Applying Machine Learning Algorithms for the Classification of Mink Infected with Aleutian Disease Using Different...

Area of Science:

• Veterinary Medicine
• Animal Science
• Machine Learning Applications

Background:

• American mink (Neogale vison) farming is economically significant but threatened by Aleutian disease (AD).
• Current AD detection relies on counterimmunoelectrophoresis (CIEP), a costly test-and-remove strategy.
• Limited application of machine learning (ML) in mink farming hinders advanced disease management.

Purpose of the Study:

• To assess the efficacy of ML algorithms in classifying AD infection in mink.
• To identify key features for accurate AD prediction, potentially reducing reliance on CIEP.
• To develop a predictive model for controlling AD in mink populations.

Main Methods:

• Collected data from 1157 mink on a Canadian farm, including 33 distinct features.
• Evaluated nine ML algorithms (e.g., Random Forest, SVM, ANN) for AD classification.
• Utilized CIEP as the gold standard for disease confirmation and model training.

Main Results:

• The Aleutian mink disease virus capsid protein-based ELISA was the most significant predictor.
• Random Forest demonstrated superior performance with high sensitivity (0.938), specificity (0.986), and accuracy (0.962).
• Cross-validation confirmed the robustness of the Random Forest model's predictive capabilities.

Conclusions:

• Machine learning, particularly the Random Forest algorithm, can accurately classify AD-infected mink.
• This approach offers a viable, potentially more efficient alternative to traditional CIEP testing.
• Further validation on diverse farm data and integration of genomic information are recommended for widespread implementation.