Catching colic early can be difficult. Signs of gut pain in horses are often subtle and easy to overlook. But what if a horse could wear a device that detected signs of digestive upset and automatically notify a veterinarian? Research conducted in Belgium may one day make that possible.

Leveraging advances in wearable technology and artificial intelligence, the Belgian researchers have developed a system that incorporates leg boots fitted with accelerometers and a diagnostic computer algorithm. “The accelerometers capture movement data, which is then entered into a computer model,” says Anniek Eerdekens, M.Sc., of Ghent University. “The computer model categorizes that data into either a typical behavior or one of 10 behaviors associated with pain. These motions are fed into another model, which looks for patterns in the data. These are used to determine whether the condition is colic and, if it is, the severity.”

The accelerometers are placed on the horse’s front limbs, but they can detect pain-related behaviors—such as shivering or kicking at the abdomen—elsewhere in the body. “When a horse kicks at his abdomen, the action isn’t entirely isolated to the hindquarters. The horse may shift his weight, adjust his stance, or otherwise move his front limbs in response to the discomfort or to maintain balance, especially when engaging in repeated or forceful actions,” she says. “The accelerometer data is sampling at 50 Hz, meaning that it collects data 50 times every second. This high sampling rate makes the device extremely sensitive to even the slightest movement.”

The experiment

To develop the system, the researchers fitted eight mares with accelerometer boots and induced colic. The collected data then was used to classify each horse’s behavior as normal or pain related. It was also used to calculate a separate “activity index.” Based on this information, the researchers developed an algorithm for identifying colic and gauging its severity. Meanwhile, veterinarians observed the mares, assigning pain scores at regular intervals, and ultimately classifying each as having no colic pain, level-1 colic pain or more severe level-2 colic pain.

To determine the accuracy of the model, the researchers compared the classifications made by the computer algorithm to the observations made by the veterinarians. “The colic  detection accuracy rate of the algorithm was 91.2 percent and the colic severity accuracy was 93.8 percent when compared to the veterinarians’ findings,” says Eerdekens.

The analysis of the severity of the colic is a critical component of the system, says Eerdekens. “Understanding the severity of colic enables veterinarians and caretakers to tailor care and treatment plans more accurately. This precision in care prevents the under-treatment of serious cases and the over-treatment of mild ones.”

Algorithm that detects colic

This is the first computer algorithm to detect colic described in scientific literature, says Eerdekens. “Other algorithms can identify lameness in horses based on video data. Also there are some that can identify infectious airway diseases in other species, such as pigs and cows.” 

Consumers may one day be able to purchase “colic detection boots,” says Eerdekens, and the underlying technology may have other applications. “The algorithms have the ability to identify pain, so they could be retrained to recognize pain in horses with different diseases,” she says. “It may be used to reveal other health issues, such as asthma.”

Reference: “Automatic early detection of induced colic in horses using accelerometer devices,” Equine Veterinary Journal, February 2024