Automated 3D MRI segmentation of eye and orbit using nnU-Net constructs first large-scale atlases

In 1245 adults, nnU-Net automated 3D segmentation of eye and orbit on T1 MRI was accurate, yielding biomarkers like axial length and volume. First large-scale unbiased atlases (female, male) created for MR-eye standardization.

• Retrospective analysis of 1245 healthy adults; segmentation accuracy reported qualitatively (exact Dice coefficients not provided in source).
• Automated extraction of morphometric biomarkers (axial length, volumetry) and correlation with BMI (exact figures not reported).
• Limitations: no external validation, restricted to healthy population, generalizability to disease not evaluated.

RadPigeon summaries are original and for information only. They are not clinical advice.