June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
An Algorithm for Detecting Leaky Retinal Microaneurysms
Author Affiliations & Notes
  • Colyn Munn
    Emagix Inc., Halifax, Nova Scotia, Canada
  • Lyna Kamintsky
    Emagix Inc., Halifax, Nova Scotia, Canada
    Medical Neuroscience, Dalhousie University Faculty of Medicine, Halifax, Nova Scotia, Canada
  • Yonatan Serlin
    Neurology Residency Training Program, McGill University, Montreal, Quebec, Canada
  • Aaron Winter
    Ophthalmology and Visual Sciences, The University of British Columbia Faculty of Medicine, Vancouver, British Columbia, Canada
  • Alan Cruess
    Ophthalmology and Visual Sciences, Dalhousie University, Halifax, Nova Scotia, Canada
  • Alon Friedman
    Emagix Inc., Halifax, Nova Scotia, Canada
    Medical Neuroscience, Dalhousie University Faculty of Medicine, Halifax, Nova Scotia, Canada
  • Jaime Levy
    Ophthalmology, Hadassah-Hebrew University Medical Center, Jerusalem, Israel
  • Oren Tomkins-Netzer
    Ophthalmology, Carmel Medical Center, Haifa, Israel
    Ruth and Bruch Rappaport Faculty of Medicine, Technion Israel Institute of Technology, Haifa, Israel
  • Footnotes
    Commercial Relationships   Colyn Munn Emagix INC., Code E (Employment), Emagix INC., Code P (Patent); Lyna Kamintsky Emagix INC., Code F (Financial Support), Emagix INC., Code P (Patent); Yonatan Serlin None; Aaron Winter None; Alan Cruess Novartis, Code C (Consultant/Contractor), Bayer, Code C (Consultant/Contractor); Alon Friedman Emagix INC., Code P (Patent), Emagix INC., Code S (non-remunerative); Jaime Levy None; Oren Tomkins-Netzer Novartis, Code C (Consultant/Contractor), Allergan, Code C (Consultant/Contractor)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 253. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Colyn Munn, Lyna Kamintsky, Yonatan Serlin, Aaron Winter, Alan Cruess, Alon Friedman, Jaime Levy, Oren Tomkins-Netzer; An Algorithm for Detecting Leaky Retinal Microaneurysms. Invest. Ophthalmol. Vis. Sci. 2023;64(8):253.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Leaky microaneurysms (MAs) are a key target of treatments aimed at limiting vision loss in patients with diabetic retinopathy. Here, we present a novel algorithm for automatic detection of leaky microaneurysms from fluorescein angiography (FA) data.

Methods : The algorithm was evaluated using a dataset of 7 control eyes, 18 eyes with non-proliferative diabetic retinopathy (NPDR), and 24 eyes with proliferative diabetic retinopathy (PDR). For each eye, the algorithm generated a map of leaky and non-leaky MAs. Maps of BRB leakage and ischemia were also generated for each eye, to evaluate the co-localization between these markers and MAs. A circular grid (6mm in diameter and centered at the fovea) was used to divide each retina into four quadrants and evaluate the presence of MAs, BRB leakage, and ischemia in each quadrant. The utility of the tool was assessed by comparing the agreement among three retina-trained clinicians when identifying MAs with or without the algorithm-generated maps. Rater agreement was evaluated using the intraclass correlation coefficient.

Results : The algorithm detected 14.40±1.28 MAs per quadrant in the PDR group, 11.99±1.48 MAs per quadrant in the NPDR group, and 0.18±0.15 in controls. The NPDR and PDR groups had significantly more MAs than controls (p<.001; Fig.1A), with no difference between NPDR and PDR. On average, ~49% of MAs were leaky in the NPDR group and ~60% in the PDR group. No leaky MAs were found in controls. The NPDR and PDR groups had significantly more leaky MAs than controls (p<.001; Fig.1B), with no difference between NPDR and PDR. MAs were found in quadrants with and without BRB leakage and/or ischemia (Fig. 2A). Leaky MAs, however, were more likely to be present in quadrants with evidence of BRB leakage (p<.001; Fig. 2B). Agreement among retina-trained clinicians improved from ‘poor’ (ICCA,1=0.32) to ‘excellent’ (ICCA,1=0.75) when provided access to the algorithm-generated MA maps. For leaky MAs, the rater agreement improved from ‘poor’ (ICCA,1=0.28) to ‘good’ (ICCA,1=0.73).

Conclusions : Our findings suggest that the developed algorithm stands to help clinicians identify leaky and non-leaky MAs in clinical settings. Our results also show that MAs can be found prior to the appearance of other hallmarks of DR, while leaky MAs tend to be co-localized with BRB leakage.

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

 

Quantification of microaneurysms, BRB leakage, and ischemia.

Quantification of microaneurysms, BRB leakage, and ischemia.

 

Leaky microaneurysms are likely to be colocalized with BRB leakage.

Leaky microaneurysms are likely to be colocalized with BRB leakage.

×
×

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.

×