Research On The Function-structure Of Individual Neurons And Establishment Of Imaging Data Optimization Method

Mapping the functions and morphologies of brain neurons is the core goal in modern neuroscience.In recent years,reconstructing axonal projections of single neurons at the whole-brain level,namely single-neuron projectome,is an emerging hot topic in the field.Up to now,thousands of single neurons from different brain regions have been reconstructed.However,the corresponding physiological functions of these individual reconstructed neurons are still unclear.Our goal is to establish a one-to-one map that contains both functional properties and complete axon projections of individual neurons.This is fundamental to understanding the logic of how function-relevant information transmits in the brain,but remains a major challenge.For monitoring the functions of single neurons,two-photon microscopy has been a powerful technique and used in many labs,however the recorded imaging data generally suffer from noises and motion artifacts.Currently various image restoration techniques have been proposed to enable the recovery of important biological information that has been subject to complex corruptions,thus aiming to push the imaging data quality to new heights.In addition,a number of image registration techniques have been proposed to enable the restoration of important biological information that has been subject to motion artifacts.However,only a few studies have developed efficient motion correction methods for real-time image processing.In the thesis,we first develop a novel method that allows for reconstructing the complete morphologies of single neurons that have been functionally identified.This method combines multiple challenging steps including two-photon Ca²⁺imaging in awake mice,two-photon-targeted plasmid electroporation and local viral injection,whole-brain dual-color imaging,and morphological reconstruction.Moreover,we propose a method for denoising two-photon Ca²⁺imaging data by model blind spatio-temporal filtering,and we propose a fast and accurate image density feature based motion correction method.The main results of this study are the followings:1.Reconstructing brain-wide morphologies of single neurons with defined functionWe developed a new method,2-photon imaging-assisted Single-cell Plasmid electroporation and Adeno-associated virus（AAV）injection for Reconstructing Single n Eurons,namely 2-SPARSE.We reconstructed the brain-wide morphologies of functionally defined single neurons by combining two-photon Ca²⁺imaging,targeted single-cell plasmid electroporation,local adeno-associated virus injection and whole-brain serial section and imaging.The major advantages of the method are that we performed precisely targeted labelling of single functionally defined neurons,and achieved a robust labelling of long-range axonal projections.The results demonstrate that 2-SPARSE successfully reconstructed brain-wide axonal projection from L2/3 in auditory cortex and from L5 in motor cortex（axonal length 212 mm）.The quality of its single-neuronal labelling is comparable to that of the currently prevalent binary AAV expression approach.The preliminary results also show that the neurons with defined functions have various axonal projection patterns.Therefore,this method will advance the brain connectome field into a single-neuron functional projectome stage and significantly extend the ability to precisely uncover the principles of the brain structure and function at single-neuron level.Importantly,this method can be readily implemented in a broad range of laboratories that have been equipped with a standard two-photon microscope and conventional electrophysiological devices.2.Spatio-temporal denoising two-photon Ca²⁺imaging data by deep neural networkWe propose a neural network combining spatio-temporal filtering and model blind learning to conduct image denoising.The performance of the proposed network was first validated with synthetically generated noisy two-photon imaging data,and then with real raw two-photon imaging data.The quantitative assessment of the restored two-photon imaging data showed that the imaging data quality was improved significantly in comparison to methods reported previously.The results demonstrate that this method is able to recover true signals in both spatial and temporal dimensions and enables the handling of complex imaging noises.Therefore,we provide an invaluable tool for denoising two-photon Ca²⁺imaging data that we hope will facilitate improved high-resolution monitoring of neuronal activities,thereby elucidating brain functions.3.Fast and accurate motion correction for two-photon Ca²⁺imagingWe propose a fast and accurate motion correction method for imaging neurons in the animal during behaviors.The performance of the proposed method was first validated with raw two-photon imaging data at both neuronal population and dendritic spine levels,then the quantitative assessment of the restored two-photon imaging data showed that the imaging data quality was improved significantly in comparison to methods reported previously.The results demonstrate that this method is able to accurately restore neuronal signals with a high computational speed on a conventional computer.Therefore,we provide a powerful tool for performing fast and accurate motion correction for two-photon Ca²⁺imaging data in behaving mice that we hope will facilitate online functional imaging and photostimulation experiment in the future.Therefore,the function-structure imaging method established in this study enable us to perform a precise one-to-one mapping of projection patterns and neurophysiological functions for individual neurons.The proposed denoising method provides an invaluable tool for denoising two-photon Ca²⁺imaging data by model blind spatio-temporal processing.At last,we provide a powerful tool to perform motion correction for two-photon Ca²⁺imaging data for on-line experiments.