After measuring the eye's aberrations with a wavefront sensor, we can correct for them using adaptive optics. Once the eye's aberrations are corrected, we can either take high resolution images of the retina or perform psychophysical experiments to test for improvements in vision. Below are the layouts of our 1st (left) and 2nd (right) generation adaptive optics systems. Our original adaptive optics system used a 37 channel deformable mirror. The 97 channel deformable mirror currently used in our 2nd generation system is so large that we need to magnify the eye's pupil by about a factor of ten in order to fill the mirror. Unfortunately, extremely long focal length optics must be used to provide this degree of magnification. Therefore, in order to fit the 2nd generation system on one optical table, we have to use parabolic mirrors instead of lenses to image the eye's pupil onto the deformable mirror. Another change in the new system includes the use of a 790 nm superluminescent diode (instead of a He Ne laser) as our wavefront sensor source that is coupled into the system right at the eye, instead of near the wavefront sensor. In addition, we no longer reject certain polarizations of light emerging from the eye. These changes have the practical benefit of improving the quality of our Shack-Hartmann images. The 790 nm SLD also has the advantage, due to its long wavelength, of being much more dim and comfortable for subjects to view.

Here is a photo showing our actual system setup while Jason Porter serves as the subject. The path of the SLD illumination is indicated by the red line. The DMD display that we use to show visual stimuli though the system is not shown.

The key improvements in our new system over the 1st generation system described by Liang, Williams, & Miller (1997) are:

• Real-time closed-loop feedback control (up to 30 Hz frame rate 0.8 Hz closed-loop bandwidth)
• A 97 actuator deformable mirror instead of a 37 actuator mirror
• A more robust and accurate wavefront sensor
• A CCD camera for retinal imaging with 2.7 times the quantum efficiency of the old system
• A completely automated experimental protocol

Resulting in:

• 3 X greater strehl ratio after correction
• 40% average increase in the contrast of cones in retinal images
• 5 X reduction in the numbers of retinal images needed to achieve the same signal to noise ratio as in the old system
• Much faster and easier operation

While an AO system operating with a closed loop bandwidth of 0.8 Hz may not seem impressive by the standards of current astronomical AO systems, it represents a critical step forward for vision science. The mirror control system does not require special purpose hardware and all computations are performed on a personal computer. In the 1st generation adaptive optics system, it took approximately 30 minutes to correct the wave aberration with major operator intervention, and the correction provided was essentially static. The 2nd generation system can completely correct the eye's aberrations automatically in 250-500 milliseconds and reduces, by a factor of 5, the number of retinal images required for the same signal to noise ratio. This means we can run experiments much more quickly and efficiently, and we can successfully obtain retinal images on a larger fraction of subjects. The system can correct the eye's static aberrations up to 8th order and can track and correct fluctuations in Zernike modes up to 5th order. The system can also reduce the eye's residual rms wavefront error over a 6.8 mm pupil to values lower than 0.1 microns.

The QuickTime movies below who how the wave aberration and pointspread function change during correction with the 2nd generation adaptive optics system. The adaptive optics system was operating at a rate of 25 Hz and adaptive correction began about halfway through the measurement. These movies show the wave aberration and pointspread function of former postdoctoral fellow Nathan Doble.

The improvement in retinal image quality provided by correction of the eye's higher order aberrations with the 2nd generation adaptive optics system is illustrated below. This figure shows two single representative images of the same retinal location take on the same day for one subject without correction of higher order aberrations (left) and with correction of higher order aberration using the 2nd generation adaptive optics system (right). This particular subject has superior optical quality, compared to the majority of subjects, yet there is still a dramatic improvement in the quality of retinal image taken with adaptive optics. Cones that are barely detectable in the first image are clearly visible in the second.

We also looked at the contribution of this new dynamic correction ability to our overall gains in system performance (Hofer et al 2001). We took a series of retinal images on 5 subjects after either a static or dynamic correction of higher order aberrations. Images for a few of the subjects are shown below. Each images is a registered sum of about 20 individual images. You can see that for some of the subjects (AP & CM), the quality of the images are subjectively better with a dynamic correction instead of only a static correction with adaptive optics. Other subjects like GYY show little, if any, subjective improvement. This is expected since some subjects exhibit much more stability in their wave aberration than others.

To quantify the improvement in retinal image quality provided by our ability to dynamically correct the eye's wave aberration, we compared image power spectra obtained after both a static and dynamic correction. The figure below shows the improvement in the retinal image power spectrum (ratio of the image power spectrum with dynamic correction to the image power spectrum with only static correction of higher order aberrations) averages across 5 subjects. The range of cone frequencies for these subjects is also indicated on the plot. At these frequencies, the average improvement in the retinal images power spectrum is almost a factor of two. The average improvement in retinal image cone contrast, due solely to correcting temporal fluctuations in the wave aberration, was calculated to be 33%.

Liang J., Williams D.R., & Miller D.T., (1997). Supernormal vision and high resolution retinal imaging through adaptive optics. Journal of the Optical Society of America A, 14, 2884-2892.

Hofer H., Chen L., Yoon G.Y., Singer B., Yamauchi Y., & Williams D.R., (2001). Performance of the Rochester 2nd generation adaptive optics system for the eye. Optics Express, 8, 631-643.

top

Next: Adaptive correction and visual performance