Structured Illumination Microscopy (SIM). Theory, Implementation and Conmparison of TIRF-SIM, SROS-SIM and MSIM

Abstract: One of the most important tools for biomedical research is fluorescence microcsopy. Selective labelling of molecules enables the observation of process kinetics within living cells but is limited due to diffraction to a maximum resolution of half the wavelength in use. To further increase resolution it is possible e.g. to use structured instead of homogeneous illumination light. The spatial frequencies of illumination and flurophore distribution will mix, thus encoding fine details of the structure, which would usually be lost, in observable spatial frequency bands. The so obtained additional information is computationally decoded afterwards to offer super-resolution. Depending on the particular structure of the illumination light the respective reconstruction algorithms vary a lot. In addition, the set-ups and acquisition procedures need to be adapted to generate the needed patterns and record the appropriate amount of different raw image data. Therefore, it is useful to compare different illumination geometries as they all have their strength, weaknesses and optimal fields of application. For example, sinusoidal illumination in a total internal reflection configuration (TIRF-SIM) allows extremely fast, high-contrast imaging with the theoretical possibility of unlimited resolution improvement via the use of non-linear effects (SSIM) but is restrained to two-dimensional imaging. In contrast, three-dimensional sinusoidal illumination (3D-SIM) requires more reconstruction and hardware efforts but is capable of removing out-of-focus light computationally and enables volumetric data at even higher imaging rates than 3d-SIM but lacks resolution improvement in the axial direction. Diffraction limited multi-focal spot pattern illumination (MSIM), similarly to SROS-SIM, does not offer super-resolution in the axial direction but enhances penetration depth and contrast and therefore is useful in thick sample imaging. To optimally compare these techniques, the standard reconstruction algorithms were implemented in Matlab and a SIM microscope was built that is switchable between TIRF-SIM, SROS-SIM and MSIM. It is therefore ideal to comparatively evaluate the respective performances regarding resolution, imaging speed, and reduction of out-of-focus-light. Additionally, a new kind of reconstruction algorithm was developed, joint Richardson-Lucy deconvolution for widefield MSIM, that improves reconstruction results from noise corrupted raw data.