Performance of Retinal Fluid Monitoring in OCT Imaging by Automated Deep Learning versus Human Expert Grading in Neovascular AMD

AI Generated Summary*

Retrospective analysis of over 41,000 SD-OCT scans from the HAWK and HARRIER neovascular AMD trials compared automated deep learning fluid quantification with reading centre grading and central subfield thickness (CSFT). Central-millimetre AUCs were 0.93 (intraretinal) and 0.87 (subretinal) in HARRIER, 0.85 and 0.87 in HAWK. Intraretinal volume correlated small-to-moderately with CSFT at baseline (rho 0.45 to 0.51) and more weakly under therapy (0.24 to 0.29); subretinal correlations stayed weak (0.09 to 0.36). Residual variability of 44 to 95 um was large beside CSFT ranges. In a 100-visit subset, automated CSFT tracked reading centre values closely (rho up to 0.94; 0.79 to 0.82 using Bruch's membrane). Thickness, the paper concludes, is a weak indicator of fluid activity, pending prospective confirmation.

*This summary was generated by AI and is published unedited. Oku does not alter these summaries. It may contain errors or omissions and is provided for general informational purposes only. Oku does not guarantee its accuracy, completeness, or reliability. For authoritative information, please refer to the original, peer-reviewed article.