Reliability of observational- and machine-based teat hygiene scoring methodologies.

Area of Science:

• Dairy Science
• Agricultural Technology
• Quality Control

Background:

• Effective premilking teat cleaning is vital for udder health and milk quality.
• Current observational methods for assessing teat cleanliness can be subjective and unreliable.

Purpose of the Study:

• To evaluate the reliability of an automated system for assessing teat-end debris compared to human raters.
• To explore the potential of machine learning for objective evaluation of premilking teat cleanliness.

Main Methods:

• 175 teat swab images were rated by 8 experienced human raters using a 4-point visual scale.
• An automated method using digital image processing and machine learning was developed to quantify teat-end debris.
• Inter-rater and intra-rater reliability were assessed using Cohen's kappa (κ) and intraclass correlation coefficients.

Main Results:

• Human raters showed low inter-rater reliability (overall κ = 0.43).
• The automated machine-based system demonstrated near-perfect reliability (Pearson r > 0.99).
• Machine-based assessment significantly outperformed observational methods in consistency.

Conclusions:

• Automated teat-end debris analysis offers a more reliable and objective method for evaluating worker performance than visual inspection.
• This technology can enhance training and quality control in dairy parlors.
• Machine-based scoring can be integrated into automated milking systems for improved efficiency.