The ability to measure and quantify things has, for many years now, deepened humanity’s understanding of the world we live in. Within the field of equine biomechanics, the measurement and quantification of forces and movements have taught us a lot about horse locomotion and how it relates to the gait, performance, and welfare of the animal. One specific example of a health-related topic is weight-bearing lameness. The unwillingness of horse to fully load the painful limb, has been linked to measurable asymmetries between the vertical ground reaction forces, exerted by the limbs, while trotting. Similarly, vertical movement asymmetries of the head, withers, and pelvis have also been correlated to this type of lameness. At first glance, it might appear obvious that if a horse uses more vertical force in a left limb compared to a right while pushing against and off the ground, the resulting vertical movement of the mass it displaces should also become asymmetric. However, the horse also has a back connecting the fore- and hindlimb pairs and on top of this a neck with a head that protrude, much like a lever arm, from the cranial part of the body. The four-legged nature of the horse and the accompanying anatomical complexity can result in complicated vertical movement asymmetries that are not as obviously connected to the vertical ground reaction forces, as one might expect at a first glance.
Advancements in technology and clinical research have brought objective measurement tools, specifically aimed at lameness evaluation, to the commercial market. Kinematic systems, e.g., inertial measurement units and marker based optical motion capture, are prevalent as they provide a relatively easier means of acquiring data compared to direct measurements of ground reaction forces, e.g., using force plates, force shoes or an instrumented treadmill. Applied objective lameness analysis requires that the measurement system is reliable, simple to use and that it outputs information that can be clinically interpreted. To cater to these needs, kinematic systems are often focused on the upper body movement, e.g., head, withers, and pelvis. While it is entirely possible to also capture limb kinematics, it can be favorable to exclude them as it results in less preparation of the horse and less equipment that could be in the way of palpation and diagnostic analgesia. However, not including limb data also makes it a challenge to automatically detect the gait and when the respective limbs have ground contact during each stride, which is essential information for calculating the related kinematic variables. Sampled signals might also contain noise and movement unrelated to the actual lameness, which must be removed to not pollute any extracted asymmetry variables. Finally, if kinetic weight-bearing asymmetry can be considered the symptom of lameness, then movement asymmetries are a symptom of the symptom. This means that figuring out how the movement is related to the forces and how the forces should then be clinically evaluated is of the essence.
In this thesis it could be determined that pre-processing the kinematic signals either with a signal decomposition method or with a digital filter tuned to the horse’s stride frequency, could minimize the negative effects of the signal noise and unwanted components on the calculated upper body asymmetry variables. The various digital filters were evaluated using artificial vertical movement signals and data from 7 trotting horses with induced fore- and hindlimb lameness. For a fourth order high-pass Butterworth filter, the optimal cut-off frequency was found to be 72% of the stride frequency.
A novel method for gait classification was presented, where the movement of the withers and the pelvis, captured with marker based optical motion capture, was used to distinguish walk from trot and left from right steps. A Quadratic Discriminant Analysis model was trained on 102 horses (60 819 steps) walking and trotting on a treadmill at varying speeds and with varying degrees of weight bearing asymmetries. Subsequently, the model was tested on 120 horses (21 845 steps) trotting and walking over ground during normal clinical evaluations on the straight line, on the circle, on soft, and on hard surfaces, resulting in a predicted accuracy of 99.98%.
The connection between subjective lameness grades and kinetic weight-bearing asymmetry was investigated in 69 horses, clinically evaluated by two experienced veterinarians while being trotted on an instrumented treadmill, simultaneously measured with a marker based optical motion capture system. It was found that there was considerable overlap, in horses graded as sound and those graded as having subtle to mild lamenesses, when observing the peak vertical force differences and the head and pelvis vertical movement asymmetry.
The association between peak vertical ground reaction force asymmetries and upper body movement asymmetries was investigated in 103 horses with varying degrees of weight-bearing asymmetries. The horses were trotted on an instrumented treadmill, simultaneously measured with a marker based optical motion capture system. It was shown that it was possible to model
both fore- and hindlimb differences in peak vertical ground reaction forces as linear combinations of vertical movement asymmetry variables
extracted from the head, withers and pelvis. The best models achieved root mean squared errors of 0.83% and 0.82% for the fore- and hindlimbs, respectively.