Automated epithelial thickness measurements on CIRRUS™ HD-OCT vs manual measurements

Abstract

Purpose : Epithelial thickness measurements (ETM) may play a role in the detection of early keratoconus (KCN), which is a contraindication for refractive surgery. Epithelial thickness also contributes to corneal power which affects the accuracy of refractive and intraocular lens surgery and may be correlated with dry eye.An ETM algorithm was developed and this study characterizes the accuracy of automated ETM from CIRRUS™ HD-OCT 5000 (ZEISS, Dublin, CA) as compared to manual measurements.

Methods : This was a prospective study where 1 eye of 11 normal subjects (Gp 1) and 11 post-LASIK (pL) or KCN subjects (5 pL+ 6 KCN eyes) (Gp 2) underwent acquisition of pachymetry scans using 3 CIRRUS HD-OCT devices.
The CIRRUS software examines B-scans and defines ETM as the closest distance between each corneal anterior surface point (center of the tear film layer) to the top of Bowman’s layer (BL). The automated algorithm generated average ETM in 25 sectors.
Manual ETM were measured by 3 graders using a custom-developed Matlab tool. The tool displayed flattened B-scans (A-scans were horizontally aligned according to the corneal front surface) in which graders identified the top surface of BL and determined manual ETM as the closest distance between the corneal anterior surface and the user selected point. The first 5 measurements from each grader in all 25 sectors were used for comparison. Some peripheral subfields had less than 5 ETM per grader due to difficulty in BL identification. Deming and linear regression, Bland-Altman analyses were performed.

Results : The Deming regression results for all sectors (Fig 1a) demonstrate strong correlation between the manual and automated ETM measurements for Gps 1&2 combined with slope 1.02 and intercept of -3.17. R² of Gp 1 was 0.68 and Gp 2 was 0.81. Bland-Altman plots (Fig 1b) demonstrate good agreement with a mean difference of -2.18 and limits of agreement of -7.23 to 2.86 for Gps 1&2 combined. There was higher variability and more outliers in Gp 2 (Fig 2) which is to be expected in a KCN and post-LASIK population subset.

Conclusions : Overall, CIRRUS automated ETM and manual have a strong correlation and good agreement for normal, post-LASIK and KCN corneas. Automated ETM may be a valuable clinical tool for refractive, intraocular lens surgery and dry eye.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.

 

Fig 1a&b Deming Regression & Bland-Altman of Gp 1&2 Combined

 

Fig 2 Deming Regression & Bland-Altman Plots of Gp 1 and 2