June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Predicting Visual Field Worsening with Longitudinal Optical Coherence Tomography Data Using a Gated Transformer Network
Author Affiliations & Notes
  • Kaihua Hou
    Johns Hopkins University, Baltimore, Maryland, United States
  • Chris Bradley
    Johns Hopkins University, Baltimore, Maryland, United States
  • Patrick Herbert
    Johns Hopkins University, Baltimore, Maryland, United States
  • Chris A Johnson
    University of Iowa Hospitals and Clinics, Iowa City, Iowa, United States
  • Michael Wall
    University of Iowa Hospitals and Clinics, Iowa City, Iowa, United States
  • Pradeep Y Ramulu
    Johns Hopkins University, Baltimore, Maryland, United States
  • Mathias Unberath
    Johns Hopkins University, Baltimore, Maryland, United States
  • Jithin Yohannan
    Johns Hopkins University, Baltimore, Maryland, United States
  • Footnotes
    Commercial Relationships   Kaihua Hou None; Chris Bradley None; Patrick Herbert None; Chris Johnson None; Michael Wall None; Pradeep Ramulu None; Mathias Unberath None; Jithin Yohannan None
  • Footnotes
    Support  K23 EY032204-02
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 978. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Kaihua Hou, Chris Bradley, Patrick Herbert, Chris A Johnson, Michael Wall, Pradeep Y Ramulu, Mathias Unberath, Jithin Yohannan; Predicting Visual Field Worsening with Longitudinal Optical Coherence Tomography Data Using a Gated Transformer Network. Invest. Ophthalmol. Vis. Sci. 2023;64(8):978.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To identify visual field (VF) worsening from longitudinal optical coherence tomography (OCT) data using a Gated Transformer Network (GTN), and to examine how GTN performance varies for different definitions of visual field worsening. We also investigated the feature importance of the GTN using Shapely analysis.

Methods : For each eye, we used three trend-based methods (MD, VFI, and PLR slope) and three event-based methods (GPA, CIGTS, and AGIS) to define VF worsening. Additionally, we created an algorithm (M6), that classifies an eye as worsening if 4 or more of the 6 aforementioned methods classify the eye as worsening. Using these 7 reference standards for VF worsening, we trained 7 GTNs that accept a series of at least 5 as input OCT scans and provide as output a probability of VF worsening. The GTN architecture is illustrated in Figure 1. GTN performance was compared to non-deep learning models with the same serial OCT input from previous studies — linear mixed-effects models (MEM) and naïve Bayes classifiers (NBC) — using the same training sets and reference standards as for the GTN. We analyzed the feature importance of the GTN model by calculating the Shapley values of test-set predictions.

Results : We included a total of 4,211 eyes (2,666 patients). The M6 algorithm labeled 63 eyes (1.50%) as worsening. The GTN achieved an AUC (95% CI) of 0.97 (0.88, 1.00) when trained with M6. GTNs trained and optimized with the other 6 reference standards had AUC ranging from 0.78 (MD slope) to 0.89 (AGIS). The 7 GTNs significantly outperformed all 7 MEMs and all 7 NBCs accordingly. According to the Shapley analysis, the top three important features for M6-trained GTN predictions are average RNFL thickness, vertical cup ratio, and age.

Conclusions : GTN models trained with serial OCT data may be used to identify VF worsening. Our GTNs demonstrated good performance at predicting VF worsening with serial OCT for all 7 reference standards. Training the GTN with event-based reference standards resulted in slightly better performance than with trend-based reference standards. Among all features, the GTN predictions are affected by average RNFL thickness, vertical cup ratio, and age the most.

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

 

Figure 1. Gated Transformer Network (GTN) architecture.

Figure 1. Gated Transformer Network (GTN) architecture.

 

×
×

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.

×