Adaptive optics is helping improve two-photon microscopy and laser-scanning ophthalmoscopes.
Hank Hogan, Contributing Editor
Originally developed to see the very large and far away, adaptive optics is now being used to see the very small and relatively near.
Astronomers perform a technique that rapidly adjusts optical elements, typically mirrors, to remove the twinkle from stars. In this case, the optics are adapting to changes in the medium and are removing distortions introduced by air currents.
Similar techniques can be employed to overcome the optical effects of tissue on microscopy and other types of biological imaging. Adaptive optics can make blurred and indistinct features sharp and clear, thereby clarifying experiments. Several research groups are using adaptive optics to improve microscopy -- and not all of them are directly improving the image.
In theory, adaptive optics should improve resolution, signal and contrast in two-photon microscopy. Adjusting the wavefront of the incoming light should cancel degradations created by refractive index variations encountered as the light traverses the specimen.
In practice, researchers must measure the wavefront first, which creates a problem. Most schemes employed so far have been fluorescence-based, but determining the needed corrections is a trial-and-error process that requires taking repeated measurements. While the measurements are being taken, the fluorophores may become photobleached and, thus, rendered invisible. There is also the possibility of tissue damage from a photodynamic process.
Now researchers at Max Planck Institute for Medical Research in Heidelberg, Germany, under the direction of Winfried Denk, have come up with a different approach. Employing a deformable mirror and a technique called coherence-gated wavefront sensing, the researchers achieved nearly diffraction-limited focus. The institute has applied for a patent on the technique.
Adaptive optics correction using coherence-gated wavefront sensing was performed in living zebra fish larvae during imaging in the forebrain (a). Wavefront aberrations (contour lines spaced by 0.1 μm) are shown as measured by coherence-gated wavefront sensing before correction (b). Single focal plane images at 50-μm depth were recorded without (c) and with (d) adaptive optics correction. For both, the background was subtracted and the effects of the scan motion were removed. Single focal plane images of blood vessels in the forebrain (depth 200 μm) were recorded without (left) and with (right) correction (e). The dark regions inside the blood vessel are blood cells. Blood-flow measurements using line scans from the positions indicated by arrows in (e) are plotted as a function of time (black, without correction; red, with [f]). Reprinted with permission of PNAS.
Split the beam
Unlike other wavefront sensing methods, the coherence-gated technique rejects scattered light that doesn’t originate from the focal region. Consequently, the parameters for wavefront correction can be determined at low laser power levels and without the need for fluorescence.
The technique requires samples that backscatter light, but the backscattering resulted in optical speckle that dominated as a noise source that then had to be averaged out.
In a proof of principle, the scientists used a low-coherence interferometer for coherence-gated wavefront sensing, creating a reference and sample arm by splitting a laser beam. The laser was a Coherent Ti:sapphire laser with a center wavelength of 930 nm, typically set at ~8 mW during two-photon imaging and only 60 nW during wavefront sensing.
In the sample arm, the beam entered lenses, bounced off a 37-channel electrostatically deflected deformable mirror from Oko Technologies of Delft, the Netherlands, and was directed into the sample. The team measured the fluorescence from the sample using a photomultiplier tube from Hamamatsu in Japan, placing the detector behind a dichroic mirror from Chroma Technology Corp. in Rockingham, Vt., that reflected the infrared light from the laser but that transmitted light at less than 525 nm.
Using a Sony CCD, they measured the wavefront, averaging 20 speckle patterns created by moving the focus a micron between each. They corrected for the aberration with the deformable mirror.
Markus Rueckel, who is a postdoctoral researcher at the institute, noted that the setup involved some mirrors in the reference arm whose only purpose was to fold the beam and make everything more compact. It was important to make the setup light-efficient so that wavefront determination was fast, which is important to real-time imaging.
Despite its complexity, the arrangement wasn’t that hard to assemble. “Although the setup contains quite a few optical elements, it is not particularly difficult to align,” Rueckel said.
As described in the Nov. 14 issue of PNAS, the researchers evaluated the performance of the equipment and the technique. They corrected for the optical aberrations introduced by the elements within the setup and imaged a glass capillary (because it has known distortions). After five iterations, they reduced the distortions in one case from an initial 283 nm to 17 nm, transforming blurred bead images into relatively sharp ones. They also imaged the forebrain of zebra fish larvae, achieving results Rueckel found surprising because only a few low-order-aberration terms needed to be corrected to get a substantial imaging improvement.
With the principle proved, the researchers plan to determine how deep within tissue the coherence-gated wavefront sensing technique will work. They also are looking to upgrade the equipment. One component they are considering changing is the deformable mirror; specifically, the stroke and speed of its actuators. The stroke determines how much the mirror can be deformed and, thus, how much of a correction can be made.
In this adaptive-optics setup, coherence-gated wavefront sensing is performed by splitting a laser beam to create a reference and a sample arm. The two are combined to make an interferometer that rejects out-of-focus light during wavefront sensing, using only nanowatts of power. The deformable mirror (DM) is adjusted to remove distortions, with the CCD performing the wavefront measurement. Imaging of the sample is done using the photomultiplier tube (PMT). (ND = neutral density filters; pBSC = polarizing beamsplitter cube; BSC = nonpolarizing beamsplitter cube; GVD-P = group velocity-dispersion compensation prisms; L = lens; M = mirror; and RP = right-angle prism). Reprinted with permission of PNAS.
However, Gleb Vdovin, CEO of deformable mirror maker OKO Technologies, noted that stroke can’t be increased without cost, in more ways than one. “For a fixed mirror size, there is a trade-off between the price, maximum stroke and number of actuators.”
Imaging the eye
Eyes are windows to the soul, it is said, but in reality, eyes are not windows. Instead, they focus light, often imperfectly. Aberrations in the human eye make imaging it difficult and degrade the performance of scanning laser ophthalmoscopes. Now investigators at the University of California, Berkeley, have developed a compact, robust adaptive optics laser scanning ophthalmoscope based upon a microelectromechanical systems (MEMS) deformable mirror from Boston Micromachines of Watertown, Mass.
Using a Shack-Hartmann wavefront sensor to determine the aberration, they operated the adaptive optics in a closed-feedback loop and corrected distortionsin most eyes tested to below 0.1 μm root mean square. That is an improvement over the 0.2 to 0.4 μm rms found without adaptive optics.
Austin Roorda, associate professor of optometric sciences at the university, noted that the deformable mirror was the enabling technology. He has worked with various types of such mirrors over the past 10 years, but the current version offers some significant advantages. “The MEMS systems are nice, though, because they keep the system small and reasonably priced,” he said.
The mirror they used had 144 actuators spaced across an area 4.4 mm on a side and with a maximum stroke of 3.5 μm. The total system developed by the researchers occupied an area of 50 cm on a side. The work was published in the May 1 issue of Optics Letters.
They used 655- and 840-nm diode lasers. The higher wavelength source was a low-coherence superluminescent laser diode from Superlum Diodes Ltd. of Moscow and was selected because it reduced interference artifacts. The researchers raster-scanned the eye with the illumination well below safe exposure levels, recording the result with a photomultiplier tube.
In operation, they measured the wavefront, calculated a best fit 10th-order Zernike polynomial to the wavefront slopes and determined what actuator deflections to apply to correct the aberrations. They did this in a closed-loop fashion 10 times a second so that the aberrations were driven to a minimum.
Using test subjects, they adjusted for distortions in each eye but only with corrective trial lenses placed to minimize defocus and astigmatism. They achieved a threefold imaging benefit by decreasing distortion to less than 0.1 μm rms. The resulting images of the eyes showed increased brightness, improved contrast and enhanced lateral resolution.
Roorda noted that these results were not entirely unexpected because systems to measure and correct the eye’s aberrations have been around for 10 years. The earlier systems, however, often used deformable mirrors that were both larger and more expensive than the Berkeley system, resulting in ophthalmoscopes that were larger and not as robust.
That is not to say that the MEMS deformable mirror technology is perfect. In particular, it requires imaging through trial lenses because the deformable mirror becomes saturated and can’t correct for all of the aberrations in the eye. There are enough actuators and control points across the surface of the mirror for the task; the problem lies with the stroke, the amount of actuator travel. “If we had a mirror with higher stroke, we could correct more eyes without the need to correct defocus and astigmatism as carefully as we do now,” Roorda said.
Paul Bierden, president and CEO of Boston Micromachines, said Roorda isn’t alone in his request. Most of those who use the company’s deformable mirrors for biological imaging would like to see larger actuator stroke, and Boston Micromachines has been responding. Plans call for an increase from today’s 3.5 μm to 6 μm in 2007 and 8 μm in 2008.
Speed-stroke trade-off
Bierden said that efforts to increase the stroke run up against manufacturing method limits. A second problem is that greater stroke comes with a price. “There is a speed-stroke trade-off,” he said.
He reported that biological imaging customers typically want less than kilohertz speeds, which is relatively low and thus not an issue. The same can’t be said for all adaptive optics applications. Astronomers, for example, want it all -- a stroke of three or more microns, speeds of 5 kHz, and 4000 actuators on a mirror.
Boston Micromachines is working to produce just such a product. If the company succeeds, the resulting technological innovations could benefit deformable mirrors used in biological imaging.
Scientists from Boston University (BU) are not using deformable mirrors to remove aberrations. Instead, they have set up the mirror to add aberrations, with the goal of improving the signal-to-background ratio of two-photon excitation fluorescence imaging in thick tissue. With this approach, they can clean up images and make items of interest stand out more clearly. They have even improved the apparent resolution.
Adding the right distortion can help focus on what is important. On the left is a two-photon microscopy image of mouse olfactory sensory neurons labeled with green fluorescent protein with a flat deformable mirror. Shown in the middle is the image with the mirror deformed into two hills and valleys, producing quadrant aberrations in the image that distort and, thus, remove the signal from the focal region but not that of the background. On the right is a corrected image obtained by subtracting the second from the first. Courtesy of Jerome Mertz, Boston University.
When imaging deep into tissue, two-photon imaging requires exponentially increasing laser power. Eventually, demand exceeds supply, limiting the imaging depth. Strategies to get around this run into the problem of out-of-focus background fluorescence generated near the surface. Removing the background would make deep imaging more feasible, which is what the researchers did.
Jerome Mertz, associate professor of biomedical engineering at BU, said that the technique developed by his group exploits the nature of the fluorescence. “Two-photon fluorescence is nonlinear, making the separation of signal from background by aberration modulation particularly clear-cut.”
When aberrations are introduced, the signal decreases, but the background remains almost unchanged. Subtracting the image with added distortion from the one without distortion extracts the signal, improving the imaging.
In their prototype setup, the researchers used a Ti:sapphire laser from Spectra-Physics of Mountain View, Calif., at 800 nm. The beam bounced off a deformable mirror from Boston Micromachines and entered an arrangement of lenses, filters and scanning mirrors. A photomultiplier from Electron Tubes of Rockaway, N.J., detected the resulting two-photon fluorescence.
Although the researchers used a high-end deformable mirror, they did so largely because it was lent to them. They only needed a λ/4 stroke and four elements or zones in the mirror. Mertz noted that deformable mirrors in general offer benefits. “The advantage of a deformable mirror is that it is potentially very fast. It can also handle relatively high power densities, making it convenient for two-photon microscopy.”
When no voltage was applied, the deformable mirror was flat to 40 nm rms, and images were aberration-free. To add aberrations to images, the scientists applied an open-loop voltage to force it into a simple pattern of two hills and two valleys.
For their demonstration, they used rhodamine dye and titanium dioxide dissolved in a water-ethanol mixture, along with some fluorescently labeled pollen grains. The combination of dye and scattering particles made the mixture optically equivalent to a thick sample, and the grains acted as fluorophores in that thick sample. They imaged a frame at a time, capturing the entire field of view with and without aberration before subtracting them. The work was published in the Oct. 30 issue of Optics Express.
Upgrading system
The subtraction of the background enhanced the visibility of the fluorophores and increased the apparent resolution, something Mertz was not that surprised to see because of earlier results. “A similar idea using differential temporal aberrations had been previously demonstrated to enhance axial resolution.”
As for the future, the researchers are upgrading their system for line-by-line differential aberration imaging, with the goal of doing so pixel by pixel. Those changes should speed up the image-capture process substantially. Fast two-photon microscopy can be used for calcium imaging on a millisecond time scale.
Another planned change involves the deformable mirror. Because a much less expensive implementation would do everything that is needed, the group is looking into the development of inexpensive mirrors specifically for this application.
However, a more capable mirror could someday be necessary. The differential aberration imaging technique rejects background but doesn’t enhance the signal. If the signal falls below the noise level, it will have to be boosted. One possible solution would be to use adaptive optics with wavefront correction. The mirror would then need to switch from a configuration that adds aberrations to one that removes them, and back, reliably, something that is possible with current technology.
According to Mertz, this approach is part of the group’s future plans, but putting it into practice won’t be simple. “This can be implemented with the same deformable mirror, but it requires a multiparameter closed-loop correction algorithm, which significantly complicates matters.”