Deep-Learning-Assisted Analysis of Early Biomarker Changes in Treatment-Naïve Patients with Neovascular AMD Under...

Artificial intelligence (AI)-driven biomarker segmentation offers an objective approach to assessing neovascular age-related macular degeneration (nAMD). In addition, faricimab, a bispecific VEGF and Ang-2 inhibitor, presents new potential in disease management. This study applies an AI-based segmentation algorithm to quantify key optical coherence tomography (OCT) biomarkers and assess the short-term efficacy of intravitreal faricimab in treatment-naïve patients.

This retrospective analysis includes 40 eyes from 38 treatment-naïve patients with nAMD treated with faricimab at LMU University Hospital Munich between January 2023 and September 2024. Patients received 4-monthly intravitreal injections. Biomarkers of disease activity, including central retinal thickness (CRT), intraretinal fluid (IRF), subretinal fluid (SRF), subretinal hyperreflective material (SHRM) and fibrovascular pigment epithelium detachment (fvPED), were quantified using a deep learning-based semantic segmentation algorithm. Best-corrected visual acuity (BCVA) and OCT imaging data were analyzed at baseline (mo0) and after 1 (mo1), 2 (mo2) and 3 months (mo3).

AI-driven analysis revealed significant reductions in key biomarkers. CRT decreased from 433.6 (IQR: 306.6) µm at mo0 to 241.5 (IQR: 130.8) µm at mo3 (p < 0.0001). IRF and SRF volumes were reduced by > 99% from mo0 to mo3 (both p < 0.0001). BCVA improved from 0.60 (IQR: 0.30) logMAR at mo0 to 0.40 (IQR: 0.33) logMAR at mo3 (p < 0.0001). Correlation analysis identified IRF and SHRM reductions as the strongest predictors of visual improvement.

This study demonstrates the potential of AI-assisted biomarker analysis for precise disease monitoring in nAMD. Faricimab significantly reduced disease activity biomarkers and improved visual acuity in treatment-naïve patients, reinforcing its efficacy in early disease control. Future studies should explore long-term outcomes and further integrate AI-driven biomarker evaluation in clinical practice.