We aimed to compare fluorescein angiography (FA)-based classification of macular neovascularisation (MNV) with optical coherence tomography (OCT)-based classification, as well as examine differences in retinal fluid among OCT MNV types. We analyzed baseline FA and OCT images from 704 eyes of neovascular AMD patients across two multicenter trials, with grading conducted at the Vienna Reading Center. Using a validated AI tool (RetInSight Fluid Monitor Version 2), we localized and quantified retinal fluid. Kappa coefficients for agreement between FA and OCT MNV types were 0.58 [0.52; 0.64] (type 1/occult), 0.46 [0.38; 0.55] (type 2/classic), and 0.53 [0.44; 0.62] (type 3/RAP). Significant differences in the volumes of intraretinal fluid (IRF), subretinal fluid (SRF), and pigment epithelial detachment (PED) were noted among MNV types (p < 0.0001). Pairwise comparisons revealed significant differences in IRF volumes across all lesion types except type 2 versus mixed type, in SRF between type 3 and other types, and in PED between type 2 and other lesions. In conclusion, there was moderate agreement between FA and OCT classifications, and notable differences in fluid distribution among OCT types, suggesting potential for AI-guided MNV recognition in clinical settings.
Keywords: AMD; Artificial intelligence; CNV lesion type; FA; MNV lesion type; OCT.
© 2025. The Author(s).