An attempt to detect lameness in galloping horses by use of body-mounted inertial sensors

Marco A. F. Lopes Department of Veterinary Medicine and Surgery, College of Veterinary Medicine, University of Missouri, Columbia, MO 65211.

Search for other papers by Marco A. F. Lopes in
Current site
Google Scholar
PubMed
Close
 DVM, PhD
,
Antonio C. O. Dearo Department of Veterinary Medicine and Surgery, College of Veterinary Medicine, University of Missouri, Columbia, MO 65211.

Search for other papers by Antonio C. O. Dearo in
Current site
Google Scholar
PubMed
Close
 DVM, MSC, DVSC
,
Allen Lee Leap Scientific LLC, 5 Hilltop Cir, Hooksett, NH 03106.

Search for other papers by Allen Lee in
Current site
Google Scholar
PubMed
Close
 MS
,
Shannon K. Reed Department of Veterinary Medicine and Surgery, College of Veterinary Medicine, University of Missouri, Columbia, MO 65211.

Search for other papers by Shannon K. Reed in
Current site
Google Scholar
PubMed
Close
 DVM, MS
,
Joanne Kramer Department of Veterinary Medicine and Surgery, College of Veterinary Medicine, University of Missouri, Columbia, MO 65211.

Search for other papers by Joanne Kramer in
Current site
Google Scholar
PubMed
Close
 DVM
,
P. Frank Pai Department of Mechanical and Aerospace Engineering, College of Engineering, University of Missouri, Columbia, MO 65211.

Search for other papers by P. Frank Pai in
Current site
Google Scholar
PubMed
Close
 PhD
,
Yoshiharu Yonezawa Department of Health Science, Hiroshima Institute of Technology, Hiroshima, 731-5193, Japan.

Search for other papers by Yoshiharu Yonezawa in
Current site
Google Scholar
PubMed
Close
 PhD
,
Hiromitchi Maki Department of Health Science, Hiroshima Institute of Technology, Hiroshima, 731-5193, Japan.

Search for other papers by Hiromitchi Maki in
Current site
Google Scholar
PubMed
Close
 PhD
,
Terry L. Morgan Eastland Thoroughbred Training Center, 5367 Bohleyville Rd, Milstadt, IL 62260.

Search for other papers by Terry L. Morgan in
Current site
Google Scholar
PubMed
Close
 DVM
,
David A. Wilson Department of Veterinary Medicine and Surgery, College of Veterinary Medicine, University of Missouri, Columbia, MO 65211.

Search for other papers by David A. Wilson in
Current site
Google Scholar
PubMed
Close
 DVM, MS
, and
Kevin G. Keegan Department of Veterinary Medicine and Surgery, College of Veterinary Medicine, University of Missouri, Columbia, MO 65211.

Search for other papers by Kevin G. Keegan in
Current site
Google Scholar
PubMed
Close
 DVM, MS

Abstract

OBJECTIVE To evaluate head, pelvic, and limb movement to detect lameness in galloping horses.

ANIMALS 12 Thoroughbreds.

PROCEDURES Movement data were collected with inertial sensors mounted on the head, pelvis, and limbs of horses trotting and galloping in a straight line before and after induction of forelimb and hind limb lameness by use of sole pressure. Successful induction of lameness was determined by measurement of asymmetric vertical head and pelvic movement during trotting. Differences in gallop strides before and after induction of lameness were evaluated with paired-sample statistical analysis and neural network training and testing. Variables included maximum, minimum, range, and time indices of vertical head and pelvic acceleration, head rotation in the sagittal plane, pelvic rotation in the frontal plane, limb contact intervals, stride durations, and limb lead preference. Difference between median standardized gallop strides for each limb lead before and after induction of lameness was calculated as the sum of squared differences at each time index and assessed with a 2-way ANOVA.

RESULTS Head and pelvic acceleration and rotation, limb timing, stride duration measurements, and limb lead preference during galloping were not significantly different before and after induction of lameness in the forelimb or hind limb. Differences between limb leads before induction of lameness were similar to or greater than differences within limb leads before and after lameness induction.

CONCLUSIONS AND CLINICAL RELEVANCE Galloping horses maintained asymmetry of head, pelvic, and limb motion between limb leads that was unrelated to lameness.

Abstract

OBJECTIVE To evaluate head, pelvic, and limb movement to detect lameness in galloping horses.

ANIMALS 12 Thoroughbreds.

PROCEDURES Movement data were collected with inertial sensors mounted on the head, pelvis, and limbs of horses trotting and galloping in a straight line before and after induction of forelimb and hind limb lameness by use of sole pressure. Successful induction of lameness was determined by measurement of asymmetric vertical head and pelvic movement during trotting. Differences in gallop strides before and after induction of lameness were evaluated with paired-sample statistical analysis and neural network training and testing. Variables included maximum, minimum, range, and time indices of vertical head and pelvic acceleration, head rotation in the sagittal plane, pelvic rotation in the frontal plane, limb contact intervals, stride durations, and limb lead preference. Difference between median standardized gallop strides for each limb lead before and after induction of lameness was calculated as the sum of squared differences at each time index and assessed with a 2-way ANOVA.

RESULTS Head and pelvic acceleration and rotation, limb timing, stride duration measurements, and limb lead preference during galloping were not significantly different before and after induction of lameness in the forelimb or hind limb. Differences between limb leads before induction of lameness were similar to or greater than differences within limb leads before and after lameness induction.

CONCLUSIONS AND CLINICAL RELEVANCE Galloping horses maintained asymmetry of head, pelvic, and limb motion between limb leads that was unrelated to lameness.

Trauma and diseases that cause lameness are the most common medical problems in horses.1–7 Especially within the racehorse industry, economic losses attributable to lameness are substantial.8 Also, catastrophic injuries in racehorses, which occur most often during a race9 and are thought to be subsequent to preexisting orthopedic disease,10,11 are particularly troublesome. News coverage of these events frequently leads to misunderstanding and negative publicity harmful to the racing industry. Elimination of all catastrophic injury may not be possible, but methods that detect lameness as early as possible and for all gaits routinely used by horses may help to reduce the incidence.12

Detection of lameness in horses is most often attempted by observing a horse in motion while trotting. Movement during trotting has been intensely studied and is well understood.13–20 Trotting in a horse without lameness is a symmetric 2-beat gait with the head and pelvis moving up and down at equivalent amplitude twice per stride. Lameness causes asymmetric vertical head and pelvic movement during trotting, which, if large enough, can be observed. Objective measures of vertical head and pelvic movement, with higher sampling frequency than that attainable with the unaided human eye, can be used to detect even smaller asymmetries of more subtle lameness.21,22 However, in many breeds, trotting is not the most common gait or the gait of maximal exercise intensity, and some orthopedic dysfunctions may not cause lameness during trotting, especially if trotting is not the gait of maximal effort. As speed of movement increases, most breeds transition to the asymmetric 3- and 4-beat gaits of the canter and gallop.23 Detection of lameness during higher speed movements during cantering or galloping could be useful when lameness is not visible or measurable during trotting. Energetics and mechanics of horses while cantering and galloping have been studied, 24–28 but surprisingly there have been no reported objective studies of lameness for these gaits.

The purpose of the study reported here was to collect movement data for Thoroughbred racehorses during galloping before and after induction of weight-bearing pain in the forelimb and hind limb and to evaluate head, limb, and pelvic movements during galloping that may indicate this pain as lameness. We hypothesized that some variables of vertical movement or rotation of the head or pelvis, timing of limb movement, stride duration, or limb lead (right or left) preference score would differ significantly during galloping after induction of forelimb or hind limb lameness.

Materials and Methods

Horses

Twelve Thoroughbreds (5 geldings and 7 mares) with a mean ± SD age of 4.3 ± 1.4 years (range, 3 to 8 years) at a racetrack were used in the study. Each horse was housed in a stall at the racetrack and exercised regularly. No lameness was seen by an experienced veterinarian (TLM) evaluating the horses while trotting with a handler running beside and lightly restraining the horse by holding the reins. The study was performed after the end of a race meet; each horse had completed at least 1 race during the meet. Permission was obtained from the owners, trainers, and jockeys for inclusion of horses in the study. The study methods and protocol were approved by the University of Missouri Animal Care and Use Committee.

Instrumentation

Each horse was instrumented with 6 inertial sensor devices (1 on the head, 1 on the pelvis, and 1 on the distal aspect of each limb). The devices comprised a modified body-mounted inertial sensor systema used to evaluate lameness in horses during trotting.15–17,20 The sensor device on the head was a vertical accelerometer (± 4 g) and uniaxial (sagittal plane) gyroscope (± 300°/s) placed in a cap. It was attached to the poll by affixing the cap to the crown piece of the bridle. The sensor device for the pelvic region, which consisted of the same accelerometer but with the gyroscope measuring rotation in the frontal plane, was attached by use of hook-and-loop tape to the most dorsal aspect of the midline of the pelvis between the tubera sacralia. Each limb was instrumented with a uniaxial sagittal plane gyroscope (± 300°/s) attached to the dorsum of the metacarpophalangeal (or metatarsophalangeal) region in a pouch sewn into a felt wrap secured with hook-and-loop strips. Devices consisted of sensors, a radio, a battery, and other electronic components contained on a single board, with the sensors mounted in the center of the device. Sensors in the devices collected samples at 200 Hz, and data were transmitted wirelessly (8-bit analogue-to-digital conversion) to a receiver connected to a computer. Each device weighed 28 g and had dimensions of 2.0 × 2.5 × 3.0 cm.

Data collection and lameness induction

Sensor data were collected from each horse during trotting in a straight line and galloping on each limb lead in a straight line before (baseline) and after various amounts of induced lameness of the forelimb and hind limb. Trotting trials were collected with each horse fully outfitted with a saddle and bridle but no rider. Gallop trials were collected with a professional jockey riding in a 3-point stance. The jockey was familiar with and had ridden each horse before the study. Track conditions were similar on each day of collection; data collection was performed after the track was dragged by a maintenance crew. Each horse was shod (all 4 feet) with a steel bar shoe that could apply pressure to the sole by means of two 0.5-inch-diameter, blunt-tipped screws.29 Screws were inserted into threaded holes located 2 cm from the medial and lateral hoof walls in the bar that crossed between the branches of the shoe at the widest part of the hoof; screws were gradually tightened to induce increasing amounts of lameness.

Data were first obtained with screws placed in the holes but without the screws touching the sole. Each horse trotted in a straight line on a fine-gravel path 90 m in length, with a handler running beside and lightly restraining the horse by holding the reins. This allowed collection of at least 25 trotting strides/trial. The horse was then mounted by the jockey and walked a short distance to the dirt track. Data were collected with the horse galloping in a straight line; at least two 200-m straight-line segments with the horse galloping on each limb lead were collected. If a horse had an initial preference for a particular limb lead, data were collected for that preferred limb lead, and the jockey then induced the horse to gallop with the other limb lead. All data collections were filmed with a 60-Hz video camera.

A forelimb or hind limb was then selected for induction of lameness. If the horse had no lameness during baseline trotting, the limb for lameness induction was selected by coin flip. If the horse had mild lameness in a limb during baseline trotting, lameness was induced in the opposite limb. Lameness was induced by tightening the screws until their tips touched the sole. Screws were short enough that the slotted end of each screw was recessed within the threaded hole. Trotting and galloping trials for both limb leads were performed again as previously described.

Lameness was then increased by tightening the screws further to increase the pressure on the sole. The intent was to collect data for a continuum of mild to moderate induced lameness in the selected limb. Lameness initially was induced by tightening the screws a fixed number of fractional turns. However, the induced lameness did not allow collection of sufficient data for some horses before fatigue from repeated galloping trials prevented further data collection. Therefore, the procedures were adjusted in an attempt to create 3 severities of lameness: one considered questionably evident during visual examination but definitively detected during trotting (approximate AAEP lameness score,30 1 or 2 [scale, 0 to 5]), one considered mild during visual examination and definitively detected during trotting (approximate AAEP lameness score, 3), and one considered moderate during visual examination and definitively detected during trotting (approximate AAEP lameness score, > 3). Two of 4 experienced equine veterinarians (MAFL, ACOD, KGK, or TLM) assessed lameness severity by observing and examining data collected for each horse trotting in a straight line. If the desired lameness severity was not achieved, screws were tightened or loosened, and the horse was again evaluated during trotting. Occasionally, it was necessary to use longer or shorter screws to achieve the desired lameness severities. Trotting and galloping trials were conducted at each lameness severity for each horse with lameness in 1 forelimb and 1 hind limb; there was at least a 3-day period between successive induced forelimb and hind limb lameness. Before the second lameness was induced, each horse was evaluated to determine whether there was residual lameness from the first induced lameness.

Analysis of trotting data

Successful lameness induction and severity was assessed by use of methods described elsewhere5–17,20 for the assessment of asymmetric vertical head (forelimb lameness) or pelvic (hind limb lameness) motion while a horse was trotting in a straight line (Figure 1). Forelimb lameness severity during trotting was measured as the VS, which is a combined measure of asymmetric head maximum position (calculated as maximum head position before stance phase of the right forelimb minus maximum head position before stance phase of the left forelimb during trotting) and asymmetric minimum head position (calculated as minimum head position during stance phase of the right forelimb minus minimum head position during stance phase of the left forelimb during trotting), taking into consideration asymmetry of upward and downward movement of the head. The VS was calculated as ([asymmetric maximum head position]2 + [asymmetric minimum head position]2)0.5. Threshold for successful induction of forelimb lameness was ± 8.5 mm, which was the approximate 95% confidence interval of repeatability for VS = 0 mm. Hind limb lameness measures were Pmax (calculated as maximum pelvic position before stance phase of the right hind limb minus maximum pelvic position before stance phase of the left hind limb during trotting) and Pmin (calculated as minimum pelvic position during stance phase of the right hind limb minus minimum pelvic position during stance phase of the left hind limb during trotting), which are measures of asymmetry of the upward and downward motion of the pelvis. Both Pmax and Pmin are independent quantities; thus, they were evaluated separately. Thresholds for successful induction of hind limb lameness for Pmax and Pmin were ± 3 mm, which was the approximate 95% confidence intervals for Pmax and Pmin = 0 mm.

Figure 1—
Figure 1—

Graphs depicting lameness measures obtained during trotting in a horse with induced forelimb lameness (A) and in a horse with induced hind limb lameness (B). In panel A, Hmax was calculated as maximum head position before stance phase of the right forelimb (RF) minus maximum head position before stance phase of the left forelimb (LF), and Hmin was calculated as minimum head position during stance phase of the RF minus minimum head position during stance phase of the LF. In panel B, Pmax was calculated as maximum pelvic position before stance phase of the right hind limb (RH) minus maximum pelvic position before stance phase of the left hind limb (LH), and Pmin was calculated as minimum pelvic position during stance phase of the RH minus minimum pelvic position during stance phase of the LH. Notice that Hmax and Pmax are calculated as the maximum nearest the right limb impact (1) minus the maximum nearest the left limb impact (2), and Hmin and Pmin are calculated as the minimum during the right limb stance (3) minus the minimum during the left limb stance (4). Gray lines represent results for single strides, and the black line represents the median value for all strides during one lameness induction.

Citation: American Journal of Veterinary Research 77, 10; 10.2460/ajvr.77.10.1121

Galloping data reduction—Data mining and exploratory statistical testing were used to evaluate potential head, pelvic, and limb movement measures that indicated lameness during galloping. Both raw (vertical acceleration and angular rate) and processed (position [double integration of the acceleration signal) and angle (single integration of the angular rate signal) measures for the head and pelvic sensor devices were analyzed. Calculation of position and angle data used an error-correcting, moving-window integration procedure that has been validated for analysis of head and pelvic movement.15–17 Raw and smoothed angular rate signals for the limb-mounted gyroscopes were also analyzed for stride timing measures.

Gallop strides for each limb lead were isolated by identifying variables in the head and pelvic vertical acceleration signals and in the right forelimb angular rate signals. Identification of right- and left-lead gallop strides was confirmed by review of video recordings. Gallop stride durations were extracted from the signals, and then each gallop stride was standardized by resampling to 300 equal time divisions.

Visual data mining and statistical analysis of galloping data—Aggregate, median, and standardized strides for each raw and processed sensor signal for each limb lead in each horse during galloping and each lameness induction were plotted in order of increasing severity of induced lameness (Figure 2). Analysis of these plots revealed stride variables that could be used to detect and measure lameness, including maximum, minimum, range (calculated as maximum value minus minimum value), and time indices of these variables within a standardized stride cycle. Differences in these variables, mean absolute stride duration, and number of strides collected for each limb lead between baseline and the most severe induced lameness in each limb for each horse were assessed by use of simple paired t tests or the paired-sample Wilcoxon signed rank test, depending on the shape of the distribution of measurements for that variable. A correction for multiple comparisons was performed (4 independent variables [maximum, minimum, and time indices of maximum and minimum] × 2 regions [head and pelvis] × 2 limb leads [left and right] + 2 stride durations [1 for each limb lead] + 2 number of strides [1 for each limb lead] = 20 independent variables). Therefore, only comparisons with a value of P < 0.0025 (ie, P = 0.05/20) were considered significant.

Figure 2—
Figure 2—

Graphs for vertical head acceleration (A), vertical head height (B), vertical pelvic acceleration (C), and vertical pelvic height (D)in a horse during galloping. Gray lines represent results for single strides, and median values for all strides are indicated for the left limb lead (solid black line) and right limb lead (dashed black line). Variables of the signals were stride maximum (gray circles), stride minimum (gray stars), and time indices within the stride (gray arrows).

Citation: American Journal of Veterinary Research 77, 10; 10.2460/ajvr.77.10.1121

Supervised neural network data mining of galloping data—Signal variables extracted from stride data obtained before and after lameness induction for all trials were used as the input to train a neural network to classify lameness.b Confirmation of each trial as not lame (ie, sound) or lame was performed by analysis of trotting data. Three classifications were tested: sound versus lame in the right forelimb, left forelimb, right hind limb, or left hind limb; forelimb lameness versus hind limb lameness; and left limb lameness versus right limb lameness given the presence of forelimb or hind limb lameness. A feed-forward network with a hidden layer (10 and 20 nodes were explored) was used; 80% of the trials were for training, and 20% were for validation. Default values for the various neural network training variables were used (Levenberg-Marquardt method,31 mean-squared error function, and early-stopping when error on validation data no longer improved). To evaluate classifier performance, a leave-one-out method was used.32 Each horse was omitted from the validation and used to test classifier performance. Performance of the neural network classifier was assessed as the performance averaged over all horses. Performance was reported as the misclassification rate (range, 0 [all classifications were correct] to 1 [all classifications were incorrect]). Classifier performance was compared to a positive control of noisy sine and cosine waves and to a naïve classifier that predicted only the most common state for each comparison.

Timing analysis of limb movement of galloping data— To perform timing analysis of limb movement during galloping, it was necessary to identify variables in the limb signals that occurred once per stride at a consistent time index. For the forelimbs, peak positive angular rotation near hoof lift off from the ground at the end of the stance phase was consistently identified. For the hind limbs, a sharply defined positive followed by negative angular rotation at hoof impact with the ground were consistently identified. Time indices of these variables were extracted for each right and left limb lead stride during galloping before and after induction of the most severe lameness. Intervals for these variables were determined. These intervals were defined as HLSD (time between impact of lead and nonlead hind limb), CLSD (time between impact of lead hind limb and nonlead forelimb at the end of the stance phase), FLSD (time between end of stance of lead forelimb and end of stance of nonlead forelimb), and NLSD (time between lead forelimb at end of stance phase and nonlead hind limb impact for subsequent stride; Figure 3). Mean values of HLSD, CLSD, FLSD, and NLSD for each limb lead and between limb leads were calculated for the most sound (before lameness induction) and most lame trial for each horse during each lameness induction experiment. Mean gallop stride frequency was determined as the reciprocal of the stride duration. Differences in limb timing intervals between sound and lame conditions were evaluated for both forelimb and hind limb lameness inductions by use of simple paired t tests or the paired-sample Wilcoxon signed rank test, depending on the shape of the distribution of measurements for that variable. A correction for multiple comparisons was performed (4 independent variables [HLSD, CLSD, NLSD, and FLSD] + stride frequency). Therefore, only comparisons with a value of P < 0.01 (ie, P = 0.05/5) were considered significant.

Figure 3—
Figure 3—

Photographs depicting reference points during a gallop stride for determination of stride timing variables. A—Impact of nonlead hind limb with track surface. B—Impact of lead hind limb with track surface. C—Takeoff of nonlead forelimb from track surface. D—Takeoff of lead forelimb from track surface. The HLSD was the interval between A and B. The CLSD was the interval between B and C. The FLSD was the interval between C and D. The NLSD was the interval between D in the current stride and A in the subsequent stride.

Citation: American Journal of Veterinary Research 77, 10; 10.2460/ajvr.77.10.1121

Signature movement analysis of galloping data— Aligned left and right limb lead signals during galloping often had different shapes that were maintained throughout the lameness experiment. Thus, differences in signals between limb leads before and after induction of lameness were compared with differences between sound and lame trials within a limb lead. Differences were calculated as the square root of the sum of squared differences at each time index between the 2 signals, and then standardized to the percentage difference relative to the total range of the signal in the sound (difference between sound and lame) or left limb lead (difference between left and right limb leads) signals. Signals for right and left limb leads were aligned to start at the time index of impact of each lead limb (Figure 4). Percentage difference between comparisons was assessed by use of a 2-way ANOVA. Values of P ≤ 0.05 were considered significant.

Figure 4—
Figure 4—

Graphs of signature movement analysis representing the difference between signals, which was calculated as the square root of the sum of squared differences at each time index (gray-shaded region). A—Difference in vertical head acceleration between left limb and right limb leads for the sound state. B—Difference in vertical head acceleration between sound and lame states for the nonlame limb lead. C—Difference in vertical head acceleration between sound and lame states for the lame limb lead. D—Difference in vertical head acceleration between left limb and right limb leads for the lame state. In each panel, the black line represents results for the left limb lead or sound state, whereas the dark gray line represents results for the right limb lead or lame state.

Citation: American Journal of Veterinary Research 77, 10; 10.2460/ajvr.77.10.1121

Preference for limb lead—Preference for limb lead was estimated by counting the number of strides for each lead limb at each evaluation time and by review of video recordings. Horses were graded for lead limb preference on a 7-point scale (+3 = strong preference for right limb lead, +2 = moderate preference for right limb lead, +1 = weak preference for right limb lead, 0 = no preference, −1 = weak preference for left limb lead, −2 = moderate preference for left limb lead, and −3 = strong preference for left limb lead). Assessment of limb lead preference was subjective and involved consideration of the number of strides for each limb lead, whether the horse switched limb leads in the middle of a trial segment, and the apparent ease or difficulty for the jockey to induce a change in limb lead to complete data collection. Assessments of preferred limb lead were performed by 2 experienced equine veterinarians (ACOD, KGK), who were not aware of the limb with induced lameness or the score provided by the other veterinarian. A change in limb lead preference toward or away from the limb with induced lameness between baseline and after the most severe lameness was induced was determined for each horse and lameness induction. Strength of association between mean limb lead preference score, number of strides for each limb lead, and lameness severity (VS, Pmax, and Pmin) was assessed by use of the Pearson product moment correlation.

Results

Success of lameness induction

Number of fractional turns of the screw required to induce lameness differed among horses, and all 3 lameness severities were not achieved for both forelimb and hind limb lameness in all 12 horses. However, at least 1 definitive severity of lameness was achieved in the desired limb in all 12 horses. Lameness was induced in the right forelimb of 7 horses and the left forelimb of 5 horses. Lameness was induced in the right hind limb of 6 horses and the left hind limb of 6 horses. One horse had residual mild lameness in the limb (right forelimb) at the time of the second induced lameness. This residual lameness was not detected subjectively, and the second lameness induction was performed in a randomly selected (coin toss) hind limb.

Before induction of forelimb lameness, the VS for 9 of 12 horses was below the threshold of 8.5 mm (mean ± SD, 7.0 ± 3.6 mm; range, 1.6 to 13.6 mm). After induction of the most severe lameness in a forelimb, VS for all 12 horses was > 8.5 mm (mean, 51.8 ± 27.4 mm; range, 9.1 to 100.0 mm), which was a > 5-fold increase in amplitude above threshold. Before induction of hind limb lameness, Pmax was < 3.0 mm for 7 of 12 horses and Pmin was < 3.0 mm for 7 of 12 horses. For all 12 horses, mean ± SD Pmax was 2.1 ± 1.9 mm (range, 0 to 5.2 mm) and mean Pmin was 3.6 ± 3.7 mm (range, 0.1 to 10.2 mm). After induction of the most severe lameness in a hind limb, Pmax was > 3.0 mm for 10 of 12 horses (mean, 9.9 ± 6.6 mm; range, 1.4 to 23.5 mm) and Pmin was > 3.0 mm for 11 of 12 horses (mean, 10.1 ± 4.7 mm; range, 2.1 to 17.9 mm), which was a > 3-fold increase for both Pmax and Pmin above threshold. Induction of the most severe lameness in a forelimb caused a significant (P < 0.001) increase in VS during trotting, and induction of the most severe lameness in a hind limb caused a significant increase in Pmax (P = 0.006) and Pmin (P = 0.011) during trotting.

Visual data mining and statistical analysis

The only variable during galloping that was significantly different between sound and lame limbs was maximum vertical pelvic acceleration for the right limb lead after induction of lameness in the right forelimb. Maximum vertical pelvic acceleration was greater in the lame (right) forelimb lead after induction of forelimb lameness (2.19 g) than before induction of forelimb lameness (1.95 g). This was also reflected in the range of vertical pelvic acceleration. Range of vertical pelvic acceleration for the lame (right) forelimb lead was greater after induction of right forelimb lameness (3.74 g) than before induction of forelimb lameness (3.46 g). Maximum vertical pelvic acceleration and range of vertical pelvic acceleration were not significantly increased for the left limb lead after induction of lameness in the left forelimb. All other variables for pelvic acceleration, head acceleration and position, and pelvic position; stride duration for each limb lead; and number of strides for each limb lead were not significantly different for the left or right limb lead before and after induction of the most severe lameness. None of the variables differed significantly for either limb lead between before and after induction of hind limb lameness. Surprisingly, even when lameness induction caused easily observable and measurable lameness during trotting, head and pelvic signals during galloping were unchanged. In some horses, head and pelvic signals for the left limb lead were different from those for the right limb lead, and the shape of the signals for each limb lead were maintained across different lameness severities and between the 2 times of data collection.

Supervised neural network training

Trained neural networks performed poorly for galloping data with regard to differentiation of sound versus lame, forelimb lameness versus hind limb lameness, and right limb lameness versus left limb lameness (Table 1). Incorrect classification for the gallop gait was higher than for the naïve predictor for all comparisons, except for differentiation of sound versus lame in the left hind limb. Classifier performance on the positive control was perfect (misclassification rate, 0%).

Table 1—

Neural network misclassification rates for determination of horses as not lame (ie, sound) or lame during trotting in a straight line.

 Misclassification rate
Classification comparisonGallopingNaïve*
Sound vs left forelimb lame0.540.43
Sound vs right forelimb lame0.460.41
Sound vs left hind limb lame0.200.28
Sound vs right hind limb lame0.470.34
Forelimb lame vs hind limb lame0.500.37
Left forelimb lame vs right forelimb lame0.650.48
Left hind limb lame vs right hind limb lame0.400.39

Naïve classifier that predicted only the most common state for each comparison.

Timing analysis of limb movement

Mean ± SD stride frequency for all horses in all trials was 1.90 ± 0.06 strides/s (range, 1.75 to 2.01 strides/s). Stride frequency was significantly less for the right limb lead during galloping after induction of lameness in the right forelimb (1.89 strides/s), compared with stride frequency before induction of lameness (1.93 strides/s), but stride frequency was not significantly different for the left limb lead during galloping before and after induction of lameness in the left forelimb. Within each limb lead during galloping, no other stride timing variables differed significantly between sound and lame states for the forelimb or hind limb (Figure 5).

Figure 5—
Figure 5—

Mean gallop stride timing durations for 12 Thoroughbreds before and after induction of lameness in a forelimb (A) or hind limb (B). Results were determined for horses with a left limb lead before induction of lameness (black bars), with a right limb lead before induction of lameness (white bars), with a nonlame lead limb after induction of lameness (crosshatched bars), and with a lame lead limb after induction of lameness (gray bars). There were no significant (P > 0.05) differences between limb leads before or after lameness induction. See Figure 3 for remainder of key.

Citation: American Journal of Veterinary Research 77, 10; 10.2460/ajvr.77.10.1121

Signature movement analysis

Except for pelvic rotation, differences in head and pelvic motion between sound and lame states for both limb leads were of the same magnitude as that between right and left limb leads before lameness induction (Figure 6). The only significant difference was found for pelvic rotation, whereby the difference in pelvic rotation between limb leads for the sound state (22.6%) was higher than that between limb leads for the hind limb lame state (17.5%) and for left (16.5%) and right (8.9%) limb leads between the sound and lame states for the hind limb.

Figure 6—
Figure 6—

Difference in head and pelvic vertical acceleration, angular rotation, vertical position, and angle between limb leads for sound and lame states and within a lead between sound and lame states for 12 Thoroughbreds during galloping before and after induction of lameness of a forelimb (A) or hind limb (B). Results represent differences between the left limb lead and right limb lead for the sound state (black bars), between the left limb lead and right limb lead for the lame state (white bars), between the sound and lame states for the nonlame limb lead (crosshatched bars), and between the sound and lame states for the lame limb lead (gray bars). a–cValues with different letters differ significantly (P ≤ 0.05). HAR = Head angular rotation in the sagittal plane. HAS = Head angle in the sagittal plane. PAF = Pelvic angle in the frontal plane. PAR = Pelvic angular rotation in the frontal plane. VHA = Vertical head acceleration. VHP = Vertical head position. VPA = Vertical pelvic acceleration. VPP = Vertical pelvic position.

Citation: American Journal of Veterinary Research 77, 10; 10.2460/ajvr.77.10.1121

Limb lead preference

Before the first lameness induction, there were approximately equal numbers of horses with right limb lead, left limb lead, and no limb lead preference; 4 horses changed limb lead preference between lameness induction days. After induction of the most severe forelimb lameness, 4 horses had no change in limb lead preference, 5 horses changed from a weak limb lead preference to no limb lead preference, and 1 horse changed from a weak left limb lead preference to a weak right limb lead preference after induction of lameness in the right forelimb (2 horses had no lead preference before induction of lameness). After induction of the most severe hind limb lameness, 6 horses had no change in limb lead preference, 3 horses changed from a weak limb lead preference to no limb lead preference, and 1 horse (left hind limb lameness) changed from a weak right limb lead preference to a weak left limb lead preference. For induction of both forelimb and hind limb lameness, some horses had a change in limb lead preference away from the limb with the induced lameness and some had a change in limb lead preference toward the limb with the induced lameness. There were no significant associations (R2 < 0.02) between subjective mean limb lead preference score or number of gallop strides collected for each limb lead with any of the objective measures of lameness severity during trotting (VS, Pmax, and Pmin).

Discussion

To the authors’ knowledge, meaningful variables for the head, pelvis, limb, or limb lead preference during galloping that consistently indicate forelimb or hind limb lameness have not been published. Maximum pelvic vertical acceleration was greater for the lame forelimb lead, but only when lameness was induced in the right forelimb. We were able to create lameness in the desired limb that was easily detected visually, detected by use of the inertial sensors during trotting in a straight line, or both. These results support a contention that detection of lameness during galloping by use of these motion variables is not sensitive. Although we are unaware of any previous attempts to examine the utility of the gallop gait to determine lameness in horses, this was suggested as early as 188833 on the basis of observations. It is possible that when asymmetric gaits are induced in horses moving in a straight line, mechanisms are no longer available to shift force from a painful limb to other limbs.

In the present study, attempted neural network training and classification of lameness during galloping was unsuccessful. Correct classification by use of data obtained during galloping was worse than that for the naïve predictor, which was little better than a guess.

The authors are not aware of any peer-reviewed reports indicating that timing of hind limb contact with the ground during galloping is altered because of lameness. However, lay34–36 and non–peer-reviewed37–39 publications as well as personal communications with practitioners experienced with equine lameness suggest that changes in hind limb cadence during cantering and galloping are useful for the detection of lameness. For example, the term bunny hopping implies simultaneous impact of the left and right hind limbs and is thought in some cases to be a manifestation of hind limb lameness or pain in the lumbar vertebral region. Ostensibly, a horse with signs of pain during weight bearing will delay the impact of that limb and hasten weight bearing onto nonlame limbs, thus changing the cadence of footfalls. However, this was not found in the present study. Bunny hopping should not be considered a sign of unilateral weight-bearing lameness of either hind limb. Also, a consistent effect on forelimb or hind limb step or interstride durations was not detected.

Forward velocity of each horse was not directly measured; however, stride frequency was determined. At low speeds, stride frequency increases linearly with speed of movement, until it reaches a plateau and remains relatively stable at higher speeds as increasing speed of movement is accomplished with increases in stride length.40 In the study reported here, mean stride frequency for all horses was 1.9 strides/s (with a small range) during all trials on both dates before and after lameness induction. Paradoxically, stride rate increased during the right limb lead for horses with induced lameness of the right forelimb, but the same effect was not evident during the left limb lead for horses with induced lameness of the left forelimb or during any limb lead with induced lameness of the right or left hind limb. Stride rates for the present study are equivalent to speed of movement substantially greater than the typical range for speed of transition between trotting and galloping41 but are near the low end of fast gallop speeds.42 Thus, many of the data collections in the present study were performed in horses that were not at a full gallop, with no overlap between ground contact of lead hind limb and nonlead forelimb; review of the video recordings confirmed this. However, video review of the trials and evaluation of limb impact timing confirmed a 4-beat gait, and speed of movement was much greater during galloping with a rider than during the period when a horse was trotting and did not carry a rider.

The study reported here was conducted with horses moving in a straight line and induction of unilateral weight-bearing lameness of the hoof. Torso orientation was not measured. Vertical torso movement, foot fall pattern, limb tracking, and limb lead preference for a horse moving in a straight line likely differ from results for that horse moving in a circle. Also, lameness causing alteration of movement as a result of pain during limb swing, pain in a more proximal location in a limb, or pain in the lumbar vertebral region, which were not encompassed by the lameness induction method used in the present study, may have yielded different results.

In the present study, limb-mounted gyroscopes were used to detect limb placement and position. This body-mounted inertial sensor systema is available to veterinarians. The gyroscopic sensor in this system detects forelimb stance and breakover as a clear positive deflection in the angular rate signal during weight bearing followed by an abrupt negative deflection as the horse rotates the limb forward. By contrast, unequivocal detection of the beginning of the stance phase requires identification of a high-frequency event at limb impact. This was reliably detected in the hind limbs by the gyroscopes, but this event was not consistently determined for the forelimb signal trajectories, possibly as a result of the damping effect of the deep track surface. Other studies40–45 in which inertial sensors were used to detect limb impact in horses at high-speed galloping gaits have involved high-sampling rate, limb-mounted accelerometers that were capable of detecting high gravitational (± 12 g) acceleration or deceleration. Use of limb-mounted gyroscopes instead of accelerometers required detection of the end of the stance phase for the forelimb for comparison with timings of other limbs. Thus, stride timing determinations for forelimb step duration were only estimates.

Differences in signals between limb leads for the sound state were as large as (or larger than) differences in signals between limb leads for the lame state and within each limb lead between sound and lame states. This was true for induced lameness of both the forelimb and hind limb. This was further evidence that there were natural differences between limb leads that were not attributable to lameness.

Limb lead preference has been reported for Thoroughbreds and Quarter Horses, but findings have been inconsistent.46,47 Before lameness was induced in the present study, some horses had a left limb lead, some had a right limb lead, and some had no limb lead preference; most did not change limb lead preference between the 2 dates of data collection. There was no consistent change in limb lead preference score with induced lameness of the forelimb or hind limb, even when considering only horses with the most severe induced lameness. Only 1 horse during 1 baseline evaluation had more than a moderate limb lead preference, and no horses had moderate to strong limb lead preferences after induction of lameness. Thus, limb lead preference was not a good indicator of pain-induced, weight-bearing lameness in Thoroughbreds galloping in a straight line in the present study. The conditions for data collection and the speed used in this study may have obscured limb lead preference for other conditions, such as moving in a circle, exiting a starting gate from a standstill, and moving at maximum speed. It is also possible because only 1 jockey was involved in the study and the order for collection of limb lead data was not randomized that the jockey caused a random limb lead preference.

A potential confounding factor in the present study was the different surfaces between baseline trotting without a rider and galloping with a rider. Trotting was performed at a slower speed on a firm, fine gravel–based surface, whereas the horses were ridden at faster speeds on a deeper track surface during galloping. We do not know whether induced lameness that was easily seen and measured during trotting on the fine gravel surface would have been as easily seen and measured during trotting on the track. However, a prerace examination by a veterinarian that includes a quick evaluation during trotting in a straight line is not likely to be conducted on a track-like surface, and a horse is unlikely to be galloping on a firm, fine gravel surface. Findings in this study (eg, lameness was easily seen and measured during trotting but not easily seen or measured during galloping on the track) should be applicable to clinical situations.

In the present study, lameness was poorly displayed during galloping. During galloping, horses maintained natural asymmetry of head, pelvic, and limb motion between limb leads that was unrelated to lameness.

Acknowledgments

Supported by the Grayson-Jockey Club Research Foundation Inc, the E. Paige Laurie Endowed Program in Equine Lameness at the University of Missouri, and Equinosis LLC through the Phase II National Science Foundation Small Business Technology Transfer (IIP-STTR) Program.

ABBREVIATIONS

AAEP

American Association of Equine Practitioners

CLSD

Contralateral limb stance duration of a gallop stride

FLSD

Forelimb step duration of a gallop stride

HLSD

Hind limb step duration of a gallop stride

NLSD

No limb stance duration of a gallop stride

Pmax

Maximum pelvic height

Pmin

Minimum pelvic height

VS

Vector sum of vertical head movement asymmetry

Footnotes

a.

Lameness Locator, Equinosis, St Louis, Mo.

b.

Matlab Toolbox, The Mathworks Inc, Natick, Mass.

References

  • 1. USDA APHIS Veterinary Services. Lameness and laminitis in horses. Fort Collins, Colo: Centers for Epidemiology and Animal Health, National Animal Health Monitoring Systems, USDA, 2000.

    • Search Google Scholar
    • Export Citation
  • 2. USDA APHIS Veterinary Services. Equine ′98 assessment survey results. Fort Collins, Colo: Centers for Epidemiology and Animal Health, National Animal Health Monitoring Systems, USDA, 1997.

    • Search Google Scholar
    • Export Citation
  • 3. Pennell JC, Engenvall A, Bonnett BN, et al. Specific causes of morbidity among Swedish horses insured for veterinary care between 1997 and 2000. Vet Rec 2005;157: 470477.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 4. Cole FL, Hodgson DR, Reid SWJ, et al. Owner-reported equine health disorders: results of an Australia-wide postal survey. Aust Vet J 2005;83: 490495.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5. Ross WA, Kaneene JB. An individual-animal-level prospective study of risk factors associated with the occurrence of lameness in the Michigan (USA) equine population. Prev Vet Med 1996;29: 5975.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 6. Dabareiner RM, Cohen ND, Carter GK, et al. Musculoskeletal problems associated with lameness and poor performance among horses used for barrel racing: 118 cases (2000–2003). J Am Vet Med Assoc 2005;227: 16461650.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 7. Kaneene JB, Ross WA, Miller RA. The Michigan Health Monitoring System. II. Frequencies and impact of selected health problems. Prev Vet Med 1997;29: 277292.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 8. Hernandez J, Hawkins DL. Training failure among yearling horses. Am J Vet Res 2001;62: 14181422.

  • 9. Estberg L, Stover SM, Gardner IA, et al. Fatal musculoskeletal injuries incurred during racing and training in Thoroughbreds. J Am Vet Med Assoc 1996;208: 9296.

    • Search Google Scholar
    • Export Citation
  • 10. Stover SM, Johnson BJ, Daft BM, et al. As association between complete and incomplete stress fractures of the humerus in racehorses. Equine Vet J 1992;24: 260263.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 11. Hayens PF, Robinson RA. Racetrack breakdown pilot study summary, in Proceedings. 34th Annu Conv Am Assoc Equine Pract 1989; 673676.

    • Search Google Scholar
    • Export Citation
  • 12. Cohen ND, Peloso JG, Mundy GD, et al. Racing-related factors and results of prerace physical inspection and their association with musculoskeletal injuries incurred in Thoroughbreds during races. J Am Vet Med Assoc 1997;211: 454463.

    • Search Google Scholar
    • Export Citation
  • 13. Keegan KG, Wilson DA, Wilson DJ, et al. Evaluation of mild lameness in horses trotting on a treadmill by clinicians and interns or residents and correlation of their assessments with kinematic gait assessments. Am J Vet Res 1998;59: 13701377.

    • Search Google Scholar
    • Export Citation
  • 14. Thomsen MH, Jensen AT, Sorensen H, et al. Symmetry indices based on accelerometric data in trotting horses. J Biomechanics 2010;43: 26082612.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 15. Keegan KG, Wilson DA, Reed SK, et al. Comparison of a body-mounted inertial sensor system–based method with subjective evaluation for detection of lameness in horses. Am J Vet Res 2013;74: 1724.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 16. Keegan KG, MacAllister CG, Gedon CA, et al. Comparison of an inertial sensor system with a stationary force plate for evaluation of horses with bilateral forelimb lameness. Am J Vet Res 2012;73: 368374.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17. Keegan KG, Kramer J, Yonezawa Y, et al. Assessment of repeatability of a wireless, inertial sensor–based lameness evaluation system for horses. Am J Vet Res 2011;72: 11561153.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 18. Audigie F, Pourcelot P, Degueurce C, et al. Fourier analysis of trunk displacements: a method to identify the lame limb in trotting horses. J Biomechanics 2002;35: 11731182.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 19. Marshall JF, Lung DG, Voute LC. Use of a wireless, inertial sensor–based system to objectively evaluate flexion tests in the horse. Equine Vet J 2012;44: 811.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 20. Kramer J, Keegan KG, Wilson DA, et al. Kinematics of the hind limb in trotting horses after induced lameness of the distal intertarsal and tarsometarsal joints and intra-articular administration of anesthetic. Am J Vet Res 2000;61: 10311036.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 21. McCracken MJ, Kramer J, Keegan KG, et al. Comparison of an inertial sensor system of lameness quantification with subjective lameness evaluation. Equine Vet J 2012;44: 652656.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 22. Ishihara A, Bertone AL, Rajala-Schultz, et al. Association between subjective lameness grade and kinetic gait parameters in horses with experimentally induced forelimb lameness. Am J Vet Res 2005;66: 18051815.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 23. Robilliard JJ, Pfau T, Wilson AM. Gait characterization and classification in horse. J Exp Biol 2007;210: 187197.

  • 24. Nauwelaerts S, Aerts P, Clayton H. Stride to stride variability in joint angle profiles during transitions from trot to canter in horses. Vet J 2013; 198: e59e64.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 25. Nauwelaerts S, Aerts, Clayton H. Spatio-temporal gait characteristics during transitions from trot to canter in horses. Zoology 2013;116: 197204.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 26. Maes L, Abourachid A. Gait transitions and modular organization of mammalian locomotion. J Exp Biol 2013;216: 22572265.

  • 27. Pfau T, Witte TH, Wilson AM. Centre of mass movement and mechanical energy fluctuation during gallop locomotion in the Thoroughbred racehorse. J Exp Biol 2006;209: 37423757.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 28. Farley CT, Taylor CR. A mechanical trigger for the trot-gallop transition in horses. Science 1991;253: 306308.

  • 29. Merkens HW, Schamhardt HC. Evaluation of equine locomotion during different degrees of experimentally induced lameness. I: lameness model and quantification of ground reaction force patterns in the limbs. Equine Vet J Suppl 1988;6: 99106.

    • Search Google Scholar
    • Export Citation
  • 30. Stashak TS. Examination for lameness: history, visual exam, palpation and manipulation. In: Baxter GM, ed. Adam's and Stashak's lameness in horses. 6th ed. Ames, Iowa: Wiley-Blackwell, 2011; 10915.

    • Search Google Scholar
    • Export Citation
  • 31. Levenberg K. A method for the solution of certain non-linear problems in least squares. Q Applied Math 1944;2: 164168.

  • 32. Picard R, Cook D. Cross-validation of regression models. J Am Statistical Assoc 1984;79: 574583.

  • 33. Liautard A. Lameness in horses; symptomology and diagnosis, the designation of the diseased limb. In: Liautard A, ed. Lameness of horses and diseases of the locomotory apparatus. New York: William R. Jenkins, Veterinary Publisher & Bookseller, 1888; 1420.

    • Search Google Scholar
    • Export Citation
  • 34. Nadeau J. Fact sheet. Preventing back pain in horses. Available at: animalscience.uconn.edu/extension/publications/backpain.htm. Accessed Sep 18, 2015.

    • Search Google Scholar
    • Export Citation
  • 35. Coumbe K. Lameness and orthopedic nursing; clinical examination, lunging. In: Coumbe K, ed. Equine veterinary nursing. West Essex, UK; Wiley-Blackwell, 2012; 298322.

    • Search Google Scholar
    • Export Citation
  • 36. Loving N. Joints: health, problems and therapy; gait considerations of specific areas of joint pain. In: Loving N, ed. All systems go: the horse owner's full-color veterinary care and conditioning resource for modern performance, sport, and pleasure horses. North Pomfret, Vt: Trafalgar Square Publishing, 2006; 117121.

    • Search Google Scholar
    • Export Citation
  • 37. Jenks A. Osteopathic treatment without sedation. In: Pusey A, Brooks J, Jenks A, eds. Osteopathy and the treatment of horses. Chichester, West Sussex, England: John Wiley and Sons Ltd, 2010; 106153.

    • Search Google Scholar
    • Export Citation
  • 38. Licka T. Differentiation of ataxic and orthopedic gait abnormalities in the horse. Vet Clin North Am Equine Pract 2011;27: 411416.

  • 39. Johnston K, Nickels FA. The Standardbred racehorse. In: Baxter GM, ed. Adam's and Stashak's lameness in horses. 6th ed. Ames, Iowa: Wiley-Blackwell, 2011; 10391046.

    • Search Google Scholar
    • Export Citation
  • 40. Parsons KJ, Pfau T, Wilson AM. High speed gallop locomotion in the Thoroughbred racehorse: I. The effect of incline on stride parameters. J Exp Biol 2008;211: 934944.

    • Search Google Scholar
    • Export Citation
  • 41. Wickler SJ, Hoyt DF, Cogger EA, et al. The energetics of the trot-gallop transition. J Exp Biol 2003;206: 557564.

  • 42. Witte TH, Hirst CV, Wilson AM. Effect of speed on stride parameters in racehorses at gallop in field conditions. J Exp Biol 2006;209: 43894397.

  • 43. Hoyt DF, Wickler SJ, Dutto DJ, et al. What are the gait relations between mechanics, gait parameters, and energetics in terrestrial locomotion? J Exp Biol 2006; 305A:912922.

    • Search Google Scholar
    • Export Citation
  • 44. Parsons KJ, Wilson AM. The use of MP3 recorders to log data from equine hoof mounted accelerometers. Equine Vet J 2006;38: 675680.

  • 45. Witte TH, Knill K, Wilson AM. Determination of peak vertical ground reaction force from duty factor in the horse (Equus caballus). J Exp Biol 2004;207: 36393648.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 46. McGreevy PD, Thomson PC. Differences in motor laterality between breeds of performance horse. Appl Anim Behav Sci 2006;99: 183190.

  • 47. Murphy J, Sutherland A, Arkins S. Idiosyncratic motor laterality in the horse. Appl Anim Behav Sci 2005;91: 297310.

  • Figure 1—

    Graphs depicting lameness measures obtained during trotting in a horse with induced forelimb lameness (A) and in a horse with induced hind limb lameness (B). In panel A, Hmax was calculated as maximum head position before stance phase of the right forelimb (RF) minus maximum head position before stance phase of the left forelimb (LF), and Hmin was calculated as minimum head position during stance phase of the RF minus minimum head position during stance phase of the LF. In panel B, Pmax was calculated as maximum pelvic position before stance phase of the right hind limb (RH) minus maximum pelvic position before stance phase of the left hind limb (LH), and Pmin was calculated as minimum pelvic position during stance phase of the RH minus minimum pelvic position during stance phase of the LH. Notice that Hmax and Pmax are calculated as the maximum nearest the right limb impact (1) minus the maximum nearest the left limb impact (2), and Hmin and Pmin are calculated as the minimum during the right limb stance (3) minus the minimum during the left limb stance (4). Gray lines represent results for single strides, and the black line represents the median value for all strides during one lameness induction.

  • Figure 2—

    Graphs for vertical head acceleration (A), vertical head height (B), vertical pelvic acceleration (C), and vertical pelvic height (D)in a horse during galloping. Gray lines represent results for single strides, and median values for all strides are indicated for the left limb lead (solid black line) and right limb lead (dashed black line). Variables of the signals were stride maximum (gray circles), stride minimum (gray stars), and time indices within the stride (gray arrows).

  • Figure 3—

    Photographs depicting reference points during a gallop stride for determination of stride timing variables. A—Impact of nonlead hind limb with track surface. B—Impact of lead hind limb with track surface. C—Takeoff of nonlead forelimb from track surface. D—Takeoff of lead forelimb from track surface. The HLSD was the interval between A and B. The CLSD was the interval between B and C. The FLSD was the interval between C and D. The NLSD was the interval between D in the current stride and A in the subsequent stride.

  • Figure 4—

    Graphs of signature movement analysis representing the difference between signals, which was calculated as the square root of the sum of squared differences at each time index (gray-shaded region). A—Difference in vertical head acceleration between left limb and right limb leads for the sound state. B—Difference in vertical head acceleration between sound and lame states for the nonlame limb lead. C—Difference in vertical head acceleration between sound and lame states for the lame limb lead. D—Difference in vertical head acceleration between left limb and right limb leads for the lame state. In each panel, the black line represents results for the left limb lead or sound state, whereas the dark gray line represents results for the right limb lead or lame state.

  • Figure 5—

    Mean gallop stride timing durations for 12 Thoroughbreds before and after induction of lameness in a forelimb (A) or hind limb (B). Results were determined for horses with a left limb lead before induction of lameness (black bars), with a right limb lead before induction of lameness (white bars), with a nonlame lead limb after induction of lameness (crosshatched bars), and with a lame lead limb after induction of lameness (gray bars). There were no significant (P > 0.05) differences between limb leads before or after lameness induction. See Figure 3 for remainder of key.

  • Figure 6—

    Difference in head and pelvic vertical acceleration, angular rotation, vertical position, and angle between limb leads for sound and lame states and within a lead between sound and lame states for 12 Thoroughbreds during galloping before and after induction of lameness of a forelimb (A) or hind limb (B). Results represent differences between the left limb lead and right limb lead for the sound state (black bars), between the left limb lead and right limb lead for the lame state (white bars), between the sound and lame states for the nonlame limb lead (crosshatched bars), and between the sound and lame states for the lame limb lead (gray bars). a–cValues with different letters differ significantly (P ≤ 0.05). HAR = Head angular rotation in the sagittal plane. HAS = Head angle in the sagittal plane. PAF = Pelvic angle in the frontal plane. PAR = Pelvic angular rotation in the frontal plane. VHA = Vertical head acceleration. VHP = Vertical head position. VPA = Vertical pelvic acceleration. VPP = Vertical pelvic position.

  • 1. USDA APHIS Veterinary Services. Lameness and laminitis in horses. Fort Collins, Colo: Centers for Epidemiology and Animal Health, National Animal Health Monitoring Systems, USDA, 2000.

    • Search Google Scholar
    • Export Citation
  • 2. USDA APHIS Veterinary Services. Equine ′98 assessment survey results. Fort Collins, Colo: Centers for Epidemiology and Animal Health, National Animal Health Monitoring Systems, USDA, 1997.

    • Search Google Scholar
    • Export Citation
  • 3. Pennell JC, Engenvall A, Bonnett BN, et al. Specific causes of morbidity among Swedish horses insured for veterinary care between 1997 and 2000. Vet Rec 2005;157: 470477.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 4. Cole FL, Hodgson DR, Reid SWJ, et al. Owner-reported equine health disorders: results of an Australia-wide postal survey. Aust Vet J 2005;83: 490495.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5. Ross WA, Kaneene JB. An individual-animal-level prospective study of risk factors associated with the occurrence of lameness in the Michigan (USA) equine population. Prev Vet Med 1996;29: 5975.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 6. Dabareiner RM, Cohen ND, Carter GK, et al. Musculoskeletal problems associated with lameness and poor performance among horses used for barrel racing: 118 cases (2000–2003). J Am Vet Med Assoc 2005;227: 16461650.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 7. Kaneene JB, Ross WA, Miller RA. The Michigan Health Monitoring System. II. Frequencies and impact of selected health problems. Prev Vet Med 1997;29: 277292.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 8. Hernandez J, Hawkins DL. Training failure among yearling horses. Am J Vet Res 2001;62: 14181422.

  • 9. Estberg L, Stover SM, Gardner IA, et al. Fatal musculoskeletal injuries incurred during racing and training in Thoroughbreds. J Am Vet Med Assoc 1996;208: 9296.

    • Search Google Scholar
    • Export Citation
  • 10. Stover SM, Johnson BJ, Daft BM, et al. As association between complete and incomplete stress fractures of the humerus in racehorses. Equine Vet J 1992;24: 260263.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 11. Hayens PF, Robinson RA. Racetrack breakdown pilot study summary, in Proceedings. 34th Annu Conv Am Assoc Equine Pract 1989; 673676.

    • Search Google Scholar
    • Export Citation
  • 12. Cohen ND, Peloso JG, Mundy GD, et al. Racing-related factors and results of prerace physical inspection and their association with musculoskeletal injuries incurred in Thoroughbreds during races. J Am Vet Med Assoc 1997;211: 454463.

    • Search Google Scholar
    • Export Citation
  • 13. Keegan KG, Wilson DA, Wilson DJ, et al. Evaluation of mild lameness in horses trotting on a treadmill by clinicians and interns or residents and correlation of their assessments with kinematic gait assessments. Am J Vet Res 1998;59: 13701377.

    • Search Google Scholar
    • Export Citation
  • 14. Thomsen MH, Jensen AT, Sorensen H, et al. Symmetry indices based on accelerometric data in trotting horses. J Biomechanics 2010;43: 26082612.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 15. Keegan KG, Wilson DA, Reed SK, et al. Comparison of a body-mounted inertial sensor system–based method with subjective evaluation for detection of lameness in horses. Am J Vet Res 2013;74: 1724.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 16. Keegan KG, MacAllister CG, Gedon CA, et al. Comparison of an inertial sensor system with a stationary force plate for evaluation of horses with bilateral forelimb lameness. Am J Vet Res 2012;73: 368374.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17. Keegan KG, Kramer J, Yonezawa Y, et al. Assessment of repeatability of a wireless, inertial sensor–based lameness evaluation system for horses. Am J Vet Res 2011;72: 11561153.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 18. Audigie F, Pourcelot P, Degueurce C, et al. Fourier analysis of trunk displacements: a method to identify the lame limb in trotting horses. J Biomechanics 2002;35: 11731182.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 19. Marshall JF, Lung DG, Voute LC. Use of a wireless, inertial sensor–based system to objectively evaluate flexion tests in the horse. Equine Vet J 2012;44: 811.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 20. Kramer J, Keegan KG, Wilson DA, et al. Kinematics of the hind limb in trotting horses after induced lameness of the distal intertarsal and tarsometarsal joints and intra-articular administration of anesthetic. Am J Vet Res 2000;61: 10311036.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 21. McCracken MJ, Kramer J, Keegan KG, et al. Comparison of an inertial sensor system of lameness quantification with subjective lameness evaluation. Equine Vet J 2012;44: 652656.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 22. Ishihara A, Bertone AL, Rajala-Schultz, et al. Association between subjective lameness grade and kinetic gait parameters in horses with experimentally induced forelimb lameness. Am J Vet Res 2005;66: 18051815.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 23. Robilliard JJ, Pfau T, Wilson AM. Gait characterization and classification in horse. J Exp Biol 2007;210: 187197.

  • 24. Nauwelaerts S, Aerts P, Clayton H. Stride to stride variability in joint angle profiles during transitions from trot to canter in horses. Vet J 2013; 198: e59e64.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 25. Nauwelaerts S, Aerts, Clayton H. Spatio-temporal gait characteristics during transitions from trot to canter in horses. Zoology 2013;116: 197204.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 26. Maes L, Abourachid A. Gait transitions and modular organization of mammalian locomotion. J Exp Biol 2013;216: 22572265.

  • 27. Pfau T, Witte TH, Wilson AM. Centre of mass movement and mechanical energy fluctuation during gallop locomotion in the Thoroughbred racehorse. J Exp Biol 2006;209: 37423757.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 28. Farley CT, Taylor CR. A mechanical trigger for the trot-gallop transition in horses. Science 1991;253: 306308.

  • 29. Merkens HW, Schamhardt HC. Evaluation of equine locomotion during different degrees of experimentally induced lameness. I: lameness model and quantification of ground reaction force patterns in the limbs. Equine Vet J Suppl 1988;6: 99106.

    • Search Google Scholar
    • Export Citation
  • 30. Stashak TS. Examination for lameness: history, visual exam, palpation and manipulation. In: Baxter GM, ed. Adam's and Stashak's lameness in horses. 6th ed. Ames, Iowa: Wiley-Blackwell, 2011; 10915.

    • Search Google Scholar
    • Export Citation
  • 31. Levenberg K. A method for the solution of certain non-linear problems in least squares. Q Applied Math 1944;2: 164168.

  • 32. Picard R, Cook D. Cross-validation of regression models. J Am Statistical Assoc 1984;79: 574583.

  • 33. Liautard A. Lameness in horses; symptomology and diagnosis, the designation of the diseased limb. In: Liautard A, ed. Lameness of horses and diseases of the locomotory apparatus. New York: William R. Jenkins, Veterinary Publisher & Bookseller, 1888; 1420.

    • Search Google Scholar
    • Export Citation
  • 34. Nadeau J. Fact sheet. Preventing back pain in horses. Available at: animalscience.uconn.edu/extension/publications/backpain.htm. Accessed Sep 18, 2015.

    • Search Google Scholar
    • Export Citation
  • 35. Coumbe K. Lameness and orthopedic nursing; clinical examination, lunging. In: Coumbe K, ed. Equine veterinary nursing. West Essex, UK; Wiley-Blackwell, 2012; 298322.

    • Search Google Scholar
    • Export Citation
  • 36. Loving N. Joints: health, problems and therapy; gait considerations of specific areas of joint pain. In: Loving N, ed. All systems go: the horse owner's full-color veterinary care and conditioning resource for modern performance, sport, and pleasure horses. North Pomfret, Vt: Trafalgar Square Publishing, 2006; 117121.

    • Search Google Scholar
    • Export Citation
  • 37. Jenks A. Osteopathic treatment without sedation. In: Pusey A, Brooks J, Jenks A, eds. Osteopathy and the treatment of horses. Chichester, West Sussex, England: John Wiley and Sons Ltd, 2010; 106153.

    • Search Google Scholar
    • Export Citation
  • 38. Licka T. Differentiation of ataxic and orthopedic gait abnormalities in the horse. Vet Clin North Am Equine Pract 2011;27: 411416.

  • 39. Johnston K, Nickels FA. The Standardbred racehorse. In: Baxter GM, ed. Adam's and Stashak's lameness in horses. 6th ed. Ames, Iowa: Wiley-Blackwell, 2011; 10391046.

    • Search Google Scholar
    • Export Citation
  • 40. Parsons KJ, Pfau T, Wilson AM. High speed gallop locomotion in the Thoroughbred racehorse: I. The effect of incline on stride parameters. J Exp Biol 2008;211: 934944.

    • Search Google Scholar
    • Export Citation
  • 41. Wickler SJ, Hoyt DF, Cogger EA, et al. The energetics of the trot-gallop transition. J Exp Biol 2003;206: 557564.

  • 42. Witte TH, Hirst CV, Wilson AM. Effect of speed on stride parameters in racehorses at gallop in field conditions. J Exp Biol 2006;209: 43894397.

  • 43. Hoyt DF, Wickler SJ, Dutto DJ, et al. What are the gait relations between mechanics, gait parameters, and energetics in terrestrial locomotion? J Exp Biol 2006; 305A:912922.

    • Search Google Scholar
    • Export Citation
  • 44. Parsons KJ, Wilson AM. The use of MP3 recorders to log data from equine hoof mounted accelerometers. Equine Vet J 2006;38: 675680.

  • 45. Witte TH, Knill K, Wilson AM. Determination of peak vertical ground reaction force from duty factor in the horse (Equus caballus). J Exp Biol 2004;207: 36393648.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 46. McGreevy PD, Thomson PC. Differences in motor laterality between breeds of performance horse. Appl Anim Behav Sci 2006;99: 183190.

  • 47. Murphy J, Sutherland A, Arkins S. Idiosyncratic motor laterality in the horse. Appl Anim Behav Sci 2005;91: 297310.

Advertisement