 m runway. Participants were asked to perform 10 over-ground running trials at speed of 3.5 m/s ± 5% while making a right foot contact on the force platform without adjusting their speed or stride. Running speed was monitored by timing gates placed six meters before and after the center of the force platform. Marker data were processed with Visual 3D software (CMotion, Bethesda, MD, USA) using a fourth-order, zero-lag, low-pass Butterworth filter with a cut-off frequency of 12 Hz. Strike index, stance time, and horizontal distance from heel to whole-body COM at initial contact were calculated. 2.3.
Continuous Wavelet Transform The resultant GRF vector and the vertical GRF component were processed with the continuous
wavelet transform (CWT) function in MATLAB (Mathworks, Inc., Natick, MA). The CWT calculates the inner product of a signal and the translated and scaled versions of an analyzing function called the mother wavelet. The mother wavelet is of a finite length and oscillates with specific time and frequency properties. With each iteration of the transform, the mother wavelet was scaled by stretching or compressing it in the time-domain, therefore altering its frequency. Each scaled wavelet represents a region of frequencies rather than a single frequency due to the uncertainty principle [27]. Lower scales
7
result in a more compressed wavelet with higher frequency content whereas higher scales result in a more stretched wavelet with a lower frequency content. Because of the linear but imprecise relationship between scale and frequency, the term pseudo-frequency is used to describe the frequency content results of the CWT. The output of the CWT is a wavelet coefficient for each scale (i.e. pseudofrequency) and time point of the original signal. The wavelet coefficients have a larger magnitude (i.e. signal power) when the scaled wavelet and the original signal are more similar. The CWT using the Mexican Hat mother wavelet was performed on the raw resultant GRF vector and the raw vertical GRF component from the entire stance phase of all trials and participants for scale values 1 – 200 (i.e. pseudo-frequency values 1.5 – 300 Hz). This mother wavelet was chosen because it has features that closely match the acceleration profiles seen during impact in running. Only positive wavelet coefficients were considered because they indicate a positive correlation between the scaled wavelet and the original signal. It was assumed that wavelet coefficients with a signal power intensity of less than 200 represented negligible power and were excluded from the analysis of each trial. The GRF data were time-normalized to 101 data points after the CWT was performed. We examined the time and frequency characteristics of the resultant and vertical GRF signal within an impact phase of each trial. The impact phase was defined in the time-domain as the time from initial contact to the instant of the first peak vertical acceleration of the leg segment COM during stance, multiplied by two. Although earlier studies have identified a frequency range of the vertical impact force generated during rearfoot running between 10 – 20 Hz, more recent evidence suggests impact frequencies of substantial power may include frequencies above 35 Hz [22, 28] and frequencies representing impact in forefoot running may be as low as 8 Hz [29]. Therefore, the frequency range defining the impact phase for both footfall patterns was 8 – 50 Hz (presented visually later in the text).
8
The mean pseudo-frequency (PFmean) weighted by signal power was calculated as the product of each wavelet coefficient and its associated pseudo-frequency divided by the total sum of all wavelet coefficients calculated within the impact phase. The sum of the signal power (Psum) and the maximum power (Pmax) were calculated as the sum of the wavelet coefficients and the maximum wavelet coefficient within the impact range, respectively. The pseudo-frequency (PFmax) and percent of stance (%Smax) in which the maximum signal power occurred was also identified.
2.4.
Statistical Analysis An independent sample student’s t-test was used to assess the differences between the RFP and
NRFP groups for participant characteristics and each of the following variables calculated from the impact phase: PFmean, Psum, Pmax, PFmax, and %Smax of the resultant and vertical GRF. Differences were considered significant if P < 0.05. Cohen’s d was calculated to determine the size of the effect of each variable between RFP and NRFP groups. A Cohen’s d of ≤0.4, 0.5–0.7, and ≥0.8 indicated a small, moderate, and large effect size, respectively. 3.
Results The RFP and NRFP groups ran with a mean±SD strike index of 13.0±5.7% and 62.1±5.7%,
respectively (P<0.001, d=8.6). At the same running speed, the RFP group had a 9.7% longer total stance time and a 30.6% greater horizontal distance from heel to whole-body COM at initial contact compared with the NRFP group (P<0.001, d>1.4). The RFP group generated an impact peak visible in the timedomain for both the vertical and resultant GRF, whereas the NRFP group generated a visible impact peak in the anteroposterior and mediolateral components (Figure 1). The impact phase ended at 18.2±3.2% and 24.4±3.6% of the stance phase in the RFP and NRFP groups, respectively (P<0.001, d=1.9; range: RFP=10.0–28.0%, NRFP=15.0–33.0%). Visual inspection of the plots for the group mean wavelet
9
coefficients in time-frequency space of the resultant (Figure 2) and vertical GRF (Figure 3) generated by the NRFP group revealed distinctive signal power occurring during the impact phase in both signals (bottom panels of Figure 2 and Figure 3). [Insert Figures 1-3 here] The statistical results of the CWT performed on the resultant GRF and the vertical GRF component were similar (Table 2). For both signals, PFmean was significantly lower in the NRFP group compared with the RFP group (P<0.001, d>1.0). The NRFP group had a greater Psum during the impact phase compared with the RFP group but the difference was not significant (P > 0.05, d=0.1). Pmax and PFmax were significantly greater in the RFP group than in the NRFP group (Pmax: P<0.001, d≤0.6; PFmax: P<0.001, d>1.0). %Smax was significantly greater in the NRFP group compared with the RFP group (P<0.001, d>1.0). [Insert Table 2 here] 4.
Discussion Non-rearfoot running has been advocated to reduce the risk of running related overuse injury
because these footfall patterns do not generate a visible impact within the time-domain GRF profile. The aim of this study was to characterize the waveform components of the GRF generated during impact by RFP runners and NRFP runners using the continuous wavelet transform. In support of our first hypothesis, the NRFP group generated a resultant GRF and vertical GRF component that contained signal power within the 10 – 20 Hz frequency range which is associated with impact in typical RFP runners [7]. Specifically, maximum signal power occurred at a pseudo-frequency of 15.4±9.1 Hz in the NRFP group, which indicates that non-rearfoot running generates an impact force that is not necessarily visible in the time-domain GRF.
10
The impact force may not appear as a prominent feature within the time-domain GRF signal during NRFP running because the timing of the maximum lower extremity segment deceleration occurs later after initial ground contact compared with RFP running. In support of our second hypothesis, the maximum signal power of the frequencies associated with impact in the NRFP group occurred ~12% later in the stance phase and with a lower frequency compared with the RFP group. These results indicate that the time course of the impact force may be delayed in NRFP running due to a reduced impact loading rate compared with the RFP group. The motion of the ankle joint and foot segment after initial contact by NRFP runners [10] may be responsible for the delay in the time course of maximum signal power of the impact energy compared with rearfoot running and may explain why a distinctive impact peak is visually absent from the resultant GRF and vertical GRF component in the time-domain. Traditional time-domain analyses have only been able to identify a visible impact peak in the anteroposterior and mediolateral GRF components generated during forefoot running [10-12]. The impact peak in these components generated during forefoot running may be a result of differences in foot COM velocity prior to foot contact and center of pressure velocity immediately after foot contact, as described by other authors [10, 12]. Differences in stride length will also contribute to time-domain differences in primarily the vertical GRF component between footfall patterns. Given the similarity in the results between the vertical GRF component and the resultant GRF in the NRFP group, it is likely that the vertical GRF component was the primary contributor to the impact frequencies observed in the resultant GRF rather than the anteroposterior and mediolateral components. However, the impact signal power in the mediolateral and anteroposterior directions should not be ignored given the potential role of shear loading on tissue health [30]. Calculating impact related variables such as vertical loading rate and impact magnitude from the time-domain representations of the vertical ground reaction force may be inappropriate because the active GRF visually obscures the impact force in non-rearfoot running. By isolating the frequencies
11
associated only with foot-ground collision, the present study found that the sum of the signal power of these frequencies within the impact phase was statistically similar between groups for both the resultant and vertical GRF. The similarity of the total impact energy observed in the present study cannot be detected from traditional time-domain analyses. The magnitude of the impact peak will be greater when it is calculated from the original time-domain GRF waveform than the peak magnitude calculated after the impact force is extracted. When a distinctive impact peak is not visually present, conventional time-domain methods for quantifying impact characteristics may not truly represent features of the signal resulting from the foot-ground collision specifically. Therefore, it may be misleading to suggest that one footfall pattern is more or less injurious than another due to differences in vertical impact parameters because these parameters are being calculated on different waveforms for each footfall pattern. Impact parameters of both footfall patterns may be more accurately assessed and compared if they are calculated independent of the active portion of the curve or in the frequency domain. It is currently unknown whether signal power, the frequency of waveform components, the duration that signal power occurs during stance, and/or some other frequency characteristic has the greatest effect on running injury development. Although the sum of the power of the impact energy was similar between groups in the present study, maximum power, the frequency that peak power occurred, and the weighted mean pseudo-frequency were greater in the RFP group than in the NRFP group. The effect of these frequency-domain characteristics on the body tissues and how that effect may differ between rearfoot and non-rearfoot runners should be examined. A limitation of the continuous wavelet transform and other frequency analyses is that individual frequency components cannot be fully isolated due to the uncertainty principle [27]. Each scaled wavelet represents a region of frequencies rather than a single frequency component, thus there is an imprecise relationship between scale and frequency. Therefore, the power and frequency of the
12
waveform components that result from the impact force overlap with those of the active force and cannot be completely isolated. This complication may cause some of the signal power resulting from the active force being included in the calculation of signal power during the impact phase and may explain why several participants had the greatest signal power occurring in the lower end of the frequency range defined for the impact phase (see Table 2). However, a benefit of the CWT analysis over the Fourier Transform is that the time-course of frequency information is maintained, which allowed for the analysis of the frequencies due specifically to impact occurring during an isolated timeperiod of the stance phase. 5.
Conclusions The results of the present study indicate that the foot-ground collision during running generates
frequencies associated with an impact transient, regardless of the portion of the foot that makes initial contact with the ground. The impact force may not appear as a prominent feature within the timedomain signal in non-rearfoot running because the timing of the maximum lower extremity segment acceleration is likely shifted relative to initial ground contact. Given that the vertical impact peak is visually absent in the time-domain with forefoot running, conventional time-domain methods for analyzing loading parameters may be inappropriate. Caution should be used when comparing timedomain impact parameters between footfall patterns without decomposing the impact and active forces into separate waveforms. Conflict of Interest None.
13
References
[1] Daoud AI, Geissler GJ, Wang F, Saretsky J, Daoud YA, Lieberman DE. Foot Strike and Injury Rates in Endurance Runners: a retrospective study. Med Sci Sports Exerc. 2012; 44:132534. [2] Willson JD, Ratcliff OM, Meardon SA, Willy RW. Influence of step length and landing pattern on patellofemoral joint kinetics during running. Scand J Med Sci Sports. 2015; 25:73643. [3] Milner CE, Ferber R, Pollard CD, Hamill J, Davis IS. Biomechanical factors associated with tibial stress fracture in female runners. Med Sci Sports Exerc. 2006; 38:323-8. [4] Pohl MB, Hamill J, Davis IS. Biomechanical and anatomic factors associated with a history of plantar fasciitis in female runners. Clin J Sport Med. 2009; 19:372-6. [5] Zadpoor AA, Nikooyan AA. The relationship between lower-extremity stress fractures and the ground reaction force: a systematic review. Clin Biomech (Bristol, Avon). 2011; 26:23-8. [6] Lieberman DE, Venkadesan M, Werbel WA, Daoud AI, D'Andrea S, Davis IS, et al. Foot strike patterns and collision forces in habitually barefoot versus shod runners. Nature. 2010; 463:531-5. [7] Bobbert MF, Schamhardt HC, Nigg BM. Calculation of vertical ground reaction force estimates during running from positional data. J Biomech. 1991; 24:1095-105. [8] Shorten M, Mientjes MI. The ‘heel impact’ force peak during running is neither ‘heel’ nor ‘impact’ and does not quantify shoe cushioning effects. Footwear Science. 2011; 3:41-58. [9] Gruber AH, Boyer KA, Derrick TR, Hamill J. Impact shock frequency components and attenuation in rearfoot and forefoot running. Journal of Sport and Health Science. 2014; 3:11321. [10] Boyer ER, Rooney BD, Derrick TR. Rearfoot and midfoot or forefoot impacts in habitually shod runners. Med Sci Sports Exerc. 2014; 46:1384-91. [11] Nordin AD, Dufek JS, Mercer JA. Three-dimensional impact kinetics with foot-strike manipulations during running. Journal of Sport and Health Science. 2015; ePub ahead of print. [12] Cavanagh PR, Lafortune MA. Ground reaction forces in distance running. J Biomech. 1980; 13:397-406. [13] Bobbert MF, Yeadon MR, Nigg BM. Mechanical analysis of the landing phase in heel-toe running. J Biomech. 1992; 25:223-34.
14
[14] Gerritsen KG, van den Bogert AJ, Nigg BM. Direct dynamics simulation of the impact phase in heel-toe running. J Biomech. 1995; 28:661-8. [15] Mercer JA, Vance J, Hreljac A, Hamill J. Relationship between shock attenuation and stride length during running at different velocities. Eur J Appl Physiol. 2002; 87:403-8. [16] Potthast W, Bruggemann GP, Lundberg A, Arndt A. The influences of impact interface, muscle activity, and knee angle on impact forces and tibial and femoral accelerations occurring after external impacts. J Appl Biomech. 2010; 26:1-9. [17] Shorten MR, Winslow DS. Spectral Analysis of Impact Shock During Running. Int J Sports Biomech. 1992; 8:288-304. [18] Stergiou N, Giakas G, Byrne JB, Pomeroy V. Frequency domain characteristics of ground reaction forces during walking of young and elderly females. Clin Biomech (Bristol, Avon). 2002; 17:615-7. [19] Nigg BM, Wakeling JM. Impact forces and muscle tuning: a new paradigm. Exerc Sport Sci Rev. 2001; 29:37-41. [20] von Tscharner V. Intensity analysis in time-frequency space of surface myoelectric signals by wavelets of specified resolution. J Electromyogr Kinesiol. 2000; 10:433-45. [21] Enders H, von Tscharner V, Nigg BM. Analysis of damped tissue vibrations in timefrequency space: a wavelet-based approach. J Biomech. 2012; 45:2855-9. [22] 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; 18:50-9. [23] Verdini F, Leo T, Fioretti S, Benedetti MG, Catani F, Giannini S. Analysis of ground reaction forces by means of wavelet transform. Clin Biomech (Bristol, Avon). 2000; 15:607-10. [24] Giakas G, Baltzopoulos V, Dangerfield PH, Dorgan JC, Dalmira S. Comparison of gait patterns between healthy and scoliotic patients using time and frequency domain analysis of ground reaction forces. Spine (Phila Pa 1976). 1996; 21:2235-42. [25] Altman AR, Davis IS. A kinematic method for footstrike pattern detection in barefoot and shod runners. Gait Posture. 2012; 35:298-300. [26] Hamill J, Selbie WS, Kepple TM. Three-Dimensional Kinematics. In: Robertson DGE, Caldwell GE, Hamill J, Kamen G, Whittlesey SN, editors. Research Methods in Biomechanics. 2 ed. Champaign, IL: Human Kinetics; 2014. p. 35-61.
15
[27] Mallat SG. A Theory for Multiresolution Signal Decomposition: The Wavelet Representation. IEEE Trans Pattern Anal Mach Intell. 1989; 11:674-93. [28] Edwards WB, Derrick TR, Hamill J. Musculoskeletal attenuation of impact shock in response to knee angle manipulation. J Appl Biomech. 2012; 28:502-10. [29] Gruber AH, Boyer KA, Derrick TR, Hamill J. Impact shock frequency components and attenuation in rearfoot and forefoot running. Journal of Sport and Health Science. 2014; 3:11321. [30] Turner CH, Wang T, Burr DB. Shear strength and fatigue properties of human cortical bone determined from pure shear tests. Calcif Tissue Int. 2001; 69:373-8. Figure Caption
16
Figr-1
17
Figr-2
18
Figr-3
19
Figr-4Figure 1: Mean stance phase ground reaction forces in the (A) vertical, (B) anteroposterior,
and (C) mediolateral directions generated by the rearfoot pattern group (black) and non-rearfoot pattern group (grey). Figure 2: Continuous wavelet transform results of the resultant ground reaction force. Values are wavelet coefficients for the rearfoot pattern group (RFP, right column) and the non-rearfoot pattern group (NRFP, left column). Coefficient magnitudes are indicated by color intensity. White space represents coefficient magnitudes less than 200 (arbitrary units). Results are plotted across the stance phase for pseudo-frequencies 0 – 100 Hz (top row) and for the impact phase within 0 – 25% of stance and for pseudo-frequencies 8 – 50 Hz (bottom row). The thick vertical black line in the plots of the bottom row represent the mean end of the impact phase in the time domain (RFP = 18.2±3.2% of stance; NRFP = 24.4±3.2% of stance). Figure 3: Continuous wavelet transform results of the vertical ground reaction force component. Values are wavelet coefficients for the rearfoot pattern group (RFP, right column) and the nonrearfoot pattern group (NRFP, left column). Coefficient magnitudes are indicated by color intensity. White space represents coefficient magnitudes less than 200 (arbitrary units). Results are plotted across the stance phase for pseudo-frequencies 0 – 100 Hz (top row) and for the impact phase within 0 – 25% of stance and for pseudo-frequencies 8 – 50 Hz (bottom row). The thick vertical black line in the plots of the bottom row represent the mean end of the impact phase in the time domain (RFP = 18.2±3.2% of stance; NRFP = 24.4±3.2% of stance).
20
Table 1: Participant characteristics of the rearfoot pattern group (RFP) and non-rearfoot pattern group (NRFP).
Group
Males/ Females (#)
Age (yrs)
Height (m)
Mass (kg)
Pref. Speed (m/s)
km/week (km)
RFP
13/7
26.3 ± 6.2
1.76 ± 0.09
69.9 ± 9.8
3.5 ± 0.9
46.3 ± 32.3
NRFP
15/5
25.6 ± 6.1
1.76 ± 0.10
70.3 ± 10.7
3.7 ± 0.3
59.5 ± 25.2
p-value
-
0.702
0.822
0.906
0.324
0.736
21
Table 2: Mean ± SD (range) frequency characteristics of the resultant ground reaction force (GRF) and the vertical GRF during the impact phase generated by the rearfoot pattern group (RFP) and nonrearfoot pattern group (NRFP).
Resultant GRF
RFP
NRFP
Vertical GRF % Difference
RFP
NRFP
(P, d) 26.6 ± 3.0 Hz
18.7 ± 5.8 Hz
(16.3–34.1 Hz)
(12.2–50.0 Hz)
93.4 ± 65.3
102.4 ± 92.0
(18.9– 460.6)
(0.2 – 426.0)
Max Power
1.5 ± 0.4
1.2 ± 0.7
(x103, arbitrary units)
(0.8–2.9)
(0.2 – 4.2)
27.4 ± 4.1 Hz
14.2 ± 7.0 Hz
(8.1–42.9 Hz)
(8.1 – 50.0 Hz)
11.6 ± 1.9%
24.0 ± 4.5%
(7.0– 28.0%)
(7.0 – 33.0%)
Weighted mean pseudofrequency
Power Sum (x103, arbitrary units)
Pseudo-frequency of Max Power
% Stance Max Power Occurred
22
35.0%
(P, d) 26.4 ± 2.8 18.1 ± 5.6 Hz Hz
(<0.001, (16.5–33.7 (12.0 – 46.4 1.8) Hz) Hz)
9.2% (0.278, 0.1)
22.8%
96.4 ± 64.3 (19.9 – 453.0)
1.5 ± 0.5
103.4 ± 92.6 (0.2 – 426.5) 1.2 ± 0.7
(<0.001, (0.8 – 2.9) (0.2 – 4.2) 0.5) 63.3%
% Differenc e
27.2 ± 3.9 15.4 ± 9.1 Hz Hz
37.5% (<0.001, 2.3)
7.0% (0.402, 0.1) 24.1% (<0.001, 0.6) 55.6%
(<0.001, (18.8–37.5 (8.1 – 50.0 2.4) Hz) Hz)
(<0.001, 1.8)
11.5 ± 1.5%
23.1 ± 6.3%
66.6%
(1.0 – 33.0%)
(<0.001, 2.9)
69.9% (<0.001, 3.8)
(8.0 – 16.0%)