FOCUS
Smart illumination for 3D-imaging of biological tissues
Figure 2. Confocal principle. The excitation light( blue) is focused to a spot that scans the sample. Fluorescence( green) passes through a small pinhole placed in a plane optically conjugate to the focal plane, so emission from above or below focus is blocked( dashed rays), producing a slice with much less out-of-focus background.
present complementary strategies that address these limits. Firstly, 3D Random Illumination Microscopy( 3D-RIM) uses speckled excitation and variance-based reconstruction to recover super-resolved, optically sectioned volumes while reassigning photons to their correct depth. Secondly, an extended-depth-of-field( EDF) mode compresses the volume onto a single exposure to capture fast dynamics with a much lower readout burden. Lastly, smart illumination allows to reduce the light dose impinging on the sample by focusing the illumination patterns only where it is required.
3D-RANDOM ILLUMINATION MICROSCOPY An alternative to point-scanning is Structured Illumination Microscopy( SIM, [ 2 ]): rather than scanning a focus, the whole field is illuminated with periodic patterns and several images are recorded with shifted phases and orientations. The patterned light down-modulates within the microscope’ s passband the information otherwise beyond the resolution limit. These images are combined numerically to yield a synthetic image with twice better resolution. In 3D-SIM, the pattern is also modulated along z, which improves axial as well as lateral resolution and provides sectioning for 3D imaging. The main drawback of 3D-SIM is that the reconstruction technique requires the knowledge of the illumination patterns inside the specimen. In embryos and other thick, heterogeneous tissues, aberrations and scattering distort the patterns in ways that are hard to measure or predict, so conventional 3D-SIM can produce artifacts or fail at depth.
In recent work [ 3 ], we proposed Random Illumination Microscopy( RIM) as an alternative structured illumination microscopy technique that avoids the knowledge of the excitation patterns. The sample is illuminated with a sequence of random speckle patterns and imaged with a camera at gentle intensity. Instead of using the patterns value at each point in the tissue, RIM reconstruction procedure exploits the statistics of the speckle ensemble. It extracts a superresolved image from the variance of the multiple low-resolution speckled images using a variance matching inversion scheme. The RIM approach requires only the knowledge of the speckle short-range correlation and the observation point spread function. In addition to a two-fold resolution gain which has been mathematically demonstrated and experimentally confirmed [ 4,5 ], it provides numerical optical sectioning. The latter can be simply explained by first noting that the speckled illuminations form random bright grains throughout the sample volume. The fluorescence light emitted by the in-focus excited fluorophores yields sharp spots at the image plane, while the light emitted by the out-of-focus fluorophores yields large smears. When the speckled patterns are changed, the fluorescence coming from in-focus structures undergo large intensity fluctuations, whereas the blurred
Figure 3. 3D-RIM. Left: a 3D structure( grey) is imaged with a speckle, which appears as a random distribution of bright spots( blue). The imaged plane appears as a dashed line. We focus on a small region of interest( black box). Right: In that small box, the imaged structure is excited by a speckle grain above the imaged plane, thus contributing a blur in the imaged plane. A 2D approach( confocal or 2D RIM, bottom right) removes this blur because it has no knowledge of the source above the imaged plane. A 3D approach( top right) will reassign the out of focus photons to their source through a deconvolution.
38 www. photoniques. com I Photoniques 134