Purchase this article with an account.
W. Wu, O. Tan, D. Huang; Assessment of Frames Averaging Algorithms for Optical Coherence Tomography. Invest. Ophthalmol. Vis. Sci. 2010;51(13):1780.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Averaging several frames of optical coherence tomography (OCT) images can improve image quality by suppressing background and speckle noises, but can also introduce artifact if the registration of frames does not adequately remove motion error. The purpose of this study is to evaluate two commercially available frame-averaging algorithms, by post processing, in two RTVue software versions (Version 220.127.116.11 and Version 18.104.22.168).
A Fourier domain OCT system (RTVue, Optovue, Inc. Fremont, CA), was used to scan 40 human subjects (20 normal and 20 glaucoma) with line scans (centered at fovea, scan length = 6mm). The RTVue postprocessing software registers and averages 16 consecutive frames for each line scan. Two versions of RTVue software (22.214.171.124 and 126.96.36.199) were evaluated. Three parameters were used to assess image quality. The noise variance evaluates noise suppression (smaller value better). The edge width evaluates edge blurring due to misregistration (smaller value better). The contrast to noise ration (CNR) evaluates contrast enhancement (larger value better). The Wilcoxon rank sum test was used to evaluate the significance of differences in image quality.
The noise variance, edge width, and CNR of frame-averaged images (both software versions) were significantly better than single frames (p < 0.001, Table 1). Frame averaging by software version 188.8.131.52 was significantly better than version 184.108.40.206 in terms of edge width (p = 0.04), but not the other 2 parameters (Table 1). The edge width parameter correlated with subjective assessment of image quality.
Both version of frame-averaging software dramatically suppressed noise and increased contrast. Software version 220.127.116.11 preserves edge sharpness better than version 18.104.22.168, indicating better registration of image frames. Registration of image frames is critical to image quality and edge width is a good parameter for its evaluation.
This PDF is available to Subscribers Only