Abstract
Purpose :
Accurate retinal fluid monitoring is essential to offer precise personalized treatment in neovascular age related macular disease (nAMD). The purpose of the study is to correlate manually measured retinal fluid heights with automatically quantified retinal fluid volumes and distributions in nAMD using an approved AI-based algorithm.
Methods :
This is a retrospective study where we included nAMD patients at baseline visit (n=130) from multicenter study data of the Vienna Reading Center (VRC). Manually measured parameters acquired from expert readers of the VRC were correlated (Pearson coefficient) with corresponding fluid volumes/areas from the automatic quantification of the RetInSight Fluid Monitor Version 2 (aFLM) in the central mm (CMM) of the ETDRS grid: (1) height of subretinal fluid (SRF) at the foveal centerpoint (CP); (2) height of pigment epithelium detachment (PED) at CP; (3) height of intraretinal cystoid fluid (IRC) at CP.
Results :
Main outcome variables are shown in Table 1. Out of 46 patients with IRC, 79 with SRF, and 100 with PED in the CMM, only 12 (IRC), 35 (SRF), and 68 (PED) cases had measurable fluid at the CP. Correlation of the manual gradings at the CP with the aFLM data of the CMM revealed an excellent correlation between manually measured SRF height with aFLM-based SRF volumes (r=0.895, P<0.01), and measured PED height with aFLM PED volume (r=0.889, P<0.01). Manually measured IRC showed a significantly lower correlation with aFLM IRC volume values (r=0.683, P<0.05). Accordingly, the aFLM-based SRF area correlation was consistent with the manual SRF area measurements (r=0.573, P<0.01), but was low with IRC and PED measurements (IRC r=0.359, P>0.05; PED r=0.224, P>0.05). Correlation between FLM volume and FLM area was r=0.834 (p<0.01), r=0.714 (p<0.01), r=0.389 (p<0.01), for IRC, SRF, and PED, respectively.
Conclusions :
aFLM analysis adds another dimension to conventional manual retinal fluid height measurements, providing comprehensive quantitative measurements of IRF, SRF, and PED volume and area and offers advantages especially in clinical routine, where large amounts of data must be evaluated, and workflow is high, and where manual measurements are neither practical nor standardized. AI-based tools provide a significant contribution to the interpretation of nAMD activity as a base for precise personalized therapies in clinical practice.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.