Detection of spinal ataxia in horses using fuzzy clustering of body position uncertainty

Reasons for performing study: Subjective neurological evaluation in horses is prone to bias. An objective method of spinal ataxia detection is not subject to these limitations and could be of use in equine practice and research. Hypothesis: Kinematic data in the walking horse can differentiate normal and spinal ataxic horses. Methods: Twelve normal and 12 spinal ataxic horses were evaluated by kinematic analysis walking on a treadmill. Each body position signal was reduced to a scalar measure of uncertainty then fuzzy clustered into normal or ataxic groups. Correct classification percentage (CCP) was then calculated using membership values of each horse in the 2 groups. Subsequently, a guided search for measure combinations with high CCP was performed. Results: Eight measures of body position resulted in CCP≥70%. Several combinations of 4-5 measures resulted in 100% CCP. All combinations with 100% CCP could be obtained with one body marker on the back measuring vertical and horizontal movement and one body marker each on the right fore- and hindlimb measuring vertical movement. Conclusions and potential relevance: Kinematic gait analysis using simple body marker combinations can be used objectively to detect spinal ataxia in horses.