Gait analysis of hoof strain data
The high value of competition horses and the risk of injury involved in sports drive the dependency on efficient veterinary support. Detection and evaluation of lameness by subjective visual assessment is a difficult medical task. The examiner needs to consider several different and rapidly changing body movement patterns. When the patterns become too complex, tools like Artificial Neural Network (ANN) can be useful. ANN can be gainfully applied to gait analysis, if the system has the ability to distinguish stride characteristics and to differentiate pathological gait. A Self-organizing Map (SOM) and Radial Basis Functions (RBF) network was trained to cluster and classify strain measurement data collected from a single hoof of moving horses. The performance of the network was examined by presenting different combinations of the available data and by varying network parameters. The method was successful in differentiating certain stride characteristics of individual horses, such as gait fingerprint, shoeing, gait, speed, and direction of movement. The neural network is still in search of the lame horse. The network's ability to distinguish the various stride characteristics encourages further investigation.