Background: Acute vestibular syndrome usually represents either vestibular neuritis (VN), an innocuous viral illness, or posterior circulation stroke (PCS), a potentially life-threatening event. The video head impulse test (VHIT) is a quantitative measure of the vestibulo-ocular reflex that can distinguish between these two diagnoses. It can be rapidly performed at the bedside by any trained healthcare professional but requires interpretation by an expert clinician. We developed machine learning models to differentiate between PCS and VN using only the VHIT. Methods: We trained machine learning classification models using unedited head- and eye-velocity data from acute VHIT performed in an Emergency Room on patients presenting with acute vestibular syndrome and whose final diagnosis was VN or PCS. The models were validated using an independent test dataset collected at a second institution. We compared the performance of the models against expert clinicians as well as a widely used VHIT metric: the gain cutoff value. Results: The training and test datasets comprised 252 and 49 patients, respectively. In the test dataset, the best machine learning model identified VN with 87.8% (95% CI 77.6%–95.9%) accuracy. Model performance was not significantly different (p = 0.56) from that of blinded expert clinicians who achieved 85.7% accuracy (75.5%–93.9%) and was superior (p = 0.01) to that of the optimal gain cutoff value (75.5% accuracy (63.8%–85.7%)). Conclusion: Machine learning models can effectively differentiate PCS from VN using only VHIT data, with comparable accuracy to expert clinicians. They hold promise as a tool to assist Emergency Room clinicians evaluating patients with acute vestibular syndrome.
Background: Acute unilateral vestibulopathy (AUVP) and posterior circulation stroke (PCS) are the two common causes of the acute vestibular syndrome. We developed and evaluated machine learning models to differentiate AUVP and PCS using combinations of patient history, examination, and vestibular tests. Methods: We recruited 294 patients presenting to the Emergency Room (ER) with acute vestibular syndrome (AUVP: n = 163, PCS: n = 131). Data from history, bedside examination and vestibular tests (video-nystagmography (VNG), video head-impulse test (VHIT), vestibular-evoked myogenic potentials (VEMP) and subjective visual horizontal) were used for machine learning model development. Different subsets of data simulated three scenarios hierarchically reflecting different levels of clinical expertise and resources: Tier 1 represented an ER with neuro-otology support (history, neuro-otological examination, VNG, VHIT, ocular VEMP), Tier 2 an ER with VHIT (history, basic examination, VHIT) and Tier 3 an ER reliant on history and basic examination only. Model performance was also compared against the HINTS test (head impulse, nystagmus, test-of-skew). Results: 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. The most important variables were the bedside head-impulse test, presence of focal neurological symptoms and spontaneous nystagmus slow phase velocity in Tier 1; focal neurological symptoms in Tier 2; and age in Tier 3. Conclusion: Machine learning models can accurately separate PCS and AUVP and hold promise as diagnostic aids for frontline clinicians.