The Design And Implementation Of Medical Image Annotation System For Deep Learning

In recent years,with the development of medical image inspection equipment and the wide application of PACS,people are able to access kinds of big data about medical images,which makes artificial intelligence medical image analysis based on deep learning be a hot area of research.At present,In the field of medical image intelligent analysis,there are only a few public data sets available,lacking high quality data.In respect of medical image annotation data acquisition,although there are a lot of domestic relevant medical software,most of them are small tools directly label on an existing single type of image or some simple annotation function in professional image processing software,lacking of medical image annotation system designed for deep learning technology,which seriously affected the research progress of deep learning technology in the field of medical.In view of this,this project designed and developed a medical image annotation system for deep learning based on common computer hardware devices,and the main tasks are as follows:1.For lacking of callable medical data,data transfer module was designed,realizing the multithreading mode of medical data migration and desensitization treatment,guaranteeing the stability and efficiency of the running of migration.2.In view of the large amounts of migration data storage,in addition to the local storage,the cloud storage way is designed based on baidu cloud and ali cloud,realizing manual,automatic uploading and downloading of the data under the multithreading mode,designing the function of encryption based on Rijndael algorithm to ensure the security of the cloud storage.3.In view of a large number of medical data statistical analysis problem,the design realizes the statistical result bar chart display and list display function under various query conditions,which makes it convenience for users to query,extract and analyze the data of multi-region,multi-disease and different age.The design are adapted for the prevention and control of frequently-occurring disease of different regions.4.In view of that current labeling software has a single function and no universality,the design implements the functions of "relational database plus XML file" and "one screen multi-display" mode for the management,reading,annotation and visualization of medical images of different types and formats,which can assist user annotation more conveniently and more accurately.By completing the above work,this project has achieved a medical image annotation system that can be used for deep learning of medical image annotation results of different types and formats.At the same time,the data migration of heterogeneous PACS database is designed,and the functions such as encryption cloud storage,statistical analysis,report browsing and annotation visualization are designed,and multi-threading technology is used to improve the system efficiency.It is of great practical value in the context of the current use of deep learning technology to promote more intelligent,more efficient and accurate diagnosis in medical field.