August 2019
Volume 60, Issue 11
Free
ARVO Imaging in the Eye Conference Abstract  |   August 2019
Artificial intelligence-based automated segmentation of subretinal fluid and subretinal pigment epithelial fluid in patients with chronic central serous chorioretinopathy (CSC)
Author Affiliations & Notes
  • Mustafa Safi
    Ophthalmology, California Pacific Medical Center, San Leandro, California, United States
  • Roger Goldberg
    Ophthalmology, California Pacific Medical Center, San Leandro, California, United States
  • Footnotes
    Commercial Relationships   Mustafa Safi, None; Roger Goldberg, None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science August 2019, Vol.60, PB096. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Mustafa Safi, Roger Goldberg; Artificial intelligence-based automated segmentation of subretinal fluid and subretinal pigment epithelial fluid in patients with chronic central serous chorioretinopathy (CSC). Invest. Ophthalmol. Vis. Sci. 2019;60(11):PB096.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Automated image analysis using artificial intelligence-powered software offers the promise of more efficient detection, identification and analysis of ocular pathology. Pegasus, an automated AI decision support platform, was used to identify both subretinal fluid (SRF) and sub-retinal pigment epithelium fluid (PED) in patients with chronic CSC (Figure 1), and the accuracy of these results was evaluated.

Methods : Optical coherence tomography (OCT) volume scans of patients with CSC who participated in the Short Term Oral Mifepristone for CSC (STOMP) study were evaluated by the Pegasus system (Visulytix, London, UK). The results were then evaluated by two graders, and scored for the entire volume scan from 0 – 4, based on the accuracy of the automated detection results (0, 0-20% of fluid detected; 1: 21-40%, 2: 41-60%; 3: 61-80%; 4: 81-100%). Scores were given to both the SRF and PED volumes and averaged; qualitative notes were obtained as well.

Results : 30 eyes were evaluated at 5 different time points; a total of 145 OCT volume scans were assessed. The mean score for the accuracy of SRF and PED detection were 3.5 and 1.9, respectively. The majority of scans were graded as a 4 for SRF detection (64%), while PED detection was scored a 4 by both graders only 32% of scans. Eyes with large SRF and serous PEDs were, for the most part, accurately detected, though as the fluid waned during the course of the STOMP study, very small slivers of SRF or PED went under-detected. The presence of sub-retinal hyperreflective material on OCT seemed to present the largest hurdle for the accurate quantification of SRF (Figure 2). Of note, there was less variability in the SRF scoring than the PED scoring (p<0.05).

Conclusions : The Pegasus AI automated AI decision support system performed well to detect SRF and PED volume in a population of CSC patients, and may be helpful as a tool to identify and follow these patients over time. Expanded data sets may help further improve the detection algorithms, especially in a dynamic disease like CSC, where the fluid status can fluctuate significantly over short time periods.

This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.

 

 

×
×

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.

×