FOCUS
WAVEFRONT control
often take hours. We observed signal improvements of about 50 %, which can translate into a comparable reduction in acquisition time.
DIGITAL ADAPTIVE OPTICS IN SINGLE MOLECULE LOCALIZATION MICROSCOPY Single molecule localization microscopy( SMLM) reaches nanometre-scale resolution by separating fluorophores in time rather than space [ 5 ]. The underlying concept was recognized with the 2014 Nobel Prize in Chemistry. A sample is prepared so that only a sparse subset of molecules emits in each camera frame. Each isolated molecule image is then compared to a computational point-spread-function( PSF) model and localized with precisions of tens of nanometres. The thousands of coordinates of individually localized molecules then form the super-resolved image.
Figure 4 illustrates this principle for nuclear pore complexes in human bone cancer cells( U2OS). Nuclear pores are small gateways through the nuclear envelope. When labelled, they appear as two parallel rings separated by about 50 nm, each measuring about 100 nm across and containing eight bright spots corresponding to labelled nucleoporins.
Aberrations affect SMLM in two related but distinct ways. First, if the PSF fitting model does not include them, fewer molecules are detected and systematic localization errors appear. This part of the problem can be corrected computationally if the aberration is known, an approach known as digital AO. However, aberrations also cause a real loss of information that cannot be regained digitally. Restoring localization precision therefore requires physical wavefront correction.
SMLM offers a remarkably photon-efficient way to measure aberrations. Each emitting molecule is, from an optical point of view, an image of the PSF. Although each single molecule image is faint, an SMLM data set contains thousands of such pointsource images. Their collective shape therefore encodes the system aberration. By extending the usual molecule-fitting model with a small set of wavefront parameters, the aberration can be retrieved directly from the data [ 6 ]. Compared with iterative optimization, this reduces bleaching and can be implemented more quickly.
CONCLUSION We showed that AO in microscopy is not a single add-on that can be transferred unchanged from one modality to another. In MP microscopy, the nonlinear signal provides a strong route to deep-tissue correction. In Raman microscopy, a spectral feature can serve as a guide star for label-free chemical imaging. In SMLM, the raw data themselves contain thousands of pointsource measurements that enable wavefront sensing. These examples show why AO in microscopy must be tailored to the contrast mechanism, photon budget and sample.
Today, most biological AO applications still compensate relatively smooth aberrations. The larger ambition is scattering compensation: measuring and correcting complex wavefronts after multiple scattering events, so that light that would otherwise be lost can form an image. This remains difficult because living samples change, fluorescent markers bleach and complex aberrations are demanding to estimate.
A second limitation is the size of the isoplanatic patch, the region over which one correction remains valid. As corrections become more complex, this region can shrink to a few micrometres or less. Extending the corrected field of view while preserving deep compensation is therefore one of the central challenges for the next generation of AO microscopy.
Figure 3. AO in confocal Raman microscopy.( a) Principle of Raman scattering and a sketch of an AO-equipped confocal Raman microscope with a DPP mounted to the objective. The drawing on top illustrates the DPP working principle.( b) Raman spectra of a red blood cell( red) and mouse brain tissue( blue) at 30 ยต m depth. AO almost doubled the haemoglobin signal and improved the brain tissue signal in close vicinity by up to 50 %. Parts of the figure are adapted from [ 4 ] under the CC BY 4.0 license.
54 www. photoniques. com I Photoniques 140