Development of oculomotor digital biomarkers using clinical examinations as "Ground Truth".

OBJECTIVE

The purpose of this study was to assess the validity of six computerized eye-tracking tests against a clinician-administered oculomotor exam.

METHODS

A total of 53 participants completed the horizontal random saccade (HRS), circular smooth pursuit (CSP), horizontal smooth pursuit (HSP), vertical smooth pursuit (VSP), horizontal saccade (HS), and vertical saccade (VS) oculomotor tests on the computerized system. A board-certified neurologist with 16 years of experience also conducted an oculomotor examination to mirror eye movement patterns.

RESULTS

Data analysis included a series of single-block logistic regressions to examine the scoring of the six eye-tracking tests (RightEye, LLC) to predict clinician-rated eye movement classifications (i.e., normal or abnormal). The computerized battery demonstrated concurrent validity for each of the six oculomotor tests as they significantly predicted the neurologist's classification. The sensitivity and specificity of the six eye-tracking tests ranged from 70.4 to 93.5% and from 84.6 to 90.5%, respectively. The diagnostic accuracy of the computerized tests ranged from good (78.8%) to excellent (92.3%). The area under the curve (AUC) analysis for the eye-tracking tests yielded values ranging from 0.734 (VSs) to 0.921 (HRSs).

CONCLUSION

The results suggest that each of the six computerized eye-tracking tests accurately distinguished between normal and abnormal oculomotor movements.