Abstract
Purpose :
This study employs Artificial Intelligence (AI) to elucidate key ocular biomarkers in neovascular age-related macular degeneration (nAMD) using Optical Coherence Tomography (OCT). The aim is to integrate AI for a comprehensive demographic and tomographic profiling in nAMD, focusing on the prevalence and reliability of biomarkers in the central 1 mm, 6 mm, and entire OCT scan areas
Methods :
A retrospective analysis of 78 nAMD patients; OCT scans (Heidelberg, Spectralis) was conducted using AI (Discovery 3.7 clinics, RetinAI) for quantification. Data included age, sex, and tomographic variables like Intraretinal Fluid (IRF), Subretinal Fluid (SRF), and Pigment Epithelial Detachment (PED), measured in nanoliters. Biomarkers included SRF, IRF, Fibroplasia PED (FPED), Hyperreflective foci (HF), Drusen, Reticular pseudodrusen (RPD), Epiretinal membrane (ERM), geographic atrophy (GA) and Outer retinal atrophy (ORA). We considered a biomarker to be present if its probability of being found by AI was equal to or greater than 90%. Statistical analyses comprised descriptive statistics and confidence interval calculations for biomarker prevalences
Results :
The cohort had a mean age of 80 years, predominantly female (65%). Notable biomarkers with high prevalence and narrow confidence intervals included FPED (100%, 95% CI [1,0-1,0] prevalence in all OCT areas), Drusen (91.03% to 98.70% across areas), and Hyperreflective Foci (HF, 89.74% to 92.31%). SRF and IRF were also significant, with prevalences of 94.87% 95% IC [0,89 – 0,99] and 76.92%, 95% IC [0,67 – 0,86] in the total OCT area, respectively. These findings are supported by narrow 95% confidence intervals, underscoring the consistency of these biomarkers in nAMD
Conclusions :
AI-assisted analysis revealed that FPED, Drusen, HF, SRF, and IRF are the most consistent and prevalent biomarkers in nAMD. The high prevalence rates coupled with narrow confidence intervals indicate these biomarkers; robust presence in the clinical presentation of nAMD. This study highlights the potential of AI in enhancing diagnostic accuracy and understanding the heterogeneity of nAMD
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.