Purpose:
To determine the reproducibility of retinal nerve fiber layer (RNFL) segmentation software of various spectral-domain optical coherence tomography (SD-OCT) systems, we determined the ‘RNFL’ thickness in a standard phantom eye model.
Methods:
We tested 3x4 SD-OCT systems (3 RTVue OptoVue systems, 3 Cirrus Zeiss systems, 3 3DOCT-1000 Topcon systems, 3 Spectralis Heidelberg systems). To exclude biological variations within human eyes, each apparatus imaged a layered phantom eye model 5 times. An average RNFL thickness was determined by the software of the apparatus using a circular scan (Ø=3.4mm) around a fixation target. We used ANOVA test to compare means for intra variability (deviation within 1 manufacturer) and inter variability (comparison of 4 manufacturers). As a model we used a developed and constructed phantom eye model with a lens, water as vitreous and silicon elastomer based layers as retina. This artificial retina consists of five layers (10x10mm2) of 50 µm thickness with different scattering properties. The top layer was used to mimic the RNFL.
Results:
Average thickness and standard deviation calculated by OptoVue (system 1: 65.1±1.1µm system 2: 66.1±1.8µm system 3: 62.8±1.6µm), Topcon (system 1: 63.9±0.5µm system 2: 66.8±1.0µm system 3: 66.5±3.6µm), Cirrus (system 1: 67.7±0.4µm system 2: 66.1±1.8µm system 3: 62.8±1.6µm), Spectralis (system 1: 69.0±0.6µm system 2: 67.9±0.3µm system 3: 68.3±2.0µm). Results are also displayed in Table 1 (see figure).
Conclusions:
Intra variability: RNFL thickness measurements are significantly different for 1 Topcon and 1 OptoVue and 1 Cirrus. Inter variability: RNFL thickness determined with OptoVue and Topcon differ significantly with Cirrus and Spectralis. Furthermore, no OCT system calculates the real thickness of our simulated RNFL
Keywords: imaging/image analysis: clinical • optic nerve • retina