Study On Image Pre-processing Of Micro-optical Sectioning Tomography

To figure out how brain works is a critical challenge to modern scientific research,and the neuroanatomical architecture is the structural basis for researches on brainfunction and dysfunction. Optical microscopy has provided massive and complexanatomical architecture image data sets for biological research, especially for the researchon brain science by virtue of the optical microscopic tools with the invention anddevelopment of the optical microscopy with large field and high resolution. To deal withthe massive and complex data sets obtained from the whole brain imaging with highresolution, automatic imaging processing methods is urgent and necessary for imageanalysis and information extraction. Due to the sample preparation and imaging method’sdefects, the acquired image quality decreased by suffering from various artifacts, thatraised new requirements for the image processing method of the massive and complexneural anatomy data sets.Based on the5TB and8TB data set of mouse brain with Nissl and Golgi stainingacquired by the Micro-optical Sectioning Tomography system respectively, this paperproposed a set of automatic artifacts removal method for images obtained by tissuestaining and optical microscopy. This method has demonstrated to get great effects on thecomplicated images containing various structures including both cells and blood vessels.Firstly, remove strip noise on sections by moving medium filtering. Secondly, takeadvantage of the extracted brain contour as mask to improve the accuracy of the extracteduneven background, and fix the uneven staining phenomenon. Thirdly, use morphologicalfiltering to remove the irregular plaque in the image. Finally, unify the inter-slice intensitydistribution by taking the extracted background as reference.Two whole brain image data sets are processed in this paper. After processing withthis method, the image brightness became uniform in all regions and the image quality improved significantly. The corrected image data set could demonstrate thecytoarchitecture, vascular topology and cell morphology of the whole brain with highresolution. Furthermore, the processed image data set with uniform brightness and highquality could be a fundamental atlas for the following image analysis as cell segmentation,vascular tracing and cellular morphology detection, and provide reliable basic data sets forthe neuroscientists to reveal the brain function mechanism.