Research On TEM Simulation And Three-dimensional Reconstruction Of Biological Images

Electron tomography is an important method to obtain internal structure of the object.A series of two-dimensional projection images were taken through electron microscope at different angles. Then these images are used to reconstruct the object by using some reconstruction algorithms. Iterative reconstruction algorithm is one of the well-known technologies using in three-dimensional reconstruction recently. It usually solving the equations by iterating until it presents a converged state. But some reconstruction results is always not clear on the marginal structure, besides, those results often contains the noise which will influence the analysis of biological structures. The method used to evaluate the reconstruction algorithm currently is generally carried out by calculating the projection error or compared with phantom. However, the phantom constructed by known methods always have large difference with real biological samples, so the range of applications of the phantom is limited.In this paper, We have proposed a three-dimensional reconstruction method based on simultaneous algebraic reconstruction algorithm, which have added the improved nonlinear anisotropic filtering with adaptive gradient threshold parameter into the reconstruction process. The filtering process has been parallelized to improve the computation speed by the using of mpi. Experimental results have shown that this reconstruction method can effectively improve the image effects and image quality.Besides, for the reason that the phantom constructed by the known methods are not sufficiently close to the actual biological samples, we have designed a virtual electron microscopy molecular modeling simulation system based on serial block face scanning electron microscope image(SBEM). Firstly, we need to prepcoess the series of SBEM images; Secondly, we will simulate the warping effect of samples which is generated by electron beam bombardment, and then can adding the virtual gold particles int the phantom. Finally, the phantom models obtained in the previous steps need to be applied in spiral projection transformation model, resulting in a series of virtual projection images at different angles. Experimental results have shown that both the phantom and simulated projection images constructed by our method is more similar to the real experiment images.We have also used the virtual projection images into the iterative reconstruction method proposed below, and the results further validates the utility and practicability of the electron microscopy simulation system.