Development and Validation of a Machine Learning Model for Detection and Classification of Vertigo

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Vertigo is a common and often complex condition with various underlying causes.
• Accurate and timely diagnosis of vertigo is crucial for effective management and patient outcomes.
• Traditional diagnostic methods can be time-consuming and may require specialized expertise.

Purpose of the Study:

• To evaluate the effectiveness of an artificial intelligence (AI) model in improving the diagnostic accuracy of vertigo-related diseases.
• To develop and validate a machine learning model for classifying specific vertigo conditions.

Main Methods:

• Clinical symptoms and laboratory data from electronic medical records were extracted based on clinical guidelines.
• A machine learning diagnostic model (XGBoost) was developed and trained on a cohort of 1789 patients.
• The model's performance was assessed in a prospective validation cohort of 1148 patients using sensitivity, precision, and AUC metrics.

Main Results:

• The AI model demonstrated high diagnostic performance, surpassing traditional methods.
• Task 1 (BPPV vs. non-BPPV) achieved 98.32% sensitivity, 87.03% accuracy, and 0.947 AUC.
• Task 2 (classifying non-BPPV into MD, VM, SSNHLV) showed high sensitivity and AUC values for each condition, with SSNHLV achieving 100% in sensitivity and AUC.

Conclusions:

• The developed AI system significantly enhances the accuracy of classifying and diagnosing vertigo diseases.
• This technology can aid in initial therapy decisions or guide referrals, especially in resource-limited settings.
• AI holds promise for improving the diagnostic workflow and patient care for vestibular disorders.