Relevance of artifact removal and number of stimuli for video head impulse test examination

Abstract

Objective: To evaluate the effect of artifacts on the impulse and response recordings with the video head impulse test (VHIT) and determine how many stimuli are necessary for obtaining acceptably efficient measurements. Methods: One hundred fifty patients were examined using VHIT and their registries searched for artifacts. We compared several variations of the dataset. The first variation used only samples without artifacts, the second used all samples (with and without artifacts), and the rest used only samples with each type of artifact. We calculated the relative efficiency (RE) of evaluating an increasingly large number of samples (3 to 19 per side) when compared with the complete sample (20 impulses per side). Results: Overshoot was associated with significantly higher speed (p = 0.005), higher duration (p < 0.001) and lower amplitude of the impulses (p = 0.002), and consequent higher saccades’ latency (p = 0.035) and lower amplitude (p = 0.025). Loss of track was associated with lower gain (p = 0.035). Blink was associated with a higher number of saccades (p < 0.001), and wrong way was associated with lower saccade latency (p = 0.012). The coefficient of quartile deviation escalated as the number of artifacts of any type rose, indicating an increment of variability. Overshoot increased the probability of the impulse to lay on the outlier range for gain and peak speed. Blink did so for the number of saccades, and wrong way for the saccade amplitude and speed. RE reached a tolerable level of 1.1 at 7 to 10 impulses for all measurements except the PR score. Conclusions: Our results suggest the necessity of removing artifacts after collecting VHIT samples to improve the accuracy and precision of results. Ten impulses are sufficient for achieving acceptable RE for all measurements except the PR score.