Agreement rates for predominantly peripheral lesion(PPL) grading assessed using qualitative and quantitative methods on ultrawide field color images(UWF-CI) and fluorescein angiography(UWF-FA) in diabetic retinopathy(DR)

Mohamed Ashraf Elmasry, Abdulrahman Rageh, Recivall Pascual Salongcay, Jennifer K Sun, Lloyd Paul Aiello, Paolo S Silva

Research output: Contribution to journalMeeting abstractpeer-review

Abstract

Purpose : To compare agreement rates of DR PPL grading using UWF-CI and UWF-FA via qualitative and quantitative methods.

Methods : UWF-CI and UWF-FA images were acquired at the same visit from eyes without panretinal photocoagulation and graded for DR severity level at a centralized reading center. UWF-CI were graded qualitatively by 2 trained graders for presence or absence of PPL (qualitative PPL). UWF-FA images were graded for qualitative PPL by 2 separate masked graders. Hemorrhages and microaneurysm (H/Ma) manual counts inside and outside the ETDRS fields were obtained from UWF-FA images. Quantitative PPL-H/MA were defined as greater H/MA counts outside than inside each ETDRS field.

Results : Images from 280 eyes of 188 patients were evaluated. Distribution of DR severity based on UWF grading was 85 eyes (30.4%) mild nonproliferative DR (NPDR), 94 (33.6%) moderate, 66 (23.6%) severe and 35 (12.5%) proliferative DR (PDR). Agreement for qualitative PPL for the entire cohort was moderate for UWF-CI (exact agreement, 77.9%, κ= 0.421) but was substantially greater for UWF-FA (87.9%, 0.702). Within DR severity levels, agreement was similar in mild NPDR (κ=0.486 UWF-FA vs κ=0.439 UWF-CI) but substantially better using UWF-FA in all other DR severity levels; moderate NPDR (κ=0.743 vs κ=0.538), severe NPDR (κ=0.553 vs κ=0.339) and PDR (κ=0.639 vs κ=0.212).
Average agreement between qualitative PPL grading for UWF-CI and UWF-FA was poor (70.2%, κ=0.228). Using a difference of 1 H/MA or more in the UWF fields, quantitative PPL-H/Ma had a moderate agreement with both qualitative PPL graders (72.8%, 0.447 and 73.8%, 0.468). Increasing the threshold difference to determine PPL-H/Ma from 2-8 did not improve the agreement rates with qualitative grading (70.3% - 72.8%, 0.330 – 0.447; 73.6% - 78.2%, 0.468 – 0.516).

Conclusions : Agreement between qualitative PPL grading on UWF-FA and UWF-CI is poor, suggesting that they are not readily interchangeable. Determining PPL qualitatively on UWF-FA is more reproducible than UWF-CI in eyes with moderate NPDR or worse. Given these differences, future studies will need to evaluate whether PPL graded on FA or CI is more associated with DR progression. If FA is deemed more prognostic, then the role of FA in assessing DR management may become more important.
Original languageEnglish
JournalInvestigative Opthalmology and Visual Science
Volume62
Issue number8
Publication statusPublished - 01 Jun 2021
Externally publishedYes
EventAssociation for Research in Vision and Ophthalmology (ARVO) 2021 -
Duration: 01 May 202107 May 2021

Fingerprint

Dive into the research topics of 'Agreement rates for predominantly peripheral lesion(PPL) grading assessed using qualitative and quantitative methods on ultrawide field color images(UWF-CI) and fluorescein angiography(UWF-FA) in diabetic retinopathy(DR)'. Together they form a unique fingerprint.

Cite this