Machine learning algorithm as a diagnostic tool for hypoadrenocorticism in dogs.

Area of Science:

• Veterinary Medicine
• Machine Learning in Diagnostics
• Canine Health

Background:

• Canine hypoadrenocorticism (CHA) is a life-threatening endocrine disorder affecting 3 in 1,000 dogs.
• CHA presents with diverse clinical signs, often mimicking kidney or gastrointestinal diseases, posing a significant diagnostic challenge.
• Delayed diagnosis of CHA carries substantial patient risk, while prompt treatment ensures an excellent prognosis.

Purpose of the Study:

• To develop and validate a machine learning model for the early diagnosis of canine hypoadrenocorticism (CHA).
• To utilize routinely collected screening diagnostics (complete blood count and serum chemistry panel) for CHA detection.
• To create an accessible tool for veterinary practitioners to improve CHA screening and patient outcomes.

Main Methods:

• A boosted tree algorithm (AdaBoost) was trained using data from 908 control dogs and 133 dogs with confirmed CHA.
• The algorithm was trained on 80% of the data and tested on the remaining 20%.
• Performance was evaluated based on sensitivity, specificity, and area under the receiver operator characteristic curve (AUC).

Main Results:

• The developed AdaBoost model achieved high diagnostic performance: 96.3% sensitivity and 97.2% specificity.
• The model demonstrated strong discriminative ability with an AUC of 0.994.
• The machine learning approach outperformed traditional logistic regression analysis in screening for CHA.

Conclusions:

• Machine learning, specifically the AdaBoost algorithm, is a highly effective tool for screening canine hypoadrenocorticism (CHA).
• The developed model, integrated into a user-friendly interface, can significantly aid practitioners in diagnosing CHA.
• Implementing this AI-driven screening method promises to improve diagnostic accuracy and patient outcomes for dogs with CHA.