Comparison of foveal microstructure imaging with different spectral domain optical coherence tomography machines

Abstract

Purpose: To investigate the reproducibility of imaging the foveal microstructures of healthy eyes with 3 spectral domain optical coherence tomography (SD-OCT) machines: Cirrus (Carl Zeiss Meditec Inc.), Spectralis (Heidelberg Engineering), and Topcon (Topcon 3D OCT-1000 Mark II).

Design: Cross-sectional, prospective, noninterventional study.

Participants: Images were obtained for 50 eyes of 50 healthy undilated volunteers without ocular pathology in a clinical setting.

Methods: The fovea of all subjects was imaged using Cirrus, Spectralis, and Topcon.

Main outcome measures: Among the 4 hyperreflective bands in the outer subfovea on SD-OCT imaging, the innermost band (external limiting membrane [ELM] band), the second innermost band (second band), and the third innermost band (third band) were classified as "continuous," "disrupted," or "none" by 2 independent raters. Weighted κ-coefficient analysis and/or Fisher exact test were used to compare interrater, intermachine, and intramachine agreement measurements. The sensitivity of each machine was also evaluated.

Results: The group of 50 subjects consisted of 22 men and 28 women, with an average age of 31.4 years (range, 21-52 years). Interrater agreement for 3 bands was high (κ = 0.876, 0.738, and 0.774) with Cirrus, Spectralis, and Topcon, respectively. The sensitivity of each machine was high for the ELM band (0.92, 0.98, and 0.96), the second band (all 1.00), and the third band (0.96, 0.94, and 0.88) with Cirrus, Spectralis, and Topcon, respectively. The sensitivity of the third band was significantly lower than the second band with Topcon (Fisher exact test, P = 0.027), but the difference was not significant with the other machines. Intermachine agreement was fair to moderate for the third band (κ = 0.65, 0.512, and 0.464) and for all bands (κ = 0.531, 0.369, and 0.362) between Cirrus-Spectralis, Spectralis-Topcon, and Topcon-Cirrus, respectively; however, it was not significant for ELM band (κ = -0.027) between Spectralis-Topcon.

Conclusions: In healthy adults with normal vision, there was almost perfect reproducibility between raters for foveal microstructural images acquired with the Cirrus, Spectralis, and Topcon devices. The machines have good sensitivity to image foveal microstructures, and the sensitivity does not differ significantly among machines; however, they are not necessarily identical or interchangeable for imaging certain structures.