June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Developing a continuous severity scale for Macular Telangiectasia
Author Affiliations & Notes
  • Yue Wu
    Ophthalmology, University of Washington, Seattle, Washington, United States
    The Roger and Angie Karalis Johnson Retina Center, Seattle, Washington, United States
  • Lea Scheppke
    Lowy Medical Research Institute, La Jolla, California, United States
    Scripps Research Institute Department of Molecular and Experimental Medicine, La Jolla, California, United States
  • Catherine A Egan
    Moorfields Eye Hospital NHS Foundation Trust, London, London, United Kingdom
    University College London, London, London, United Kingdom
  • Abraham Olvera-Barrios
    Moorfields Eye Hospital NHS Foundation Trust, London, London, United Kingdom
    University College London, London, London, United Kingdom
  • Tunde Peto
    Queen's University Belfast, Belfast, Belfast, United Kingdom
  • Cecilia S. Lee
    Ophthalmology, University of Washington, Seattle, Washington, United States
    Queen's University Belfast, Belfast, Belfast, United Kingdom
  • Emily Y. Chew
    National Eye Institute, Bethesda, Maryland, United States
  • Martin Friedlander
    Scripps Health, San Diego, California, United States
    Scripps Research Institute Department of Molecular and Experimental Medicine, La Jolla, California, United States
  • Aaron Y Lee
    Ophthalmology, University of Washington, Seattle, Washington, United States
    The Roger and Angie Karalis Johnson Retina Center, Seattle, Washington, United States
  • Footnotes
    Commercial Relationships   Yue Wu None; Lea Scheppke None; Catherine Egan Heidelberg, Code C (Consultant/Contractor), LMRI, Code F (Financial Support); Abraham Olvera-Barrios None; Tunde Peto Allergan (Abbvie), Apellis, Boehringer Ingleheim, Novartis, Roche, OPTOS, Oxurion, Heidelberg, Roche, Alimera, Code C (Consultant/Contractor); Cecilia Lee None; Emily Chew None; Martin Friedlander None; Aaron Lee Genentech, Code C (Consultant/Contractor), Verana, Code C (Consultant/Contractor), US Food and Drug Administration, Code E (Employment), Santen, Code F (Financial Support), Carl Zeiss Meditec, Code F (Financial Support), Novartis, Code F (Financial Support), Topcon, Code R (Recipient)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 1287. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Yue Wu, Lea Scheppke, Catherine A Egan, Abraham Olvera-Barrios, Tunde Peto, Cecilia S. Lee, Emily Y. Chew, Martin Friedlander, Aaron Y Lee; Developing a continuous severity scale for Macular Telangiectasia. Invest. Ophthalmol. Vis. Sci. 2023;64(8):1287.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Deep learning (DL) models have transformed medical image analysis by achieving state-of-the-art diagnosis classification accuracy. However, the models may be limited by discrete diagnosis training labels and could potentially yield more informative continuous diagnosis scales. We sought to develop a novel continuous severity scaling system for Macular Telangiectasia Type II (MacTel) by combining a DL classification model with Uniform Manifold Approximation and Projection (UMAP).

Methods : First, we trained a classifier on 2003 optical coherence tomography (OCT) volumes from 1089 participants in the MacTel Project to learn a discrete 7 step MacTel severity scale from Chew et al. Specifically, the classifier was a modified EfficientNet-B0 that took the central 20 B-scans from each macular volume as input. Next, a representative vector was extracted from the last feature layer of the trained classifier as input for the UMAP. The UMAP embedded these features into a continuous 2D manifold. We assessed the multi-view classifier in terms of test accuracy and the UMAP embedding in terms of disease progression in the subset of 120 patients with multi-year longitudinal OCT volume scans in manifold space.

Results : The multi-view classifier achieved top-1 accuracy of 63.3% (186/294) on held-out test OCT volumes. The test set confusion table is given in Figure 2a. The 2D UMAP embedding is shown in Figure 2b and color-coded by Chew et al. severity grades where available. The 2D embedding shows a clear continuous gradation of MacTel severity that has a Spearman Rank Correlation of 0.84 with Chew et al. grades. In addition, we performed a paired signed rank test of the 120 longitudinal MacTel patient eyes on the UMAP scores of their first scans versus that of their final scans. This test had a p-value of 0.00007, showing that eyes significantly worsen over time as expected for MacTel.

Conclusions : Our multi-view classifier and UMAP embedding generated a continuous severity scale for MacTel, without requiring continuous training labels. This technique can be applied to other diseases, which may lead to more accurate diagnosis, as well as improve understanding and analysis of disease progression.

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

 

Fig 1a. Chew et al. 7 stage MacTel grades as a decision tree. 1b. Schematic of Multi-view classifier and extraction of its last feature layer to learn a UMAP embedding.

Fig 1a. Chew et al. 7 stage MacTel grades as a decision tree. 1b. Schematic of Multi-view classifier and extraction of its last feature layer to learn a UMAP embedding.

 

Fig 2a. Test confusion table 2b. Continuous 2D UMAP embedding for all patients.

Fig 2a. Test confusion table 2b. Continuous 2D UMAP embedding for all patients.

×
×

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.

×