Automatic tear film and tear meniscus parameter assessment in healthy subjects with ultrahigh-resolution OCT

Abstract

Purpose : To investigate the applicability of ultrahigh-resolution OCT (UHR-OCT) and fully automatic segmentation algorithms for the evaluation of tear parameters in healthy subjects.

Methods : Measurements were obtained from a study including ten eyes of ten healthy subjects (5 female, 5 male, age 30.9 ± 9.53 years, mean ± SD) with no signs of dry eye disease (DED). One measurement cycle consisted of a central tear film measurement, a lower tear meniscus measurement - both acquired with an UHR-OCT system - and a lipid layer thickness (LLT) measurement acquired with a commercial LipiView II Ocular Surface Interferometer (Tear Science Inc, NC, USA). Seven cycles were completed at 10 minute intervals. After the last cycle, fluorescein break up time (FBUT) and Schirmer I test values were measured. A custom Matlab (Matlab R2017a, The MathWorks, Inc., MA, USA) algorithm for the automatic tear meniscus segmentation was developed and added to an already existing tear film thickness (TFT) estimation algorithm. The new algorithm estimates tear meniscus area (TMA), height (TMH), depth (TMD) and radius of curvature (TMR) without any operator input from 3D OCT data (Fig. 1).

Results : Table 1 shows mean values and standard deviations automatically computed from the UHR-OCT measurements. Partial correlation coefficients were calculated to adjust for repeated measurements in the same subject. A highly significant correlation (p < 0.001) between all tear meniscus parameters (TMA, TMH, TMD, TMR) was observed after correction for multiple testing.

Conclusions : The good agreement between the measured parameters and the literature values for healthy subjects as well as the highly significant correlations between the tear meniscus parameter pairs indicate the suitability of UHR-OCT and automatic segmentation algorithms for the assessment of the tear film meniscus. In the future, the technology might allow for a better understanding of tear film dynamics and tear film alterations in DED.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

 

Fig. 1 - Automatic segmentation of the lower tear meniscus in a healthy subject. Estimated parameters are (A) tear meniscus area, (B) height, (C) depth and (D) radius of curvature. Green crosses indicate the points used for the estimation of the radius of curvature.