Gait analysis tools for use in horses have been the subject of interest in the equine industry for many years. Use of such tools for the purposes of lameness investigation and objective assessment of gait quality is quickly becoming widespread in equine practice. For example, the aim of dressage is the development of movement quality, which is partly defined by symmetry, rhythm, and regularity in all gaits1 and could be objectively assessed with gait analysis tools. Quality of movement has also been studied as a means of determining the gait quality potential of young horses or the expected performance of horses in training.2
Historically, objective gait analysis in horses was largely limited to research laboratories because of the high cost of and stringent requirements for use of the necessary equipment. To capture steady and representative movement during overground locomotion with optical systems for analysis requires the collection of data from 3 to 5 consecutive strides.3 To do so necessitates the availability of a large amount of space with sophisticated infrastructure. Alternatively, the use of treadmills was considered a promising option for capture of many strides with reductions in space and equipment requirements. However, treadmills significantly alter the way a horse moves and have specific limitations in gait evaluation.4,5 In the face of these limitations, the use of trunk-mounted IMUs for identification of gait abnormalities is becoming more popular; because of their ability to capture many strides, affordability, and accuracy6 in addition to their portability, IMUs appear to be superior to the use of optical systems in overground conditions. The use of IMUs for clinical lameness assessments in horses has been reported.7,8 In recent years, several head-, pelvis-, and trunk-mounted IMUs have been developed and marketed to evaluate the symmetry of movement in horses as an aid in lameness diagnosis in field situations.7,8 In comparison, extremity-mounted sensors provide spatiotemporal variables of the extremities and allow the analysis of extremity motion in horses.9,10
The objective of the study reported here was to determine the repeatability of gait variables measured by use of extremity-mounted IMUs in nonlame horses during trotting under controlled conditions of treadmill exercise. One of the output variables of interest was limb phasing.
Limb phasing in horses was conceptually described initially by Hildebrand11 on the basis of the Muybridge experiments.12 Limb phasing in horses and humans has been described,9,13–15 and the system of IMUs investigated has patent validation for both species.14 Limb phasing is based on the concept that each limb has a similar cyclic or sinusoidal motion, and the system determines the temporal relationships among limbs through signal processing and a cross-correlation approach.13,14 The intent of our investigation was to acquire kinematic data from extremity-mounted IMUs and determine the repeatability of measurements over time for selected gait variables as a step toward determining the units’ clinical applicability for equine gait analysis.
Materials and Methods
Horses
The study was approved by the Swiss Vaud cantonal authority (VD protocol 3041) and followed institutional guidelines for humane animal treatment. This prospective study was performed in a period of 3 weeks.
Ten horses were randomly selected from a herd of Franches Montagne stallions at the Haras National Stud Center in Avenches, Switzerland. Horses were nonlame as determined by a veterinary evaluation, and were excluded if they were visually lame, unhealthy, or receiving medication. The horses used in the experiment had a mean ± SD body mass of 537.5 ± 28.3 kg and a height (from the ground to the highest point of the shoulders [withers]) of 157.4 ± 1.4 cm. The selected horses, which ranged from 6 to 19 years of age, were in good physical condition and had good conformation. All horses were preconditioned to work on a treadmilla according to standard procedure.16
Inertial measurement sensor system
Data collection was performed with commercially available IMUs.b An IMU records 6 degrees of freedom (outputs of acceleration and angular rotation along the 3 orthogonal axes). Each unit was mounted into a brushing boot or tibia strap (Figure 1). As described in the patent, each IMU (total mass, 54 g; dimensions, 73 × 36 × 19 mm) contained a tri-axial 5 × g accelerometer and 3 single-axis, 1,200-degrees/s gyroscopes followed by anti-aliasing hardware filters with a cutoff frequency of approximately 50 Hz.14 Data sampling was done with a 12-bit analog-to-digital converter at a frequency of 102.4 Hz.14 Each IMU was factory set to within 1 ppm (3.6 ms/h) of a reference time, achieving relative drift between units after synchronization of < 10 ms/h.14 Each IMU was time stamped and synchronized at the start of each horse's trials by a simultaneous pulse sent to the respective units with specifically written software.c This software was also used for the recording and automated processing of the data. Each IMU contained a precision clock and a memory storage service card (SD card).
Photograph of 1 of 10 horses instrumented in readiness for use in a study to determine the repeatability of gait variables measured by use of an inertial measurement sensor system involving extremity-mounted IMUs in nonlame horses during trotting under controlled conditions of treadmill exercise. The system's 6 sensors are in place. An IMU (in a brushing boot with a small custom-fitted pouch) is located on the distolateral aspect of each metacarpal and metatarsal region. An IMU is also located (by means of a custom-made soft strap) on the distolateral aspect of each tibia, just dorsal to the calcanean tendon.
Citation: American Journal of Veterinary Research 78, 9; 10.2460/ajvr.78.9.1011
Variables of interest
The system used was capable of simultaneously capturing the entire motion cycle of the metacarpal and metatarsal regions and distal portions of the horses’ tibias. The sensors were programmed to determine the cycle associated with each stride via a proprietary algorithm. To define some aspects of the kinematic characteristics of trotting, the IMU system reported 15 temporal variables and 19 spatial variables. The temporal variables included stride duration and calculated variables of stride frequency and length, trotting speed (predetermined), limb phasing (defined as the percentage difference of time within a stride of each extremity [left and right forelimb and right hind limb, relative to the left hind limb]); the time point within a stride at which maximal protraction and retraction of the right and left third metacarpal and metatarsal bones occurred was expressed as a percentage of the stride duration. The diagonal asymmetry (%) was calculated as the difference between the diagonal limb phasing timing couplets as follows: (left forelimb-right hind limb) – (right forelimb-left hind limb), where the left hind limb was used as the reference limb and assigned a value of zero. The spatial variables included sagittal ROM (ie, sagittal angles) of the left and right tarsal joints and segment angles of the left and right metacarpi, metatarsi, and tibias and the coronal ROM of the left and right metacarpi, metatarsi, and tibias. The symmetry (%) of each segment or joint ROM was calculated as the difference between the left and right limb values divided by the mean of both measurements.
With the exception of the tarsus, the other angles that defined the ROMs were considered segment angles. A segment angle is the resulting angle that the segment subtends from its maximum retracted position to its maximum protracted position (Figure 2). The tarsus angle is a joint angle, which is the angle subtended between 2 segments that rotate about the joint. The coronal angle was the maximum range of the segment movement during the stride in the frontal plane (Figure 3). The sensors calculated the orientation in 2 planes. Because the sensors were mounted on the side of each limb, these 2 planes were presumed to be the sagittal and coronal planes of the limb movement.
Photograph of a horse in motion to illustrate movement in the sagittal plane, the sagittal ROM of the tarsal joints, and the segment angle (Θ) that results from the point of maximal protraction to maximal retraction of the metacarpi and metatarsi.
Citation: American Journal of Veterinary Research 78, 9; 10.2460/ajvr.78.9.1011
Photograph of a standing horse to illustrate movement in the coronal plane and the ROM of the metatarsi. The ROM has a positive value if the movement is lateral (abduction) and negative value if the movement is medial (adduction). Dotted line represents the midline axis.
Citation: American Journal of Veterinary Research 78, 9; 10.2460/ajvr.78.9.1011
Data collection protocol
The treadmill speed was calibrated at the beginning of the experiment by use of a magnetic speedometer systemd attached to a racing bicycle by performing 3 consecutive trials for which results for the treadmill settings and the speedometer readings were identical. Prior to the beginning of each recording period, each horse was warmed up by walking outdoors for 20 minutes. The horse was then exercised by walking for 10 minutes and trotting for 5 minutes on the treadmill to reach steady-state locomotion.16 Additionally, prior to every data collection session for each horse, the IMUs were synchronized and time stamped. Following this period, brushing bootse equipped with a small custom-fitted pouch to accommodate the sensor were secured at a standard location on the distolateral aspect of the metacarpal and metatarsal regions on each of the 4 limbs. In addition, 2 custom-made soft straps equipped with a pouch where the sensor was tightly inserted were placed on the distolateral aspect of each tibia in the groove just dorsal to the calcanean tendon (Figure 1). In total, 6 sensors were used on each horse during each data collection session. The sensors were all aligned visually to the line of the limb segment. Each horse was walked and trotted after the straps and boots had been placed until the gait appeared visually normal (usually within 2 minutes) after which the sensors were switched on and placed in their respective locations. The horse then remained stationary for 10 seconds; a minimum stationary period of 10 seconds was a prerequisite for the sensors to calibrate. This calibration enabled the system to define the gravitational vector. If the horse moved within this 10-second period, then another 10-second stationary period was required to ensure proper system calibration. For all trials, each horse was halter led with a loose hanging rope by the same handler on the left side of the horse. External factors, such as noise or moving objects, which could have influenced the measurement results, were eliminated when at all possible. If a horse's level of distraction or excitement was noticeable, the measurement trial was discarded and repeated once the conditions were optimal. Each horse was then walked for a minimum of 20 strides at a constant speed of 1.8 m/s prior to data collection, and then trotted for a minimum of 20 strides at 3.33 m/s during which data collection took place. The speed was preselected on the treadmill speedometer. From these collected strides, a minimum of 10 strides, avoiding the first and last strides of the data collection period, were selected for the final analysis.
Strides were counted visually, and data were collected in triplicate (3 repetitions coded for analysis as measurements 1, 2, and 3) with a minimum 10-second waiting period between repetitions. This process was repeated once a week for 3 weeks. At the end of each horse's data collection period, the sensors were removed from their pockets, switched off, and then connected to a laptop computer onto which the information from the data collection period was downloaded.
Data collection and analysis by the inertial measurement sensor system
The inertial measurement sensor system included patented software,c 6 IMUs, and a laptop.14 The data collected by the sensors were downloaded onto this specific laptop and analyzed by the software to compute and display the temporal and orientation data obtained from processing of the accelerometer and gyroscope signals. From this display of temporal and orientation output, we manually and visually selected a data window of steady-state locomotion that included ≥ 10 strides (avoiding the beginning and the end of the trial where acceleration and deceleration may have affected steady-state locomotion) for analysis. For the purpose of the study, steady-state locomotion was characterized by a regular signal from each sensor as well as regular stride duration, which could be seen in the user's interface of the system. By means of a proprietary algorithm,14 the system processed the accelerometer and gyroscope signals for the selected data window yielding orientation and temporal data. The system calculated the orientation and temporal events of each segment, then calculated the joint angles from the relationship of 2 adjacent segments and determined the limb phasing from the timing of segments relative to each other.14
Statistical analysis
The descriptive data analysis was performed with commercial software.f Data were examined for the presence of outliers. Values ≥ 97.5th or ≤ 2.5th percentiles, were considered outliers, and their origin was investigated. All variables under investigation were tested for normality (Shapiro-Wilk test). Repeatability coefficients were calculated as 1.96•√2• within-horse SD, as previously described.17 Within-horse SD in turn was obtained by computing a 1-way ANOVA of each outcome variable with respect to the variable horse. The number of observations was 30 (10 horses each assessed 3 consecutive times) for each of the 3 weeks. According to Bartlett and Frost,18 an “estimated repeatability coefficient of 1 means that the absolute difference between any two future measurements made on a particular horse are estimated to be no greater than 1 on 95% of occasions.”
Results
Weekly mean and SD of each gait analysis variable for the 10 horses were summarized (Tables 1 and 2), and the repeatability coefficients were calculated (Tables 3 and 4). The repeatability coefficients that inform on the absolute difference expected between any 2 future measures on the same horse (in the units of each gait analysis variable) indicated that temporal variables and limb phasing variables were highly repeatable (low repeatability coefficients). The repeatability coefficients for timing variables ranged from 2% to 8.5%. The repeatability coefficients for ROMs were approximately 5 degrees, with a maximum value in week 3 for the metatarsal region of the right hind limb ROM (16 degrees). The repeatability coefficients for symmetry variables were somewhat higher, ranging from as low as 1.4% to 1.6% (hind limb asymmetry in all 3 weeks) up to 23% to 42%; this last maximum value was estimated in week 3 for the metatarsal symmetry (between left and right limbs) and was attributable to a single measurement that was out of the norm. Higher repeatability coefficients were often caused by 1 of the 3 measurements being different from the other 2 measurements.
Temporal gait variables (mean [SD]) measured by use of extremity-mounted IMUs during trotting of 10 nonlame horses under controlled conditions of treadmill exercise (3 measurements/d on 3 d/wk each week over a 3-week period).
| Week | ||||
|---|---|---|---|---|
| Temporal variable (mean [SD]) | 1 | 2 | 3 | All weeks |
| Stride duration (s) | 0.72 (0.02) | 0.73 (0.03) | 0.74 (0.02) | 0.73 (0.02) |
| Stride frequency (No. of strides•min−1) | 82.94 (2.99) | 82.72 (3.00) | 81.28 (2.62) | 82.31 (2.94) |
| Stride length (m) | 2.39 (0.08) | 2.40 (0.08) | 2.44 (0.08) | 2.41 (0.08) |
| Speed (m•s−1) | 3.30 (0) | 3.30 (0) | 3.30 (0) | 3.30 (0) |
| Limb phasing* | ||||
| Left forelimb (%) | 63.37 (1.37) | 63.42 (1.63) | 62.68 (1.24) | 63.16 (1.45) |
| Right forelimb (%) | 13.72 (1.56) | 13.71 (1.31) | 12.77 (1.59) | 13.4 (1.54) |
| Right hind limb (%) | 50.04 (0.63) | 49.47 (0.79) | 49.34 (0.72) | 49.62 (0.77) |
| Protraction-retraction | ||||
| Maximal left metatarsal protraction (%) | 47.8 (1.92) | 47.87 (2.16) | 47.53 (2.56) | 47.73 (2.21) |
| Maximal right metatarsal protraction (%) | 48.87 (1.94) | 49.00 (2.39) | 48.40 (3.04) | 48.76 (2.48) |
| Maximal left metacarpal protraction (%) | 52.87 (1.87) | 53.53 (2.21) | 52.73 (2.20) | 53.04 (2.10) |
| Maximal right metacarpal protraction (%) | 53.47 (1.89) | 53.20 (2.20) | 52.27 (2.50) | 52.98 (2.25) |
| Maximal left metatarsal retraction (%) | 4.87 (2.08) | 4.60 (2.53) | 4.47 (2.81) | 4.64 (2.47) |
| Maximal right metatarsal retraction (%) | 5.00 (3.10) | 5.00 (2.67) | 5.40 (2.36) | 5.13 (2.70) |
| Maximal left metacarpal retraction (%) | 18.67 (1.77) | 18.87 (1.63) | 18.73 (1.44) | 18.76 (1.60) |
| Maximal right metacarpal retraction (%) | 19.00 (1.72) | 18.53 (2.10) | 17.87 (2.03) | 18.47 (1.99) |
| Symmetry | ||||
| Diagonal asymmetry (%) | 0.39 (1.08) | −0.25 (0.94) | −0.56 (1.22) | −0.14 (1.15) |
Data collection was performed with a commercially available system of 6 IMUs, each of which recorded 6 degrees of freedom (outputs of acceleration and angular rotation along the 3 orthogonal axes), and proprietary software. A brushing boot equipped with a small custom-fitted pouch to accommodate the sensor was secured at a standard location on the distolateral aspect of the metacarpal and metatarsal regions on each of the 4 limbs. In addition, 2 custom-made soft straps equipped with a pouch that tightly fitted the sensors were placed on the distolateral aspect of each tibia in the groove just dorsal to the calcanean tendon. The treadmill speed was calibrated at the beginning of the experiment. Prior to the beginning of each recording period, each horse was warmed up by walking outdoors for 20 minutes. The horse was then exercised by walking for 10 minutes and trotting for 5 minutes on the treadmill to reach steady-state locomotion with a gait that appeared visually normal (usually within 2 minutes); thereafter, the sensors were switched on and placed in their respective locations (6 IMUs/horse). The horse then remained stationary for 10 seconds to allow the sensors to calibrate. For all trials, horses were halter led with a loose hanging rope by the same handler on the left side of the horse. Each horse was then walked for a minimum of 20 strides at a constant speed of 1.8 m/s prior to data collection, and then trotted for a minimum of 20 strides at 3.33 m/s during which data collection took place. The speed was preselected on the treadmill speedometer. From these collected strides, a minimum of 10 strides, avoiding the first and last strides of the data collection period, were selected for the final analysis. Strides were counted visually and data were collected in triplicate with a minimum 10-second waiting period between repetitions. This process was repeated once a week for 3 weeks; the number of observations was 30 (10 horses each assessed 3 consecutive times) for each of the 3 weeks. Among the temporal variables of interest, limb phasing was defined as the percentage difference of time within a stride of each extremity (left and right forelimb and right hind limb, relative to the left hind limb) and maximal protraction and retraction of the right and left third metacarpal and metatarsal bones were expressed as a percentage of the stride duration. The diagonal asymmetry was calculated as the difference between the diagonal limb phasing timing couplets as follows: (left forelimb-right hind limb) – (right forelimb-left hind limb), where the left hind limb was used as the reference limb and assigned a value of zero.
Spatial gait variables (mean [SD]) measured by use of extremity-mounted IMUs during trotting of the 10 nonlame horses in Table 1.
| Spatial variable | 1 | 2 | 3 | All weeks |
|---|---|---|---|---|
| Sagittal ROM | ||||
| Left tarsal region (°) | 41.46 (7.00) | 42.18 (3.10) | 40.98 (3.78) | 41.54 (4.90) |
| Right tarsal region (°) | 37.59 (4.76) | 37.38 (3.37) | 36.90 (4.51) | 37.29 (4.22) |
| Left tibia (°) | 49.44 (4.51) | 45.60 (3.84) | 45.64 (3.08) | 46.89 (4.22) |
| Right tibia (°) | 47.39 (2.64) | 43.97 (4.13) | 44.51 (4.08) | 45.29 (3.94) |
| Left metacarpal region (°) | 84.41 (3.32) | 84.87 (2.26) | 83.91 (2.38) | 84.40 (2.69) |
| Right metacarpal region (°) | 85.6 (4.76) | 85.47 (4.48) | 84.6 (3.70) | 85.22 (4.31) |
| Left metatarsal region (°) | 55.59 (3.92) | 54.08 (2.42) | 52.76 (3.20) | 54.14 (3.41) |
| Right metatarsal region (°) | 54.93 (4.58 | 53.65 (2.75 | 51.41 (5.88 | 53.33 (4.76) |
| Coronal ROM | ||||
| Mediolateral left tibia (°) | 19.28 (3.21) | 20.92 (4.06) | 21.87 (3.30) | 20.69 (3.67) |
| Mediolateral right tibia (°) | 19.6 (3.30) | 21.72 (1.95) | 21.41 (4.17) | 20.91 (3.36) |
| Mediolateral left metacarpal region (°) | 25.32 (9.56) | 12.50 (4.25) | 13.97 (4.46) | 17.26 (8.68) |
| Mediolateral right metacarpal region (°) | 29.34 (6.90) | 19.24 (8.52) | 14.90 (5.58) | 21.16 (9.29) |
| Mediolateral left metatarsal region (°) | 15.64 (4.87) | 13.54 (2.22) | 17.86 (7.13) | 15.68 (5.39) |
| Mediolateral right metacarpal region (°) | 13.21 (4.57) | 13.46 (2.53) | 14.95 (6.94) | 13.87 (5.02) |
| Symmetry | ||||
| Tarsal region (%) | 9.00 (14.53) | 12.14 (12.44) | 10.78 (13.64) | 10.64 (13.47) |
| Tibia (%) | 3.96 (8.55) | 3.70 (7.71) | 2.69 (11.94) | 3.45 (9.49) |
| Metacarpal region (%) | −1.33 (4.61) | −0.61 (3.99) | −0.76 (2.88) | −0.90 (3.86) |
| Metatarsal region (%) | 1.31 (4.29) | 0.83 (3.82) | 3.17 (13.77) | 1.77 (8.58) |
| Hind limb asymmetry (%) | −0.04 (0.63) | 0.53 (0.79) | 0.68 (0.73) | 0.39 (0.78) |
The spatial variables of interest included sagittal ROM (ie, sagittal angles) of the left and right tarsal joints and segment angles of the left and right metacarpi, metatarsi, and tibias and the coronal ROM of the left and right metacarpi, metatarsi, and tibias. The symmetry of each segment or joint ROM was calculated as the difference between the left and right limb values divided by the mean of both measurements. A segment angle was defined as the resulting angle that the segment subtended from its maximum retracted position to its maximum protracted position. The tarsus angle was a joint angle defined as the angle subtended between 2 segments that rotated about the joint. The coronal angle was the maximum range of the segment movement during the stride in the frontal plane. The sensors calculated the orientation in 2 planes. Because the sensors were mounted on the side of each limb, these 2 planes were presumed to be the sagittal and coronal planes of the limb movement.
See Table 1 for key.
Repeatability coefficients calculated from temporal gait variables measured by use of extremity-mounted IMUs during trotting of the 10 nonlame horses in Table 1.
| Week | |||
|---|---|---|---|
| Temporal variable | 1 | 2 | 3 |
| Stride duration (s) | 0.02 | 0.01 | 0.02 |
| Stride frequency (No. of strides•s−1) | 1.97 | 1.39 | 2.27 |
| Stride length (m) | 0.05 | 0.04 | 0.07 |
| Speed (m•s−1) | — | — | — |
| Limb phasing | |||
| Left forelimb (%) | 2.05 | 1.47 | 2.08 |
| Right forelimb (%) | 1.39 | 1.66 | 1.66 |
| Right hind limb (%) | 1.39 | 1.66 | 1.44 |
| Protraction-retraction | |||
| Maximal left metatarsal protraction (%) | 4.05 | 3.66 | 4.52 |
| Maximal right metatarsal protraction (%) | 4.63 | 5.07 | 4.85 |
| Maximal left metacarpal protraction (%) | 4.05 | 3.49 | 3.35 |
| Maximal right metacarpal protraction (%) | 3.35 | 3.04 | 2.69 |
| Maximal left metatarsal retraction (%) | 7.29 | 6.24 | 6.71 |
| Maximal right metatarsal retraction (%) | 5.74 | 9.54 | 8.54 |
| Maximal left metacarpal retraction (%) | 3.66 | 3.80 | 3.66 |
| Maximal right metacarpal retraction (%) | 2.69 | 2.69 | 3.35 |
| Symmetry | |||
| Diagonal asymmetry (%) | 2.19 | 2.05 | 2.13 |
— = Not applicable.
Repeatability coefficients were calculated as 1.96•√2•within-horse SD. Within-horse SD was obtained by 1-way ANOVA of each outcome gait analysis variable with respect to horse. Each horse was assessed 3 times every week (n = 30 assessments/wk). The premise used was that an estimated repeatability coefficient of 1 meant that the absolute difference between any 2 future measurements made on a particular horse was estimated to be no greater than 1 on 95% of occasions.15
See Table 1 for key.
Repeatability coefficients calculated from spatial gait variables measured by use of extremity-mounted IMUs during trotting of the 10 nonlame horses in Table 1.
| Week | |||
|---|---|---|---|
| Spatial variable | 1 | 2 | 3 |
| Sagittal ROM | |||
| Left tarsal region (°) | 5.07 | 3.88 | 5.79 |
| Right tarsal region (°) | 5.29 | 4.21 | 8.98 |
| Left tibia (°) | 4.32 | 2.94 | 4.43 |
| Right tibia (°) | 5.38 | 3.94 | 9.54 |
| Left metacarpal region (°) | 3.69 | 3.44 | 4.05 |
| Right metacarpal region (°) | 8.34 | 5.49 | 4.74 |
| Left metatarsal region (°) | 5.49 | 4.02 | 6.13 |
| Right metatarsal region (°) | 5.88 | 5.18 | 15.99 (5.02)* |
| Coronal ROM | |||
| Mediolateral left tibia (°) | 5.41 | 5.49 | 4.32 |
| Mediolateral right tibia (°) | 5.04 | 2.72 | 3.71 |
| Mediolateral left metacarpal region (°) | 5.41 | 3.27 | 3.69 |
| Mediolateral right metacarpal region (°) | 5.24 | 7.26 | 4.49 |
| Mediolateral left metatarsal region (°) | 4.05 | 4.43 | 4.93 |
| Mediolateral right metacarpal region (°) | 3.19 | 3.27 | 3.74 |
| Symmetry | |||
| Tarsal region (%) | 17.41 | 16.08 | 34.45 (18.60)* |
| Tibia (%) | 8.68 | 8.93 | 23.09 |
| Metacarpal region (%) | 10.20 | 7.43 | 3.91 |
| Metatarsal region (%) | 9.06 | 10.81 | 41.86 (6.51)* |
| Hind limb asymmetry (%) | 1.39 | 1.66 | 1.41* |
Values in parentheses are the recalculated repeatability coefficients after removing 1 data point (first trot) of 1 horse, which was out of the norm.
See Tables 1 and 3 for key.
Discussion
Results of the present study highlighted the repeatability of gait patterns within day of measurements obtained by use of extremity-mounted IMUs in nonlame horses during trotting under controlled conditions of treadmill exercise. We believe that this finding is an important contribution toward the clinical implementation of this gait analysis tool.
The repeatability for most of the temporal and spatial variables was high, and the repeatability coefficients were consistent from week to week, with a few exceptions. Therefore, one can infer that the horses trotted in a repeatable manner from week to week. The repeatability coefficients for some spatial variables, and more specifically the symmetry variables, were higher (lower repeatability) in the third week of the study. This last week was not necessarily different from the preceding 2 weeks in terms of the conduct of the experiments. The abnormally high values for the right metatarsal region ROM and tarsal and metatarsal symmetry appeared to have been generated by the contribution of 1 horse to the dataset for the first trotting session in week 3. The high repeatability coefficient of 23% for tibia symmetry in week 3 was attributable to another horse for which all 3 tibia symmetry values in this week were higher than those of the remainder of the horses (symmetry, approx 25% to 26%). Interestingly, this horse had very low tibia symmetry values in week 1 and values of approximately 11% to 13% in week 2. It is possible that something occurred that affected these 2 horses specifically in week 3. Perhaps the sensors were not securely fastened and we failed to identify such a problem or the horse moved abnormally for some reason during the week 3 trials. These anomalous findings seemed to be isolated incidents, and it was not possible to discern their cause a posteriori, even though an irregular signal could be seen on the displayed graphs. Removal of these specific data points (1 set of measurements from the total of 90 [ie, first trotting session of 1 horse in week 3]) from the analyzed data vastly improved the repeatability coefficient (from 41.86 to 6.51 [Table 2]) for this variable, thereby highlighting the importance of controlling the conditions of measurement and repeating the trials.
The coronal ROM of the metacarpal regions reflected the lateromedial movement of the metacarpal bone segment, and we believe that it has a greater variability because of the paddling action of some horses and the effect of carpal instability during the swing phase that allowed freer movement of the segment of the extremity distal to it.19 We considered this a normal and expected finding on the basis of our clinical experience evaluating gait in horses.
All symmetry variables were calculated variables and therefore dependent on the fluctuations of other variables. When evaluating percentages of change in temporal variables, it is important to consider that they occurred over a total time of approximately 720 milliseconds, which is the duration of 1 stride. Therefore, a 5% change would represent an absolute timing change of 36 milliseconds. The same rationale would apply to spatial variables; the differences in percentages of these variables may represent small changes in absolute values, which could be within the limits of expected biological variation.
Prior to use in clinical practice, further validation of the inertial measurement sensor system in horses in addition to what has already been done13 would be advantageous to ensure full comparison of the system to optical kinematic systems (eg, investigation of gait variables simultaneously with IMUs and an optical 3-D kinematic system). The use of lightweight, low-power, inertial sensors to measure acceleration or angular velocity is now widespread in human clinical science in developed countries.9,20,21 In humans, inertial sensor data have been used to infer activity type and intensity, falls and fall risk, muscle activity, and gait events.22,23 Furthermore, joint angle determination in humans has been investigated with IMUs and findings compared against those obtained with an optical system. Results indicated that highly accurate knee angle estimates can be obtained by placing an IMU on each side (proximal and distal) of the knee joint, under the assumption that the knee behaves as a pure hinge joint.17 In that study, the knee angle was estimated from angular velocities and linear accelerations measured in the 2 IMU reference frames. Recently, initial studies were undertaken to support the use of IMUs in horses. A pilot study13 in horses compared the sagittal ROM of metacarpal and metatarsal bones measured by use of an optical system and an IMU system, and found that the accuracy and precision of the IMU measurements were very good. In another study,10 the accuracy for detecting stance phase (bias, 0.01 to 1.3 milliseconds) in horses with IMUs placed on the metacarpal and metatarsal regions (similar to the placement in the present study) was good.
Ideally, a database of IMU-obtained kinematic data in nonlame horses representative of the target population should be developed to assess possible benchmarks for future clinical use. In addition, it will be necessary to determine the range of the measured variables in lame horses through a case-control study.
Although it would be optimal if an inertial measurement sensor system could assess precisely and with high accuracy the duration of stance and swing phases, because they are considered important lameness variables,24 it is also possible that some of the variables measured by this system could be equally if not more useful in determining lameness in horses. Such possibility must be evaluated in additional studies with lame horses.
In general, the inertial measurement sensor system evaluated in the present study was considered easy to use and quick to implement. One trial of 3 measurements to generate a gait profile can be carried out in < 10 minutes. Presently, the only drawback of this system is that visualization of the recorded data is not possible in real time because the sensors have to be removed from the horse and connected to the computer for data visualization. Therefore, we recommend to always obtain at least 3 measurements or even better 5 measurements and, whenever possible, to transfer the data to the computer for visualization while it is still possible to carry out additional measurements. Alternatively, the sensors can be left in place while interventions are performed (eg, diagnostic nerve blocks) because they have sufficient battery power. However, to identify the nature of a specific gait abnormality, the data need to be extracted, which requires connection of the sensors to the computer. Ensuring repeatability of measurements potentially reduces the chances of introducing erroneous measurements when removing the sensors repeatedly in the course of an examination. Real-time data transfer would increase and facilitate the system's range of applications (eg, lameness examination) beyond the generation of a gait analysis profile. In the present study, a proper recording of the stationary period was needed for the system to work appropriately and avoid an error message, and this was accomplished by ensuring that each horse remained still for at least 10 seconds immediately after sensor placement. It is also important that the selection of the window of strides to be analyzed by the operator with the dedicated software is carried out by a properly trained person to ensure that the window does not contain any abnormal signals resulting from a misplaced or not securely fastened sensor or a tripping horse. Overall, results of the present study have highlighted the repeatability of an inertial measurement sensor system involving extremity-mounted IMUs that can potentially be applied for gait evaluation of horses.
Acknowledgments
Supported by the Institute Suisse du Médicine Équine.
The authors thank European Technology for Business for the loan of the equipment used in the study, Ms. Marie Mayerat and Mr. Christian Herren for technical assistance, and Prof. Vinzenz Gerber and Alessandra Ramseyer for use of the facilities at Haras National Reproduction Center.
ABBREVIATIONS
| IMU | Inertial measurement unit |
| ROM | Range of motion |
Footnotes
Kagra, Graber AG, Fahrwangen, Switzerland.
Pegasus GaitSmart, European Technology for Business, ETB, Codicote, Hertfordshire, England.
Poseidon version 4.0 European Technology for Business, ETB, Codicote, Hertfordshire, England.
Anima +, TwoNav, Arenys de Mar, Barcelona, Spain.
Woof Wear, Bodmin, Cornwall, England.
NCSS 10, NCSS Statistical Software, Kaysville, Utah.
Function “loneway,” Stata statistical software: release 14, StataCorp LP, College Station, Tex.
References
1. FN – Deutsche Reiterlicher Vereinigung. (2014). Bundesverband für Pferdesport und Pferdezucht. Available at: www.pferd-aktuell.de/pferdesport-pferdezucht/deutsche-reiterliche-vereinigung. Accessed Jun 11, 2014.
2. Back W, Barneveld A, Bruin G, et al. Kinematic detection of superior gait quality in young trotting warmbloods. Vet Q 1994;16:S91–S96.
3. Drevemo S, Dalin G, Fredricson I, et al. Equine locomotion: 1. The analysis of linear and temporal stride characteristics of trotting standardbreds. Equine Vet J 1980;12:60–65.
4. Buchner HHF, Savelberg HHCM, Schamhardt HC. Kinematics of treadmill vs overground locomotion in horses. Vet Q 1994;16:S87–S90.
5. Barrey E, Galloux P, Valette JP, et al. Stride characteristics of overground versus treadmill locomotion in the saddle horse. Acta Anat (Basel) 1993;146:90–94.
6. Keegan KG, Yonezawa Y, Wilson DA, et al. Evaluation of a sensor-based system of motion analysis for detection and quantification of forelimb and hind limb lameness in horses. Am J Vet Res 2004;65:665–670.
7. Keegan KG. Evidence-based lameness detection and quantification. Vet Clin North Am Equine Pract 2007;23:403–423.
8. Practical assessment of movement symmetry in horses: use of inertial sensors during clinical lameness exams. Available at: www.befred.org/docs/2423.pdf. Accessed Sep 8, 2015.
9. Cuesta-Vargas AI, Galán-Mercant A, Williams JM. The use of inertial sensors system for human motion analysis. Phys Ther Rev 2010;15:462–473.
10. Olsen E, Andersen PH, Pfau T. Accuracy and precision of equine gait event detection during walking with limb and trunk mounted inertial sensors. Sensors (Basel) 2012;12:8145–8156.
11. Hildebrand M. Symmetrical gaits of horses. Science 1965;150:701–708.
12. Muybridge E. Animals in motion. New York:Dover Publications, 1957.
13. Roepstorff L, Wiestner T, Weishaupt MA, et al. Comparison of microgyro-based measurements of equine metatarsal/metacarpal bone to a high speed video locomotion analysis system during treadmill locomotion. Vet J 2013;198:e157–e160.
14. Whalley A, Hodgins D. Assessment of gait. Patent number US 8,715,208 B2, May 6, 2014, United States of America.
15. Cooper G, Sheret I, McMillan L, et al. Inertial-sensor based knee flexion/extension angle estimation. J Biomech 2009;42:2678–2685.
16. Bartlett JW, Frost C. Reliability, repeatability and reproducibility: analysis of measurement errors in continuous variables. Ultrasound Obstet Gynecol 2008;31:466–475
17. Clayton HM, Sha D, Stick JA, et al. Three-dimensional carpal kinematics of trotting horses. Equine Vet J 2004;36:671–676.
18. Buchner HH, Savelberg HHCM, Schamhardt HC, et al. Habituation of horses to treadmill locomotion. Equine Vet J Suppl 1994;17:13–15.
19. Lynch JA, Clayton HM, Mullineaux DR. The reliability of force platform data from trotting horses. Equine Comp Exerc Physiol 2005;2:129–132.
20. Monda M, Goldberg A, Smitham P, et al. Use of inertial measurement units to assess age-related changes in gait kinematics in an active population. J Aging Phys Act 2015;23:18–23.
21. Kavanagh JJ, Menz HB. Accelerometry: a technique for quantifying movement patterns during walking. Gait Posture 2008;28:1–15.
22. Orizio C. Muscle sound: bases for the introduction of a mechanomyographic signal in muscle studies. Crit Rev Biomed Eng 1993;21:201–243.
23. Plasqui G, Westerterp KR. Physical activity assessment with accelerometers: an evaluation against doubly labeled water. Obesity (Silver Spring) 2007;15:2371–2379.
24. Galisteo AM, Cano MR, Morales JL, et al. Kinematics in horses at the trot before and after an induced forelimb supporting lameness. Equine Vet J Suppl 1997;23:97–101.
