Algorithm for predicting treatment outcomes in vascular pigment epithelial detachment in neovascular age-related macular degeneration

• Published in: Vestnik oftal'mologii Vol. 141; no. 4; p. 12
• Main Authors: Kozina, E V, Samoylov, A N, Aksenov, K D, Aksenova, L E
• Format: Journal Article
• Language: English; Russian
• Published: Russia (Federation) 2025
• Subjects: Aged; Algorithms; Angiogenesis Inhibitors - administration & dosage; Female; Fluorescein Angiography - methods; Humans; Intravitreal Injections; Male; Retinal Detachment - diagnosis; Retinal Detachment - drug therapy; Retinal Detachment - etiology; Retinal Pigment Epithelium - diagnostic imaging; Retinal Pigment Epithelium - pathology; Tomography, Optical Coherence - methods; Treatment Outcome; Wet Macular Degeneration - complications; Wet Macular Degeneration - diagnosis; Wet Macular Degeneration - drug therapy; neovascular age-related macular degeneration; vascular pigment epithelial detachment; OCT biomarkers; automated analysis; prediction algorithm; pigment epithelial tear
• ISSN: 0042-465X
• DOI: 10.17116/oftalma202514104112

Automated analysis of optical coherence tomography (OCT) biomarkers improves the prediction of results of loading anti-VEGF therapy of vascular pigment epithelial detachment (PED) associated with neovascular age-related macular degeneration (nAMD). This study evaluated the effectiveness of OCT biomarker analysis algorithm in predicting the anatomical outcomes of loading anti-VEGF therapy for vascular PED in nAMD. OCT scans performed prior to loading anti-VEGF therapy were analyzed using the algorithm in 69 treatment-naïve nAMD patients (70 eyes) with vascular PED exceeding 200 µm in height. Qualitative biomarkers included subretinal fluid (SRF), intraretinal fluid (IRF), hyperreflective material beneath the PED, PED defects, and hyperreflective foci in the outer retinal layers. Quantitative parameters - height, width, and area of PED - were manually measured and segmented using a U-NET-based neural network. The algorithm predicted flattening of higher (499.2±198.1 μm) and wider (3205.6±734.2 μm) PEDs containing large hyperreflective foci in the mid-retinal layers and defects in PED ( =35 eyes). Resistance to loading therapy was predicted in lower (430.1±126.4 μm) and narrower (2824.1±732.8 μm) PEDs with sub-PED hyperreflective material without IRF ( =31 eyes). Risk of retinal pigment epithelium (RPE) tear was predicted for PEDs higher than 600 μm (mean 587.4±193.6 μm) in the presence of SRF but without PED defects ( =4 eyes). The automated OCT biomarker analysis algorithm effectively predicted resistance of PEDs with the lowest height, width, and area. Greater PED area and width were predictive of flattening, while PED height over 600 μm, calculated using the algorithm, was a predictor of RPE tear. The developed OCT biomarker analysis algorithm enables effective prediction of anatomical outcomes following loading anti-VEGF therapy in vascular PED in nAMD.