Design And Implementation Of Medical Document Recognition System

Document recognition technology has been a hot research issue in the field of word processing and character recognition research,such as card identification,invoice recognition,etc.have been widely used.However,in the area of medical document processing,there are no mature commercial products in the relevant research market.Medical documents are common documents used by patients for medical treatment,as proof of medical condition and follow-up treatment.Due to the different information formats of medical documents in different hospitals,it is very difficult to utilize medical documents in a comprehensive way and share them among hospitals.In this paper,we design and implement a medical document recognition system,which has the ability to recognize unstructured medical document information,such as medical record image information,and transform it into structured information for archiving and sharing of medical documents.The system consists of functional modules such as image pre-processing,text detection and text recognition,and document management.Since the images of the collected medical documents have noise,uneven illumination and incorrect spatial field of view,image pre-processing techniques such as binarization,document spatial field of view correction and non-uniform illumination correction can be used to improve the image quality.The system adopts the differentiable binarization algorithm to locate the text area in the image,and uses RCNN model to extract text sequences from the text area to achieve document recognition functions such as text detection and text recognition.The system also provides interactive image text extraction function to solve the logical problem of information structuring when image text is automatically extracted.The system is designed with B/S design architecture,uses My SQL database to save structured data,and provides UI-based interactive operation function.The system can be used independently or as an API to provide calls to other related systems.Through system testing,the test results show that the system runs stably,can accurately extract structured information from medical document images,and can meet the actual needs of users.