| The brain is a complex organ with multiple structures.Deciphering the connections of brain network consisting of neurons,glial cells and blood vessels is essential for studying brain function.In recent years,several whole-brain optical imaging techniques have provided various solutions to resolve the three-dimensional fine structure of the brain.Among them,the high-definition fluorescence micro-optical sectioning tomography(HDfMOST)technique developed by the applicant's research group is capable of obtaining highresolution whole-brain data sets.Due to its excellent data quality,it is expected to become an important tool for brain science research.However,the continuous data acquisition of a single mouse brain for more than 100 hours and the need to image thousands of samples for brain atlas rendering place high demands on the stability,reliability and efficiency of the long-term operation of HD-fMIOST.To meet this demand,this paper has carried out systematic research and development work on the workflow of data acquisition,data storage optimization,stability optimization of the data acquisition and control software during the development of the engineering prototype of HD-fMOST.In order to accelerate data acquisition,this paper proposes a workflow where the image compression and storage run in parallel with image acquisition.Starting from the data acquisition control of the system,the process of acquiring the strip image data is analyzed and a feasible scheme of process optimization is designed.The speed-up effect is calculated theoretically,along with a specific optimization effect of different strip images size.Finally,the result concludes that the optimization scheme has important practical significance for data acquisition of large volume biological samples.Three aspects of the system are studied in terms of two-dimensional storage format,reduction of local data volume and three-dimensional storage format,and the storage optimization strategy is proposed.The usage and format of the system's 2D data storage are analyzed.LZ4 format with higher compression and decompression efficiency is chosen for storing the native resolution stitching image.A method of automatic modification imaging range and a cytoarchitecture images down-sampling scheme are designed and implemented to reduce redundant data storage.The 3D data real-time conversion module is designed to avoid the time-consuming format conversion after data acquisition for the 2D image data is not conducive to characterize 3D biological sample information.For factors affecting the stability of data acquisition,this paper proposes corresponding processing schemes.For camera's image capturing exception,the processing solution of exception catching and restarting strategy is proposed.As for the image dislocation caused by camera memory error during long-time acquisition,a solution of regularly releasing camera memory is designed and implemented.For the problem of preview image loss,the solution to monitor the preview image generation and reset the preview image stitching thread automatically are proposed.The recovery scheme of acquisition parameters after data acquisition exception is designed and implemented,which is used to ensure the spatial consistency of 3D data set after restarting data acquisition.In summary,this paper optimizes the data acquisition control software of the engineered HD-fMOST from the perspective of data acquisition.The software can run efficiently and stably for large-scale three-dimensional fine structure data acquisition of the mouse brain,which is of great practical significance for systematic and large-scale wholebrain optical imaging. |
PDF Full Text Request