Next Article in Journal
Weld Feature Extraction Based on Semantic Segmentation Network
Previous Article in Journal
Comparative Study of Structural Anomaly Diagnosis Based on ANN Model Using Random Displacement and Acceleration Responses with Incomplete Measurements
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sensors
Volume 22
Issue 11
10.3390/s22114129
sensors-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Between-Day Reliability of Commonly Used IMU Features during a Fatiguing Run and the Effect of Speed

by
Hannah L. Dimmick
1,*,
Cody R. van Rassel
1,
Martin J. MacInnis
1 and
Reed Ferber
1,2,3
1
Faculty of Kinesiology, University of Calgary, Calgary, AB T2N 1N4, Canada
2
Faculty of Nursing, University of Calgary, Calgary, AB T2N 1N4, Canada
3
Cumming School of Medicine, University of Calgary, Calgary, AB T2N 1N4, Canada
*
Author to whom correspondence should be addressed.
Sensors 2022, 22(11), 4129; https://doi.org/10.3390/s22114129
Submission received: 24 April 2022 / Revised: 20 May 2022 / Accepted: 26 May 2022 / Published: 29 May 2022
(This article belongs to the Section Wearables)
Browse Figures
Versions Notes

Abstract

:
The purpose of this study was to determine if fatigue-related changes in biomechanics derived from an inertial measurement unit (IMU) placed at the center of mass (CoM) are reliable day-to-day. Sixteen runners performed two runs at maximal lactate steady state (MLSS) on a treadmill, one run 5% above MLSS speed, and one run 5% below MLSS speed while wearing a CoM-mounted IMU. Trials were performed to volitional exhaustion or a specified termination time. IMU features were derived from each axis and the resultant. Feature means were calculated for each subject during non-fatigued and fatigued states. Comparisons were performed between the two trials at MLSS and between all four trials. The only significant fatigue state × trial interaction was the 25th percentile of the results when comparing all trials. There were no main effects for trial for either comparison method. There were main effects for fatigue state for most features in both comparison methods. Reliability, measured by an intraclass coefficient (ICC), was good-to-excellent for most features. These results suggest that fatigue-related changes in biomechanics derived from a CoM-mounted IMU are reliable day-to-day when participants ran at or around MLSS and are not significantly affected by slight deviations in speed.

1. Introduction

Inertial measurement units (IMUs) have gained popularity in recent years due to their ability to measure gait patterns in an accessible, inexpensive, and portable manner. IMUs have been used in clinical contexts [1,2], team sports [3,4], activity recognition [5,6], and to investigate various aspects of walking and running gait [7,8,9,10]. Moreover, a growing number of studies have employed research-grade and commercial IMUs to detect fatigue-related changes in running, using various experimental setups and sensor locations [11,12,13].
Although running is a popular modality of physical activity [14], the injury rate is extremely high [15,16]. Most running-related injuries (RRIs) are due to overuse rather than acute incidents [17,18], and the risk of overuse injury may be heightened in a fatigued state, as the resultant “atypical” mechanics lead to atypical stresses on the musculoskeletal system [19,20,21]. Bones, tendons, ligaments, cartilage, and muscles may have a lower tensile limit for the direction and magnitude of the altered mechanics, causing an accumulation of microdamage “beyond the capabilities of the specific structure” that can lead to injury [22,23]. Thus, fatigue-related alterations in biomechanics have been suggested as a potential factor in the development of RRIs [19,24,25].
IMUs placed at the center of mass (CoM) have proven particularly useful for gaining insights into fatigue-related changes in running, due to their unobtrusive positioning and ability to capture deviations related to the whole body [12]. Using a CoM-mounted IMU, a variety of statistical, regularity, dynamic, and symmetry features have demonstrated sensitivity to fatigue-related changes, both in basic statistical comparisons [12,26,27,28] and as inputs into machine learning algorithms [29,30,31]. However, most studies only perform single-day data collections, and thus it has not been determined whether these features (with the exception of root mean square (RMS) [32,33,34]) possess good day-to-day reliability during either non-fatigue or fatigue conditions.
In order to elucidate the mechanism of fatigue-related injury, it is important to first determine whether running biomechanics are consistent day-to-day, to determine if internal and external circumstances (e.g., stress, sleep, diet, time of day, and running speed) influence this dynamic. For example, Kawabata et al. [33] demonstrated that changes in speed (~±15%) had a significant effect on the RMS of the acceleration signal during non-fatigue conditions, and McGregor et al. [34] similarly demonstrated significant correlations between speed and acceleration RMS/RMS ratio (RMSR) during an incremental test.
If running gait biomechanics during fatigue are consistent day-to-day, it would support the suggestion that fatigue-related changes to gait patterns are potentially related to injury, as tissues would be regularly subjected to the same atypical forces [20]. However, if fatigued biomechanics are variable, the value of using single-trial studies to analyze fatigue-related alterations may not be particularly high and deriving meaning from these experimental designs about possible mechanisms of injury may not be possible. To our knowledge, no prior study has sought to determine the day-to-day reliability of IMU-based running biomechanics during a fatigued state. Further, few studies have effectively considered the relative running intensity when evaluating changes in running biomechanics from fatigue, particularly with respect to the maximal lactate steady state (MLSS). The MLSS threshold demarcates the heavy from the severe exercise intensity domains and represents the boundary between sustainable and unsustainable intensities of exercise [35,36,37,38]. To set exercise intensity, most prior research designs have used time trial speed [11,12,24], previous race performance [30,39], or a percentage of the maximal speed obtained during an incremental treadmill test [9], which can produce heterogeneous responses in metabolic strain [40,41]. We contend that in order to better discern the effect of fatigue on biomechanical alterations in IMU data, the employed exercise stimulus should be prescribed relative to the physiological capabilities of the individual.
Therefore, the primary purpose of this study is to determine the reliability of commonly derived IMU features used for the detection of biomechanical alterations in treadmill running fatigue. We hypothesize that most features will demonstrate good-to-excellent reliability during both non-fatigue and fatigue conditions. The secondary purpose of this study is to determine if adjustments in running speed around the MLSS boundary affect measures of reliability, and we hypothesize that reliability decreases as running speed increases and/or decreases.

2. Materials and Methods

2.1. General

Sixteen recreationally and competitively trained runners (7 female, 9 male, age = 30.1 ± 4.2 yrs, height = 174.3 ± 9.1 cm, weight = 70.5 ± 10.5 kg) provided informed consent to participate in this study, which was approved by the Ethics Board at the University of Calgary (REB20-0111). Participants were included if they were between the ages of 18 and 45 and had a recent 10 km performance of ≤50 min or ≤55 min for men and women, respectively. All participants were familiar with treadmill running, were free of medical conditions and injuries that could interfere with metabolic and cardiorespiratory exercise responses, and completed the Physical Activity Readiness Questionnaire (PAR-Q+) prior to exercise to ensure there were no identifiable contraindications to exercise.
Participants visited the lab five or six times. For all visits, participants were instructed to refrain from smoking, eating, and consuming caffeine at least 2 h prior to testing and to refrain from strenuous exercise at least 6 h prior. Participants used their own running shoes but were required to use the same shoes for each visit. Participants were weighed and measured at the beginning of each visit and familiarized with the Borg rating of perceived exertion (RPE, 6–20) scale [42]. The RPE scale has previously been used in running-related fatigue literature to detect subjective and holistic changes in an individual’s capacity to continue a task [11,25,29].

2.1.1. Visit 1: Step-Ramp-Step Test

On the first visit, a modified “Step-Ramp-Step” (SRS) exercise test was performed [41]. A metabolic cart (Quark, CPET, Cosmed, Rome, Italy) was used to measure ventilatory and gas exchange variables. Participants were fitted with an Oro-Nasal 7450 V2 mask connected to a two-way non-rebreathing valve (Hans-Rudolph Inc., Kansas City, MO, USA), and a plastic hose with expired air was directed into a 7 L mixing chamber, from which ventilatory parameters were measured or derived and reported as 10 s averages. Prior to the testing session, gas analyzers were calibrated using a gas mixture of known composition (5% CO2, 16% O2, and N2 balance), and turbine flow meters were calibrated using a 3 L syringe, according to manufacturer guidelines. Heart rate was measured by a chest strap placed just below the sternum (H10; Polar Electro Inc., Bethpage, NY, USA). Blood lactate concentration ([BLa]) measurements were made using a capillary blood sample from the fingertip to determine resting and peak blood lactate responses to incremental exercise. [BLa] was analyzed using the Biosen C-Line lactate analyzer (EKF Diagnostics, Cardiff, Wales) for the first 8 runners. Due to supply chain issues mid-way through the course of the study, [BLa] analyses were transitioned to the Lactate Plus (Nova Biomedical, Waltham, MA, USA) system for the remaining eight runners. Data from the SRS test was used to estimate the speed associated with maximal lactate steady state (MLSS) for experimental trials.

2.1.2. Visits 2–5: Experimental Trials

Prior to the start of each experimental trial, participants were outfitted with an IMU (Blue Trident, Vicon, Oxford, UK; tri-axial accelerometer sampling rate 1125 Hz, range ±16 g) positioned between the posterior superior iliac spines with the top border of the sensor positioned on a line coincident with the inferior aspect of the iliac crest. The Blue Trident IMU is a research-grade device capable of collecting data at high frequencies and providing high-quality data. The X, Y, and Z axes were oriented in the vertical (VT; + superior), medial–lateral (ML; + to the left), and anterior–posterior (AP; + to the posterior) directions, respectively. Incline of the treadmill was set at 1% to best match the energetic cost of outdoor running [43].
Each experimental trial began with the participant performing a 5 min warmup at 1.92 m/s before increasing to the target speed. The trial was terminated when the participant reached volitional exhaustion, or at 45 min, whichever occurred first. All runners performed the initial constant-speed test at the running speed estimated to be MLSS by the SRS protocol. Follow-up visits were conducted at either 5% faster or 5% slower treadmill speeds until running tests were conducted at MLSS speed, 5% above MLSS speed, and 5% below MLSS speed. The final constant-speed treadmill running visit was a repeat trial at the running speed associated with MLSS. Each visit was separated by at least 48 h, and participants were not informed of the speed until after all experimental trials were completed. The MLSS for each runner was identified as the highest treadmill speed where at least 30 min of exercise was performed and the difference between [BLa] at 10 and 30 min was < 1 mmol·L−1 [35], using [BLa] measurements taken from a capillary blood sample from the fingertip. Participants provided RPE measures every 5 min and at trial termination, and verbal encouragement was provided throughout.

2.2. Data Processing

Overall, four experimental trials at three target speeds were included for final analyses—two trials at the estimated speed at MLSS (MLSS1/MLSS2), one at 5% faster than estimated speed at MLSS (F), and one at 5% slower than estimated speed at MLSS (S).
All data processing was performed using custom MATLAB software (version 2021b, Mathworks, Inc., Natick, MA, USA).
A static attitude correction was performed on acceleration data to align the sensor with a global three-dimensional position to enable comparisons between trials and participants [10,44,45].
Initial contact (IC) for each step was identified using methods described in Benson et al. [8]. A step was defined as the duration between consecutive ICs from contralateral feet, and a stride was defined as the duration between consecutive ICs from the ipsilateral foot. Mean and standard deviation of the number of data points in each step were calculated, and those ±2 standard deviations from the mean were labeled as improperly segmented and excluded [8].
Samples were created by selecting five consecutive strides [7,8]. Non-fatigue (NF) and fatigue (FT) conditions were considered the first 5 min of the trial and the final complete 5 min segment (to correspond to RPE sampling), respectively [9]. For example, if a participant terminated the test at 38 min, the period from 30 to 35 min would be defined as the FT condition using a single RPE value (taken at 35 min) to represent the segment. The RPE label for each condition was the value provided at the end of the 5 min segment.

2.3. Feature Extraction

Acceleration data were not filtered, and raw signals were maintained for feature creation [9,11,34]. Features were chosen based on previous literature [8,29] and were extracted from the three primary axes (VT, ML, AP) and the resultant (RES) of the acceleration signal: mean, standard deviation, median, 25th percentile, 75th percentile, RMS [5,6,46], maximum, minimum [5,11], and sample entropy (SE) [11,12,47,48,49]. The SE signal stabilizes over 2000 data points [50], and the minimum number of data points in each of our samples was ~3400 (corresponding to a cadence of 180 steps per minute), indicating that there were sufficient data for the calculation. A gait cycle approach, rather than a time sample approach, was used to derive the measurement [12,51]. Additionally, the ratio of single-axis RMS (RMSVT, RMSML, RMSAP) to RMSRES (RMSR) [34] was extracted. This resulted in 39 features extracted from each sample.

2.4. Statistical Analysis

One-way analysis of variance (ANOVA) was performed to compare the following variables across trials: trial time, RPENF, RPEFT, change in RPE from NF to FT (ΔRPE), number of strides analyzed for NF, and number of strides analyzed for FT. Additionally, paired t-tests were performed between RPENF and RPEFT for each trial.
The IMU data were analyzed using two methods: first by comparing MLSS1 and MLSS2 (MM, where speed was the same in both trials) and subsequently comparing all 4 experimental trials (ALL). In both methods, a two-way ANOVA comparing fatigue state and trial (i.e., fatigue state (NF vs. FT) × trial (MLSS1 vs. MLSS2) or fatigue state (NF vs. FT) × trial (MLSS1 vs. MLSS2 vs. F vs. S)) was performed for each feature. Additionally, within-fatigue state, between-trial reliability for each feature was evaluated by intraclass correlation coefficients (ICC 2,k). Values of <0.5, 0.5–0.75, 0.75–0.9, and >0.9 were interpreted as poor, moderate, good, and excellent, respectively [52]. This was calculated for both MM and ALL methods.
The level of significance was set at p < 0.05 for the statistical tests. Statistical analyses were performed using SPSS version 26 (IBM Corp., Armonk, NY, USA).

3. Results

Participants reported a recent training volume of 31.0 ± 21.0 km/week for the previous 3 months and had a recent performance best 10 km time of 44.5 ± 6.3 min. Trial details are presented in Table 1. Except for the F trial, all participants completed at least 30 min of treadmill running during the MLSS1, MLSS2, and S trials (Table 1).

3.1. Between-Trial Analysis

There were no significant fatigue state × trial interactions for the MM method. For the ALL method, only 25th percentile (RES) demonstrated a significant fatigue state × trial interaction (p < 0.001). There were no main effects for trial for either the MM or ALL methods. The main effects for the fatigue state are summarized in Table 2. Specific values for each feature during each trial and fatigue state are contained in Supplementary Materials Table S1.

3.2. Reliability Analysis

Most ICCs were considered good-to-excellent, with ALL values slightly higher than MM values, and NF values slightly higher than FT values (NFMM = 0.839 ± 0.14, FTMM = 0.812 ± 0.24, NFALL = 0.934 ± 0.06, FTALL = 0.927 ± 0.08). Results for ICC(2,k) values for each axis are summarized in Figure 1, Figure 2, Figure 3 and Figure 4.

4. Discussion

The primary finding of the present study was that, in general, the IMU features included in this analysis can be considered highly reliable for both NF and FT conditions. The ICCs reported were mostly good or excellent (ICC > 0.75), and there were no main effects for the trial, indicating that features were similar across days in both NF and FT conditions, even with differing speeds. These results do not support our hypothesis that minor (±5%) variations in speed would reduce the day-to-day reliability of a feature. Although the F trial was significantly shorter than the other trials, there was no main effect for the trial, indicating that the directionality and pattern of fatigue-related alterations are consistent even if the time to onset is different. Therefore, researchers can be reasonably confident that single-trial studies using IMUs to detect fatigue-related changes are representative of the participants’ typical running gait biomechanics irrespective of fatigue state.
Interestingly, for the MM method, 6 of 8 features with ICC values <0.75 were in the ML axis. Although these same features were considered good or excellent when analyzed using the ALL method (except for median), the mean ICC values in the ML axis (NFALL = 0.88 ± 0.09, FTALL = 0.88 ± 0.13) were still lower than those in the AP (NFALL = 0.94 ± 0.03, FTALL = 0.93 ± 0.04) and VT (NFALL = 0.95 ± 0.03, FTALL = 0.94 ± 0.03) axes. The median in the ML axis demonstrated particularly low ICC values (NFMM = 0.21, FTMM = 0.58, NFALL = 0.63, FTALL = 0.51); although, there were no significant differences across trials. Previous studies have found the greatest levels of running fatigue-related deviation [28] and sensitivity to disease-related gait changes [53] in the ML axis. Taken with the present results, this may suggest that the musculature controlling ML movements is the most susceptible to fatigue, and this increased variability may contribute to lower reliability.
The results from the NF state are consistent with previous findings on the reliability of the acceleration RMS signal, which have reported ICC values of 0.89–0.98, 0.95–0.97, 0.97, and 0.90–0.99 for the RMSVT, RMSML, RMSAP, and RMSRES, respectively [32,34]. These values align with our findings, where ICC values using the ALL method were 0.97, 0.93, 0.96, and 0.97 for RMSVT, RMSML, RMSAP, and RMSRES, respectively. However, the values from the previous studies and the ALL method were somewhat higher than those of the MM method (RMSVT = 0.89, RMSML = 0.84, RMSAP = 0.92, RMSRES = 0.88). To our knowledge, no other accelerometer-derived statistical features have been previously analyzed for reliability, so it is not possible to compare other features to previous results.
Most features demonstrated a significant main effect for fatigue state, confirming that these measures are largely sensitive to fatigue-related changes in biomechanical alterations during running, and that the high reliability during FT was not only due to a lack of changes from NF to FT. The observed changes in biomechanics with fatigue were also generally consistent with previous literature. For example, the acceleration RMS values in the current study were observed to increase in all axes, which is consistent with Schütte et al. [12], and somewhat consistent with Le Bris et al. [28], who reported increases in RMSML only. A change in the RMSR was observed in all axes. RMSRVT decreased, while RMSRML and RMSRAP values increased; although, for RMSRAP, only the ALL method was significant. These results are similar to previous results [47,54], such as Winter et al. [55], who reported decreased RMSRVT and increased RMSRML. Schütte et al. [48] also reported an increase in RMSRML, and in another study, [12] reported increased RMSRML and RMSRAP. These slightly different results may be due to differences in protocol and statistical methodology. For example, while Winter et al. [55] used an all-out 8 km run, the others used exhaustive protocols that lasted only 4–20 min [12,28,48]. These shorter protocols involved running to exhaustion, indicating subjects were likely running well above MLSS (i.e., unsustainable intensity), which could have produced more variability in their responses [38]. The relatively consistent observation of increased RMSML and RMSRML supports the earlier suggestion that the reliability of ML features may be lower due to increases in variability with fatigue.
There were also main effects for fatigue state for SE, with SEVT and SERES increasing significantly, SEML decreasing significantly, and no changes observed in SEAP. These results are similar to Pla et al. [47], who demonstrated increases in SEVT with fatigue, but also observed an increase in SEAP. However, Schütte et al. [12] reported contrary results, with no fatigue-related changes in SEVT or SEML, but significant increases in SEAP. Interestingly, in another study, these same authors [48] reported fatigue-related decreases in SEML, but only for those with a history of medial tibial stress syndrome. Higher SE indicates lower predictability in the signal, which may represent neuromuscular re-organization [51]. An increase in SE with running fatigue has been interpreted as a protective neuromuscular mechanism to avoid pain [12], whereas lower SE has been associated with unhealthy and frail states [56]. Whether the present decrease in SEML portends specific injury risk or reflects injury history is beyond the scope of the study but should continue to be investigated.
Specific changes in RMS, RMSR, and SE are discussed here given their previous use in the literature, but caution is warranted in over-interpreting any single significant feature here. Between the MM and ALL methods, 78 features were compared; given that α = 0.05, ~4 effects would be significant due to chance [57]; although, the general pattern of significant differences between fatigue states indicates that biomechanical adjustments were made over the course of the trial. However, single feature interpretation is not necessarily required, as most of these features are used as inputs for machine learning algorithms in practice [29,30,31]. Therefore, the reliability of the features is more important than the individual features’ sensitivity to fatigue.
Additional limitations are acknowledged. First, no differences were found between trials when speeds were different, which contradicts previous findings from studies investigating non-fatigued biomechanics [33,34]. The absence of speed-related differences is likely due to smaller perturbations in speed than previously reported. While this protocol only altered speed ±5%, others used ranges of 8–16 km/h (2.22–4.44 m/s) [34], or ~±15% changes [33]. Future studies should investigate the effect of larger changes in speed on fatigue-related biomechanical adjustments. Second, Benson et al. [58] and Ahamed et al. [59] previously reported that a consistent running pattern can be determined using IMU data from 4 to 5 runs, whereas the current study only compared between-day reliability. Third, the study design was highly controlled, and data were collected in a laboratory setting. Although the treadmill was useful for maintaining speed, multiple studies have demonstrated differences in treadmill and overground running biomechanics [60,61], as well as differences in fatigue-related changes [9,57] so these results may not be generalizable to outdoor running. Fourth, the features investigated here were not comprehensive. Future studies should investigate the fatigue-related reliability of discrete features (e.g., ground contact time, step length) and features derived from the other instruments in the IMU (e.g., gyroscope). Fifth, the RPE at trial termination (14.1–16.8) was lower than others have previously defined as “fatigue” (i.e., 17) [62,63,64]. However, our RPEFT values were consistent with other studies reporting running fatigue [25,29] and represented a significant increase from the RPENF in each trial. Sixth, possible signal drift is an important limitation of IMUs. However, the current study only utilized the accelerometer readings, which are less affected by drift than the other sensors, and the temperature was kept relatively constant throughout each trial [65]. Finally, some trials were terminated due to volitional exhaustion, while others were terminated due to the stipulated maximum duration. These two “types” of trials may have generated different fatigue-related adjustments; although, this likely would have shown up as a main effect of the trial, as only one participant was able to complete 45 min for all trials.
In summary, to our knowledge, this is the first study to present evidence that fatigue-related alterations to biomechanics are consistent day-to-day. The results of the current study suggest that the commonly used features extracted from a CoM-mounted IMU during treadmill running are reliable in both NF and FT conditions, are not sensitive to slight (±5%) alterations in speed, and are generally sensitive to changes in fatigue state; although, features extracted from the ML axis should be interpreted with greater caution. Further study is needed to determine if these results are generalizable to overground running, or if larger changes in speed have different effects.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/s22114129/s1, Table S1: Mean values used for reliability and between-trial analysis.

Author Contributions

Conceptualization, C.R.v.R., M.J.M. and H.L.D.; methodology, C.R.v.R.; software, H.L.D., formal analysis, H.L.D., investigation, C.R.v.R.; resources, M.J.M., R.F., C.R.v.R. and H.L.D.; data curation, C.R.v.R. and H.L.D.; writing—original draft preparation, H.L.D. and R.F.; writing—review and editing, H.L.D., C.R.v.R., M.J.M. and R.F.; visualization, H.L.D.; supervision, M.J.M. and C.R.v.R.; project administration, M.J.M. and C.R.v.R.; funding acquisition, M.J.M., C.R.v.R. and R.F. All authors have read and agreed to the published version of the manuscript.

Funding

This work was partially funded by the NSERC CREATE Wearable Technology and Collaboration (We-TRAC) Training Program (Project No. CREATE/511166-229 2018) and two NSERC Discovery Grants (Project Nos. RGPIN/04079-2014, RGPIN-2018-06424) and supported by funding from an NSERC Vanier Canada Graduate Scholarship, an Alberta Innovates Graduate Studentship in Health Innovation, an Alberta Innovates Graduate Student Scholarship for Data-Enable Innovation, and a University of Calgary Eyes High Doctoral Recruitment Scholarship.

Institutional Review Board Statement

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Conjoint Health Research Ethics Board at the University of Calgary (REB-20-0111 on 29 July 2020).

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study.

Data Availability Statement

The data generated and/or analyses for to the current study are available from the corresponding author upon reasonable request.

Acknowledgments

The authors thank the students who assisted in data collection, Kate M. Sales and Oluwatimilehin Olamilekan Ajayi.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Mannini, A.; Trojaniello, D.; Cereatti, A.; Sabatini, A. A Machine Learning Framework for Gait Classification Using Inertial Sensors: Application to Elderly, Post-Stroke and Huntington’s Disease Patients. Sensors 2016, 16, 134. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  2. Chang, H.-C.; Hsu, Y.-L.; Yang, S.-C.; Lin, J.-C.; Wu, Z.-H. A Wearable Inertial Measurement System With Complementary Filter for Gait Analysis of Patients With Stroke or Parkinson’s Disease. IEEE Access 2016, 4, 8442–8453. [Google Scholar] [CrossRef]
  3. van den Tillaar, R.; Bhandurge, S.; Stewart, T. Detection of Different Throw Types and Ball Velocity with IMUs and Machine Learning in Team Handball. In Proceedings of the International Society of Biomechanics in Sport annual conference. International Society of Biomechanics in Sport (ISBS), Virtual Conference, Virtual, 21–25 July 2020; pp. 184–187. [Google Scholar]
  4. Allard, P.; Martinez, R.; Deguire, S.; Tremblay, J. In-Season Session Training Load Relative to Match Load in Professional Ice Hockey. J. Strength Cond. Res. 2022, 36, 486–492. [Google Scholar] [CrossRef] [PubMed]
  5. Kwapisz, J.R.; Weiss, G.; Moore, S.A. Activity recognition using cell phone accelerometers. ACM SIGKDD Explor. Newsl. 2011, 12, 74–82. [Google Scholar] [CrossRef]
  6. Preece, S.J.; Goulermas, J.Y.; Kenney, L.P.J.; Howard, D. A Comparison of Feature Extraction Methods for the Classification of Dynamic Activities From Accelerometer Data. IEEE Trans. Biomed. Eng. 2009, 56, 871–879. [Google Scholar] [CrossRef]
  7. Wundersitz, D.W.T.; Gastin, P.B.; Richter, C.; Robertson, S.J.; Netto, K.J. Validity of a Trunk-Mounted Accel-erometer to Assess Peak Accelerations during Walking, Jogging and Running. Eur. J. Sport Sci. 2015, 15, 382–390. [Google Scholar] [CrossRef]
  8. Benson, L.C.; Clermont, C.A.; Osis, S.T.; Kobsar, D.; Ferber, R. Classifying running speed conditions using a single wearable sensor: Optimal segmentation and feature extraction methods. J. Biomech. 2018, 71, 94–99. [Google Scholar] [CrossRef]
  9. Strohrmann, C.; Harms, H.; Kappeler-Setz, C.; Troster, G. Monitoring Kinematic Changes With Fatigue in Running Using Body-Worn Sensors. IEEE Trans. Inf. Technol. Biomed. 2012, 16, 983–990. [Google Scholar] [CrossRef]
  10. Kobsar, D.; Ferber, R. Wearable Sensor Data to Track Subject-Specific Movement Patterns Related to Clinical Outcomes Using a Machine Learning Approach. Sensors 2018, 18, 2828. [Google Scholar] [CrossRef] [Green Version]
  11. Op De Beéck, T.; Meert, W.; Schütte, K.; Vanwanseele, B.; Davis, J. Fatigue Prediction in Outdoor Runners Via Machine Learning and Sensor Fusion. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining ACM, London, UK, 19 July 2018; pp. 606–615. [Google Scholar]
  12. Schütte, K.H.; Maas, E.A.; Exadaktylos, V.; Berckmans, D.; Venter, R.; Vanwanseele, B. Wireless Tri-Axial Trunk Accelerometry Detects Deviations in Dynamic Center of Mass Motion Due to Running-Induced Fatigue. PLoS ONE 2015, 10, e0141957. [Google Scholar] [CrossRef] [Green Version]
  13. Clermont, C.A.; Benson, L.C.; Edwards, W.B.; Hettinga, B.A.; Ferber, R. New Considerations for Wearable Technology Data: Changes in Running Biomechanics During a Marathon. J. Appl. Biomech. 2019, 35, 401–409. [Google Scholar] [CrossRef] [PubMed]
  14. Hulteen, R.M.; Smith, J.J.; Morgan, P.J.; Barnett, L.M.; Hallal, P.C.; Colyvas, K.; Lubans, D.R. Global Partici-pation in Sport and Leisure-Time Physical Activities: A Systematic Review and Meta-Analysis. Prev. Med. 2017, 95, 14–25. [Google Scholar] [CrossRef] [PubMed]
  15. Videbæk, S.; Bueno, A.M.; Nielsen, R.O.; Rasmussen, S. Incidence of Running-Related Injuries Per 1000 h of running in Different Types of Runners: A Systematic Review and Meta-Analysis. Sports Med. 2015, 45, 1017–1026. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  16. van Gent, R.N.; Siem, D.; van Middelkoop, M.; van Os, A.G.; Bierma-Zeinstra, S.M.A.; Koes, B.W. Incidence and Determinants of Lower Extremity Running Injuries in Long Distance Runners: A Systematic Review. Br. J. Sports Med. 2007, 41, 469–480. [Google Scholar] [CrossRef] [Green Version]
  17. Lopes, A.D.; Hespanhol, L.C., Jr.; Yeung, S.S.; Costa, L.O.P. What Are the Main Running-Related Musculoskeletal Injuries? Sports Med. 2012, 42, 891–905. [Google Scholar] [CrossRef]
  18. Knobloch, K.; Yoon, U.; Vogt, P.M. Acute and Overuse Injuries Correlated to Hours of Training in Master Running Athletes. Foot Ankle Int. 2008, 29, 671–676. [Google Scholar] [CrossRef]
  19. Voloshin, A.S.; Mizrahi, J.; Verbitsky, O.; Isakov, E. Dynamic loading on the human musculoskeletal system—Effect of fatigue. Clin. Biomech. 1998, 13, 515–520. [Google Scholar] [CrossRef]
  20. Mizrahi, J.; Verbitsky, O.; Isakov, E.; Daily, D. Effect of fatigue on leg kinematics and impact acceleration in long distance running. Hum. Mov. Sci. 2000, 19, 139–151. [Google Scholar] [CrossRef]
  21. Christina, K.A.; White, S.C.; Gilchrist, L.A. Effect of Localized Muscle Fatigue on Vertical Ground Reaction Forces and Ankle Joint Motion during Running. Hum. Mov. Sci. 2001, 20, 257–276. [Google Scholar] [CrossRef]
  22. Grimston, S.K.; Zernicke, R.F. Exercise-Related Stress Responses in Bone. J. Appl. Biomech. 1993, 9, 2–14. [Google Scholar] [CrossRef]
  23. Hreljac, A. Etiology, Prevention, and Early Intervention of Overuse Injuries in Runners: A Biomechanical Perspective. Phys. Med. Rehabil. Clin. N. Am. 2005, 16, 651–667. [Google Scholar] [CrossRef] [PubMed]
  24. Derrick, T.R.; Dereu, D.; McLean, S.P. Impacts and kinematic adjustments during an exhaustive run. Med. Sci. Sports Exerc. 2002, 34, 998–1002. [Google Scholar] [CrossRef] [PubMed]
  25. Clansey, A.C.; Hanlon, M.; Wallace, E.S.; Lake, M.J. Effects of Fatigue on Running Mechanics Associated with Tibial Stress Fracture Risk. Med. Sci. Sports Exerc. 2012, 44, 1917–1923. [Google Scholar] [CrossRef] [PubMed]
  26. Hoenig, T.; Hamacher, D.; Braumann, K.-M.; Zech, A.; Hollander, K. Analysis of running stability during 5000 m running. Eur. J. Sport Sci. 2019, 19, 413–421. [Google Scholar] [CrossRef] [PubMed]
  27. Meardon, S.A.; Hamill, J.; Derrick, T.R. Running injury and stride time variability over a prolonged run. Gait Posture 2011, 33, 36–40. [Google Scholar] [CrossRef] [PubMed]
  28. Le Bris, R.; Billat, V.; Auvinet, B.; Chaleil, D.; Hamard, L.; Barrey, E. Effect of Fatigue on Stride Pattern Continu-ously Measured by an Accelerometric Gait Recorder in Middle Distance Runners. J. Sports Med. Phys. Fit. 2006, 46, 227–231. [Google Scholar]
  29. Marotta, L.; Buurke, J.; van Beijnum, B.-J.; Reenalda, J. Towards Machine Learning-Based Detection of Running-Induced Fatigue in Real-World Scenarios: Evaluation of IMU Sensor Configurations to Reduce Intrusiveness. Sensors 2021, 21, 3451. [Google Scholar] [CrossRef]
  30. Buckley, C.; O’Reilly, M.A.; Whelan, D.; Farrell, A.V.; Clark, L.; Longo, V.; Gilchrist, M.D.; Caulfield, B. Binary Classification of Running Fatigue Using a Single Inertial Measurement Unit. In Proceedings of the 14th International Conference on Wearable and Implantable Body Sensor Networks (BSN) IEEE, Eindhoven, The Netherlands, 9–12 May 2017; pp. 197–201. [Google Scholar]
  31. Tang, S.; Loh, W.P. Classification predictive models of running- and cycling-induced fatigue. AIP Conf. Proc. 2019, 2138, 050031. [Google Scholar] [CrossRef]
  32. Lindsay, T.R.; Yaggie, J.A.; McGregor, S.J. A wireless accelerometer node for reliable and valid measurement of lumbar accelerations during treadmill running. Sports Biomech. 2016, 15, 11–22. [Google Scholar] [CrossRef]
  33. Kawabata, M.; Goto, K.; Fukusaki, C.; Sasaki, K.; Hihara, E.; Mizushina, T.; Ishii, N. Acceleration patterns in the lower and upper trunk during running. J. Sports Sci. 2013, 31, 1841–1853. [Google Scholar] [CrossRef]
  34. McGregor, S.J.; Busa, M.A.; Yaggie, J.A.; Bollt, E.M. High Resolution MEMS Accelerometers to Estimate VO2 and Compare Running Mechanics between Highly Trained Inter-Collegiate and Untrained Runners. PLoS ONE 2009, 4, e7355. [Google Scholar] [CrossRef] [PubMed]
  35. Beneke, R.; von Duvillard, S.P. Determination of Maximal Lactate Steady State Response in Selected Sports Events. Med. Sci. Sports Exerc. 1996, 28, 241–246. [Google Scholar] [CrossRef] [PubMed]
  36. Billat, V.L.; Sirvent, P.; Py, G.; Koralsztein, J.-P.; Mercier, J. The Concept of Maximal Lactate Steady State: A Bridge Between Biochemistry, Physiology and Sport Science. Sports Med. 2003, 33, 407–426. [Google Scholar] [CrossRef] [PubMed]
  37. Faude, O.; Kindermann, W.; Meyer, T. Lactate Threshold Concepts: How Valid Are They? Sports Med. 2009, 39, 469–490. [Google Scholar] [CrossRef]
  38. Poole, D.C.; Jones, A.M. Oxygen Uptake Kinetics. In Comprehensive Physiology; Terjung, R., Ed.; Wiley: Hoboken, NJ, USA, 2012; pp. 933–996. [Google Scholar]
  39. Gerlach, K.E.; White, S.C.; Burton, H.W.; Dorn, J.M.; Leddy, J.J.; Horvath, P.J. Kinetic Changes with Fatigue and Relationship to Injury in Female Runners. Med. Sci. Sports Exerc. 2005, 37, 657–663. [Google Scholar] [CrossRef]
  40. Scharhag-Rosenberger, F.; Meyer, T.; Gäßler, N.; Faude, O.; Kindermann, W. Exercise at given percentages of VO2max: Heterogeneous metabolic responses between individuals. J. Sci. Med. Sport 2010, 13, 74–79. [Google Scholar] [CrossRef]
  41. Iannetta, D.; Inglis, E.C.; Pogliaghi, S.; Murias, J.M.; Keir, D.A. A ‘Step–Ramp–Step’ Protocol to Identify the Maximal Metabolic Steady State. Med. Sci. Sports Exerc. 2020, 52, 2011–2019. [Google Scholar] [CrossRef]
  42. Borg, G.A.V. Pyschophysical Bases of Perceived Exertion. Med. Sci. Sports Exerc. 1982, 14, 377–381. [Google Scholar] [CrossRef]
  43. Jones, A.M.; Doust, J.H. A 1% treadmill grade most accurately reflects the energetic cost of outdoor running. J. Sports Sci. 1996, 14, 321–327. [Google Scholar] [CrossRef]
  44. Tundo, M.D.; Lemaire, E.; Baddour, N. Correcting Smartphone Orientation for Accelerometer-Based Analysis. In Proceedings of the IEEE International Symposium on Medical Measurements and Applications (MeMeA) IEEE, Gatineau, QC, USA, 4–5 May 2013; pp. 58–62. [Google Scholar]
  45. Kobsar, D.; Osis, S.T.; Phinyomark, A.; Boyd, J.E.; Ferber, R. Reliability of gait analysis using wearable sensors in patients with knee osteoarthritis. J. Biomech. 2016, 49, 3977–3982. [Google Scholar] [CrossRef]
  46. Pirttikangas, S.; Fujinami, K.; Nakajima, T. Feature Selection and Activity Recognition from Wearable Sensors; International Symposium on Ubiquitious Computing Systems; Springer: Berlin, Germany, 2006; pp. 516–527. [Google Scholar]
  47. Pla, G.A.; Hollville, E.; Schütte, K.; Vanwanseele, B. The Use of a Single Trunk-Mounted Accelerometer to Detect Changes in Center of Mass Motion Linked to Lower-Leg Overuse Injuries: A Prospective Study. Sensors 2021, 21, 7385. [Google Scholar] [CrossRef]
  48. Schütte, K.H.; Seerden, S.; Venter, R.; Vanwanseele, B. Influence of outdoor running fatigue and medial tibial stress syndrome on accelerometer-based loading and stability. Gait Posture 2018, 59, 222–228. [Google Scholar] [CrossRef] [PubMed]
  49. Richman, J.S.; Moorman, J.R. Physiological Time-Series Analysis Using Approximate Entropy and Sample Entropy. Am. J. Physiol.-Heart Circ. Physiol. 2000, 278, H2039–H2049. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  50. Yentes, J.M.; Hunt, N.; Schmid, K.K.; Kaipust, J.P.; McGrath, D.; Stergiou, N. The Appropriate Use of Ap-proximate Entropy and Sample Entropy with Short Data Sets. Ann. Biomed. Eng. 2013, 41, 349–365. [Google Scholar] [CrossRef]
  51. Preatoni, E.; Ferrario, M.; Donà, G.; Hamill, J.; Rodano, R. Motor variability in sports: A non-linear analysis of race walking. J. Sports Sci. 2010, 28, 1327–1336. [Google Scholar] [CrossRef] [Green Version]
  52. Koo, T.K.; Li, M.Y. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J. Chiropr. Med. 2016, 15, 155–163. [Google Scholar] [CrossRef] [Green Version]
  53. Sekine, M.; Tamura, T.; Yoshida, M.; Suda, Y.; Kimura, Y.; Miyoshi, H.; Kijima, Y.; Higashi, Y.; Fujimoto, T. A gait abnormality measure based on root mean square of trunk acceleration. J. Neuroeng. Rehabil. 2013, 10, 118. [Google Scholar] [CrossRef] [Green Version]
  54. Schütte, K.H.; Seerden, S.; Venter, R.; Vanwanseele, B. Fatigue-Related Asymmetry and Instability during a 3200-m Time-Trial Performance in Healthy Runners. In Proceedings of the 34 International Conference of Biomechanics in Sport, Tsukuba, Japan, 18–22 July 2016; p. 4. [Google Scholar]
  55. Winter, S.C.; Gordon, S.; Brice, S.; Lindsay, D.; Barrs, S. A Multifactorial Approach to Overuse Running Injuries: A 1-Year Prospective Study. Sports Health Multidiscip. Approach 2020, 12, 296–303. [Google Scholar] [CrossRef]
  56. Lipsitz, L.A.; Goldberger, A.L. Loss of “complexity” and aging. Potential applications of fractals and chaos theory to senescence. JAMA 1992, 267, 1806–1809. [Google Scholar] [CrossRef]
  57. Williams, K.R.; Snow, R.; Agruss, C. Changes in Distance Running Kinematics with Fatigue. Int. J. Sport Biomech. 1991, 7, 138–162. [Google Scholar] [CrossRef]
  58. Benson, L.C.; Ahamed, N.U.; Kobsar, D.; Ferber, R. New considerations for collecting biomechanical data using wearable sensors: Number of level runs to define a stable running pattern with a single IMU. J. Biomech. 2019, 85, 187–192. [Google Scholar] [CrossRef] [PubMed]
  59. Ahamed, N.U.; Benson, L.C.; Clermont, C.A.; Pohl, A.J.; Ferber, R. New Considerations for Collecting Biome-chanical Data Using Wearable Sensors: How Does Inclination Influence the Number of Runs Needed to Determine a Stable Running Gait Pattern? Sensors 2019, 19, 2516. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  60. Benson, L.C.; Clermont, C.A.; Ferber, R. New Considerations for Collecting Biomechanical Data Using Wearable Sensors: The Effect of Different Running Environments. Front. Bioeng. Biotechnol. 2020, 8, 86. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  61. Sinclair, J.; Richards, J.; Taylor, P.J.; Edmundson, C.J.; Brooks, D.; Hobbs, S.J. Three-dimensional kinematic comparison of treadmill and overground running. Sports Biomech. 2013, 12, 272–282. [Google Scholar] [CrossRef] [Green Version]
  62. Dierks, T.A.; Manal, K.T.; Hamill, J.; Davis, I.S. Proximal and Distal Influences on Hip and Knee Kinematics in Runners With Patellofemoral Pain During a Prolonged Run. J. Orthop. Sports Phys. Ther. 2008, 38, 448–456. [Google Scholar] [CrossRef]
  63. Benson, L.C.; O’Connor, K.M. The Effect of Exertion on Joint Kinematics and Kinetics During Running Using a Waveform Analysis Approach. J. Appl. Biomech. 2015, 31, 250–257. [Google Scholar] [CrossRef]
  64. Maas, E.; De Bie, J.; Vanfleteren, R.; Hoogkamer, W.; Vanwanseele, B. Novice runners show greater changes in kinematics with fatigue compared with competitive runners. Sports Biomech. 2017, 17, 350–360. [Google Scholar] [CrossRef]
  65. Burland, J.P.; Outerleys, J.B.; Lattermann, C.; Davis, I.S. Reliability of wearable sensors to assess impact metrics during sport-specific tasks. J. Sports Sci. 2021, 39, 406–411. [Google Scholar] [CrossRef]
Sensors 22 04129 g001 550
Figure 1. ICC(2,k) values for all analyses in the vertical (VT) axis of the acceleration signal. MM = comparison performed between first (MLSS1) and second (MLSS2) trials at MLSS speed, ALL = comparison performed between all four trials; NF = non-fatigued state, FT = fatigued state, SD = standard deviation, RMS = root mean square, RMSR = ratio of root mean square, SE = sample entropy.
Figure 1. ICC(2,k) values for all analyses in the vertical (VT) axis of the acceleration signal. MM = comparison performed between first (MLSS1) and second (MLSS2) trials at MLSS speed, ALL = comparison performed between all four trials; NF = non-fatigued state, FT = fatigued state, SD = standard deviation, RMS = root mean square, RMSR = ratio of root mean square, SE = sample entropy.
Sensors 22 04129 g001
Sensors 22 04129 g002 550
Figure 2. ICC(2,k) values for all analyses in the mediolateral (ML) axis of the acceleration signal.
Figure 2. ICC(2,k) values for all analyses in the mediolateral (ML) axis of the acceleration signal.
Sensors 22 04129 g002
Sensors 22 04129 g003 550
Figure 3. ICC(2,k) values for all analyses in the anterior–posterior (AP) axis of the acceleration signal.
Figure 3. ICC(2,k) values for all analyses in the anterior–posterior (AP) axis of the acceleration signal.
Sensors 22 04129 g003
Sensors 22 04129 g004 550
Figure 4. ICC(2,k) values for all analyses in the resultant (RES) of the acceleration signal.
Figure 4. ICC(2,k) values for all analyses in the resultant (RES) of the acceleration signal.
Sensors 22 04129 g004
Table
Table 1. Values for trial variables. MLSS1 = first trial at speed at MLSS, MLSS2 = second trial at speed at MLSS, F = trial at 5% faster than speed at MLSS, S = trial at 5% slower than speed at MLSS, NF = non-fatigued state, FT = fatigued state.
Table 1. Values for trial variables. MLSS1 = first trial at speed at MLSS, MLSS2 = second trial at speed at MLSS, F = trial at 5% faster than speed at MLSS, S = trial at 5% slower than speed at MLSS, NF = non-fatigued state, FT = fatigued state.
MLSS1MLSS2FS
Treadmill speed (m/s) 3.35 ± 0.43.35 ± 0.43.53 ± 0.43.17 ± 0.4
Number of strides used (NF)376.8 ± 119.2390.5 ± 69.5411.5 ± 127395.5 ± 109.5
Number of strides used (FT)375.5 ± 71.5387.5 ± 53.5366 ± 77362.5 ± 61
Trial time (min) *39.8 ± 6.6 @37.7 ± 7.3 @28.8 ± 6.942.0 ± 5.6 @
RPENF11.5 ± 1.711.4 ± 1.812.5 ± 1.511.0 ± 1.5
RPEFT *15.6 ± 1.6 #,^15.3 ± 1.6 #,^16.8 ± 1.3 †,‡,#,^14.1 ± 1.7 ^
ΔRPE4.1 ± 2.03.9 ± 2.04.3 ± 1.7 3.1 ± 1.6
* Significant main effect (p < 0.05); @ significantly greater than F trial (p < 0.05); † significantly greater than MLSS1 trial (p < 0.05); ‡ significantly greater than MLSS2 trial (p < 0.05); # significantly greater than S trial (p < 0.05), ^ significantly greater than NF state (p < 0.05).
Table
Table 2. Main effects for fatigue state in each axis of the acceleration signal. Direction of change indicates differences from NF to FT. Significant p-values in bold. VT = vertical, ML = mediolateral, AP = anterior–posterior, RES = resultant; MM = comparison performed between MLSS1 and MLSS2, ALL = comparison performed between all four trials. RMS = root mean square, RMSR = ratio of root mean square, SE = sample entropy.
Table 2. Main effects for fatigue state in each axis of the acceleration signal. Direction of change indicates differences from NF to FT. Significant p-values in bold. VT = vertical, ML = mediolateral, AP = anterior–posterior, RES = resultant; MM = comparison performed between MLSS1 and MLSS2, ALL = comparison performed between all four trials. RMS = root mean square, RMSR = ratio of root mean square, SE = sample entropy.
AxisFeatureMMALL
p ValueDirection of Changep ValueDirection of Change
VTMean0.022-0.009-
St. deviation0.032+0.006+
25th percentile0.091 0.003-
Median0.014+0.001+
75th percentile0.741 0.479
Max0.003+0.002+
Min0.281 0.037-
RMS0.075 0.023+
RMSR<0.001-<0.001-
SE0.001+0.001+
MLMean0.098 0.115
St. deviation<0.001+<0.001+
25th percentile0.152 0.025-
Median0.207 0.123
75th percentile0.492 0.212
Max<0.001+<0.001+
Min0.004-<0.001-
RMS<0.001+<0.001+
RMSR<0.001+<0.001+
SE<0.001-<0.001-
APMean0.001-<0.001-
St. deviation0.007+<0.001+
25th percentile0.002-<0.001-
Median0.699 0.241
75th percentile0.157 0.027+
Max0.123 0.514
Min0.009-<0.001-
RMS0.008+0.001+
RMSR0.078 0.028+
SE0.136 0.07
RESMean<0.001+<0.001+
St. deviation0.016+0.009+
25th percentile0.001+0.06+
Median0.169 0.005+
75th percentile0.38 0.346
Max0.003+0.004+
Min0.136 0.601
RMS0.001+<0.001+
SE0.017+0.01+
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Dimmick, H.L.; van Rassel, C.R.; MacInnis, M.J.; Ferber, R. Between-Day Reliability of Commonly Used IMU Features during a Fatiguing Run and the Effect of Speed. Sensors 2022, 22, 4129. https://doi.org/10.3390/s22114129

AMA Style

Dimmick HL, van Rassel CR, MacInnis MJ, Ferber R. Between-Day Reliability of Commonly Used IMU Features during a Fatiguing Run and the Effect of Speed. Sensors. 2022; 22(11):4129. https://doi.org/10.3390/s22114129

Chicago/Turabian Style

Dimmick, Hannah L., Cody R. van Rassel, Martin J. MacInnis, and Reed Ferber. 2022. "Between-Day Reliability of Commonly Used IMU Features during a Fatiguing Run and the Effect of Speed" Sensors 22, no. 11: 4129. https://doi.org/10.3390/s22114129

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop