Soon after the invention of the adaptive optics ophthalmoscope [1
J. Liang, D. R. Williams, and D. T. Miller, J. Opt. Soc. Am. A 14, 2884 (1997). [CrossRef]
], a scanning version of the ophthalmoscope was introduced [2
A. Roorda, F. Romero-Borja, W. Donnelly III, H. Queener, T. Hebert, and M. Campbell, Opt. Express 10, 405 (2002).
]. Since then, the adaptive optics scanning laser ophthalmoscope (AO-SLO) has been heavily and successfully used in many productive studies [3
P. Godara, A. M. Dubis, A. Roorda, J. L. Duncan, and J. Carroll, Optometry Vision Sci. 87, 930 (2010). [CrossRef]
]. The instrumentation of the AO-SLO has been improved with modern, sophisticated technologies. It has been equipped with a retinal tracker to compensate for eye movement, optics that observe a wide field of view (FOV) [4
S. A. Burns, R. Tumbar, A. E. Elsner, D. Ferguson, and D. X. Hammer, J. Opt. Soc. Am. A 24, 1313 (2007). [CrossRef]
], and spectral domain optical coherence tomography [5
M. Mujat, R. D. Ferguson, A. H. Patel, N. Iftimia, N. Lue, and D. X. Hammer, Opt. Express 18, 11607 (2010). [CrossRef]
]. An advanced system can even resolve rod photoreceptors [6
A. Dubra and Y. Sulai, Biomed. Opt. Express 2, 1757 (2011). [CrossRef]
]. Another trend has been the development of practically useful clinical instruments from inexpensive devices, such as line scan mechanisms [7
M. Mujat, R. D. Ferguson, N. Iftimia, and D. X. Hammer, Opt. Express 17, 10242 (2009). [CrossRef]
] or microelectro mechanical systems (MEMS) mirrors [8
Y. Zhang, S. Poonja, and A. Roorda, Opt. Lett. 31, 1268 (2006). [CrossRef]
To our knowledge, most AO-SLOs are built with reflective concave and convex mirrors. Some of them use refractive lenses, but some optical elements are still reflective because the use of refractive elements can cause unwanted reflections, which eventually become noise in wavefront sensor and retinal images. One limitation of reflex concave and convex mirrors is that they reduce aberrations over a small FOV.
We have built an adaptive optics dioptric scanning ophthalmoscope (AO-DSO) without using any concave or convex mirrors; instead, we used refractive lenses for all of the power optics. One of our interests was developing small and simplified AO scanning ophthalmoscopes for clinical needs. The obstacles to their clinical use are their high cost, large size, limited clinical applications, and small FOV. The device typically responsible for the high cost is the deformable mirror. However, we believe that the price of deformable mirrors is decreasing with advances in MEMS technology.
The optics of the AO-DSO introduced in this study had an area of . The AO-DSO is smaller than a commercially available ophthalmoscope (TRC-50LX, Topcon, Tokyo, Japan), which has an area of . Using the advantages of dioptric optics, we designed and built an AO scanning system with a high-resolution FOV (FOV1.5), a medium resolution FOV (FOV3), and a normal scanning system with a FOV (FOV10). The maximum FOV was determined by an objective lens, for which we used an objective lens from a commercially available autorefractometer (KR8100, Topcon, Tokyo, Japan).
Using the AO-DSO, we investigated images with FOV1.5 and FOV10 on normal eyes and eyes with occult macular dystrophy (OMD). We also investigated imaging of the nerve fiber layer (NFL) and photoreceptor layer. Additionally, one important concern addressed here is the effectiveness of the AO for FOV10 and FOV1.5 images.
A schematic of the AO-DSO optics is shown in Fig. 1
. The light source for the imaging and wavefront sensing was an 840 nm superluminescent diode (SLD, AS8E210GP30M, Anritsu, Kanagawa, Japan). The spectral half width of the SLD was 15 nm. The radiant power at the cornea was 80 μW. The beam diameter at the pupil was 6 mm. If optics are diffraction limited, the optical resolution is 0.58 arcmin. Our system was designed so that FOV1.5 is diffractive limited for the 740–1000 nm spectral region and for the vergence range from
. The SLD was coupled with a single-mode optical fiber, and the light emitted from the fiber impinged on a deformable mirror (Mirao 52d, Imagine Eyes, Orsay, France), a vertical scanner, and then a horizontal scanner. The deformable mirror and scanners were optically conjugated with the pupil of the eye. The vertical scan was driven by a galvanic scanner (
, Cambridge Technology, Lexington, Massachusetts), and the horizontal scan was driven by an 8 kHz resonant scanner (8 kHz CRS, Cambridge Technology, Lexington, Massachusetts). The vertical and horizontal scanners can change the scanning range of the retina from 0.92 to 6.9 deg and from 1.25 to 9.6 deg, respectively. After the scanners, the light passed through a focusing prism and objective lens to the retina.
Fig. 1. AI, light source for imaging the anterior part of the eye; OL, objective lens; FL, focusing lens; OE, organic electroluminescence diode display; AC, CCD for anterior imaging; P, focusing prism; HS, horizontal scanner; VS, vertical scanner; DM, deformable mirror; PBS, polarized beam splitter; HM, half-mirror; PH, pinhole; H2, 2 mm diameter hole; PMT, photomultiplier; SH, Shack–Hartmann wavefront sensor; SF, single mode fiber; SLD, superluminescent diode. Most lenses are doublet-type lenses.
An anti-reflection (AR) coating was applied to all of the refractive lenses’ surfaces, and the reflectance after the coating was less than 0.02% at 840 nm (Fig. 2
). The AR coating was developed especially for our wavefront sensor product (KR-1W, Topcon, Tokyo, Japan) using ion-assisted deposition technology. We used the focusing prism to correct the spherical errors (
) of the eyes, such that the retina and optical fiber were optically conjugated. The reflected light from the retina was passed from the retina to the deformable mirror. After the deformable mirror, the light reached a photomultiplier (PMT; H7732-10, Hamamatsu Photonics, Hamamatsu, Japan) for imaging. The sampling rate of the electronic circuit for imaging was 15 MHz. The pixel number of the image was
. The acquisition frame rate was 30 frames per second. Scanning ranges were changed for three FOVs. The pixels per arc minute of FOV10, FOV3, and FOV1.5 were 0.83, 0.22, and 0.11, respectively. The ratio of penetration and reflection of a half-mirror after the deformable mirror was 9 to 1 because the light returned from the retina should retain as much of its intensity as possible. The pinhole in front of the PMT was optically conjugated with the fiber of the light source through the layer of the retina that we wanted to image. The optical magnification between the retina and pinhole was 3.5. The pinhole was five times (17.5 times in real dimensions) as large as the Airy disk. Ten percent of the light reflected at the half-mirror went to a polarized beam splitter (PBS), and then polarized light went to a Shack—Hartmann wavefront sensor (SHWS). The PBS was used to form cross-polarized isolation blocking noise light from optics of the instrument and eye to the SHWS. We also inserted a 2 mm diameter hole in front of the SHWS, which is optically conjugated with the retina. This tactic effectively blocks reflection noise from the anterior part of the eye.
Fig. 2. Anti-reflection coating of the objective lens.
The SHWS consisted of a lens array and charge coupled device (CCD). The image obtained by the CCD was analyzed with Zernike polynomials. The coefficients from the second-to sixth-order terms were sent to the control circuit of the deformable mirror to compensate for eye aberrations. Two 940 nm light emitting diodes and a CCD were used to image the anterior part of the eye. An organic electroluminescent diode display (OELD) was used for the fixation target.
The eyes of the subjects were dilated before imaging. Using an electrically adjustable chin rest and the image of the anterior part of the eye, the center of the pupil was aligned to the optical axis of the objective lens of the AO-DSO. The focusing prism for imaging and the focusing lens for the fixation target were adjusted to correct for the eye’s spherical error. The OELD fixation target was shown at a certain position on the OELD to rotate the eye, allowing any desired position of the retina within 5 deg from the fovea to be imaged. After the alignment was completed, the imaging SLD was turned on to begin imaging. The focusing prism was precisely controlled to image a particular layer of the retina. The feedback loop of the wavefront sensing and compensation was 10 Hz. The protocol for this study was approved by the Institutional Review Board (IRB) of the Osaka University Medical School, and the procedures conformed to the tenets of the Declaration of Helsinki.
We performed imaging on the eye of a normal male (36 years old) and on an eye with OMD. We confirmed that the AO function worked for normal eyes when focused at the photoreceptor layer of the retina to obtain FOV1.5 images, as in previous studies with catoptric and catadioptric systems [Fig. 3(c)
]. We also successfully imaged the same areas of the retina with FOV3 and FOV10.
Fig. 3. Imaging of the photoreceptor layer of a normal retina at three different magnifications. The images were taken at 5 deg nasal from the fovea. All the images were obtained by averaging 10 images. (a) FOV10, (b) FOV3, and (c) FOV1.5.
We performed imaging of the NFL with FOV10. The NFL could be imaged by moving the focus 68 μm toward the cornea by moving the focusing prism. The difference in focusing is shown in Fig. 4(a)
, and a montage NFL image is shown in Fig. 4(b)
. We experimentally confirmed that the same NFL imaging was not possible when the size of the beam for imaging was set to 2 mm.
Fig. 4. Optical sectioning between retinal photoreceptor layers and nerve fiber layers. (a) A montage FOV10 image of the photoreceptor layer. (b) A montage FOV10 image of the nerve fiber layer. The white bar represents 3.5 deg. (c) The high-resolution image was taken at the location of 1 deg eccentricity from the center of the fovea with FOV1.5 (four images average).
We calculated modulation transfer functions (MTFs) for on-axis FOV1.5 imaging with and without AO. We also calculated MTFs for on and off-axis (5 deg) FOV10 over 6 mm pupil imaging with and without AO and over 2 mm pupil imaging without AO (Fig. 5
). We found that the optics were almost perfect in FOV1.5 imaging with AO. However, the optics were not ideal in FOV10 imaging, even with AO. Although the MTF over a 2 mm pupil is better than that over a 6 mm pupil, as expected, the MTF over a 6 mm pupil with AO was better than that over a 2 mm pupil without AO for 50 to 80 cycles per degree.
Fig. 5. MTFs. (a) FOV1.5; (b) FOV10 at the center of the fovea; (c) FOV10 at 5 deg nasal from the fovea.
FOV1.5 images and FOV10 images of the retina of an OMD patient with and without AO are shown in Fig. 6
. The patient was a 51-year-old woman who visited Osaka University hospital with the complaint of decreased vision in both eyes. Her best corrected visual acuity was
in both eyes. The fundus picture was normal. However, multifocal electroretinography showed decreased responses in the macular areas of both eyes, which is typical in eyes with OMD. Fourier domain optical coherence tomography imaging has shown that the inner and outer segment (IS/OS) junction of the photoreceptors in the foveal area was disrupted, which was consistent with the previous report [9
Y. Kitaguchi, S. Kusaka, T. Yamaguchi, T. Mihashi, and T. Fujikado, Clin. Ophthalmol. 5, 345 (2011). [CrossRef]
Fig. 6. Images of the photoreceptor layer of occult macular dystrophy. AO-OFF imaging (a) FOV10; (b) and (c) FOV1.5. AO-ON (d) FOV10; (e) and (f) FOV1.5. Images correspond to the arrows in the FOV10 images.
The FOV1.5 images of both small areas with AO were clearer than those without AO. The photoreceptors in the intact area were healthy, with a complete mosaic pattern. However, fewer photoreceptors were found in the diseased area. From these findings, we confirmed that the AO-DSO revealed the photoreceptor disruption in the eye with OMD with FOV1.5 images.
In the FOV10 images, the photoreceptors were not resolved, even with AO. Although we could not see the photoreceptors, the border of the normal and abnormal areas was much clearer in the image with AO than in that without AO. The AO-DSO allows for identification of the diseased area in the retina at a glance from the FOV10 images, reducing examination time.
AO-DSO was effective in the production of both FOV1.5 and FOV10 images. We were not troubled by noise from the refractive lenses during wavefront sensing and imaging. This demonstration that AO scanning ophthalmoscopes are useful for normal SLOs and very high-resolution imaging suggests that they may be useful in the clinical setting.