Title

Machine learning to predict pupillary dynamics in conscious visual perception

Website

http://medicine.yale.edu/lab/blumenfeld/

Description

Many recent studies have attempted to isolate the neural correlates of consciousness. A promising paradigm involves contrasting the mechanisms involved in the conscious perception of a visual stimulus with those at work when the same stimulus is presented but not consciously seen by the subject. One difficulty these studies often present is that they rely on the subject’s report (usually via button presses) to gauge his or her conscious awareness of the stimulus––the act of reporting upon one’s experience likely induces extra cognitive activity beyond what naturally occurs during conscious perception, such as memorizing some feature of the stimulus or executing a button press. To address this fundamental issue, we investigated whether pupillary responses could be used as a reliable covert indicator of conscious perception during a visual stimulus detection task. Pupil dilation has been previously observed during perceptual switching in binocular rivalry, and at the moment of object or facial recognition. We measured pupil diameter, gaze location, and eye-blinks continuously as subjects were presented with facial stimuli, calibrated to a low opacity such that it would be perceived during approximately 50% of trials. Significant diameter changes were observed in response to faces that were perceived by participants––on average, a peak dilation response occurred between 1 and 2 seconds post-stimulus, an effect that was not observed in the opposing “not-perceived” trials. In addition, we implemented a machine learning algorithm with a goal of using the pupillary data to predict whether conscious perception occurred on a trial-by-trial basis, which has reached about 71% accuracy, and continues to improve as we refine our results. In conclusion, our findings suggest that pupil diameter changes may be a good candidate for a physiological marker of conscious perception, which could be used both to further isolate the neural correlates of consciousness and to help assess the state of patients who are unresponsive, but possibly conscious.

This document is currently not available here.

Share

COinS
 

Machine learning to predict pupillary dynamics in conscious visual perception

Many recent studies have attempted to isolate the neural correlates of consciousness. A promising paradigm involves contrasting the mechanisms involved in the conscious perception of a visual stimulus with those at work when the same stimulus is presented but not consciously seen by the subject. One difficulty these studies often present is that they rely on the subject’s report (usually via button presses) to gauge his or her conscious awareness of the stimulus––the act of reporting upon one’s experience likely induces extra cognitive activity beyond what naturally occurs during conscious perception, such as memorizing some feature of the stimulus or executing a button press. To address this fundamental issue, we investigated whether pupillary responses could be used as a reliable covert indicator of conscious perception during a visual stimulus detection task. Pupil dilation has been previously observed during perceptual switching in binocular rivalry, and at the moment of object or facial recognition. We measured pupil diameter, gaze location, and eye-blinks continuously as subjects were presented with facial stimuli, calibrated to a low opacity such that it would be perceived during approximately 50% of trials. Significant diameter changes were observed in response to faces that were perceived by participants––on average, a peak dilation response occurred between 1 and 2 seconds post-stimulus, an effect that was not observed in the opposing “not-perceived” trials. In addition, we implemented a machine learning algorithm with a goal of using the pupillary data to predict whether conscious perception occurred on a trial-by-trial basis, which has reached about 71% accuracy, and continues to improve as we refine our results. In conclusion, our findings suggest that pupil diameter changes may be a good candidate for a physiological marker of conscious perception, which could be used both to further isolate the neural correlates of consciousness and to help assess the state of patients who are unresponsive, but possibly conscious.

http://elischolar.library.yale.edu/dayofdata/2016/posters/12