FOCUS
WAVEFRONT control
architecture. Thus, there is no universal‘ best’ AO solution. The wavefront element, sensing strategy and quality metric must be carefully chosen for the respective modality. Here we discuss this through three examples: multi-photon microscopy, confocal Raman microscopy and single molecule localization microscopy.
Adaptive optics in multi-photon microscopy
Multi-photon( MP) excitation fluorescence microscopy was developed to access deeper tissue regions and remains central to in vivo brain imaging [ 2 ]. Instead of one visible photon, MP microscopy uses the combined energy of two or three near-infrared( NIR) photons arriving almost simultaneously. NIR light is less scattered, and nonlinear excitation confines fluorescence to the focal volume, creating optical sectioning without a confocal pinhole.
This geometry is particularly attractive for AO: only the excitation laser requires correction, because the emitted fluorescence does not need to be imaged back through the tissue onto a pinhole. Since the laser is quasi-monochromatic and polarised, either a high-pixel-count liquid crystal SLM or a recently developed piston-phase light modulator( PLM) can be used to impose highly complex phase patterns on the incoming wavefront.
Different wavefront-sensing strategies are possible. Direct sensing can use, for example, a Shack-Hartmann sensor, exploiting the fact that most of the fluorescence originates from the focus and thus forms a kind of guide star. Alternatively, the wavefront can also be measured indirectly. Because of nonlinear excitation a well corrected focus produces the highest signal and therefore the integrated signal itself can guide an iterative optimization. Such routines make it possible to compensate even strong aberrations where photons are multiply scattered. Figure 2 shows such a demonstration: fluorescently labelled microspheres hidden below 0.5 mm of chicken breast tissue are invisible as distinct objects without AO. Using DASH [ 3 ], a wavefront-sensing method developed in our group, the wavefront can be measured in a few seconds and corrected so that one bead becomes discernible.
This example points to the larger promise of scattering compensation. At a depth where photons are scattered three times on average, scattered light already exceeds ballistic light by roughly a factor of 20. In two-photon microscopy this can translate into a signal loss of about 400. Even partial recycling of the scattered light field could therefore open imaging depths that are inaccessible to purely ballistic approaches.
Figure 1. Left: Working principle of a deformable membrane mirror. A thin, flexible metal-coated membrane is shaped by electrostatic or magnetic forces applied via an array of actuators. In the example shown, a voltage V is applied to a single electrode, causing the membrane to“ dent” inwards. Right: Principle of correcting optical phase aberrations. A turbid medium alters the paths of light rays, causing the image to blur. A dynamic SLM can correct the ray directions.
ADAPTIVE OPTICS FOR LABEL-FREE CONFOCAL RAMAN MICROSCOPY Raman scattering offers chemical contrast without fluorescent labels. When a sample is illuminated with a single wavelength, a small fraction of photons is inelastically scattered and loses energy to molecular vibrations. The resulting spectrum acts like a chemical fingerprint. The effect is very weak, but it can identify molecular composition without perturbing the sample with fluorescent dyes.
Confocal Raman microscopy combines this spectroscopy with raster scanning. At each sampling position, Raman-shifted light is collected through a confocal aperture, often an optical fibre, and analysed by a spectrometer. The output is a multi-dimensional data set in which each voxel contains a full Raman spectrum. AO must handle two directions: the excitation beam has to be pre-compensated on its way in, and the broadband Raman signal has to be corrected on its way out.
An achromatic wavefront element is therefore useful. In our recent work we used a Deformable Phase Plate( DPP), an optofluidic transmissive modulator that can be attached to a commercial Raman microscope [ 4 ]. The DPP shapes the wavefront by locally deforming a fluid layer between a transparent electrode array and a flexible membrane. Its transmissive geometry makes it practical where a reflective deformable mirror would be difficult to integrate.
The second challenge is wavefront sensing. Confocal microscopy usually lacks a bright point-like guide star, because the signal arises throughout the focal volume. We addressed this by using indirect wavefront sensing: a series of test wavefronts is applied, and the correction is optimized with respect to a chosen quality metric. For Raman imaging, the metric can be spectral rather than spatial. In the example shown in Fig. 3, a red blood cell acted as a“ spectral guide star”: maximizing a narrow haemoglobin band
52 www. photoniques. com I Photoniques 140