For
most mares, foaling is a straightforward process that ends with the arrival of
a healthy foal. However, when 
complications occur, they can escalate rapidly.
Problems such as an abnormal foetal presentation can quickly become
life-threatening for both mare and foal, making rapid veterinary intervention
essential. Having experienced staff on hand at the right moment can mean the
difference between a successful birth and a tragic outcome.
The challenge is knowing exactly when that moment will arrive. Mares often show remarkably few obvious signs that foaling is imminent, forcing stud farms to maintain overnight watches throughout the breeding season. On large commercial studs, this means dedicated staff monitoring mares around the clock, while owners with just one or two mares frequently endure several sleepless nights waiting for labour to begin.
A reliable system capable of predicting foaling before labour starts would therefore be invaluable, improving welfare while reducing both labour demands and the risk of missing a difficult birth.
Researchers from the School of Veterinary Medicine at Kitasato University in Japan, working with engineers from Noritsu Precision Co., Ltd., have now tested a fully non-contact monitoring system that combines thermal imaging with artificial intelligence to predict the onset of foaling under commercial breeding conditions.
The study involved 115 pregnant Thoroughbred mares on 13 breeding farms in Hokkaido during the 2024 foaling season. Each mare was housed individually in a foaling box overnight between 3 p.m. and 7 a.m. and turned out during the day. A camera unit, combining a thermal infrared camera with a conventional visible-light camera, was mounted on the stall wall so the mare always remained within view.
Unlike wearable sensors or invasive monitoring systems, the camera required no contact with the mare. Instead, artificial intelligence analysed images every five minutes, measuring locomotor activity and body surface temperature from the thermal images while simultaneously identifying posture changes and tail-raising behaviour from the visible-light footage.
The researchers discovered that mares exhibit two distinct phases before giving birth. The earliest indicators appeared around 70 to 90 minutes before foaling, when both movement and body surface temperature relative to the surrounding environment increased significantly. More obvious behavioural changes, including frequent posture adjustments and repeated tail raising, occurred much closer to delivery, typically within the final 25 to 45 minutes
These sequential changes allowed the researchers to develop a prediction model that reflected the natural progression towards labour. Using only the early indicators of movement and temperature, the system correctly detected impending foaling in 80% of mares, providing an average warning time of just over three hours
When the later behavioural signs were incorporated into the model, predictive accuracy increased dramatically. The complete system correctly predicted foaling in 94.8% of mares, with an average alert approximately 89 minutes before delivery, providing staff with sufficient time to prepare for the birth or summon veterinary assistance if necessary.
The findings suggest that combining thermal imaging with AI-based behavioural analysis could offer breeders a practical, reliable and welfare-friendly method of monitoring mares during the foaling season. Because the system operates entirely without physical contact, it avoids the limitations of wearable devices while remaining suitable for routine use on commercial breeding farms.
While further studies will be needed to evaluate its performance in different breeds and management systems, the technology represents a promising step towards smarter foaling surveillance. For breeders, earlier warning of impending birth could reduce overnight monitoring, improve staff efficiency and, most importantly, increase the chances of providing timely assistance when mare and foal need it most.
For more details, see:
Nabenishi, Hisashi, Nagisa Taki, Shoji Nishibayashi, and Tomoyuki Ishii.
Noninvasive Two-Phase Foaling Prediction in Thoroughbred Mares Using Thermal Imaging and AI-Based Behavioral Analysis
Animals (2026) 16, no. 14: 2221.
No comments:
Post a Comment