Prevalence and Associated Factors of Segmentation Errors in the Peripapillary Retinal Nerve Fiber Layer and Macular Ganglion Cell Complex in Spectral-domain Optical Coherence Tomography Images.
Atsuya Miki, Miho Kumoi, Shinichi Usui, Takao Endo, Rumi Kawashima, Takeshi Morimoto, Kenji Matsushita, Takashi Fujikado, Kohji Nishida
Summary
Although segmentation failure was common in both RNFL and GCC scans, it was less frequently observed in GCC scans.
Abstract
PURPOSE
To determine the prevalence of errors in segmentation of the peripapillary retinal nerve fiber layer (RNFL) and macular ganglion cell complex (GCC) boundary in spectral-domain optical coherence tomography (SDOCT) images, and to identify factors associated with the errors.
MATERIALS AND METHODS
Peripapillary RNFL circle scans and macular 3-dimensional scans of consecutive cases imaged with SDOCT (RS-3000 Advance; Nidek, Gamagori, Japan) were retrospectively reviewed by a glaucoma specialist. Images with signal strength index (SSI)<6 were excluded. Threshold for segmentation failure was determined as 15 degrees in the RNFL scans and 1/24 of the scanned area in the GCC scans. Relationships between segmentation failure and clinical factors were statistically evaluated with univariable and multivariable analyses.
RESULTS
This retrospective cross-sectional study included 207 eyes of 117 subjects (mean age, 58.5±16.5 y). Segmentation failure was found in 20.7% of the peripapillary RNFL scans, 16.6% of the 9 mm GCC scans, and 6.9% of the 6 mm GCC scans in SDOCT images. In multivariable logistic regression analyses, low SSI, large disc area, and disease type significantly correlated with RNFL segmentation failure, whereas SSI was the only baseline factor that was significantly associated with GCC segmentation failure.
CONCLUSIONS
Although segmentation failure was common in both RNFL and GCC scans, it was less frequently observed in GCC scans. SSI, disc area, and disease type were significantly associated with segmentation failure. Predictive performance of baseline factors for failure was poor, underlining the importance of reviewing raw OCT images before using OCT parameters.
More by Atsuya Miki
View full profile →Using Deep Learning and Transfer Learning to Accurately Diagnose Early-Onset Glaucoma From Macular Optical Coherence Tomography Images.
Investigating the usefulness of a cluster-based trend analysis to detect visual field progression in patients with open-angle glaucoma.
Predicting the Glaucomatous Central 10-Degree Visual Field From Optical Coherence Tomography Using Deep Learning and Tensor Regression.
Top Research in Optic Nerve & Disc
Browse all →Efficacy of a Deep Learning System for Detecting Glaucomatous Optic Neuropathy Based on Color Fundus Photographs.
Relationship between Optical Coherence Tomography Angiography Vessel Density and Severity of Visual Field Loss in Glaucoma.
Inflammation in Glaucoma: From the back to the front of the eye, and beyond.
Discussion
Comments and discussion will appear here in a future update.