Deep Learning-Based Automated Segmentation and Quantification of the Ellipsoid Zone and the RPE-Bruch's Membrane...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate segmentation and quantification of outer retinal layers are crucial for diagnosing and monitoring retinal diseases like geographic atrophy (GA).
• Manual analysis of spectral-domain optical coherence tomography (SD-OCT) is time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To validate a deep learning algorithm (NMI-Outer Retina Analyzer) for automated segmentation and quantitative assessment of the ellipsoid zone (EZ) and the retinal pigment epithelium (RPE)-Bruch's membrane (BM) complex.
• To compare the algorithm's performance against manual segmentation by expert graders in both healthy and GA eyes.

Main Methods:

• Retrospective analysis of SD-OCT volume scans from 30 healthy eyes and 30 GA eyes.
• Automated segmentation of EZ, RPE, and BM using the NMI-Outer Retina Analyzer.
• Calculation of average thicknesses (EZ-RPE, EZ-BM, RPE-BM) across ETDRS sectors and comparison with manually corrected segmentations using Dice coefficients, Pearson correlation, and absolute thickness differences.

Main Results:

• The algorithm achieved high segmentation accuracy (mean Dice coefficient 0.995-0.998) in both healthy and GA eyes.
• Minimal differences were observed between automated and manual measurements in healthy eyes (1.88-3.05%).
• Strong correlations (r=0.89-0.97) were found between automated and manual measurements in GA eyes, despite greater variability.

Conclusions:

• The NMI-Outer Retina Analyzer provides accurate and automated segmentation and quantification of outer retinal layers.
• The algorithm's performance is comparable to that of expert graders, offering a reliable tool for retinal analysis.