Automatic segmentation using artificial intelligence of baseline anatomical parameters of patients starting anti-VEGF injections for neovascular age-related macular degeneration

Abstract

Purpose : The baseline morphology of patients due to commence anti-VEGF therapy for neovascular age related macular degeneration (nAMD) has been shown to correlate with patients' anatomical and visual outcomes. We present a descriptive overview of segmented data generated by an artificial intelligence (AI)-derived system previously published by De Fauw et al. (2018) that provides volumes and thickness measurements for anatomical parameters from baseline OCT scans.

Methods : Data was extracted from the Moorfields Eye Hospital AMD database for all patients that started treatment for AMD in one eye between June 2012 and June 2017. This data included segmentation data from baseline OCT scans (Topcon 3D OCT-2000) of patients due to start anti-VEGF treatment. Only first treated eyes were included in this analysis. The mean volumes and thicknesses at baseline were converted from the automatically calculated segmentation on a voxel-count basis by an AI-derived segmentation algorithm. The volumetric parameters analysed for this cohort include intra-retinal fluid (IRF), subretinal fluid (SRF), pigment epithelium detachment (PED), and intraretinal hyperreflective foci (IHF). The thickness parameters analysed were neurosensory retina (NR) and retinal pigment epithelium (RPE).

Results : Baseline scans from a total of 1996 treatment naive eyes were analysed. Before commencing treatment, 91.6 % of the eyes presented IRF, 98.4% SRF, 59.9% had serous PED and 99.3% fibrovascular PED. Automatic volumetric calculations for retinal fluid show a mean IRF volume of 0.119mm3 and a mean SRF volume of 0.463mm3 at baseline for first treated eyes. Mean PED volume at baseline was 0.759mm3. Hyperreflective foci was also segmented and the mean volume was 0.003mm3. The mean neurosensory retinal thickness at baseline was 262um while the mean RPE thickness was 23.9um.

Conclusions : The importance of a detailed evaluation of baseline OCT morphological parameters of patients due to start VEGF injections may have a profound impact on the future management protocols for AMD. Better understanding of the structural anatomy of retinal tissues provides valuable information that may suggest novel personalized treatment protocols and decrease nAMD disease burden.

This is a 2020 ARVO Annual Meeting abstract.