An Adaptive Optics System for Retinal Imaging using a Pyramid Wavefront Sensor
Investigators: Stéphane Chamot*, Elizabeth Daly, Sabine Chiesa, and Chris Dainty.
* now at Ecole Polytechnique Federale de Lausanne (Switzerland)
Collaborators: Simone Esposito, Osservatorio Astrifisico di Arcetri, Firenze, Italy.
Funding: Science Foundation Ireland , Sharp-Eye European Research Training , EU Marie Curie Early-Stage Training .
Pyramid Wavefront Sensing
Pyramid Wavefront Sensing (PWS) was first proposed in 1996 by Ragazzoni [1] for astronomical applications, and has since been used to measure [2], and correct for the aberrations in the human eye [3]. PWS is a modified version of the Foucault knife-edge test, where a refractive pyramid is used to split the light from a pupil into four sub-beams. These are then re-imaged to form four images of the pupil. A straightforward calculation carried out on the intensity distribution among the four pupil images yields the x and y wavefront gradient maps.
Figure 1 . Principal of PWS.
A schematic diagram of our optical system showing the core components is shown in Figure 2. A detailed discussion of the design and performance of the PWS operating in vivo can be found in Chamot et al. [3]. The setup incorporates a 19-channel OKO piezoelectric deformable mirror and can correct ocular aberrations to λ/5 residual RMS over a 6 mm pupil in closed-loop at a rate of 55 Hz. The steering mirror modulation employed allows flexibility in control of the aberration measurement sensitivity and linear range through adjustment of the mirror drive voltage. A custom LabVIEW interface controls the loop and allows one to easily change, for example, the pupil sampling; coarser pupil sampling has the advantage of reducing the effect of speckle and allows faster operation of the loop.
Figure 2 . Schematic of experimental setup
Typical in vivo results showing the evolution of the rms wavefront error on going from open- to closed-loop operation is shown in Figure 3. A recording of the double-pass point spread function shows the benefit to be gained by correcting ocular aberrations.
Figure 3: Evolution of the RMS wavefront error
Movie : Double-pass point spread function film for open/closed loop control
ConorL_PSF_OL-CL.wmv Movie of double-pass psf from CL on 061206
Retinal imaging
The system has been modified to incorporate retinal imaging. Illumination of the fundus is provided by a Luxeon high-power LED (centre wavelength 530nm, bandwidth ~50nm) which can be driven in flash-mode to provide short flashes (typically 10 – 100ms) of variable duty cycle. A Köhler-type illumination scheme is used to ensure that none of the LED structure is imaged at the retinal plane, and a mask with a central obscuration placed in a conjugate pupil plane ensures off-axis illumination thus minimizing back reflections. The field of illumination on the retina is approximately 3˚ to provide a relatively flat intensity profile on imaging over 1˚ - the angle over which we expect good correction with a single Deformable Mirror. Typically, a sequence of 10 images, coinciding with 10 flashes from the LED, is gathered and analysed. A fixation target mounted in the illumination arm allows repeated imaging of the same patch of retina for each flash. The magnification of factor of ~8 from retina to imaging camera provides good sampling of the retina given the choice of camera (Retiga 1300, QImaging, Burnaby, Canada). A very preliminary result showing a single raw image (fov ~ 2°), taken about 2° off-axis, is shown below. Some small capillaries are visible in the image.
Figure 4 . Single raw frame for t=100ms, and after flat-fielding.
Future work
Recent studies [4,5] have suggested that PWS performance could be improved through use of a mirror with a larger stroke and an increased number of actuators than the DM currently in use. Simulations carried out using a magnetic DM (Mirao52, Imagine Eyes) indicate that it should provide much better correction on fitting typical ocular wavefronts, as shown in Figure 5.
Figure 5 . Simulated performance of various DMs on fitting typical ocular wavefronts.
In the near future we plan to upgrade the DM in the system and characterize both sensor and fundus imaging performance. Our aim is to image photoreceptors close to the fovea of the human retina.
Parallel work
We have done some work on the linearity of the pyramid as a wavefront sensor [6], and are also investigating the influence of speckle in the PWS.
[1] Ragazzoni, R., Pupil plane wavefront sensing with an oscillating prism, JOURNAL OF MODERN OPTICS, 1996, Vol.43, No.2, (289-293).
[2] Iglesias, I., Rgazzoni, R., Julien, Y. and Artal, P., Extended source pyramidwave-front sensor for the human eye, OPTICS EXPRESS, 2002, Vol.10, No.9, (419-428).
[3] Chamot, S. R. and Dainty, J. C., Adaptive optics for ophtalmic applications using a pyramid wave-front sensor, OPTICS EXPRESS, 2006, Vol.14, No.2, (518-526).
[4] Dalimier, E., Dainty, J.C., Comparative analysis of deformable mirrors for ocular adaptive optics, OPTICS EXPRESS, 2005, Vol.13, no.11, (4275-4285)
[5] Daly, E., Dainty, J.C., Requirements for MEMS mirrors for adaptive optics in the eye, Proc. SPIE, 2006, Vol.6133, 611309.
[6] Burvall, A., Daly, E., Chamot, S.R., and Dainty, J.C., Linearity of the pyramid wavefront sensor, OPTICS EXPRESS, 2006, Vol.14, No.5, (11925-11934)