July 2018
Volume 59, Issue 9
Free
ARVO Annual Meeting Abstract  |   July 2018
Artificial intelligence to predict optimal retreatment intervals in treat-and-extend (T&E)
Author Affiliations & Notes
  • Hrvoje Bogunovic
    Department of Ophthalmology, Medical University of Vienna, Vienna, Austria
  • Sebastian M Waldstein
    Department of Ophthalmology, Medical University of Vienna, Vienna, Austria
  • Amir Sadeghipour
    Department of Ophthalmology, Medical University of Vienna, Vienna, Austria
  • Bianca S Gerendas
    Department of Ophthalmology, Medical University of Vienna, Vienna, Austria
  • Ursula Schmidt-Erfurth
    Department of Ophthalmology, Medical University of Vienna, Vienna, Austria
  • Footnotes
    Commercial Relationships   Hrvoje Bogunovic, None; Sebastian Waldstein, Bayer (F), Bayer (C), Genentech (F), Novartis (C); Amir Sadeghipour, None; Bianca S Gerendas, Roche (C); Ursula Schmidt-Erfurth, Bayer (C), Boehringer (C), Carl Zeiss Meditec (C), Novartis (C)
  • Footnotes
    Support  Austrian Federal Ministry of Science, Research and Economy; National Foundation for Research, Technology and Development; Novartis AG.
Investigative Ophthalmology & Visual Science July 2018, Vol.59, 1620. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Hrvoje Bogunovic, Sebastian M Waldstein, Amir Sadeghipour, Bianca S Gerendas, Ursula Schmidt-Erfurth; Artificial intelligence to predict optimal retreatment intervals in treat-and-extend (T&E). Invest. Ophthalmol. Vis. Sci. 2018;59(9):1620.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Treat-and-extend (T&E) is an increasingly popular treatment regimen in anti-VEGF therapy in neovascular age-related macular degeneration (nAMD). We developed and evaluated a predictive model based on machine learning to determine a priori the optimum interval in a T&E regimen using clinical information and optical coherence tomography (OCT) based biomarkers.

Methods : SD-OCT volume scans (512x128x1024 voxels, Cirrus, Zeiss, or 512x49x496 voxels, Spectralis, Heidelberg Engineering) were processed at baseline and after the first anti-VEGF injection (Month 1). First, automated segmentations were performed (Figure 1) in which intraretinal (IRF) and subretinal (SRF) fluids were segmented using a deep learning convolutional neural network. Retinal layers were segmented with a graph-theoretic approach. Second, a set of quantitative features from the segmented layers and fluid regions was computed across an Early Treatment Diabetic Retinopathy Study (ETDRS) grid to describe the retinal pathomorphology both quantitatively and spatially. Finally, using the computed set of OCT features and clinical info, a predictive model of future treatment intervals was built using machine learning and was evaluated with cross-validation.

Results : Clinical trial data of 210 evaluable patients receiving standardized ranibizumab T&E therapy for 1 year according to the protocol specified in the TREND study were used. The maximum treatment intervals ranged from 4 to 12 weeks; 107/210 patients reached and maintained long (8, 10, 12 weeks) and other 103/210 had short (4, 6 weeks) treatment intervals. The model predicted (Figure 2) the maximum treatment interval in an individual patient within 4.7 weeks (95% confidence interval (CI)) and it identified short vs. long interval groups with a mean accuracy of 0.72 (CI: 0.64-0.79) area under the curve (AUC). The amount of SRF remaining at Month 1 was found to be the most important predictive feature.

Conclusions : We proposed and evaluated machine learning methodology to predict the T&E treatment intervals. The results are a promising step toward image-guided prediction of optimal treatment intervals in nAMD therapy and may help to personalize anti-VEGF therapy.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.

 

Segmented morphological features: Total retinal thickness (TRT), intraretinal fluid (IRF), and subretinal fluid (SRF).

Segmented morphological features: Total retinal thickness (TRT), intraretinal fluid (IRF), and subretinal fluid (SRF).

 

(a) Scatter plot of maximum interval predictions. (b) Receiver operating curve for interval group identification.

(a) Scatter plot of maximum interval predictions. (b) Receiver operating curve for interval group identification.

×
×

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.

×