Integrating Human Expertise with Artificial Intelligence (AI) Models for Optical Coherence Tomography (OCT) Retinal Fluid and Pathology Quantification: A Systematic Review

AI Generated Summary*

A systematic review examines collaborative human and machine workflows for quantifying retinal fluid and pathology on optical coherence tomography. Searches of Scopus, PubMed and Web of Science covering 2021 to 2025 yielded nine studies, with cohorts of 16 to 1,097 patients. U-Net variants and custom convolutional models produced initial segmentations or risk maps that clinicians reviewed and corrected. Model accuracy was highest for retinal layers (Dice 0.94), moderate for intraretinal and subretinal fluid (0.61 to 0.67) and poor for sub-RPE lesions (0.11). Reported outcomes include expert-level reliability for 11 of 13 biomarkers, agreement with manual grading near Pearson r 0.85, larger volumetric errors in atrophic areas, and processing time cut by over half, though cohorts were small and heterogeneous.

*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.