June 2022
Volume 63, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2022
Predicting 5-year progression to vision-threatening complications using machine learning
Author Affiliations & Notes
  • Dimitrios Damopoulos
    F Hoffmann-La Roche AG, Basel, Basel-Stadt, Switzerland
  • Luís Mendes
    Associacao para a Investigacao Biomedica e Inovacao em Luz e Imagem, Coimbra, Coimbra, Portugal
  • Torcato Santos
    Associacao para a Investigacao Biomedica e Inovacao em Luz e Imagem, Coimbra, Coimbra, Portugal
  • Ales Neubert
    F Hoffmann-La Roche AG, Basel, Basel-Stadt, Switzerland
  • Daniela Ferrara
    Genentech Inc, South San Francisco, California, United States
  • Jose G Cunha-Vaz
    Associacao para a Investigacao Biomedica e Inovacao em Luz e Imagem, Coimbra, Coimbra, Portugal
  • Fethallah Benmansour
    F Hoffmann-La Roche AG, Basel, Basel-Stadt, Switzerland
  • Footnotes
    Commercial Relationships   Dimitrios Damopoulos F Hoffmann-La Roche, Code C (Consultant/Contractor), HAYS plc, Code E (Employment); Luís Mendes None; Torcato Santos None; Ales Neubert F Hoffmann-La Roche, Code E (Employment); Daniela Ferrara Genentech Inc, Code E (Employment), F Hoffmann-La Roche, Code I (Personal Financial Interest); Jose Cunha-Vaz Carl Zeiss Meditec, Alimera Sciences, Allergan, Bayer, Gene Signal, Novartis, Pfizer, Oxular, Roche, Sanofi, Code C (Consultant/Contractor); Fethallah Benmansour F Hoffmann-La Roche, Code E (Employment)
  • Footnotes
    Support  F. Hoffmann-La Roche Ltd., Basel, Switzerland, provided support for the study and participated in the study design; conducting the study; and data collection, management, and interpretation.
Investigative Ophthalmology & Visual Science June 2022, Vol.63, 2085 – F0074. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Dimitrios Damopoulos, Luís Mendes, Torcato Santos, Ales Neubert, Daniela Ferrara, Jose G Cunha-Vaz, Fethallah Benmansour; Predicting 5-year progression to vision-threatening complications using machine learning. Invest. Ophthalmol. Vis. Sci. 2022;63(7):2085 – F0074.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To develop and evaluate a method for predicting 5-year risk of progression to a vision-threatening complication (VTC) in patients with diabetes and absent or mild diabetic retinopathy (DR).

Methods : Systemic and retinal imaging–based measurements (optical coherence tomography [OCT] and color fundus photographs) were collected from a cohort study (NCT03010397) of 172 patients with diabetes (32% female; mean age at baseline, 63 years). Per inclusion criteria, maximum level of DR at baseline was 35, as classified by the Diabetic Retinopathy Severity Scale. Patients were followed up annually for up to 5 years or until they developed a VTC (OCT-defined centrally involved macular edema, clinically significant macular edema, or proliferative DR). Of 172 patients, 27 (15.7%) progressed to ≥ 1 VTC. Linear regression models were trained and evaluated using 5-fold cross-validation for predicting VTC progression events. Features were either specified via an unsupervised hierarchical clustering method that reduced dimensionality of the feature space or selected manually guided by expert domain knowledge. The selected features of the manual approach consisted of 3 systemic and 5 imaging-based features, summarized in Table 1. Area under the receiver operating characteristic curve (AUC) was used as the performance metric.

Results : The trained models predicted progression to VTC with a mean (SD) AUC of 0.76 (0.04) when using the unsupervised dimensionality reduction approach and 0.78 (0.09) when using the 8 manually selected features. The models performed with a mean (SD) AUC of 0.72 (0.11) when using only the 5 selected imaging–based features versus 0.64 (0.05) when using only the 3 selected systemic features.

Conclusions : In this cohort of patients with absent or mild DR, the models predicted progression events with a mean AUC of 0.78 despite the low progression rate to a VTC and with comparable performance for both the manual and unsupervised approaches. Our results also indicate that imaging features alone are more powerful than systemic features alone in predicting progression to a VTC. With validation from larger and more diverse patient sets, predictive models of VTC in patients with DR could be valuable for informing personalized monitoring and follow-up in both clinical development and clinical practice.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

 

Table 1. Mean (SD) Values of the Employed Features of the Manual Approach at the Baseline Visit

Table 1. Mean (SD) Values of the Employed Features of the Manual Approach at the Baseline Visit

×
×

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.

×