Objective determination of optimal number of spectral-domain optical coherence tomographic images of retina to average.

Abstract

PURPOSE: To determine by objective methods the minimum number of spectral-domain optical coherence tomographic (SD-OCT) images to average to obtain the clearest retinal image. METHODS: SD-OCT Images were obtained from 9 healthy eyes and also from a phantom eye model. The SD-OCT images were obtained by averaging 1, 5, 20, 60, and 100 B-scan images. The reflectivity (mean gray value) of the different retinal layers was evaluated in these images. The image quality was evaluated by the size of the standard deviations (SDs) and the contrast-to-noise ratios (CNRs). A phantom eye model made by TiO2 silicone plates was also examined. RESULTS: The SDs decreased significantly when the number of images averaged increased from 1 to 5 and also from 5 to 20 (P<0.05, post hoc Tukey's honestly significant difference tests). The SD of the automatic real time averaging of 1 (ART = 1) and ART = 5 were significantly larger than the SD of ART = 100 (P<0.05). The SDs of all other averaged numbers were not significantly larger than that of ART = 100. The CNR increased with an increase in the number of images averaged, and there was a significant increase between ART = 1 to 5 and between ART = 5 to 20 (P<0.05). No significant differences in the CNR was observed between ART = 5, ART = 20 and ART = 60. Similar results were obtained with the phantom eye model. CONCLUSIONS: Although the image quality of the SD-OCT images of the retina improved with an increase in the number of images averaged, it does not improve significantly by averaging more than 20 images.