August 2021
Volume 62, Issue 11
Open Access
ARVO Imaging in the Eye Conference Abstract  |   August 2021
Automated Ultra-widefield Angiographic Early-Late Phase Frame Selection Using a Deep Learning-Based Phase Value Score and Optic Disc-Based Laterality and Field of View Classifier
Author Affiliations & Notes
  • Duriye Damla Sevgi
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Jon Whitney
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Margaret O'connell
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Danielle Burton
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Cindy Chen
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Phuoc-Hanh Le
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Jasmin Bhangu
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Sunil K. Srivastava
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • sumit sharma
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Kimberly Baynes
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Jamie Reese
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Justis P. Ehlers
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Footnotes
    Commercial Relationships   Duriye Damla Sevgi, None; Jon Whitney, None; Margaret O'connell, None; Danielle Burton, None; Cindy Chen, None; Phuoc-Hanh Le, None; Jasmin Bhangu, None; Sunil Srivastava, Abbvie (C), Allergan (F), Allergan (C), Eyepoint (F), Eyepoint (C), Eyevensys (F), Eyevenysy (C), Gilead (C), Leica (P), Novartis (C), Regeneron (F), Regeneron (C), Santen (F), Zeiss (C); sumit sharma, Alimera (C), Allergan (C), Bausch and Lomb (C), Clearside (C), Eyepoint (C), Genentech (C), Regeneron (C); Kimberly Baynes, None; Jamie Reese, None; Justis Ehlers, Adverum (C), Aerpio (F), Aerpio (C), Alcon (F), Alcon (C), Allergan (F), Allergan (C), Boehringer-Ingelheim (F), Genentech (F), Genentech/Roche (C), Leica (C), Novartis (F), Novartis (C), Regeneron (F), Regeneron (C), Santen (C), Stealth (C), Thrombogenics/Oxurion (F), Thrombogenics/Oxurion (C), Zeiss (C)
  • Footnotes
    Support  RPB Cole Eye Institutional Grant for support and NIH/NEI K23-EY022947
Investigative Ophthalmology & Visual Science August 2021, Vol.62, 76. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Duriye Damla Sevgi, Jon Whitney, Margaret O'connell, Danielle Burton, Cindy Chen, Phuoc-Hanh Le, Jasmin Bhangu, Sunil K. Srivastava, sumit sharma, Kimberly Baynes, Jamie Reese, Justis P. Ehlers; Automated Ultra-widefield Angiographic Early-Late Phase Frame Selection Using a Deep Learning-Based Phase Value Score and Optic Disc-Based Laterality and Field of View Classifier. Invest. Ophthalmol. Vis. Sci. 2021;62(11):76.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Automated optimal ultra-widefield fluorescein angiography (UWFA) frame selection is essential for fully automated UWFA analysis as a clinical decision support tool and could facilitate more efficient clinical review of UWFA studies. We evaluated the performance of deep learning augmented early and late phase selection application (ELSA) on eyes within the PANTHER study.

Methods : UWFA sequences from 122 visits for 42 eyes (21 patients) from the IRB-approved PANTHER clinical study for inflammatory eye disease were used to evaluate the performance of ELSA. A deep learning algorithm was developed to calculate the phase value of each frame using a regression algorithm trained by the angiographs and their time-stamps. Frames were categorized as early and late phases using the mean phase value of the angiographic session as the threshold. A U-Net architecture was trained to detect optic nerve location and categorize the frames based on their laterality and field of view (central, non-central). Retinal vessel areas were extracted by a deep learning algorithm from all central frames. The frames with maximum retinal vessel area per phase were selected for each eye as the optimal early and late phase frames. Two trained image analysts manually selected the subjective optimal frames and subsequently compared those to automated selections.

Results : ELSA successfully identified comparable quality early frames in 91.8% (112/122) of the sessions. The success rate for identifying late frames was 77.9% (95/122). The most common cause of algorithm failure to identify the optimal early image was due to incorrect detection of the optic disc in 72.7% and incorrect phase classification (i.e., incorrect phase value assignment) of an early image in 27.3%. In contrast, the most common cause of suboptimal late frame detection was due to incorrect phase classification of an early image (67.7%) compared to 33.3% due to incorrect localization of the optic disc.

Conclusions : Automated optimal frame selection using angiographic imaging features including optic disc location and retinal vessel area is feasible. Late frame selection is more susceptible to phase errors as vascular details are more prominent in the early phase images. Further model refinement will focus on optimizing the late phase scoring.

This is a 2021 Imaging in the Eye Conference abstract.

×
×

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.

×