Purchase this article with an account.
M. Kraus, M. A. Mayer, R. Bock, B. Potsaid, V. Manjunath, J. S. Duker, J. Hornegger, J. G. Fujimoto; Motion Artifact Correction in OCT Volume Scans Using Image Registration. Invest. Ophthalmol. Vis. Sci. 2010;51(13):4405.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Artifacts resulting from eye motion during 3D-OCT volume scanning degrade image quality and are a source of reproducibility error in quantitative measurements. We present a pure software based correction method that corrects motion in all three dimensions. It uses two consecutive 3D-OCT volume scans with orthogonal fast scan axis directions.
The two 3D-OCT data sets are registered in order to find the unknown eye motion. Registration is performed by optimizing a global objective function with two displacement fields and special regularization based on the time structure of the volume sampling process. To improve both speed and solution quality, a multi-resolution approach is used. After optimization, each volume is undistorted and a single merged volume is constructed as an adaptive weighted sum of both registration results.
Data sets of the macula and optic disc from normal subjects and patients with retinal diseases were registered. The algorithm is able to correct for motion in all three dimensions. It can handle both slow drifting as well as saccadic movement. Quantitative assessment of registration performance using mutual information and visual inspection indicate very good registration results. The resulting volumes do not show visible motion artifacts. The algorithm produces stable results over a wide range of parameters. Images show single corresponding en face slices, from left to right: Fast horizontal, fast vertical input, registered and merged result.
The proposed method constitutes a robust, software based, motion artifact correction solution able to remove distortion introduced by movement in all three dimensions.These methods promise to enable the acquisition of large 3D-OCT data sets and improve measurement reproducibility without the need for active eye tracking.
This PDF is available to Subscribers Only