Study Of Brain-wide Neuron Tracing In Golgi-and XFP-labeled Mouse Brain

The complex functions of brain fundamentally depend on the neural network constructed by hundreds of millions of neurons. Generally, specific brain function depends on specific neural circuit, which contains local connections and long-distance connections. Electronic and traditional light microscopy can only acquire small, high-resolution local connnections. Magnetic Resonance Imaging can only acquire coarse, long-distance connections. In order to get a overall understanding of brain structure, it’s necessary to get large-scale and precise neural network information, which has always been a major challenge in neuroscience community. Until recently, a new imaging technique called Micro-Optical Sectioning Tomograpy made great breakthrough for the first time, reporting the first high-resolution datasets of Golgi-staining and fluorescence-labeled whole mouse brain. These basic datsets make it possible for people to do research on precise neural circuit in whole brain. However, these datasets cann’t be used by neuroscientists directly. Quantitative anaysis cann’t be used to reveal the relation between structure and function until neural network structures are extracted from these datasets by neuron tracing.In the last decades, considerable progress has been made on neuron tracing technique. But most of them could be suitable only for their specific datset without general application. The datasets above show their unique characteristics, such as large amount of dataset, the hollow somas and dense dendrites in Golgi dataset, and the long-distance neuron projection in fluorescence dataset. These unprecedented challenges make it urgent to develop accurate and efficient neuron tracing method for these new datasets.In this thesis, we establish effective neuron tracing method for our dataset, and do some quantitative analysis on the tracing result to address some biological problems.(1) Local neuron tracing. For Golgi dataset, an enhanced2.5D morphological method is provided for soma location; a new gradient-based Rayburst algorithm is presented for the morphological detection of hollow soma’s irregular surface; via improving the3D interactive performance of OpenSnake tracing algorithm, and optimising tracing results based on neuronal connection, a accurate and semi-automatic method for neuron tracing from Golgi datasets is proposed by integrating our soma detection algorithm. The accuracy for soma localization and soma surface detection is greater than93%and84%, respectively. For neurite tracing, the accuracy of automatic tracing is more than86%, and the average deviation of traced centerline is about0.5μm.(2) Long-distance neuron tracing. For fluorescent mouse brain dataset, we firstly provides the efficient solution based on Amira software to long-distance neuron tracing. Compared with traditional and inefficient (several days) block tracing and stitching, our method only consumes2to3hours to trace a single neuron projection.(3) Quantitative analysis on the morphologies of neuronal in different brain regions. Based on our local neuron tracing method, several pyrimidal neurons in different cortex layers were traced and some morphological features such as soma surfce area, volume and distribution of neurite surface are analyzed. The results show these neurons are embedded with certain differences on these morphological features. These differences might have laid a great foundation for the research on neuron classification.(4) Location of long-distance neuron projection at nuclei level. The dataset of sparsely fluorescence-labeled mouse brain (Postnatal day14, P14) was used for long-distance neuron tracing. Since no brain atlas for P14mouse is available now, the P14and adult mouse brain datasets from MRI imaging are chosen as reference for registering all traced long-distance neurons projection to The Mouse Brain in Stereotaxic Coordinates atlas regionally, thus realizing the location of long-distance neuron projection at nuclei level. The results not only confirm some previous report, such as the classic bidirectional projection in the somatosensory pathway, but also highlight some unreported but putative projection pathways, such as the projection from DpG to RtTg. these new findings will provide reference for the research on brain-wide information processing functions.