Data from: Human walking biomechanics on sand substrates of varying foot sinking depth
Data files
Oct 17, 2024 version files 765.75 MB
-
Additional_data.zip
309.22 KB
-
EMG.zip
599.62 MB
-
Kinematics.zip
165.79 MB
-
README_Sand.txt
16.80 KB
-
README.md
17.29 KB
Abstract
Our current understanding of human gait is mostly based on studies using hard, level surfaces in a laboratory environment. However, humans navigate a wide range of different substrates every day, which incur varied demands on stability and efficiency. Several studies have shown that when walking on natural compliant substrates there is an increase in energy expenditure. However, these studies report variable changes to other aspects of gait such as muscle activity. Discrepancies between studies exist even within substrate types (e.g. sand), which suggests that relatively ‘fine-scale’ differences in substrate properties exert quantifiable influences on gait mechanics. In this study, we compare human walking mechanics on a range of sand substrates that vary in overall foot sinking depth. We demonstrate that variation in the overall sinking depth in sand is associated with statistically significant changes in joint angles and spatiotemporal variables in human walking but exerts relatively little influence on pendular energy recovery and muscle activations. Significant correlated changes between gait metrics are frequently recovered, suggesting a degree of coupled or mechanistic interaction in their variation within and across substrates. However, only walking speed (and its associated spatiotemporal variables) correlate frequently with absolute foot sinkage depth within individual sand substrates, but not across them. This suggests a causative relationship between walking speed and foot sinkage depth within individual sand substates is not coupled with systematic changes in joint kinematics and muscle activity in the same way as is observed across sand substrates.
Gait analysis data (motion capture and muscle activation data) for a set of healthy young adults (n=21) walking at self-selected speeds over four different surfaces. One hard surface (level, hard floor) and three compliant
substrates: play sand, wet build sand and dry build sand.
Associated publication: Grant, B. F., Charles, J. P., D’Août, K., Falkingham, P. L., & Bates, K. T. (2024). Human walking biomechanics on sand substrates of varying foot sinking depth. The Journal of experimental biology. doi:10.1242/jeb.246787
For all code and data, participants are allocated numbers based on our overall dataset (combined with our previous publication (Grant et al, 2022)).
Grant, B., Charles, J., Geraghty, B., Gardiner, J., D’Août, K., Falkingham, P.L. and Bates, K.T., 2022. Why does the metabolic cost of walking increase on compliant substrates?. Journal of the Royal Society Interface, 19(196), doi:10.1098/rsif.2022.0483
Participant numbers include 1, 2, 4, 7 , 9, 10, 15, 16, 19, 23 , 27, 28 (all took part in previous foam study) and 32-40. In the associated publication, they are referred to as participants 1-21.
In the additional data folder, there is an file called Sand_subject_number_match.xlsx which states what the original/overall dataset participant number is compared to the new/this publication participant number.
List of all software used
12-camera Qualisys Oqus 7 motion capture system (Qualisys Inc., Götenborg, Sweden) (3D kinematics - data collection)
Qualisys Track Manager 2.15 (2017) (3D kinematics - data collection and export to .c3d format for greater data accessibility)
Delsys EMGworks (EMG - data collection and export as .csv files)
Visual 3D v.6 (C-Motion Inc., Germantown, MD, USA) (Kinematics data analysis, export gait events, joint angles and spatiotemporal variables)
MATLAB v.2019a - v. 2023a (Mathworks, Natick, USA) (data processing, data analysis)
R v. 3.6.1 - 4.3.0 (R Core Team)
List and description of data included
Kinematics.zip folder
Includes all raw 3D kinematics as .c3d files for every subject (1-21) as well as .txt text files showing the XYZ co-ordinates for each marker.
Motion capture data includes a full-body 69 marker set (see appendix at bottom of README file for location list).
This data was recorded at 200Hz using a 12-camera Qualisys Oqus 7 motion capture system (Qualisys Inc., Götenborg, Sweden) with Qualisys Track Manager (https://www.qualisys.com/software/qualisys-track-manager/). For the publication, data labelling was done using Qualisys Track Manager and then imported into C-Motion Visual3D for data analysis (https://c-motion.com/). These applications require a subscription to use. For free alternatives, please see: https://www.c3d.org/c3dapps.html. The authors would recommend using Mokka: Motion kinematic and kinetic analyser which is an open-source and cross-platform software (https://biomechanical-toolkit.github.io/mokka/).
EMG.zip folder
Includes all raw EMG files (.csv) that were recorded using surface electromyography for the lower limbs (all lower limb muscles used were from the left side due to availability of sensors) using wireless Trigno EMG (Delsys, MA, USA) with Delsys EMGworks (https://delsys.com/emgworks/) at a sampling rate of 1110 Hz. Muscles measured were: Tibialis Anterior (TA), Rectus Femoris (RF), Lateral Gastrocnemius (LG), Medial Gastrocnemius (LG), Soleus (SOL), Vastus Lateralis (VL), Vastus Medialis (VM), Biceps Femoris longus (BFL).
Also included in the files are measurements from sensors on the torso, which were not included in the associated publication due to concerns regarding the reliability of this data (left and right External Obliques (L/REO), Internal Obliques (L/RIO) and Erector Spinae (L/RES)).
The raw EMG data are included as .csv files for 20 participants (not including participant 19 in original files / participant 9 in manuscript (due to an issue with recording)).
Within the EMG.zip folder, is a .xlsx file which states which .csv raw EMG data files correspond to which trials/substrates.
There is also an additional data folder which includes important data post-processing which were used for analysis. These files were used as input tables for further analysis and the linear mixed effect statistical models (LMMs) and correlation analyses. For the LMMs, the specified fixed effects are ‘substrate’, ‘speed’ and ‘sex’ and random effects set as ‘participants’.
Additional data folder
The files presented in the additional data folder (and corresponding coding files) include some or all of the following common variables which are detailed below.
For further information on the methodology and details on some of these variables and calculations, please see publications (Grant et al, 2022; Grant et al, 2024).
Variables description
Participant = Participant number as determined by full data set allocation (1-40)
Substrate = Hard floor, play sand, build sand wet, build sand dry
Trial number = 5 individual trials on each sand substrate and 3 on the hard floor substrate
Lowest sinking depth values (cm) (lowest z-position of markers on the participants foot in relation to the substrate). Lab is calibrated as Z=0 and markers on the end of each of the sand walkways are used to calculate the Z-value of the sand substrates.
Calcaneus (CALC) = Lowest z-position of the marker on the calcaneus (cm)
Hallux = Lowest z-position of the marker on the hallux (cm)
Speed (m/s) = Average speed per trial calculated based on the 3D position of the kinematic markers throughout the trial (XYZ)
Spatiotemporal variables. Gait events are marked in Visual3d to state timings of left and right heel-strike and toe-off in the participant’s gait cycle. These gait events and the 3D positions of the markers are used to calculate a range of spatiotemporal variables. These variables include stride length (m), stride width (m), cycle time (gait cycle)(s), stance time (duration of the time between heel-strike and toe-off of same foot)(s), swing time (duration of the time between toe-off and next heel-strike of the same foot)(s), double limb support time (duration of time when both feet are in contact with the ground)(s), duty factor (fraction of the duration of a stride for which reach foot remains on the ground).
Kinematic / joint angle variables. These are calculated in Visual3D based on the 3D position of the kinematic markers assigned to different segments of the body (feet, shanks, thighs, upper arms, forearms (all of these bilaterally) and head, trunk and pelvis). Joint angles (degrees) throughout the trial were exported and maximum range of motions were calculated based on the max and min angles measured for each joint for the ankle, knee and hip joint throughout the trial.
Mechanical energy exchange variables. From the labelled 3D kinematic data and assignmed segments in Visual3D, the position of the segmental CoM positions in relation to the laboratory were calculated using mechanical principle patterns. These were then exported to Matlab to calculate the gravitational potential energy (Epot)(J), kinetic energy (Ekin)(J) and total mechanical energy (Etot)(J). From this, the energy exchange variables were calculated.
R = Recovery of mechanical energy (expressed as a percentage)(%)
RA = Relative amplitude (of potential and kinetic energy)
CO = Congruity (the time when potential energy and kinetic energy are moving in the same direction)(%)
Normalised EMG values (nEMG) = Processed EMG values- raw EMG signals were high pass filtered at 12Hz using a second-order Butterweoth filter, full-wave rectified and cropped to stride. For each muscle, th data were normalised to maximum amplitude during all walking trials for that participant.
Integrated EMG values (iEMG) = Integrated values of the nEMG values were calculated for each stride.
Included files in the additional data folder
Sand_all_variables_correlations.xlsx = This file includes data from all variables used for the correlation analyses (average value per trial). Included data is participant number, substrate type, lowest z-value (sinking depth) of calcaneus and hallux, speed, stride length, stride width, cycle time, stance time, swing time, double limb support time, duty factor, ankle max Range of Motion (RoM), Knee max RoM, Hip max RoM, integrated EMG values for TRA, RF, LG, MG, SOL, VL, VM, BFL and R (recovery of mechanical energy)
Sand_EE_table.xlsx = This file includes data from the mechanical energy exchange variables- R, RA and CO as well as participant number, sex and substrate.
Sand_control_depth_measurements_cm.csv = Final values taken before data collection to determine the suitability of the sand substrates- these include measurements at the heel, midfoot and forefoot of each footprint created by the lead investigator (and the mean of these values), and shear strength and force gauge measurements (5 measurements were taken throughout the length of the sand walkways). Empty cells are due to more footprint measurements recorded (e.g. 6 footprints) than for the shear strength and force gauge (5 measurements). Shear strength and force gauge measurements do not necessarily match up with each footprint in the same row but were taken from between footprint positions.
Sand_IVT_EMG_table.csv = iEMG values for all muscles per trial. Also included is participant number, sex, substrate and average speed.
Sand_JA_ROM.xlsx = RoM for ankle, knee and hip joint per trial. Also included is participant number, sex, substrate and average speed.
Sand_match_kinematics_EMG_data.xlsx = This includes details of which EMG .csv file corresponds to which kinematics .c3d file (due to an issue with synchronous labelling) and substrate type.
Sand_meanfootdepth_speed.xlsx = Mean foot sinking depth values on each sand substrate (as calculated by z-position of the hallux and calcaneus markers) and average walking speed for each trial. Also included is participant number.
Sand_spatiotemporal_variables.xlsx = All spatiotemporal variables as stated above, speed, substrate, sex and participant number. Empty cells due to some variables only calculating one value as an average per stride, or average per trial whereas some variables calculate per step. This file includes all available data for each variable.
Additional code folder
Included in the code folder are all the main code needed for data analysis and statistical tests for spatiotemporal data, EMG, energy exchange, SPM analysis on joint angles and correlation analyses. For the analysis and results presented in the associated paper, there were some additional data processing and code used.
.m files were created using MATLAB\
.R files were created using RStudio
For .R files.\
Raincloud plot scripts require these packages:
readxl, ggplot2, dplyr, ggpubr, gridExtra, grid, ggdist
LMM’s require these packages:
readxl, ggplot2, lme4, lmerTest, sjPlot, gridExtra
ANOVA require these packages:
readxl, ggplot2, tidyverse, remotes, report, multcomp, sjPlot
Correlation analysis require these packages:
readxl, ggplot2, ggcorrplot, ggstatsplot, lares, dplyr, Hmisc, corrplot
For MATLAB, additional packages and functions are specified below with the corresponding code.
Included files in the additional code folder:
ShadedErrorBar.m
- Additional function that allows user to create shaded error bars (e.g. standard deviations, confidence intervals) as used to create figures in below matlab scripts.
1. Spatiotemporal variables
1.1 Spatiotemporal_raincloudplots.R
- creates raincloud plots for each spatiotemporal variable and creates combined figure
1.2 LMM_Spatiotemporal.R
- runs LMMs for each spatiotemporal variable with participant as a random effect and substrate, sex and speed as fixed effects
2. Mechanical Energy exchange
2.1 EE_analysis.m
- calculates potential energy (Epot), kinetic energy (Ekin) and total energy (Etot) based on the 3D centre of mass positions exported from V3D
- calculates the mechanical energy exchange variables R, RA and CO.
2.2 EE_raincloudplots.R
- creates raincloud plots for each energy exchange variable (R, RA, CO) and creates combined figure
2.3 LMM_E.R
- runs LMMs on the mechanical energy exchange variables R, RA and CO with participant as a random effect and substrate, speed and sex as fixed effects
3. Joint Angles
SPM = Statistical Parametric Mapping. Anknowledgements to Todd Pataky et al. for use of the spm1d package.
spm1d is a package used for one-dimensional Statistical Parametric Mapping. It uses random field theory to make statistical inferences regarding registered (normalised) sets of 1D measurements.
For further information see https://spm1d.org/ and https://spm1d.org/_downloads/793453c1fd63a0f274ac3bd67fbd99cb/Slides-ISB2017.pdf
3.1 CI_SPM_analysis.m
- allows for statistical testing on joint angles throughout the whole gait cycle
- Calculates SPM stats (paired t-tests with bonferroni corrections) between all four substrates for ankle, knee and hip angles in the sagittal plane.
- Creates SPM figures over the whole gait cycle (mean and confidence intervals)
For CI_SPM_analysis to work, you also need to have the spm1dmatlab-master.zip folder (available in additional code folder) installed in your directory.
Joint Angles calculated as maximum joint range of motion
3.2 LMM_Joint_ROM.R
- runs LMMs on the maximum joint range of motion values with participant as a random effect and substrate, speed and sex as fixed effects
4. EMG
Code names were labelled in order of alphabet (A - C). Further analysis (after A-C and some additional processing were labelled in order of alphabet (AA - CC)).
4.1 AB_EMG_Hilbert_Crop.m
(combines A and B)
- Crops .csv files
- Hilbert wavelet transform data
- Filter and smooth emg data
- Requires additional filter functions below (4.1.1 butfilt.m, 4.1.2 envelop_hilbert.m snf 4.1.3 emg_filter.m)
4.1.1 butfilt.m
Butter filter function (required by 4.1 AB_EMG_Hilbert_Crop.m)
4.1.2 envelop_hilbert.m
Hilbert wavelet envelope function (required by 4.1 AB_EMG_Hilbert_Crop.m)
4.1.3 emg_filter.m
Additional filter function (required by 4.1 AB_EMG_Hilbert_Crop.m)
4.2 C_EMG_Stride.m
- crops filtered data to gait cycles (using gait events exported from V3D)
-- Participant data was then combined for further analyses
4.3 AA_normalise_EMG_figs.m
- calculate normalised EMG values
- create figures using normalised data (mean and std)
4.3 BB_CI_normalised_EMG.m
- create figures using normalised data (confidence intervals)
4.4 CC_integrated_normalised_values.m
- calculate integrated emg values
- create bar chart using integrated and normalised data
4.5 LMM_iEMG.R
- runs LMMs on the integrated normalised EMG values with participant as a random effect and substrate, speed and sex as fixed effects.
5. Footprint depths
5.1 Footprint_depth_raincloudplots.R
- creates raincloud plots for minimum calcaneus and hallux z positions and creates combined figure
5.2 Footprint_depth_ANOVA.R
- performs ANOVA on footprint depth values
6. Correlation analyses
6.1 Correlation_gait_variables.R
- runs correlations on gait variables for each substrate individually and all substrates combined and ordered by First Principal Component (FPC).
6.2 Correlation_foot_sinking_depth.R
- runs correlations on foot sinking depths (lowest z-positions for calcaneus and hallux) and gait variables for each substrate individually and all substrates combined
Appendix
Appendix 1: Locations of the 69 3D kinematic reflective markers:
Trunk: 5 markers
LACR RACR Acromion (left and right)
JUG Jugular notch
XYPH Xyphisternal joint
C7 Spine of the 7th cervical vertebra
Head: 4 markers
HEADF HEADB HEADL HEADR Band with four markers (1 front, 1 back, 2 side)
Pelvis: 6 markers
LASIS RASIS Anterior superior iliac spine
LPSI RPSI Posterior superior iliac spine
LICR RICR Iliac crest tubercle
Upper leg: 7 markers (x2)
LGTR RGTR Greater trochanter
LLEPI RLEPI Lateral epicondyle
LMEPI RMEPI Medial epicondyle
LTHPA LTHPP LTHDA LTDP
RTHPA RTHPP RTHDA RTDP Left and Right THIGH plates: proximal/distal and anterior/posterior
Lower leg: 8 markers (x2)
LFIB RFIB Fibular head
LLMAL RLMAL Lateral malleolus
LMMAL RMMAL Medial malleolus
LSHPA LSHPP LSHDA LSHP
RSHPA RSHPP RSHDA RSHDP Left and Right SHANK plates: proximal/distal and anterior/posterior
LTUB RTUB Tibial tuberosity
Foot: 10 markers (x2)
LLCA RLCA Lateral calcaneus
LCAL RCAL Back of Heel
LSTL RSTL Sustentaculum tail
LNAV RNAV Navicular
LP1M RP1M Metatarsal I base
LP5M RP5M Metatarsal V base
LD1M RD1M Metatarsal I head
LD5M RD5M Metatarsal V head
LTOE RTOE Between metatarsal I and II heads
LHALL RHALL Hallux (tip)
Arms: 2 markers (x2)
LHUM RHUM Lateral humeral epicondyle
LULNA RULNA Ulnar head (distal epiphysis)
Matlab - data processing
R - statistics
Qualisys Track Manager - Kinematic data collection
Visual3D - Kinematic data processing
Delsys - EMG data collection