Evaluating impaired consciousness after acquired brain injury using a virtual-reality-based eye-tracking system

Background Virtual reality (VR) can provide an experimental basis for inferring consciousness using information obtained from the responses of persons with disorders of consciousness (DOC) to denoised exogenous stimuli. Although integration of eye-tracking technologies has been proposed for evaluating levels of DOC, the calibration process poses substantial challenges or may be infeasible for persons with DOC. We aimed to demonstrate the validity and clinical utility of biomarkers obtained from pupil movements in response to visuoauditory stimuli presented in a VR environment using eye-tracking technology, while addressing the limitations of uncalibrated pupil trajectories. Methods We enrolled persons with prolonged DOC caused by acquired brain injury who showed continuous eye-opening for at least 15 min, along with healthy individuals. Participants were shown nine visuoauditory stimuli in a three-dimensional VR space while pupil movements were measured using an eye-tracking system. We calculated the relative pupil tracking length for visual and auditory stimuli (RPTL-V and RPTL-A). We established their cut-off values based on their congruence with the evoked potential test result to ascertain the presence of a “visuoauditory response by the RPTL.” Based on these results and the “visuoauditory response by Coma Recovery Scale-Revised (CRS-R),” the individuals were classified into “overt tracking,” “covert tracking,” and “no sign of tracking” groups. After 1 year, we assessed whether the participants could obey a simple command. Results Fifteen persons with prolonged DOC (median age, 67 [interquartile range {IQR}, 64.5–72.5] years; 9 [60%] women) and six healthy individuals (median age, 55 [IQR, 52.3–58.3] years; 3 [50%] women) participated. The RPTL-V and RPTL-A distribution varied according to the level of DOC and integrity of the visual or auditory pathway. The RPTL-V and RPTL-A cut-off values were 14.737 and 30.019, respectively. Frequencies of simple command obeying in the groups were: overt tracking, 5/8 (62.5%); covert tracking, 2/4 (50%); and no sign of tracking, 0/3 (0%). In 1/15 persons with DOC, a visuoauditory response not detected through the CRS-R was identified via the RPTL. Conclusions A VR-based eye-tracking system can quantitatively assess DOC, offering valid and clinically useful support for diagnosis and prognosis in conjunction with the CRS-R.