Machine learning models help differentiate between causes of recurrent spontaneous vertigo.

Area of Science:

• Neurology
• Medical Diagnostics
• Machine Learning in Healthcare

Background:

• Vestibular migraine (VM) and Menière's disease (MD) are primary causes of recurrent vertigo.
• Accurate differentiation between VM and MD is clinically significant but challenging.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for distinguishing VM from MD.
• To assess model performance across different data availability tiers simulating clinical settings.

Main Methods:

• Utilized 100 features from history, video-nystagmography, and audiovestibular tests (vHIT, VEMP, caloric, audiogram).
• Applied ten ML algorithms, with models trained and validated using tenfold cross-validation.
• Simulated three clinical data availability tiers: neuro-otology clinic (all data), general neurology (history, caloric, audiogram), and primary care (history only).

Main Results:

• Models achieved high accuracy (85.77-97.81%) in differentiating VM and MD across all tiers.
• AdaBoost and Random Forest algorithms demonstrated superior performance, with the highest accuracy reaching 97.81% in Tier 1.
• Patient history emerged as the most valuable single feature subset for diagnosis.

Conclusions:

• Machine learning models offer a robust and accurate method for differentiating VM and MD.
• These ML tools can support clinicians with varying levels of diagnostic expertise and available resources.
• The findings highlight the potential of AI to enhance diagnostic accuracy in neuro-otology.