Differentiating vestibular migraine from posterior circulation ischemia initially presenting with dizziness/vertigo: a machine learning approach
Introduction Vestibular migraine (VM) and posterior circulation ischemia (PCI) share initial dizziness or vertigo as chief manifestations with substantial clinical overlap, and the limited diagnostic utility of available examinations complicates their differentiation, creating an urgent clinical demand for multidimensional assessment strategies to reduce misdiagnosis rates and improve diagnosis and treatment. Methods This retrospective study consecutively recruited 571 patients presenting with initial dizziness or vertigo, including 382 VM cases and 189 PCI cases. The dataset was randomly split into a 70% training set and a 30% test set. Optimal predictive features were screened by combining univariable area-under-the-curve analysis, LASSO regression and the Boruta algorithm, and five machine-learning algorithms were further compared in terms of classification performance and clinical interpretability. Results Nine core predictive features were finally identified, and all five models achieved favorable discriminatory performance in validation. Among these models, XGBoost and Random Forest demonstrated strong and comparable diagnostic performance. Discussion This multidimensional machine-learning model can accurately distinguish VM from PCI in patients with new-onset dizziness or vertigo, and may serve as a potential auxiliary tool for differential diagnosis after initial clinical evaluation.
Authors
- Weimin Yang (ORCID: https://orcid.org/0000-0003-2265-7532)
- Xuemei Xiao
- Xinhang Wen
- Yongping Wang
- Wendi Wang
- Xikai Wang
- Xiaoyu Liu
Institutions
- First Affiliated Hospital of Zhengzhou University (CN)
Publication Details
- Journal
- Frontiers in Neurology
- Published
- 2026-11-19
- DOI
- https://doi.org/10.3389/fneur.2026.1994159
- Primary Topic
- Vestibular and auditory disorders
- Type
- article
- Field-Weighted Citation Impact
- 0.00