SSVE(PLR): Comparing target classification via pupillary light responses to standard EEG-based SSVEP

The pupillary light response (PLR) is a predictable physiological response occurring whenever there is a sudden and sufficient brightness change in the environment, and is characterized by a canonical waveform shape. This study leveraged the predictable timing of PLR onset latencies to classify fixations to target objects, analogous to BCI approaches using steady-state visual-evoked potentials (SSVEP). In the same sessions and subjects, we aimed to directly compare classification accuracy of our PLR-based approach to a classic SSVEP approach. In our VR experiment (n=12) with simultaneous EEG and eye-tracking, four targets were presented along two cardinal axes (up, down, left, right) relative to the center fixation cross. Stimuli were cubes with frequency-modulated high contrast flicker, each at different frequencies (9, 10, 11, 12 Hz) to modulate the EEG signal for SSVEP analysis, as well as amplitude-modulated flicker with a stable frequency of 1 Hz to modulate the PLR at four different temporal phases (separated by 250ms). First, we developed a velocity-based algorithm to detect the occurrence of PLR events in a baseline task with a single flashing stimulus to empirically measure PLR onset latency for each subject. Second, we confirmed that PLRs only occurred when subjects actually fixated on a target, even though multiple flashing stimuli were in the field-of-view. Third, we applied the PLR detection algorithm and predicted which target was fixated by looking back in time, using the baseline-estimated PLR latency, to identify the nearest-in-time target that flashed at that moment. Using this classification scheme, we found that the timing of PLR events could correctly classify ~70% of targets (chance = 25%), which was comparable to SSVEP results in the same data set (~70% accuracy). These results highlight measurement reliability and target-specificity of PLRs in complex environments, and the potential for interesting BCI-like applications using eye-tracking in VR.