Research Of Super Resolution Microscopy Through Back Propagating Errors Structured Detection

Optical microscopy has always played an important role in the field of industrial precision measurement and bioimaging,especially in the field of bioimaging.The development of molecular fluorescent probe-specific labeling technology makes optical microscopes become more and more irreplaceable,as its advantages of non-contact,non-destructive and providing dynamic real-time observations are difficult to be replaced by other non-optical microscopicg measuring methods.However,the well-known optical diffraction limit severely limits its spatial resolution.Therefore,improving the spatial resolution of optical microscopes is an urgent need in many frontier fields such as industrial precision measurement,precise production,bio-pharmacy,and live cell imaging.The research of super resolution microscopy through back propagating errors structured detection is mainly aimed at the field of bioimaging and is aimed to achieve super resolution of bio-fluorescence samples by structured detection methods.The structured detection method can increase the lateral resolution of the wide field system by 2 times without changing the structure of the existing laser scanning microscopy system.Compared with the wide field structured illumination microscopy,this structured detection method realized in the detection arm of the laser scanning microscopy system has chromatographic ability.Also it can obtain the resoluti on of short-wave illumination under long-wave illumination condition.At the same time,the longer excitation wavelength means greater detection depth,less phototoxicity and photobleaching,which help to realize biological cell imaging in a better way.However,developed from structured illumination microscopy,the traditional structured detection technology still need to use the cosine function to detect and process spectrum reconstruction.But the actual optical microscopy system is not completely ideal,and any device parameters or positional deviation will lead to the difference between detection function and theoretical function.In addition,spectrum reconstruction process will bring out new errors,resulting in a reduction in the resolution of the structured detection super resolution imaging.In addition,the structured detection technology must first collect the spot image point by point through CCD/CMOS,and then realize the“virtually”detection and image reconstruction by digital image processing in computer.The point-by-point acquisition mode of the original spot image makes imaging speed of the traditional structured detection technology extremely low,far from meeting the needs of fast imaging.In this study,the structured non-uniform detection theory in the detection arm of laser scanning microscopy system is analyzed,and it is proved that the structured detection method in the laser scanning microscopy system and the structured illumination method in the wide field microscopy system also have the band-based expansion ability.Then a super-resolution model of structured detection based on single neuron is established,and the structured detection function is solved by error back propagation algorithm.In addition,this research focuses on the limitation of imaging speed of the original image acquisition method of traditional structured detection technology,and studies the rapid imaging method of array parallel structured detection based on the principle of spatiotemporal modulation,which significantly improves the structured detection imaging speed while achieving super resolution.In this research,the following aspects are analyzed:(1)For the optimization problem of structured detection function,the structured detection super resolution microscopic imaging method based on the principle of back propagating errors is proposed and studied.Based on the structured detection super resolution theory of laser scanning microscopy system,a super resolution model of structured detection based on single neuron is established.Based on this structured detection function,the super resolution image is directly reconstructed by structured detection process of the detected spot image only for one time.Also the multiple detection and spectrum reconstruction process of the traditional structured detection method is eliminated.The error back propagation method is introduced in the structured detection microscopy system for the first time and the function optimization theory is established correspondingly based on this method.The deviation between the sample reference image and the actual structured detection and reconstruction image realizes the adjustment of the weight in the structured detection function by back propagation.This method solves the problem of resolution degradation caused by the error of the cosine structured detection function.The high-performance structured detection function calculating software based on Visual Studio MFC was written.The raw sample data were provided by the simulat ion of laser scanning microscope.And this simulation data proved that the lateral resolution of the system was doubled,compared with the wide field microscope system.(2)To improve the image acquisition efficiency in traditional structured detection technology,an array-based parallel structured detection imaging method is proposed and studied based on the principle of spatiotemporal modulation.In laser scanning microscopy system,CCD/CMOS is used to collect spot in laser scanning microscopy system.The image,based on the principle of spatiotemporal modulation,modulates the intensity of the light source in time and spatially scans the position to form an array spot image.Through the high-speed modulation of the laser diode and the scanning galvanometer,the time division multiplexing of the low-speed camera is realized,and the high-speed array spot image acquisition is achieved.In the experiment,the N×N acquisition array is set to reduce the original total number of collected images to 1/N~2.What's more,by reducing the number of image acquisitions,the data acquisition speed can also be improved without increasing the system components,reducing the scanning accuracy,or losing the resolution.And the imaging speed of the structured detection method is significantly improved.Finally,the array-based parallel structured detection imaging system was built based on the spatiotemporal modulation,and the super resolution imaging test was performed on the standard resolution test pattern.The lateral resolution improvement of more than 1.6 times was obtained.In the study,super resolution imaging tests were performed on fluorescently labeled human cervical cancer cells(HeLa cells and SiHa cells),resulting in an approximately 1.8 times increase in lateral resolution.Based on the principle of error back propagation,the effectiveness of the structured detection super resolution method was verified by the imaging of bio-fluorescence samples,and the rapid structured detection super resolution imaging was realized by the laser scanning microscopy system based on spatiotemporal modulation.