Reproducibility of Spatially-Defined Structural Cluster Analysis Using Automated Superpixel Metrics on Swept-Source Optical Coherence Tomography

Authors

• Sahil Shah, Thomas Jefferson UniversityFollow
• Fabio Lavinsky, Thomas Jefferson UniversityFollow
• Bo Young Kim, Wills Eye Hospital
• Darron King, Thomas Jefferson UniversityFollow
• Ronald Zambrano, Thomas Jefferson UniversityFollow
• Fernanda Mari Fujihara
• Dilru Amarasekera, Thomas Jefferson UniversityFollow
• Lauren E. Hock, Thomas Jefferson UniversityFollow
• Ping Huang, Thomas Jefferson UniversityFollow
• Natasha Kolomeyer, Thomas Jefferson University
• Reza Razeghinejad, Thomas Jefferson UniversityFollow
• Jonathan Myers, Thomas Jefferson UniversityFollow
• Gadi Wollstein, Thomas Jefferson UniversityFollow
• Joel Schuman, Thomas Jefferson UniversityFollow

Document Type

Abstract

Publication Date

2-11-2026

Academic Year

2025-2026

Comments

Presented at the 2026 Scholarly Inquiry (SI) Research Project Symposium.

Abstract

INTRODUCTION: Optical Coherence Tomography (OCT) is essential to detect and monitor structural damage in glaucoma. However, current practices focus on global and sectoral values, not localized changes. The study explores a novel approach using spatially determined superpixel clusters from swept- source OCT (SS-OCT) to assess localized changes in both healthy and glaucomatous eyes. We hypothesize that this method will provide reproducible results comparable with standard OCT parameters.

METHODS: Healthy and glaucoma subjects with ≥ 2 qualified same-visit optic nerve head (ONH) and macular Triton SS-OCT scans were included. Automated segmentation generated superpixel thickness values from 3D disc and macula arrays. Based on Garway-Heath mapping, six clusters were created for the ONH retinal nerve fiber layer (RNFL) and five clusters for the macular ganglion-cell inner plexiform layer (GCL). Reproducibility was determined using within-subject coefficient of variation (WSCV) of the highest quality same-visit pair of RNFL and macular scans and compared to standard parameters.

RESULTS: 58 eyes (30 glaucomatous, 28 healthy) were included for ONH, and 88 eyes (47 glaucoma, 41 healthy) for the macula. RNFL clusters, average and sectoral circumpapillary, and ONH standard structural parameters showed good reproducibility (WSCV ≤ 5.87%). The reproducibility analysis for the macular clusters and standard macular parameters also showed excellent reproducibility (WSCV ≤ 1.82%).

CONCLUSION: The ONH and macular cluster analysis employing SS-OCT superpixel measurements demonstrated excellent reproducibility. The cluster values are comparable to the standard parameters in both healthy and glaucomatous eyes. These findings support the utility of clusters for future glaucoma detection and monitoring.

Recommended Citation

Shah, Sahil; Lavinsky, Fabio; Kim, Bo Young; King, Darron; Zambrano, Ronald; Fujihara, Fernanda Mari; Amarasekera, Dilru; Hock, Lauren E.; Huang, Ping; Kolomeyer, Natasha; Razeghinejad, Reza; Myers, Jonathan; Wollstein, Gadi; and Schuman, Joel, "Reproducibility of Spatially-Defined Structural Cluster Analysis Using Automated Superpixel Metrics on Swept-Source Optical Coherence Tomography" (2026). Phase 1. Paper 20.
https://jdc.jefferson.edu/si_ctr_2028_phase1/20

Language

English