Machine learning-assisted screening for canine Cushing's syndrome.

Area of Science:

• Veterinary Medicine
• Endocrinology
• Machine Learning in Diagnostics

Background:

• Cushing's syndrome (CS) is a prevalent endocrine disorder in dogs, often presenting with variable clinical signs.
• Accurate diagnosis of CS is challenging due to its complex presentation, hindering timely intervention.
• Current diagnostic pathways can be resource-intensive and may not always identify suitable candidates for advanced testing.

Purpose of the Study:

• To develop and validate a machine learning model for assisting in the diagnosis of canine Cushing's syndrome.
• To utilize routinely available screening diagnostics for improved diagnostic accuracy.
• To enhance the efficiency and accessibility of Cushing's syndrome diagnosis in veterinary practice.

Main Methods:

• A boosted tree algorithm (gradient boosting) was trained on a dataset comprising complete blood count, serum chemistry panel, and urinalysis parameters.
• Data included 153 control dogs and 152 dogs with confirmed Cushing's syndrome.
• The model was trained on 80% of the data and validated on the remaining 20%.

Main Results:

• The machine learning model achieved an overall accuracy of 88.5% (95% CI: 80.5-96.5%).
• Sensitivity was 83.3% (95% CI: 70.7-96.7%) and specificity was 93.5% (95% CI: 84.9-100%).
• The area under the receiver operating characteristic curve was 0.912 (95% CI: 0.835-0.988), indicating excellent diagnostic discriminatory ability.

Conclusions:

• Machine learning algorithms can effectively assist in diagnosing canine Cushing's syndrome using standard screening diagnostics.
• The developed model demonstrates high accuracy and discriminatory power, offering a valuable tool for veterinarians.
• Implementation of a user-friendly interface can improve diagnostic efficiency and potentially enhance owner satisfaction.