Improving gait classification in horses by using inertial measurement unit (IMU) generated data and machine learning.

Area of Science:

• Equine biomechanics
• Animal locomotion analysis
• Machine learning applications in veterinary science

Background:

• Traditional horse gait classification relies on subjective visual assessment.
• Objective, real-time methods for gait classification (GC) in horses are needed for research and breeding.
• Understanding equine gaits is crucial for welfare, performance, and genetic studies.

Purpose of the Study:

• To develop and validate an automated system for accurate, real-time gait classification in horses.
• To leverage wireless motion sensors and machine learning for objective GC.
• To enable advanced biomechanical analysis and precise phenotyping of equine gaits.

Main Methods:

• Utilized a network of seven wireless, high-sampling-rate motion sensors on 120 horses across four breeds.
• Collected and analyzed 7576 strides encompassing eight distinct equine gaits.
• Trained machine learning models using both feature-extracted and raw sensor data for gait classification.

Main Results:

• The developed automated gait classification system achieved a high accuracy of 97%.
• The machine learning models demonstrated effectiveness in classifying various equine gaits from sensor data.
• The approach proved robust, utilizing both raw and processed sensor information.

Conclusions:

• Accurate, automated gait classification in horses is feasible using sensor technology and machine learning.
• This method provides objective data for in-depth biomechanical research and genetic phenotyping.
• The developed technique shows potential for application in other quadrupedal species without specialized algorithms.