High Fidelity Image Reconstruction of Optical Sectioning Structured Illumination Microscopy

In the research fields such as biomedicine and material science, researchers need to observe the Three-dimensional (3D) structure of samples. This promotes the development of 3D optical microscopic techniques, such as Laser Scanning Confocal Microscopy (LSCM), Light Sheet Fluorescence Microscopy (LSFM), Optical Sectioning Structured Illumination Microscopy (OS-SIM). Among them, OS-SIM has the capability of extracting the in-focus target information from the out-of-focus background of the sample to enable 3D optical imaging. The quality of the optical sectioning image is directly related to the reconstruction algorithm. Although the traditional RMS algorithm is simple, the reconstructed image is often poor when the signal-to-noise ratio and the fringe contrast of the original image are not high, and the 3D reconstructed image is not ideal. To overcome the deficiencies of the RMS algorithm, a number of methods have been proposed, such as the Fast and Adaptive Bi-Dimensional Empirical Mode Decomposition-Hilbert Spiral Transform (FABMED-HS) method, Sequence Hilbert Transform (SHT) method, Fourier-OS-SIM method. All these methods provide different ideas for realizing 3D microscopic imaging. In this paper, we propose a new method, which can obtain high fidelity optical sectioning images. This method combines background removal and deconvolution processing, and finally obtains the optical sectioning image using standard deviation operation. Compared to the traditional RMS algorithm, the proposed method can effectively reduce the residual fringes and improve the visibility of minute details. Even in the low contrast of structured illumination where the RMS algorithm works abnormally, the STD algorithm can still perform well. Because the reconstruction formula of this method is similar to the standard deviation formula, the proposed method is named" STD (Standard Deviation) algorithm". Experimentally, a Digital Micro-mirror Device (DMD)based structured illumination microscope is built. In this microscope, a Laser Diode Illuminator(LDI) is used as light source that provides illumination of seven wavelength channels. The DMD has a resolution of 1 920x1 080 pixels, with a pixel size of 7.56 mu mx7.56 mu m. The SCOMS camera has a resolution of 2 048x2 048 pixels, with a pixel size of 6.5 mu mx6.5 mu m. Firstly, we compare the reconstructed images of STD algorithm and RMS algorithm using mouse kidney cells and Bovine Pulmonary Artery Endothelial (BPAE) cells as samples. The experimental results demonstrate that RMS algorithm has better optical sectioning capability. The STD algorithm is also applied to previously collected data with a mite as the sample. The experimental results again suggest the robustness of the STD algorithm. And then, we find that changing the illumination wavelength has little effect on the imaging position, which makes it possible to optical sectioning at multiple wavelengths. We obtain dual-wavelength fluorescence images by using 470 nm and 555 nm to excite the mouse kidney cells samples. Finally, 3D imaging experiments is performed with pollen samples. The field-of-view of the image is 163.84 mu mx163.84 mu m. We took 125 layers of images, each thickness is 200 nm. Using the STD algorithm, we get sharp 3D images. All the above experimental results demonstrate that the STD method can obtain better optical sectioning 3D images compared to the RMS method.