An AI-Driven Multimodal Monitoring System for Early Mastitis Indicators in Italian Mediterranean Buffalo

Area of Science:

• Veterinary Medicine
• Artificial Intelligence
• Animal Science

Background:

• Mastitis poses a significant economic and health challenge in buffalo farming.
• Current diagnostic methods for mastitis can be invasive or lack real-time capabilities.
• Non-invasive, automated monitoring is crucial for improving udder health management in buffalo.

Purpose of the Study:

• To develop and validate a fully automated AI-driven thermal imaging system for real-time, non-invasive monitoring of udder health in Italian Mediterranean buffalo.
• To integrate this system with robotic milking for synchronized data acquisition.
• To establish a correlation between thermal imaging data and key indicators of mastitis, such as somatic cell count.

Main Methods:

• Development of an AI system utilizing a transformer-based neural network (SegFormer) for udder segmentation from thermal images.
• Synchronized acquisition of thermal images during robotic milking, with compensation for environmental variables using a calibrated weather station.
• Extraction of maximum udder skin surface temperature (USST) and correlation analysis with somatic cell count (SCC).

Main Results:

• Demonstrated feasibility of the automated thermal imaging system in operational farm settings.
• Established a significant correlation between USST and SCC, indicating potential for subclinical mastitis detection.
• Successful segmentation of the udder area using SegFormer for precise temperature analysis.

Conclusions:

• The AI-driven thermal imaging system offers a scalable, precision diagnostic tool for early subclinical mastitis detection in buffalo.
• This technology can significantly improve animal welfare and reduce antibiotic usage in the dairy industry.
• The system represents a critical advancement towards intelligent, automated animal health monitoring.