A probabilistic neural network model for lameness detection

Area of Science:

• Animal Science
• Veterinary Medicine
• Agricultural Engineering

Background:

• Lameness is a significant welfare and economic issue in dairy farming.
• Early detection of lameness is crucial for timely intervention and treatment.
• Automated monitoring systems can improve herd health management.

Purpose of the Study:

• To develop and validate a system for measuring leg-load distribution in dairy cows during milking.
• To create an expert system for automatic lameness detection using probabilistic neural networks.
• To evaluate the system's accuracy and potential for on-farm application.

Main Methods:

• A 4-balance system was implemented to record leg weights during robotic milking.
• Locomotion scoring and clinical inspection were used for lameness assessment.
• A classifying probabilistic neural network model was trained and validated on collected data.

Main Results:

• The system successfully recorded leg weights from 73 cows over nearly 10,000 milkings.
• The probabilistic neural network model achieved 96.2% correct classification of sound and lame cows.
• 100% of lameness cases were identified in the validation set with only 1.1% false alarms.

Conclusions:

• The developed leg-load monitoring system effectively detects dairy cow lameness.
• The probabilistic neural network model shows high accuracy for real-time lameness monitoring.
• This technology offers potential as an on-farm decision aid for dairy farmers.