September 2014
Spotlight Summary by Robert J. Zawadzki
Closed-loop optical stabilization and digital image registration in adaptive optics scanning light ophthalmoscopy
The field of clinical in vivo retinal imaging has recently entered into a new and exciting stage, thanks to the recent advancements in adaptive optics based systems, including instrument design as well as data acquisition, processing and analysis. However, one of the limitations for its widespread use is the presence of involuntary eye movements during imaging. Increased lateral resolution offered by the implementation of AO (Adaptive Optics) magnifies the effect of eye motion making the correction of motion artifacts essential to achieving successful retinal imaging of cellular structures. The recently published Biomedical Optics Express article by Qiang Yang et al. describes the latest advancement in AOSLO (Adaptive Optics Scanning Light Ophthalmoscope) retinal imaging stabilization and image registration that culminates almost ten years of research and collaboration between key players in this field.
Efficient eye motion tracking and correction is essential not only for successful imaging but also for precise delivery of visual stimuli. This is why it has been an active area of research even before the first demonstration of retinal AOSLO. With application of AOSLO the common strategies implemented to achieve retinal image stabilization include active correction of retinal movements using a retinal tracking system or real time registration of acquired images allowing inter-frame correction of eye-motion artifacts. The authors provide a detailed historical overview of retinal motion tracking and image stabilization methods that were previously developed by groups working in this field.
The framework presented in this manuscript describes a novel real-time solution to image stabilization by implementing both optical stabilization and digital image registration in a single AOSLO system. The main improvement compared to previous reports include application of two-axis tip/tilt mirror as slow scanning mirror allowing efficient implementation of Closed-loop image based optical stabilization, followed by real-time digital image registration. This dual correction offered very accurate and robust real time stabilization of retinal images. The system performance was tested on healthy volunteers and optical stabilization was successful in all but one of the 183 trials with 85% of all frames successfully stabilized. Detailed evaluation of the tracking data revealed that tracking efficiency decreased as imaging duration increased, a result that was expected due to the subject fatigue associated with long time imaging.
In summary, the system presented in this article offers excellent real-time retinal motion artifact correction on AO-SLO images by combining optical stabilization and digital registration, resulting in a measured RMS error below 0.25 μm or 0.05 arcmin (about a tenth of a diameter of the foveal cone). The article also includes discussions on future directions for further improvement of system performance, including increased tracking range, correction of microsaccades and eye movement beyond the reference frame as well as resetting of the position of the tracing mirror.
You must log in to add comments.
Efficient eye motion tracking and correction is essential not only for successful imaging but also for precise delivery of visual stimuli. This is why it has been an active area of research even before the first demonstration of retinal AOSLO. With application of AOSLO the common strategies implemented to achieve retinal image stabilization include active correction of retinal movements using a retinal tracking system or real time registration of acquired images allowing inter-frame correction of eye-motion artifacts. The authors provide a detailed historical overview of retinal motion tracking and image stabilization methods that were previously developed by groups working in this field.
The framework presented in this manuscript describes a novel real-time solution to image stabilization by implementing both optical stabilization and digital image registration in a single AOSLO system. The main improvement compared to previous reports include application of two-axis tip/tilt mirror as slow scanning mirror allowing efficient implementation of Closed-loop image based optical stabilization, followed by real-time digital image registration. This dual correction offered very accurate and robust real time stabilization of retinal images. The system performance was tested on healthy volunteers and optical stabilization was successful in all but one of the 183 trials with 85% of all frames successfully stabilized. Detailed evaluation of the tracking data revealed that tracking efficiency decreased as imaging duration increased, a result that was expected due to the subject fatigue associated with long time imaging.
In summary, the system presented in this article offers excellent real-time retinal motion artifact correction on AO-SLO images by combining optical stabilization and digital registration, resulting in a measured RMS error below 0.25 μm or 0.05 arcmin (about a tenth of a diameter of the foveal cone). The article also includes discussions on future directions for further improvement of system performance, including increased tracking range, correction of microsaccades and eye movement beyond the reference frame as well as resetting of the position of the tracing mirror.
Reader Comments
Add Comment
You must log in to add comments.
Kaccie L.
09/25/2014 11:47 AM
Closing the loop on this difficult is no simple feat. This is in my opinion the most impressive work in the field for the last 8-10 years. Congrats!