Interactive System Of Medical Image Segmentation Based On Cloud Platform

The segmentation of medical image anatomical structure plays an important role in the auxiliary medical diagnosis.In the past 30 years,a large number of interactive 3D segmentation methods have been developed.At present,most interactive segmentation algorithms operate directly in the 3D data space and make use of all the spatial information of the image.With the standard results obtained by full manual loop-by-loop diagrams,an overlap ratio of more than 90% can often be obtained.However,it is difficult to improve the segmentation accuracy for the direct segmentation of 3D volume data.Especially for organ/tissue boundaries where the contrast is not obvious,although compared with the gold standard,only to increase the Dice value of less than 10%,but often need to be revised layer by layer.Therefore,the cumulative total time relative to full manual segmentation is not optimized.Essentially,many images,especially medical images,are in three or more dimensions.The segmentation algorithm can run in 3d environment,but because of the limitation of screen and mouse,it can only interact with the image in 2d environment and produce the result.In addition,two-dimensional image annotation can achieve arbitrary accuracy due to the user's overall control ability in two-dimensional.Secondly,most of the existing interactive segmentation algorithms can only run on single workstation.This requires the whole body data all located in this workstation,can not guarantee the image data security,easy to leak patient's privacy.In order to solve these problems,this paper first designs a low-dimensional interactive segmentation method to reconstruct high-dimensional segmentation.On the one hand,it can make full use of the whole three-dimensional or higher-dimensional volume data,but on the other hand,it also considers the complete controllability of the results of two-dimensional segmentation.If you can build a cloud platform in a health care facility for doctors to use,volume data is in the cloud,and Algorithms are in the cloud.This cloud platform distributes low-dimensional data to different clients when splitting is required.After the interactivesegmentation,the clients are merged into the cloud platform,and the algorithm is used to fuse the results of the low-dimensional segmentation.On the other hand,the same individual data can be sent to different clients,making it impossible for any client to reconstruct the whole data.This also brings better security for medical data security.Following the above idea,this paper studies the method of Interactive Medical Image Analysis based on cloud platform.The specific work includes the following aspects: firstly,this paper presents a method of low-dimensional interactive segmentation to reconstruct high-dimensional segmentation,the user marks the 2D section to guide the segmentation of Galway,and uses the registration method based on multi-map to register the 2D section to the rest of the section to get the segmentation of the whole high-dimensional volume data.Second,we will implement the Grow Cut interactive segmentation algorithm through Web GL in the browser pages to maximize the use of client-side hardware and software resources,and to minimize reliance on client-side software configuration.Thirdly,on the basis of the above algorithms and implementation,we implement a collaborative and interactive segmentation method based on cloud computing using Python's Web framework Flask.Finally,different parts of medical image data are segmented for different organs,and good results are obtained,and compared with other interactive segmentation tools in terms of execution time,the analysis shows that this tool has some advantages in the segmentation of complex organizational structure.The proposed algorithms and the construction of software architecture are of practical significance for collaborative interactive image segmentation and more data-secure medical image analysis and annotation.