The Design And Implementation Of Medical Image Data Acquisition And Standardization System

In order to solve the problems of uneven distribution of medical image resources and inadequate sharing of medical data in China,a medical image data acquisition and standardization system is designed and implemented in this paper in the context of hierarchical diagnosis and treatment.Firstly,it starts from the study of medical image on its remote retrieval and standardized display and browsing,to realize that the medical images can be accessed at anytime,anywhere and at any terminal.Furthermore,it studies the standard methods of image data acquisition,annotation and other preprocessing methods to provide support for the preparation of high-quality annotation data for deep learning.In particular,aiming at the problem of radiologist shortage while the rapid growth of medical image data,an artificial intelligence auxiliary diagnosis module which can be invoked in the diagnosis process is designed to improve the efficiency and level of a doctor's diagnosis.The main work completed in this paper includes:1.A remote image diagnosis module is designed for the remote retrieval of regional image data.Canvas drawing technique is utilized on the standard display and browse for medical image.Meanwhile,the accessed DICOM file is stored on the sever in the standardized structure by the back-end,and the file structure information is also saved in the configuration file.The multi-threading Webworkers is adopted in the front end,to realize the rapid access to hundreds of medical images in a precise and orderly way,enabling the speed of image retrieval increased by 3 times.2.In view of the lack of high-quality annotation data in deep learning algorithm,which leads to the limited performance of artificial intelligence auxiliary diagnostic products,a method combining the medical image diagnosis and the annotation is designed,so that doctors can obtain annotation data at the same time of diagnosis.It provides multiple annotated methods such as rectangle,polygon and free drawing,to support the image annotation collection,visual loading,multi-disease preservation,data export and other functions.In the process of image diagnosis,the physician user only needs to annotate and save the lesions in two steps,then the image annotation can be completed conveniently.3.Aiming at the serious shortage of radiologists compared with the needs of imaging examination,an artificial intelligence auxiliary diagnosis module is designed which can be invoked in the diagnosis process.Through image caching,container isolation deployment,high-performance message queue and other technologies,the detection module of deep learning model can be invoked in parallel,which can increase the detection speed by more than 2 times,and also supports the expansion of deep learning model types.The medical image data acquisition and standardization system in this paper,including remote access of medical images,standardized acquisition,annotation,auxiliary detection of lesions and other functions,can solve the problem of uneven distribution of regional medical resources and the lack of high-quality annotated data to a certain extent.Meanwhile,the solution of intelligence auxiliary diagnosis proposed in this paper can effectively promote the development of medical imaging field toward a more efficient,more intelligent and more accurate direction.