Noninvasive Two-Phase Foaling Prediction in Thoroughbred Mares Using Thermal Imaging and AI-Based Behavioral Analysis

Area of Science:

• Equine reproduction
• Artificial intelligence in animal science
• Thermal imaging applications

Background:

• Accurate foaling prediction is crucial for reducing foal mortality and managing labor on breeding farms.
• Current methods may be invasive or lack comprehensive predictive capabilities.

Purpose of the Study:

• To evaluate a noninvasive system integrating thermal imaging and AI-based behavioral analysis for predicting foaling in Thoroughbred mares.
• To assess the system's predictive performance under commercial breeding conditions.

Main Methods:

• A system combining thermal imaging and visible-light analysis was used to monitor 115 Thoroughbred mares.
• AI-powered neural network model analyzed locomotor activity, body surface temperature, posture changes, and tail-raising behavior at 5-minute intervals.
• Data from the 5 hours preceding foaling were analyzed to identify predictive patterns.

Main Results:

• Locomotor activity and surface temperature increased significantly 70-90 minutes before foaling.
• Posture changes and tail-raising behavior increased markedly 25-45 minutes before foaling, indicating a two-phase pre-foaling process.
• The integrated AI system achieved 94.8% detection accuracy with a mean lead time of 89 minutes.

Conclusions:

• The developed noninvasive system accurately predicts foaling using AI and thermal imaging.
• The system demonstrates high predictive performance in commercial settings without invasive devices.
• This technology offers a valuable tool for improving equine reproductive management and reducing perinatal risks.