June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Differentiating Multiple Sclerosis and Glaucoma with Partial Least Squares Discriminant Analysis of Peripapillary Retinal Nerve Fiber Layer Thickness Patterns
Author Affiliations & Notes
  • Po-Han Yeh
    Ophthalmology, Oregon Health & Science University School of Medicine, Portland, Oregon, United States
  • Elizabeth Sarah White
    Ophthalmology, Oregon Health & Science University School of Medicine, Portland, Oregon, United States
  • Ou Tan
    Ophthalmology, Oregon Health & Science University School of Medicine, Portland, Oregon, United States
  • Dongseok Choi
    Ophthalmology, Oregon Health & Science University School of Medicine, Portland, Oregon, United States
  • Aiyin Chen
    Ophthalmology, Oregon Health & Science University School of Medicine, Portland, Oregon, United States
  • Eliesa Ing
    Ophthalmology, Oregon Health & Science University School of Medicine, Portland, Oregon, United States
  • Elizabeth Silbermann
    Neurology Multiple Sclerosis, Portland VA Medical Center, Portland, Oregon, United States
  • David Huang
    Ophthalmology, Oregon Health & Science University School of Medicine, Portland, Oregon, United States
  • Footnotes
    Commercial Relationships   Po-Han Yeh None; Elizabeth White None; Ou Tan Visionix/Optovue, Code P (Patent), Visionix/Optovue, Code R (Recipient); Dongseok Choi None; Aiyin Chen None; Eliesa Ing None; Elizabeth Silbermann None; David Huang Boeringer Ingelheim, Code C (Consultant/Contractor), Visionix/Optovue, Code F (Financial Support), Visionix/Optovue, Code P (Patent), Visionix/Optovue, Code R (Recipient)
  • Footnotes
    Support   NIH grants R01EY023285, R21 EY032146, P30 EY010572, Unrestricted grant from Research to Prevent Blindness
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 365. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Po-Han Yeh, Elizabeth Sarah White, Ou Tan, Dongseok Choi, Aiyin Chen, Eliesa Ing, Elizabeth Silbermann, David Huang; Differentiating Multiple Sclerosis and Glaucoma with Partial Least Squares Discriminant Analysis of Peripapillary Retinal Nerve Fiber Layer Thickness Patterns. Invest. Ophthalmol. Vis. Sci. 2023;64(8):365.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : Multiple sclerosis (MS) and glaucoma (GL) both result in peripapillary nerve fiber layer (pNFL) thinning. This research aims to identify differentiating patterns of thinning with the aid of partial least squares discriminant analysis (PLS-DA).

Methods : In this cross-sectional observational study MS patients were diagnosed with the 2017 McDonald Criteria, GL patients had disc rim thinning or a nerve fiber layer defect confirmed on stereo photography, with or without perimetric defect. Healthy control eyes had normal eye examination, including disc and perimetry. Disc scans were performed on the participants using Solix (Visionix/Optovue, CA), a spectral-domain OCT system. The pNFL thickness was measured in 8 modified Garway-Heath sectors. Normal sectoral reference values and 2.5 percentile cutoffs were obtained from healthy control eyes. GL and MS eyes with any sectoral pNFL defect (thickness below the 2.5 percentile cutoff) were selected for PLS-DA. The sectoral pNFL thickness values were age and axial length adjusted, then normalized by the average values in control eyes with a cap of 100%. PLS-DA was conducted with these normalized sector patterns. The first three components were used to classify each subject as normal, MS, or GL. Classification accuracy was evaluated with five-fold cross-validation.

Results : We enrolled 55 control subjects (55 eyes), 46 MS subjects (92 eyes), and 80 GL subjects (80 eyes). All control eyes were analyzed. Thirty eyes from 20 MS patients and 61 eyes (38 with perimetric defect) from 61 GL patients had at least one sectoral pNFL defect and were selected for analysis. The global pNFL thickness (mean±SD) was 92.7±8.1µm for control eyes, 78.5±10.7µm for MS eyes, and 66.4±13.0µm for GL eyes. The distribution of components (Figure 1) showed that the first component mainly differentiated normal eyes from diseased eyes, while the second and third components mainly differentiated between MS and GL eyes. The PLS-DA model’s classification accuracy was excellent for normals, good for GL, and fair for MS (Table 1). The overall accuracy was 86%.

Conclusions : It is possible for a machine learning method to differentiate between MS and GL with good accuracy using sectoral pattern of pNFL thinning.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

Figure1. Relative Distribution of Components

Figure1. Relative Distribution of Components

 

Table1. Classification Accuracy of Partial Least Squares Discriminant Analysis

Table1. Classification Accuracy of Partial Least Squares Discriminant Analysis

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×