2025 American Academy of Neurology Abstract Website

Chao Wang¹, Kunal Chaturvedi³, Benjamin Nham⁴, Nicole Reid⁵, Andrew Bradshaw⁵, Sally Rosengren¹, Deborah Black², Kendall Bein⁶, Michael Halmagyi¹, Ali Braytee³, Mukesh Prasad³, Gnana Bharathy³, Miriam Welgampola¹
¹Central Clinical School, ²Faculty of Medicine and Health, University of Sydney, ³School of Computer Science, University of Technology Sydney, ⁴St George and Sutherland Clinical School, University of New South Wales, ⁵Institute of Clinical Neurosciences, Royal Prince Alfred Hospital, ⁶Royal Prince Alfred Green Light Institute for Emergency Care

Objective:

To develop and evaluate machine learning models for differentiation of vestibular neuritis and posterior circulation stroke using history, examination and vestibular function tests.

Background:

Vestibular Neuritis (VN) and Posterior Circulation Stroke (PCS) are the two common causes of the Acute Vestibular Syndrome, which is characterised by sudden, severe and persistent vertigo and/or imbalance. Expert clinicians separate the two diagnoses using information from the history, examination, and laboratory tests. Machine learning models capable of performing expert-level classification can expand the availability of diagnostic expertise.

Design/Methods:

We recruited patients who presented to the Emergency Room (ER) with acute vestibular syndrome and received a final diagnosis of VN or PCS. Data from the clinical history, bedside examination and four laboratory tests (videonystagmography, video head impulse test (VHIT), vestibular-evoked myogenic potentials (VEMP) and subjective visual horizontal) were used for model development. Different subsets of data simulated three scenarios: Tier 1 represented an ER with access to neuro-otology expertise (history, neuro-otological examination, videonystagmography, VHIT, ocular VEMP), Tier 2 an ER with VHIT (history, bedside examination, VHIT) and Tier 3 an ER reliant on history and bedside examination only. Model performance was also compared against the HINTS test (head impulse, nystagmus, test-of-skew).

Results:

Our dataset consisted of 163 VN and 131 PCS patients. Our best performing models used the CatBoost or XGBoost algorithms and identified PCS with accuracies of 96.6% (95% CI: 93.3-99.9%), 94.6% (95% CI: 90.5-98.6%) and 88.8% (95% CI: 86.0-91.6%) for Tiers 1, 2 and 3. HINTS by experts achieved 94.6% accuracy.

Conclusions:

Machine learning models can distinguish between posterior circulation stroke and vestibular neuritis with high accuracy, demonstrating their potential as a diagnostic aid to improve the differential diagnosis of acute vestibular syndrome in the Emergency Room by non-expert physicians.