A Questionnaire-Based Ensemble Learning Model to Predict the Diagnosis of Vertigo: Model Development and Validation...

Area of Science:

• Vestibular System Disorders
• Machine Learning in Medicine
• Diagnostic Tools

Background:

• Questionnaires have been utilized for two decades to predict vertigo diagnoses and support clinical decisions.
• A machine learning (ML) model based on questionnaires offers potential to enhance the diagnostic efficiency of vestibular disorders.

Purpose of the Study:

• To develop and validate a questionnaire-based ML model for predicting vertigo diagnoses.

Main Methods:

• A multicenter prospective study enrolled 1693 patients with vertigo across 7 tertiary referral centers.
• A diagnostic questionnaire was administered, and data from patients with a single final diagnosis were used for model development, cross-validation, and external validation.
• Nine ML methods were compared, with a recalibrated light gradient boosting machine (LGBM) selected for optimal performance.

Main Results:

• The LGBM model demonstrated high predictive accuracy, achieving an area under the curve (AUC) of 0.937 in cross-validation and 0.954 in external validation.
• The model effectively predicted common vestibular disorders among the 1693 enrolled patients.

Conclusions:

• The questionnaire-based LGBM model shows promise in predicting vestibular disorders and aiding clinical decision-making in ENT and vertigo clinics.
• Further research with larger sample sizes and neurologist involvement is recommended to assess the model's generalization and robustness.