Purchase this article with an account.
Koji Nozato, Qiang Yang, Kenichi Saito, Kei Suzuki, Jie Zhang, Lisa R. Latchney, David R Williams, Ethan A Rossi; Compact adaptive optics scanning light ophthalmoscope (AOSLO) with real-time optical stabilization and digital registration. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):5977.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Eye motion is a major impediment to clinical deployment of small field of view imaging systems such as the AOSLO. To obtain high signal-to-noise ratio (SNR) images, multiple images must be registered and averaged. Reference frame based image registration is usually employed for this purpose, but this fails when the eye moves outside the field of view of the reference frame. In diseased eyes with poor vision, eye motion is increased and this problem is exacerbated, greatly reducing the efficiency of imaging. We recently showed that eye motion could be efficiently optically stabilized in a large research instrument using an expensive tip/tilt mirror (Yang et al., 2014). Here we show that our approach can be implemented in a much more compact instrument using low-cost, higher speed, one-dimensional galvanometric scanners, permitting real-time optical stabilization and digital registration.
We modified a compact AOSLO (21”×29”) prototype by adding an additional galvanometric scanner to provide optical stabilization in the fast scanning direction. The traditional ‘slow’ scanner provided optical stabilization in the orthogonal direction. A real-time image based registration algorithm calculated eye motion at ~500Hz. Mirror angle was updated using closed-loop control at the same rate. Residual eye motion was eliminated with real-time digital registration. Performance was evaluated by calculating the RMS residual image motion for each 5 sec-long movie using offline software that calculated the displacement between the reference frame and each subsequent frame with a cross-correlation algorithm.
We verified our method by imaging a model eye, 5 normal eyes and 2 diseased eyes. Model eye testing shows that this system can track 1.6x faster movement than the previous system. Before optical stabilization, eye motion was 12-22 µm in normal eyes and 14-31 µm in diseased eyes. After optical stabilization, it decreased to 0.9-2.8µm in normal and 1.3-3.1µm in diseased eyes. Digital registration compensated for residual motion with RMS error of less than 1 pixel resolution.
Optical stabilization was implemented in a compact AOSLO prototype by adding just one additional galvanometric scanning mirror. Sub-pixel accuracy can be achieved after real-time digital registration, permitting image averaging during acquisition and eliminating post-processing time.
This PDF is available to Subscribers Only