Abstract
Purpose: :
To measure tear film surface quality (TFSQ) in healthy normal and dry eye subjects with three noninvasive techniques and to investigate their potential to predict dry eye.
Methods: :
Thirty four subjects, aged from 20 to 68 years were recruited for this study. Clinical assessment of the dry eye was performed using McMonnies questionnaire, FTBUT, corneal (fluorescein), and conjunctival (lissamine green) staining (NEI grading). All clinical measurements were performed by one experienced clinician who was masked with respect to the measurements performed with the noninvasive methods, which included dynamic-area high speed videokeratoscopy (HSV), dynamic wavefront sensing (DWS), and lateral shearing interferometry (LSI). The dynamic-area HSV relies on measuring changes in the recorded Placido disc pattern projected on the cornea. In DWS, temporal changes of higher order aberrations (HOA), total comatic terms (ComaV, and ComaH) and the Zernike polynomial RMS fit error were used as indicators of TFSQ. In LSI, temporal changes of the frequency characteristics of interferograms were used as a TFSQ indicator. Noninvasive measurement of TFSQ was performed in both natural and suppressed blinking conditions. Measures of TFSQ were used to calculate receiver operating characteristics (ROC) to show the capability of each technique to discriminate between dry eye and normal subjects.
Results: :
In the suppressed blinking conditions, the LSI method showed the best performance in terms of the area under the ROC curve, AUC=0.80, followed by HSV with AUC=0.72. The DWS achieved lower values: AUC[ComaV]=0.64, AUC[ComaH]=0.56, AUC[RMS fit]=0.57, and AUC[HOA]=0.60. In the natural blinking condition, the LSI method again showed the best performance, AUC=0.73 followed closely by HSV with AUC=0.71. Methods based on DWS showed much lower AUC values indicating poor detection performance.
Conclusions: :
Noninvasive techniques of tear film surface assessment can be used for predicting dry eye and this can be achieved in natural blinking as well as suppressed blinking conditions. In our study, LSI showed the best detection performance, closely followed by the dynamic-area HSV. Wavefront sensing techniques are less powerful, particularly in natural blinking conditions.
Keywords: cornea: tears/tear film/dry eye • imaging/image analysis: clinical • image processing