Glaucoma Classification Through SSVEP-Derived ON- and OFF-Pathway Features

Abstract

Purpose: This work aims to evaluate the relative contribution of the amplitude and phase of both ON-and OFF-pathway biased steady-state visually evoked potentials (SSVEPs) to the classification of patients with glaucoma from healthy controls. Methods: SSVEPs were recorded for sawtooth luminance increments (ON-biasing) and decrements (OFF-biasing), modulating at a temporal frequency of 2.73 Hz. SSVEP data from 98 adults with glaucoma and 71 controls were used to train a set of logistic regressions. Data were partitioned prior to training to investigate the relative contribution to classification for amplitude and phase features derived from ON-versus OFF-pathway stimulation. Results: We report moderate overall classification accuracy (area under the curve ∼0.7). Classification based solely on signal phase features significantly outperformed classification based solely on signal amplitude features. Classification using OFF-pathway biasing features produced a statistically significant improvement in classification only when training on signal amplitude features. This OFF advantage was not conserved in a dataset with low signal-to-noise eyes removed. Conclusions: Our findings highlight the informational value of signal phase, a metric often omitted in applications of the SSVEP to glaucoma and other optic neuropathies. Additionally, our results suggest that OFF-pathway amplitude features may be less vulnerable to the limitations imposed by a low signal-to-noise ratio. However, they are not indicative of a gross difference in glaucoma classification performance between ON-and OFF-pathway biased features. Translational Relevance: Electrophysiological estimates of visual signal delay should be considered in future clinical diagnostic tools as they make a material contribution to the classification of glaucomatous eyes.