Automated Segmentation Errors When Using Optical Coherence Tomography to Measure Retinal Nerve Fiber Layer Thickness...

Area of Science:

• Ophthalmology
• Medical Imaging
• Glaucoma Research

Background:

• Accurate measurement of retinal nerve fiber layer (RNFL) thickness is critical for diagnosing and monitoring glaucoma.
• Automated segmentation algorithms in optical coherence tomography (OCT) aim to streamline RNFL thickness analysis.
• Potential errors in automated segmentation may impact clinical decision-making.

Purpose of the Study:

• To quantify the error introduced by automated RNFL thickness segmentation without manual refinement in OCT scans.
• To assess the impact of automated segmentation errors on glaucoma classification.
• To identify factors influencing the accuracy of automated RNFL segmentation.

Main Methods:

• Cross-sectional study involving 3490 spectral domain OCT scans from 213 individuals with glaucoma or glaucoma suspect.
• RNFL thickness was measured using automated segmentation and compared with results after manual refinement.
• Differences in RNFL thickness and glaucoma classification (normal, borderline, outside normal limits) were analyzed.

Main Results:

• Automated segmentation alone resulted in significantly thinner global RNFL thickness (1.6 μm thinner, P < .001) compared to manual refinement.
• Errors increased with thinner RNFL thickness, lower scan quality, and older age.
• Manual refinement altered the glaucoma classification of 8.5% of scans, with 23.7% of borderline cases reclassified as normal.

Conclusions:

• Automated segmentation without manual refinement leads to underestimation of RNFL thickness and overestimation of glaucoma.
• The accuracy of automated segmentation is compromised in eyes with thinner RNFL, poorer scan quality, and in older individuals.
• Manual inspection and refinement of OCT RNFL segmentation are essential for reliable glaucoma management.