Lameness is one of the most common reasons that a horse is examined by a veterinarian.1,2 However, lameness diagnosis and evaluation, particularly in horses with mild lameness, can be difficult, and examiners do not always agree on the affected limb.3 Mild lameness can be a precursor to a more severe problem and can also result in suboptimal performance, leading to losses in prizes and potential sales.4 Furthermore, successful treatment of an injury requires accurate diagnosis. For these reasons and more, improvements are critically needed in techniques for lameness detection, evaluation, and reevaluation.
Over the past several decades, interest has increased in the development of objective lameness detection systems to supplement the subjective lameness examination. Several systems that can be used to accurately detect lameness have been evaluated.5–9 Many of these systems are more easily used in a gait analysis laboratory than in the field, given that the equipment is massive or requires a complicated setup.6,8 A current movement is to develop horse-mounted gait analysis systems, with quicker setup times and equipment that does not require a gait analysis laboratory.
Various horse-mounted systems have been investigated and are accurate for the diagnosis of mild lameness.7,9–12 One of these inertial sensor systemsa has achieved widespread clinical use. Its benefits include ease of use, portability, and quick setup,13 and a study11 revealed that this system could be used to identify mild lameness in horses at a lower severity of lameness than could be detected by a group of experienced veterinarians.11 This system is also marketed to provide objective data to supplement the subjective lameness examination.12
Guidelines have been created for placement of the system's sensors on the basis of certain anatomic landmarks. However, the effect of exact location of each sensor on kinematic data obtained with the system has not been reported to the authors' knowledge. The 2 sensors that have the greatest opportunity to be positioned incorrectly on a horse are those placed on the right forelimb and pelvic region; the head sensor is attached to a bonnet and generally is fairly easily positioned on the midline. Clinically, we have noticed that the sensor on the right forepastern occasionally rotates medially during data collection sessions. This malposition may also occur if the pastern wrap is not positioned appropriately, placing the sensor either slightly lateral or medial to midline. The recommended placement of the pelvic sensor for this particular inertial sensor system is on dorsal midline between the tubera sacrale. For some horses, palpation of the tubera sacrale can be more difficult owing to the gluteal musculature. In addition, sensor placement by inexperienced users or on fractious horses may result in malposition of the pelvic sensor.
The objective of the study reported here was to examine the effect of a change in location of the right forelimb and pelvic sensors on the kinematic data for the forelimbs and hind limbs measured in trotting horses by use of a particular inertial sensor system. We hypothesized that moving the location of the right forelimb sensor medially or laterally would not significantly influence the associated forelimb or hind limb kinematic data. Second, we hypothesized that the pelvic sensor location would not have a significant effect on forelimb-associated kinematic data but that hind limb–associated data would be significantly altered.
Materials and Methods
Horses
Twelve Quarter Horse crosses (6 mares and 6 geldings) between 2 and 5 years of age were used in the study. Mean ± SD body weight was 432 ± 34 kg (950 ± 75 lb). These horses were part of a separate investigation involving the creation of full-thickness cartilage defects within the femoropatellar joint. All had been trained to trot on the high-speed treadmillb and received treadmill exercise 5 d/wk. Horses were tied into the treadmill with straps attached to the left and right halter rings, but head movement was not restricted. All data collection was performed with horses during their daily exercise routine on the high-speed treadmill. All study procedures were approved by the Colorado State University Institutional Animal Care and Use Committee.
Subjective assessment of lameness
Prior to instrumentation of horses and data collection, an examiner (VJM) assessed all horses for lameness by having them walk and trot in a straight line over ground. A lameness grading scale from 0 to 5 was used to assign a grade to each limb, by which 0 indicated no lameness; 1 indicated intermittent, inconsistent lameness while trotting; 2 indicated mild but consistent lameness while trotting; 3 represented moderate but consistent lameness while trotting; 4 indicated lameness while walking and trotting; and 5 represented minimal to non–weight-bearing lameness while walking. All horses had mild to moderate lameness in at least 1 limb. No horse had a lameness grade > 3 in any limb, and lameness did not subjectively change during data collection. Half grades were assigned, when appropriate.
Inertial sensor system
All horses had previously been instrumented with the inertial sensor system,a which consisted of 3 sensors. A uniaxial accelerometer was attached to a felt head bonnet with hook-and-loop tape,c a pelvic uniaxial accelerometer was mounted between the tubera sacrale by means of the same tape and reinforced with duct tape, and a gyroscope was fastened by pastern wrap to the right forelimb. In addition to the microelectrical-mechanical device (accelerometer or gyroscope), each sensor also contained a radio transceiver with antenna, a battery, and a previously described microcontroller.13 Each sensor sampled data at 200 Hz and communicated wirelessly with a previously described portable computer.11
Inertial sensor variables
Values of kinematic variables were calculated by means of commercial software as previously described.13,14 Briefly, stride number and rate were determined by the right forelimb gyroscope. Vertical displacement of the head and pelvis was determined by double integration of the accelerometer data through the use of specially designed algorithms, which have been described elsewhere.5,11,13,14 Kinematic output variables captured by use of this inertial sensor system included stride number, stride rate, HDMax, HDMin, VS, PDMax, and PDMin. The variables HDMax, HDMin, PDMax, and PDMin represented the calculated differences in the maximum or minimum head or pelvis position between the left and right portions of stride during a single trot trial.
Experimental protocol
Horses were examined while trotting on a high-speed treadmill, and data collection sessions occurred during 2 separate daily treadmill sessions. The right forelimb sensor was tested at 3 anatomic locations in random order by use of a random-number generator: dorsal midline, 2 cm medial to the dorsal midline, and 2 cm lateral to the dorsal midline. Acrylic paint was used to mark the dorsal aspect of the right forelimb at the level of the metacarpophalangeal joint. The medial and lateral locations were measured from this midline location and were also marked with acrylic paint (Figure 1). The pelvic sensor was tested at 5 anatomic locations in random order: dorsal midline between the tubera sacrale, 2 cm to the right and left of the midline, and 2 cm cranial and caudal to the tubera sacrale on the midline. Location of midline (normal sensor position) between the tubera sacrale was determined by palpation, and this location was marked with acrylic paint. From this location, all other locations were measured and marked with the same acrylic paint. The sensor was positioned centered on the acrylic paint mark.
Photograph of a sensor applied to the pastern region of a horse in preparation for kinematic gait analysis by use of an inertial sensor system. Acrylic paint was used to mark the medial, midline, and lateral locations for sensor application. In this photograph, the sensor appears centered at the medial location. Additional white tape was placed distal to the sensor to ensure that sensor location was consistent throughout the gait trials.
Citation: Journal of the American Veterinary Medical Association 250, 5; 10.2460/javma.250.5.548
When the location of the right forelimb or pelvic sensor was changed, the treadmill was stopped but the horse remained on the treadmill. After repositioning of the sensor, the treadmill was restarted and returned to the same velocity (4.0 to 4.1 m/s) as used in the previous trials. Prior to data collection for each sensor location, all horses were allowed to reacclimate to the treadmill velocity for 15 to 30 seconds. Following this acclimation, two 20- to 30-second trot trials were recorded to collect data for 25 strides/trial. In total, data for 6 trials were collected for the 3 right forelimb sensor locations (2 trials/location), and data for 10 trials were collected for the 5 pelvic sensor locations (2 trials/location). For each horse, a similar treadmill velocity was used for all gait trials.
Statistical analysis
Statistical softwared was used for all statistical analyses. Normality of value distributions for HDMax, HDMin, VS, PDMax, and PDMin was evaluated by use of the Shapiro-Wilk test. Because of nonparametric distributions, values of kinematic variables were ranked. Mixed-model ANOVA was performed to identify differences in values of kinematic variables among the right forelimb sensor locations and among the pelvic sensor locations. In those models, horse was entered as a random variable, and sensor position was entered as a fixed categorical variable, with the midline position set as the control value. For the pelvic sensor trials, laterality of the pelvic sensor was analyzed in 2 ways: sensor toward the right or left and sensor toward or away from the lame or lamest hind limb. Values of P < 0.05 were considered significant.
Results
All 12 horses had mild to moderate lameness (subjective grade,8,9 1 to 2.5/5) in at least 1 limb. Over the data collection period, no subjective change in lameness was identified. Eleven horses had unilateral forelimb lameness (9 with lameness of the right forelimb and 2 with lameness of the left forelimb). Median lameness grade for the 11 lame forelimbs was 1/5 (range, 1 to 2/5). All horses had lameness of at least 1 hind limb, for a total of 20 lame hind limbs as determined both subjectively and objectively. Four horses had unilateral hind limb lameness (2 with lameness of the right hind limb and 2 with lameness of the left hind limb), and the remainder had bilateral lameness. Median lameness grade for the 20 lame hind limbs was 1.25/5 (range, 1 to 2.5/5).
When the location of the right forelimb sensor was changed, no significant differences in kinematic data were identified among locations for forelimbs or hind limbs (Table 1). When the location of the pelvic sensor was changed, no differences in kinematic data were identified among locations for the forelimbs, but significant differences in PDMin were identified among locations for the hind limbs. Specifically, when data were analyzed with the pelvic sensor location defined as right or left of midline, significant differences in PDMin were identified when the sensor was moved to the right (P < 0.001) and to the left (P = 0.01) as well as cranially (P = 0.009), compared with values for the midline location (Table 2). When data were analyzed with sensor movement defined as toward or away from the lame hind limb, significant differences in PDMin were also identified between values for the midline location and for sensor location away from the lame hind limb (P = 0.01) and for movement of the sensor cranially relative to the midline location (P = 0.03; Table 3).
Median (interquartile range) values of kinematic variables by location of the right forelimb inertial sensor for 12 mildly to moderately lame horses evaluated while trotting on a high-speed treadmill.
| Variable | Midline | Medial | Lateral | P value* |
|---|---|---|---|---|
| HDMax (mm) | 6.85 (−2.93 to 17.47) | 4.35 (−4.15 to 10.67) | 7.43 (−2.29 to 15.05) | 0.25 |
| HDMin (mm) | 0.74 (−10.36 to 6.37) | 3.70 (−4.40 to 6.51) | 1.20 (−0.95 to 9.82) | 0.14 |
| VS (mm) | 15.44 (9.49 to 29.93) | 13.23 (8.75 to 22.95) | 12.76 (8.51 to 24.51) | 0.15 |
| PDMax (mm) | 0.31 (−4.53 to 3.80) | −1.46 (−4.64 to 2.83) | 0.01 (−4.01 to 3.92) | 0.16 |
| PDMin (mm) | −3.14 (−5.27 to 0.39) | −1.48 (−8.51 to 0.27) | −2.77 (−5.50 to 0.12) | 0.88 |
Represents results of overall F test from ANOVA of differences among positions; P values < 0.05 were considered significant.
The right forelimb sensor was tested at 3 anatomic locations in random order: dorsal midline, 2 cm medial to the dorsal midline, and 2 cm lateral to the dorsal midline. Two 20- to 30-second trot trials were recorded to collect data for 25 strides/trial. In total, data for 6 trials were collected for the 3 right forelimb sensor locations (2 trials/location), and data for 10 trials were collected for the 5 pelvic sensor locations (2 trials/location). For each horse, a similar treadmill velocity was used for all gait trials.
Median (interquartile range) values of kinematic variables by location of the pelvic inertial sensor for the horses in Table 1, with laterality of the sensor analyzed as sensor toward the right or left side.
| Variable | Midline | Right | Left | Cranial | Caudal | P value* |
|---|---|---|---|---|---|---|
| HDMax (mm) | 6.83 (0.28 to 22.06) | 9.15 (2.24 to 18.42) | 6.68 (−2.25 to 25.34) | 9.29 (3.10 to 20.80) | 6.65 (−0.74 to 17.99) | 0.26 |
| HDMin (mm) | 3.43 (−1.77 to 8.87) | 3.14 (−7.75 to 11.56) | 1.37 (−4.67 to 11.91) | 1.32 (−0.55 to 11.19) | 6.64 (0.03 to 13.12) | 0.06 |
| VS (mm) | 15.22 (9.16 to 24.60) | 15.40 (12.13 to 23.20) | 15.98 (10.64 to 30.97) | 17.26 (14.64 to 24.56) | 11.39 (8.59 to 18.80) | 0.07 |
| PDMax (mm) | −2.06 (−6.66 to 2.26) | 0.10 (−3.65 to 1.21) | −0.77 (−4.59 to 3.63) | −2.11 (−5.56 to 3.21) | −0.70 (−4.69 to 3.52) | 0.13 |
| PDMin (mm) | −1.31 (−3.36 to 0.48) | 2.22† (−0.34 to 5.36) | −4.83† (−6.91 to 2.23) | 0.81† (−2.12 to 3.19) | −1.82 (−4.26 to 0.25) | < 0.001 |
Value differs significantly from the corresponding midline value.
The pelvic sensor was tested at 5 anatomic locations in random order: dorsal midline between the tubera sacrale, 2 cm to the right and left of the midline, and 2 cm cranial and caudal to the tubera sacrale on the midline.
See Table 1 for remainder of key.
Median (interquartile range) values of kinematic variables by location of the pelvic inertial sensor for the horses in Table 1, with laterality of the pelvic sensor analyzed as toward or away from the lame or lamest hind limb.
| Variable | Midline | Toward lame limb | Away from lame limb | Cranial | Caudal | P value* |
|---|---|---|---|---|---|---|
| HDMax (mm) | 6.83 (0.28 to 22.06) | 6.63 (−0.90 to 25.34) | 8.57 (1.66 to 18.42) | 9.29 (3.10 to 20.80) | 6.65 (−0.74 to 17.99) | 0.50 |
| HDMin (mm) | 3.43 (−1.77 to 8.87) | 1.29 (−7.08 to 12.23) | 4.0 (−5.34 to 11.51) | 1.32 (−0.55 to 11.19) | 6.64 (0.03 to 13.12) | 0.06 |
| VS (mm) | 15.22 (9.16 to 24.60) | 15.98 (11.61 to 30.97) | 15.16 (12.75 to 23.20) | 17.26 (14.64 to 24.56) | 11.39 (8.59 to 18.80) | 0.07 |
| PDMax (mm) | −2.06 (−6.66 to 2.26) | 0.25 (−4.44 to 2.15) | −1.01 (−4.74 to 1.96) | −2.11 (−5.56 to 3.21) | −0.71 (−4.70 to 3.52) | 0.14 |
| PDMin (mm) | −1.31 (−3.46 to 0.48) | −3.07 (−5.79 to 1.84) | 0.67† (−2.23 to 5.36) | 0.81† (−2.12 to 3.20) | −1.82 (−4.26 to 0.25) | 0.003 |
Discussion
The results of the present study supported our hypothesis that moving the location of the right forelimb sensor medially or laterally would have no influence on data recorded by a particular inertial sensor system for the forelimbs and hind limbs of trotting horses with mild to moderate lameness. Findings suggested that shifting this sensor a small amount (up to 2 cm) medially or laterally before or during data collection should not result in acquisition of erroneous forelimb or hind limb kinematic data from the system used. Because data from this sensor are used by the system to determine stride rate and right versus left halves of stride, it is logical that the sensor does not have to be perfectly positioned for the system to function accurately. In addition, changing the location of the right forelimb sensor had no significant effect on stride rate. Because the velocity of individual horses remained consistent throughout data collection, stride rate could also be expected to remain consistent. Because we did not move this sensor to a true medial or lateral position, we are unable to make any conclusions regarding results to be expected should the sensor be rotated farther than 2 cm from midline. However, during clinical use, as long as the pastern wrap is tightened and positioned adequately prior to data collection, the right forelimb sensor should not subjectively appear to rotate more than a few centimeters away from midline.
Altering the location of the pelvic sensor had no significant effect on the associated forelimb kinematic variables (HDMax and HDMin). This would be expected, given that the head and pelvic accelerometers work independently to measure symmetry in head and pelvic motion, respectively. The hind limb variable PDMax was also not significantly affected by change in sensor location, which indicated that PDMax may be more stable in response to changes in pelvic sensor location, at least in the group of horses used in the study. In another study,15 PDMax, but not PDMin, changed significantly in horses following a positive response to a full-limb flexion test. Although the nature of lameness in that study was not reported, the horses that were used likely differed in that respect. The group of horses in the present study had surgically created cartilage lesions within the femoropatellar joint. Although full lameness examinations were not performed to identify the site of the responsible lesion, all horses were known to have developed lameness following surgical induction of lesions, thereby making the femoropatellar joint the likely site. It remains unknown whether sensor location influences PDMax data in horses with hind limb lameness attributable to lesions other than those of the femoropatellar joint.
The PDMax measures the maximum upward motion of the pelvis, which occurs at the end of the stance phase.13 This variable would indicate symmetry of pelvic motion during push-off, with asymmetry indicating a difference in the push-off phase of stance. Conversely, PDMin measures the difference in minimum displacement of the pelvis during the first half of stance, which would be a measure of the impact phase. In the horses of the present study, a predominance of a primary push-off–type lameness was identified, with the primary hind limb lameness classified as push-off in 9 horses and impact in the other 3 horses (data not shown). In this group of horses, it appeared that movement of the sensor changed the symmetry measurement at impact. In a group of horses with other sources of hind limb lameness, particularly in horses with a primary impact-type lameness, the results reported here may not be applicable. Additional investigation involving horses with other sources of hind limb lameness may be warranted to determine whether PDMax data obtained with the inertial sensor system are also affected by pelvic sensor location.
Altering the location of the pelvic sensor on the horses in the present study had a significant effect on the variable PDMin. This output variable of the inertial sensor system was significantly influenced by sensor position regardless of whether the data were analyzed by movement of the pelvic sensor to the right or left side of midline or the sensor location was defined as toward or away from the lame limb. The median PDMin value from the pelvic sensor in a midline location had a slightly negative value (Tables 1 and 2), indicating predominantly more left hind limb lameness than right hind limb lameness in this group of horses. When the sensor was moved to the right of midline or away from the lame limb, this value became more positive, indicating a predominance of right hind limb lameness. Alternatively, when the sensor was moved toward the left, PDMin values became significantly (P = 0.01) more negative, indicating more left hind limb lameness. However, the individual comparison of PDMin values with the sensor toward the lame limb versus the midline location was not significant. When the pelvic sensor was moved cranially, the median PDMin value changed significantly from a negative to a positive value, again indicating a change from predominantly left-sided to predominantly right-sided lameness. Because a more cranial location would have placed the sensor over the caudal lumbar vertebrae, the change in PDMin may have been secondary to differences between lumbar vertebral motion and sacral or pelvic motion. Moving the sensor caudally had no effect on PDMin, likely because the sensor would have still been positioned over the pelvis and therefore would have been subject to motion similar to that of the sacrum.
A limitation of the present study was that horses were evaluated while trotting on a treadmill rather than over ground. A major benefit of using a treadmill is the ability to maintain a constant velocity both within and between trials. Although another investigation16 showed that changes in velocity had no significant effect on results of objective lameness detection, we believed it best to control velocity in the present study to avoid introduction of a confounding effect. The nature of lameness can change over time, so we also believed that use of the treadmill was justified because it allowed expedient collection of data, thereby minimizing the potential for changes in lameness during data collection. Additionally, the order of sensor position was randomized for both the right forelimb and pelvic sensor trials to minimize any potential confounding of minor changes in lameness. Subjectively, all horses maintained approximately the same degree of lameness throughout the data collection period, so we do not believe that changes in lameness influenced these results. Previously, values of several equine kinematic variables, including hoof height, limb retraction, and stride and stance duration, were shown to differ significantly between trotting over ground versus on a treadmill.17 Another investigation18 revealed a significant decrease in lateral range of motion of the lumbar portion of the vertebral column but no significant difference in vertical range of motion of the thoracolumbar portion when overground and treadmill motion were compared. To our knowledge, no study has been reported regarding differences in head and pelvis motion between overground and treadmill trotting, thereby leaving an avenue for future research.
Results of the present study suggested that the location of pelvic sensor placement is an important consideration when using the evaluated inertial sensor system for lameness detection in horses. We only examined the placement of the sensor at 2-cm distances from the midline location, and additional research is needed to determine whether movement of the pelvic sensor by a shorter distance would have less impact on PDMin values. Regardless, it is important to consider pelvic sensor placement if this system is used for repeated assessment of a horse with reapplication of the system, either for short- or long-term reevaluation, because small changes in the location of the pelvic sensor could influence these results. Findings suggested that the inertial sensors should be applied by someone experienced in palpation of the sacrum to ensure adequate positioning of the pelvic sensor on horses. In addition, findings supported those of a previous study12 that indicated the results of this system should be used in conjunction with findings from the subjective lameness assessment. If the output of the evaluated inertial sensor system does not fit with clinical findings of the veterinarian, then repeated palpation of the tubera sacrale and potential repositioning of the pelvic sensor may be indicated.
Acknowledgments
Horses used for this investigation were a part of a larger study funded by the National Institutes of Health (NIH grant No. AR047702-07A1).
Presented in abstract form at the 2015 American College of Veterinary Surgery Summit, Nashville, Tenn, October 2015, and at the 61st Annual Convention of the American Association of Equine Practitioners, Las Vegas, December 2015.
The authors thank Dr. Francisco Olea-Popelka for assistance with statistical analysis and Dr. Chris Morrow for his generous contribution of the inertial sensor system.
ABBREVIATIONS
| HDMax | Maximum difference in head motion |
| HDMin | Minimum difference in head motion |
| PDMax | Maximum difference in pelvic motion |
| PDMin | Minimum difference in pelvic motion |
| VS | Vector sum |
Footnotes
Lameness Locator, Equinosis LLC, St Louis, Mo.
EquiGym high-speed treadmill, EquiGym LLC, Lexington, Ky.
Dual-lock tape, 3M, Saint Paul, Minn.
STATA, version 13.1, StataCorp, College Station, Tex.
References
1. Kaneene JB, Ross WA, Miller RA. The Michigan equine monitoring system. II. Frequencies and impact of selected health problems. Prev Vet Med 1997; 29:277–292.
2. USDA. National economic cost of equine lameness, colic, and equine protozoal myeloencephalitis (EPM) in the United States. Information sheet No. N348.1001. Fort Collins, Colo: USDA APHIS Veterinary Services Health Monitoring System, 2001.
3. Keegan KG, Dent EV, Wilson DA, et al. Repeatability of subjective evaluation of lameness in horses. Equine Vet J 2010; 42:92–97.
4. Gaughan EM. Skeletal origins of exercise intolerance in horses. Vet Clin North Am Equine Pract 1996; 12:517–535.
5. Keegan KG, Yonezawa Y, Pai PF, 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.
6. Ishihara A, Bertone AL, Rajala-Schultz PJ. Association between subjective lameness grade and kinetic gait parameters in horses with experimentally induced forelimb lameness. Am J Vet Res 2005; 66:1805–1815.
7. Pfau T, Robilliard JJ, Weller R, et al. Assessment of mild hind limb lameness during over ground locomotion using linear discriminant analysis of inertial sensor data. Equine Vet J 2007; 39:407–413.
8. Moorman VJ, Reiser RF II, Peterson ML, et al. Effect of forelimb lameness on hoof kinematics of horses at a trot. Am J Vet Res 2013; 74:1183–1191.
9. Moorman VJ, Reiser RF II, Mahaffey CA, et al. Use of an inertial measurement unit to assess the effect of forelimb lameness on three-dimensional hoof orientation in horses at a walk and trot. Am J Vet Res 2014; 75:800–808.
10. Thomsen MH, Persson AB, Jensen AT, et al. Agreement between accelerometric symmetry scores and clinical lameness scores during experimentally induced transient distension of the metacarpophalangeal joint in horses. Equine Vet J Suppl 2010; 42:510–515.
11. 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; 44:652–656.
12. Keegan KG, Wilson DA, Kramer J, 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; 74:17–24.
13. 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; 72:1156–1163.
14. Keegan KG, MacAllister CG, Wilson DA, 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; 73:368–374.
15. Marshall JF, Lund DG, Voute LC. Use of a wireless, inertial sensor-based system to objectively evaluate flexion tests in the horse. Equine Vet J Suppl 2012; 43:8–11.
16. Starke SD, Raistrick KJ, May SA, et al. The effect of trotting speed on the evaluation of subtle lameness in horses. Vet J 2013; 197:245–252.
17. Buchner HHF, Savelberg HHCM, Schamhardt HC, et al. Kinematics of treadmill versus overground locomotion in horses. Vet Q 1994; 16 (suppl 2): S87–S90.
18. Gómez Alvarez CB, Rhodin M, Byström A, et al. Back kinematics of healthy trotting horses during treadmill versus over ground locomotion. Equine Vet J 2009; 41:297–300.
