Invited review: Toward a common language in data-driven mastitis detection research

Area of Science:

• Animal Science
• Biomedical Engineering
• Data Science

Background:

• Sensor technologies generate vast data for mastitis detection.
• Research increasingly relies on data-driven modeling over biological assumptions.
• Inconsistent terminology for similar methods hinders interdisciplinary understanding.

Purpose of the Study:

• To provide a framework (filtering, transformation, classification) for describing methods in sensor data-based mastitis detection.
• To review and categorize existing scientific literature on mastitis detection methods.
• To promote coherent terminology and clear method descriptions in future research.

Main Methods:

• A framework was developed based on filtering, transformation, and classification steps.
• 40 scientific publications from 1992-2020 applying sensor data for mastitis detection were identified and reviewed.
• Publications were categorized based on data processing techniques (filtering, transformation) and classification methods.

Main Results:

• Most publications (34/40) used filtering or transformation, or both, before classification.
• Simple thresholding was the most common classification method (19 publications).
• Significant variation in terminology for similar methods was identified across publications.

Conclusions:

• A standardized framework is crucial for understanding and communicating methods in sensor-based mastitis detection.
• Clearer, coherent terminology is needed to avoid confusion and facilitate interdisciplinary collaboration.
• This work serves as a reference and encourages improved reporting standards for future sensor-based animal disease detection research.