Automated Retinal Layer Segmentation Using Spectral Domain Optical Coherence Tomography: Evaluation of Inter-Session...

Area of Science:

• Ophthalmology
• Biomedical Imaging
• Medical Devices

Background:

• Optical coherence tomography (OCT) systems have varying software and scaling, impacting retinal thickness measurements.
• Automated segmentation of intra-retinal layers is crucial for OCT analysis but lacks standardization across devices.

Purpose of the Study:

• To evaluate the device-independent Iowa Reference Algorithms for automated intra-retinal layer segmentation and image scaling.
• To assess the repeatability and agreement of the Iowa Reference Algorithms across three different OCT systems.

Main Methods:

• Healthy participants (n=25) underwent macular scans on Zeiss, Topcon, and a long-wavelength OCT.
• Iowa Reference Algorithms were used for segmenting 10 intra-retinal layers and scaling using axial eye length (AEL).
• Inter-session repeatability and agreement were compared between algorithms, on-board software, and scaling methods.

Main Results:

• Iowa Reference Algorithms showed higher inter-session repeatability (CoR: 3.43-5.98μm) than on-board software (CoR: 4.63-7.63μm).
• The algorithms reliably segmented all 10 intra-retinal layers.
• AEL-dependent scaling minimized discrepancies between OCT systems, unlike fixed-AEL scaling which introduced bias and AEL correlations.

Conclusions:

• The Iowa Reference Algorithms are viable for clinical and research use in healthy eyes across multiple OCT devices.
• Accurate quantification of OCT images necessitates ocular biometry for precise scaling.