Horse Jumping and Dressage Training Activity Detection Using Accelerometer Data

Automatic early detection of induced colic in horses using accelerometer devices

BACKGROUND

To seek appropriate veterinary attention for horses with colic, owners must recognise early signs. Direct observation of horse behaviour has several drawbacks: it is time-consuming, hard to see subtle and common behavioural signs, and is based on intuition and subjective decisions. Due to recent advances in wearables and artificial intelligence, it may be possible to develop diagnostic software that can automatically detect colic signs.

OBJECTIVES

To develop a software algorithm to aid in the detection of colic signs and levels of pain.

STUDY DESIGN

In vivo experiments.

METHODS

Transient colic was induced in eight experimental mares with luteolytic doses of prostaglandin. Veterinarians observed the horses before and throughout the interventions and assigned pain scores which were used to separate colic episodes into none (pain score ≤5), level 1 (pain score 6-10) or level 2 (pain score ≥11). Accelerometric data and videos were collected throughout the experiments and using accelerometric data, the horse's behaviour was classified into normal and 10 pain-related behaviours and an activity index was calculated. Models were designed that utilised behaviour and activity index characteristics both detecting the presence of colic and assessing its severity. To determine the accuracy of the model, the ground truth, that is the veterinarians' observation of colic signs and assessment of pain level, was compared with the automatic detection system.

RESULTS

The cross-validation analysis demonstrated an accuracy of 91.2% for detecting colic and an accuracy of 93.8% in differentiating between level 1 colic and level 2 colic. The model was able to accurately classify 10 pain-related behaviours and distinguish them from normal behaviour with a high accuracy.

MAIN LIMITATIONS

We included a limited number of horses with severe pain related behaviours in the dataset. This constraint affects the accuracy of categorising colic severity rather than limiting the algorithms' capacity to identify early colic signs.

CONCLUSIONS

Our system for early detection of colic in horses is unique and innovative, and it can distinguish between colic of varying severity.

Longitudinal Training and Workload Assessment in Young Friesian Stallions in Relation to Fitness: Part 1

Friesian stallions have to complete a 10-week studbook approval test (70-day test) consisting of dressage and driving training for acceptance as a breeding stallion. Training response of the horses to this approval program was evaluated. External workload (EW) of 16 stallions (3.2 ± 0.4 years old) was registered during the 9 weeks preceding the 70-day test (preparation period) and during 70-day test. Ridden indoor standardized exercise tests (SET), measuring heart rate (HR) and plasma lactate concentration (LA) at walk, trot and canter, were performed in week 1 (SET-I), 6 (SET-II) and 10 (SET-III) of the 70-day test. Linear mixed effect models were used to analyze HR and LA related to SETs and EW related to the phase of the study, using Akaike's information criterion for model reduction. EW was higher in the 70-day test compared to the preparation period, with longer duration of training sessions. Compared to SET-I, horses showed increased HR and LA after canter in SET-II and SET-III, indicating that they were more fatigued. The fact that the fitness of these Friesian stallions decreased during the 70-day test suggests overtraining. To optimize training effects and welfare of these horses, the workload of the training program needs to be adjusted.

Efficient Development of Gait Classification Models for Five-Gaited Horses Based on Mobile Phone Sensors

Automated gait classification has traditionally been studied using horse-mounted sensors. However, smartphone-based sensors are more accessible, but the performance of gait classification models using data from such sensors has not been widely known or accessible. In this study, we performed horse gait classification using deep learning models and data from mobile phone sensors located in the rider's pocket. We gathered data from 17 horses and 14 riders. The data were gathered simultaneously from movement sensors in a mobile phone located in the rider's pocket and a gait classification system based on four wearable sensors attached to the horse's limbs. With this efficient approach to acquire labelled data, we trained a Bi-LSTM model for gait classification. The only input to the model was a 50 Hz signal from the phone's accelerometer and gyroscope that was rotated to the horse's frame of reference. We demonstrate that sensor data from mobile phones can be used to classify the five gaits of the Icelandic horse with up to 94.4% accuracy. The result suggests that horse riding activities can be studied at a large scale using mobile phones to gather data on gaits. While our study showed that mobile phone sensors could be effective for gait classification, there are still some limitations that need to be addressed in future research. For example, further studies could explore the effects of different riding styles or equipment on gait classification accuracy or investigate ways to minimize the influence of factors such as phone placement. By addressing these questions, we can continue to improve our understanding of horse gait and its role in horse riding activities.