Clinical Validation for Automated Geographic Atrophy Monitoring on OCT Under Complement Inhibitory Treatment

AI Generated Summary*

A deep learning system that maps retinal pigment epithelium loss from spectral-domain OCT volumes onto an en-face area was assessed for quantifying geographic atrophy in age-related macular degeneration. Fivefold cross-validation used 967 volumes from 100 routine-care patients; external testing used baseline and month 12 scans from 113 eyes in the FILLY phase 2 pegcetacoplan trial. Dice scores against manual annotation averaged 0.86 internally and 0.91 externally, within the variability of two experts on a 12-volume subset; mean Dice for the 12-month growth area was 0.46. Automated and manual square-root growth rates correlated at r = 0.81 without significant difference; automated rates were slower with monthly dosing than sham (0.20 versus 0.28 mm, p = 0.030), matching the trial's autofluorescence-based result.

*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.