Study On The Registration Method To Align 3D Fluorescence Images Of TDI-fMOST Mouse Brains

Neurons are the basic structural and functional units of the brain and are the key to the brain research.Time delay integration-fluorescence micro-optical sectioning tomography(TDI-fMOST)is a microscopic optical imaging system independently developed by our laboratory.Combining the chemical reactivation of fluorescent proteins,high-precision,high-throughput biological tissues imaging provides a wealth research material for neuroscience research.To study the complete morphology of a single neuron and its projection by mouse brain fluorescence imaging is one of the directions of our laboratory research.In order to facilitate the tracking and analysis of individual neurons,a sparse-labeling strategy was adopted in the mouse brain sample.But the sparse-labeling resulted in insufficient number of specific types of neurons of interest in individual mouse brains,so the in-depth morphological and projection analogy analysis was not easy.Therefore,it is very important to register 3D mouse brain fluorescence images and map the labeled neurons of different mouse brains to the same reference mouse brain for corresponding visualization and analogy analysis.In this paper,a complete set of 3D mouse brain fluorescence image registration method is designed for TDI-fMOST system datasets: firstly image preprocessing of 3D mouse brain fluorescence images,including image downsampling and projecting,manual removal of impurity interference signals,Rolling Ball background normalization and so on.Based on the similarity measure of mutual information,a linear global rigid registration of the mouse brain fluorescence image was performed based on the contour extraction,and a nonlinear local detailed registration of the mouse brain fluorescence image was performed based on the Demons symmetric diffeomorphisms algorithm.At the same time,this paper objectively evaluates the registration quality of 3D mouse brain fluorescence images from the perspective of informatics and biology.First,we use similarity measure analysis to analyze the image similarity before and after registration,and then use the regional proportion method to quantitative evaluate the quality of segmentation of significant brain regions.On the basis of the above,the parameters of image registration are used to map the same type of neurons in different brains into the same reference brain,which provides a basis for further neuronal visualization and analog analysis.Through the research of this paper,a set of feasible 3D mouse brain fluorescence image registration method has been successfully designed.This method can achieve better registration of the brain’s outer contour and significant brain regions,and it has been applied in the neuron mapping and 3D visualization analysis.