Study On Fourier Ptychographic Microscopy With High Fidelity

With the development of microscopes,we can explore the microscopic world,gain insight into the root of disease and understand the cellular mechanism.The capabilities of optical microscopy imaging systems have been continuously updated as research progresses,and important breakthroughs have been made in the field of high-resolution(HR)imaging.However,because the imaging mechanism has not undergone a fundamental change,the contradiction between large field-of-view(FOV)and HR imaging has not been overcome to meet the needs of observation and detection of samples with large FOV.Therefore,the ability to simultaneously achieve high resolution,large FOV and quantitative phase measurement is a goal that is still being pursued in the development of optical microscopes.Fourier ptychographic microscopy(FPM)is a recently developed quantitative phase imaging method that overcomes the physical spatial bandwidth product limitation of low numerical aperture(NA)imaging systems,sharing the roots of synthetic aperture concept and other phase retrieval methods.FPM replaces tungsten or halogen light source with a light emitting diode(LED)array,and effectively avoids the return error and hysteresis effect caused by mechanical scanning.A low magnification objective lens is used to acquire low-resolution(LR)images with different angles of illumination,then the spectrum is computationally synthesized in the Fourier domain,and finally a complex amplitude image with large field of view(FOV)and high-resolution is reconstructed.Although significant progress has been made in the imaging performance of FPM in recent years,there are still some key issues to be solved,such as errors in the simplification of the optical transfer function,reconstructed distortion of edge region,insufficient sampling in central low-middle frequency region brought about by periodic LED array illumination,and low data acquisition efficiency.To address the above issues,the thesis has carried out the following research:1、The current research status of FPM is reviewed in detail,and the imaging principle and reconstruction algorithm of FPM are introduced in detail.The reconstruction algorithm includes the method based on iterative phase restoration and the method based on deep learning.The requirements of sampling rate and spectral overlapping ratio are discussed,and the coherence and imaging resolution of the FPM imaging system with LED illumination are analyzed.In addition,a FPM experimental system based on LED array is designed and built.Quantitative phase imaging with a large FOV and high resolution with 0.68 NA was achieved using a 4×/0.1NA microscopic objective.The performance of FPM in imaging resolution enhancement,quantitative phase measurement,and phase distortion correction was verified experimentally using resolution targets and biological samples.2、Research on FPM method with double-height illumination was carried out.The double-height illuminated FPM(DHI-FPM)method is further developed to increase the sampling rate of the central low-frequency region,and the corresponding reconstruction algorithm is optimized.The LED array is moved from one plane to another to achieve the sample illuminations at two different heights and the lowresolution intensity images are totally utilized to reconstruct the high-resolution complex amplitude image.In addition,the physical insight of the method is analyzed from the perspective of data redundancy,as well as the rule for selecting the two illumination heights.The effectiveness of the method in noisy environments is also demonstrated by simulation and biological sample experiments.Finally,the imaging efficiency of the DHI-FPM method is further discussed.3、The study of FPM with high fidelity is performed.The coherence of FPM systems based on LED array is an analyzed,a Butterworth function weighted transfer function(B-TF)is proposed as the frequency domain constraint in the reconstruction process.Simulations and experiments demonstrate that this method can better fit the the actual FPM physical model and improve the quality of the reconstructed image.Besides,a wave vector correction method is proposed to solve the problem of inconsistent illumination angles in the full FOV caused by spherical wave illumination,which significantly reduces the artifacts in the reconstructed image in the edge region of the FOV.Finally,the FPM imaging method based on deep learning is studied.A model based on U-net was built,after training,only one LR image is needed to obtain a high-resolution image,which greatly reduces the amount of data and improves the reconstruction performance.