Menu

  1. Home
  2. /
  3. External Disease
  4. /
  5. FP597 : BOWMAN’S SURFACE...

Share this post

Dr.Jyoti Matalia, Dr.Himanshu Matalia,Dr.Nandini C,Dr.Shetty Bhujang K

Introduction:

Allergic eye is one of the commonest and most important associations with keratoconus (KC).As a result of extensive eye rubbing, patients with allergic eye disease (AED) might show some corneal topographic changes.Distribution of anterior surface aberrations of AED might overlap with KC, hence studying them can give new-insights towards their associations.Early stromal changes can be masked by epithelium thickness distribution, which can be studied using Bowman’s topography.

Purpose:

To assess the bowman’s surface topography in normal, AED and KC eyes and to derive an Artificial Intelligence (AI) classifier for diagnosing AED eyes

Methods:

This study was approved by ethics committee of Narayana Nethralaya Eye Hospital, Bangalore, India. The research adhered to the tenets of the Declaration of Helsinki. Written informed consent was taken from all the subjects.

This study included 100 normal, 20 AED and 20 KC eyes. Three consecutive measurements of the anterior segment optical coherence tomography (OCT) were obtained for each subject using the high speed Swept Source (SS)-OCT (DRI Triton, Topcon Inc., Japan). Exclusion criteria included any prior ocular surgery, cataract, scarring, or any other corneal degenerations. Also, scans with significant blinking errors were excluded.In this study, high speed SS-OCT acquired 12 radial frames. This high resolution SS-OCT uses a light source of 1050nm with A-scan rate of 100 KHz. OCT scans with good signal strength of 40 and above were included for the analysis. Further, the scans with visually significant deviation from center of the cornea was excluded. The distortion corrected B-frames of 6 mm were then utilized to segment air-epithelium (A-E) and epithelium-bowman’s (E-B) interface.1, 2The curvature was calculated using elevation data obtained from these segmented interfaces.

Statistical analysis:

All the variables were reported as mean ± SEM. Average Keratometry (Kmax) [diopters (D)] and aberrations (µm) for each measurement on OCT (A-E and E-B interface) were analyzed. Ray tracing method and Zernike analysis was used to calculate the aberrations, which included root mean square (RMS) of lower order aberrations (LOA) and higher order aberrations (HOA)& Coma. MedCalc v18.2.1(MedCalc, Ostend, Belgium) software and for artificial Intelligence. A p-value less than 0.05 was considered to be statistically significant.

A decision tree classifier (artificial intelligence) using Orange software was built using the following parameters:

  1. RMS of Coma,
  2. RMS of HOA
  3. RMS of LOA
  4. RMS of total aberrations at the Anterior surface
  5. RMS of total aberration of Bowman’s surface
  6. RMS of total aberrations of Anterior and Bowman’s surface combined RMS of coma.

Results:

The mean age of 100 normal, 20 AED and 20 KC eyes was 31.56 ± 0.89 years, 10.86 ± 0.53, 29.45 ± 1.18 years respectively.

Keratometry was significantly different between the normal, AED and the KC groups (p<0.001) with the AED group having relatively steeper corneas than the normal eyes and the KC group having the steepest corneas. Similarly, the RMS of Aberrations (µm) between the 3 groups was also statistically different (p<0.001) with the aberrations least in the normal group and relatively more in the AED group and highest in the KC group.

When the decision tree (AI classifier) was built around the above parameters, it provided the following classification:

(a) if RMS of total aberration at E-B interface is ≤ to 1.73 µm, then the eye was classified as normal;

(b) if RMS of total aberration is more than > 1.73 µm, RMS of coma at E-B interface ≤ 1.44 µm and RMS of HOA at A-E interface >1.56 µm, eye was classified as AED;

(c) RMS of total aberrations at E-B > 1.73 µm and RMS of coma at E-B interface > 1.44, eye was classified as KC.

AI classifier predominantly used Bowman’s surface parameters.

The classifier achieved an accuracy of 99% for normal, 80% for AED and 91.9% for KC eyes.

Conclusion: A decision tree classifier using A-E and E-B interfaces was a potent discriminator between AED and KC eyes. AI and OCT tomography detected unique pattern in AED eyes with a very good prediction accuracy. This study provided new insights in evaluating AED by employing E-B interface topography. Further study with larger sample size is needed.

References:

  1. Matalia H, Francis M, Gangil T, et al. Noncontact Quantification of Topography of Anterior Corneal Surface and Bowman’s Layer With High-Speed OCT. J Refract Surg 2017;33(5):330-336.
  2. Chiu SJ, Li XT, Nicholas P, et al. Automatic segmentation of seven retinal layers in SDOCT images congruent with expert manual segmentation. Opt Express 2010;18(18):19413-28.

 

Leave a Comment