Evaluation of a fully automated 2-dimensional imaging system for real-time cattle lameness detection using machine...

Area of Science:

• Veterinary Medicine
• Animal Science
• Artificial Intelligence in Agriculture

Background:

• Proactive lameness management in dairy cows is essential for animal welfare and herd productivity.
• Early detection and prompt treatment of lameness are critical for effective management strategies.
• Traditional lameness scoring relies on subjective human assessment, which can be variable.

Purpose of the Study:

• To evaluate the efficacy of a fully automated 2-dimensional imaging system utilizing artificial intelligence (AI) for real-time lameness detection in dairy cows.
• To compare the performance of the AI system against experienced human mobility scores (HMS) in identifying lame cows and cows with foot lesions.
• To assess the system's ability to detect early signs of lameness before the development of severe lesions.

Main Methods:

• Collected mobility data from 11 UK dairy farms, generating 40,116 paired human mobility scores (HMS) and AI mobility scores (AIMS).
• Utilized Gwet's and Cohen's coefficients to assess agreement between AIMS and HMS on 4-grade and 2-grade lameness scales.
• Calculated sensitivity, specificity, and accuracy for both AIMS and HMS in detecting cows with foot lesions, analyzing longitudinal data up to 30 days pre-trimming.

Main Results:

• Agreement between AIMS and HMS ranged from fair to almost perfect depending on the metric used (κw: 0.24-0.34, AC2: 0.81-0.93 for 4-grade; PA >80%, κ: 0.23-0.38, AC1: 0.76-0.83 for 2-grade).
• The AI system demonstrated comparable sensitivity and specificity to human assessors in detecting cows with severe foot lesions.
• AI system detected increased mobility scores in cows with severe lesions up to 23 days prior to foot trimming, indicating potential for early detection.

Conclusions:

• The automated AI imaging system performs comparably to experienced human assessors in detecting lameness and foot lesions in dairy cows.
• The system's ability to identify subtle mobility changes preceding severe lesions highlights its potential for early intervention.
• This technology can significantly enhance proactive lameness management strategies, improving dairy herd health and welfare.