The value of saccade metrics and VOR gain in detecting a vestibular stroke.

Abstract

Objective: A normal video Head Impulse Test is the gold standard in the emergency department to rule-in patients with an acute vestibular syndrome and a stroke. We aimed to compare the diagnostic accuracy of vHIT metrics regarding the vestibulo-ocular reflex gain and the corrective saccades in detecting vestibular strokes.

Methods: Prospective cross-sectional study (convenience sample) of patients presenting with acute vestibular syndrome in the emergency department of a tertiary referral centre between February 2015 and May 2020. We screened 1677 patients and enrolled 76 patients fulfilling the inclusion criteria of acute vestibular syndrome. All patients underwent video head impulse test with automated and manual data analysis. A delayed MRI served as a gold standard for vestibular stroke confirmation.

Results: Out of 76 patients, 52 were diagnosed with acute unilateral vestibulopathy and 24 with vestibular strokes. The overall accuracy of detecting stroke with an automated vestibulo-ocular reflex gain was 86.8%, compared to 77.6% for cumulative saccade amplitude and automatic saccade mean peak velocity measured by an expert and 71% for cumulative saccade amplitude and saccade mean peak velocity measured automatically. Gain misclassified 13.1% of the patients as false positive or false negative, manual cumulative saccade amplitude and saccade mean peak velocity 22.3%, and automated cumulative saccade amplitude and saccade mean peak velocity 28.9% respectively.

Conclusions: We found a better accuracy of video head impulse test for the diagnosis of vestibular strokes when using the vestibulo-ocular reflex gain than using saccade metrics. Nevertheless, saccades provide an additional and important information for video head impulse test evaluation. The automated saccade detection algorithm is not yet perfect compared to expert analysis, but it may become a valuable tool for future non-expert video head impulse test evaluations.