Purchase this article with an account.
C. A. Reisman, Q. Yang, Z. Wang, A. Tomidokoro, M. Araie, M. Hangai, N. Yoshimura, Y. Fukuma, K. Chan; Enhanced Visualization and Layer Detection via Averaging Optical Coherence Tomography Images. Invest. Ophthalmol. Vis. Sci. 2010;51(13):3859.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To enhance visualization and image quality of OCT images by speckle suppression and signal to noise ratio improvement via image averaging, and to improve layer detection reproducibility.
B-scans were consecutively acquired using Topcon OCT-1000 or OCT-2000 systems while synchronously agitating the galvanometer mirror in the slow axis dimension. Scan capture times were 2.5 seconds maximum. Image frames were registered using a proprietary technique consisting of a subpixel, multi-resolution algorithm. A conditional frame matching filter was applied in the averaging process to minimize blur.Reproducibility of macular segmentation using 3D scan data of 43 eyes with three repetitions each was evaluated with and without averaging. The registration algorithm was applied as a moving average over sets of three frames. The segmentation algorithm was a proprietary algorithm capable of detecting nine intra-retinal boundaries. Thickness maps were calculated from the ILM to the eight remaining boundaries. Standard deviation and correlation statistics were calculated on the thickness data to evaluate repeatability.
The attached figure shows a typical averaged optic disc image, where 38 of the 50 collected B-scan frames were utilized. Layer detection reproducibility improved on the whole. The largest percentage improvements in standard deviation were associated with the ELM and the NFL/GCL and GCL/IPL boundaries. The RPE was the only boundary to see a statistically significant decrease in reproducibility after averaging. Correlation statistics demonstrated the same trends as standard deviation analysis.
The presented registration and averaging technique produces sharp, detail-rich images in which the appearance of speckle noise has been largely removed. The algorithm was shown to have a synergistic effect towards improving layer detection reproducibility.
This PDF is available to Subscribers Only