Machine learning-based gait classification and genome-wide association identify a QTL for gait type in Colombian Paso Horses.
Description
The Dataset related to horse gait, sex, age, locomotion and kinematic measurements using inertial motion sensors, and AI gait classification. It consists of an Excel spreadsheet containing 217 rows.
Files
Steps to reproduce
Objective kinematic data were collected from 217 horses between 2018 and 2020 using a wireless network of seven Inertial Measurement Units (IMUs; ProMove-mini, Inertia Technology, Enschede, The Netherlands). Sensors were securely affixed to the poll, withers, pelvis (sacrum), and the dorsal aspect of all four limbs, following previously validated protocols. Data were recorded while horses were ridden over straight, hard-surface segments for durations of 3-5 min. Movements were performed at gaits designated by the rider or owner, including walk, paso fino, Colombian trot, Colombian trocha, and Colombian gallop. To ensure precise gait identification, all IMU data streams were time-synchronized with high-definition video recordings. Expert evaluators specializing in Colombian Paso Horse (CPH) locomotion performed slow-motion video analysis to isolate straight, regular, and representative locomotion sequences for downstream analysis. Stride parameters were extracted from these curated raw IMU segments using a validated gait-analysis algorithm 85. To maintain high data quality, kinematic variables were subjected to outlier detection and removal on an individual horse basis prior to neural network integration. The processed kinematic segments were utilized to train and validate an Artificial Intelligence-Long Short-Term Memory (AI-LSTM) neural network architecture27. The classification algorithm was trained using a "gold standard" reference cohort of 39 horses (14 Colombian trot and 25 Colombian trocha). The gait phenotype for each reference horse was verified via consensus by CPH experts through slow-motion video inspection, ensuring strict adherence to the locomotor standards established by the national breed authority, Fedequinas. This AI-driven approach allowed for a high-resolution, objective reclassification of the broader study population.