| Organs are important functional units of higher animals.The study of organ structure helps us understand the functions of organs and supports related disease research.Using the microscopic optical imaging technology developed in recent years,three-dimensional imaging of cell resolution at the organ level can be achieved.Based on high-resolution threedimensional data,digital slicing technology can obtain slices of the whole organ with cell resolution at any angle.However,a single set of data acquired by such high-resolution imaging technology can reach terabytes or even more than 10 TB,which brings challenges of big data to digital slices.Existing digital slicing method tools can only handle data of gigabytes.Obtaining slices of more than 10 TB data and shortening the processing time are problems that need to be solved urgently.Based on the fine-grained data block organization format,this paper solves the two key problems of redundant I/O and the calculation of batch3 D data digital slices,and a navigation slice tool for 3D images of complete organs with sub-micron resolution is further constructed.The details are as follows:An optimized blocking storage format for 3D data slicing is achieved by using a finegrained blocking strategy.The 3D big data is divided into fine-grained small blocks,which solves the problem of loading too much redundant data in the existing large-size blocking storage format.For massive small data files generated by fine-grained blocking,combined storage is used to reduce the number of files.Based on the fine-grained blocking storage format,the amount of data loaded in the slicing calculation process is reduced by 87%,and the slicing performance is increased to more than 2 times.Combined with JPEG compression and contour constraint optimization,the amount of data loaded in the slicing process is further reduced by more than 70%,and the slicing performance is improved to about 7 times.A high-throughput parallel 3D image digital slicing calculation method is established,which solves the problem of batch tasks of 3D block slicing.This method uses a highperformance parallel method to construct a thread-level distributed parallel digital slicing method,which improves the efficiency of parallel digital slicing distributed on multiple nodes.Use the addition stitching method,the complete slice image of big data can be obtained by stitching all the slice units of multiple small data,and the correctness of the slicing method was verified on the model data and biological image data.In terms of performance,this slicing method can complete the slice acquisition of three-dimensional images with cell resolution of 4.8 TB in 5 seconds.On the basis of theoretical research,a three-dimensional image interactive digital slicing system was further constructed.The system can conveniently and intuitively locate the target angle and position in the three-dimensional image big data,and obtain the highresolution slice image of the whole-organ in the corresponding orientation.It has been used in the application of obtaining standard azimuth slices and multi-angle anatomy research,showing the ability of generating multi-angle slices of large-volume three-dimensional image data.In summary,this paper uses a fine-grained block organization storage solution and a high-performance parallel computing method to reduce data loading,speed up slicing calculation.Finally,an efficient digital slicing method for three-dimensional image data of cellular-resolution is created,and an interactive slice tool is provided to quickly obtain sliced images of the target orientation. |
PDF Full Text Request