Research And Realization Of Clinical Form Auto-Recognition System Based On Machine Vision

As the medical innovation environment improves,domestic hospitals’ demands for medical efficiency and high-quality clinical data show a rapidly growing trend.The purpose of this project is to study an automatic identification system system which realizes the standardization and structuring of clinical data,and the electronic case report form(ECRF)can be customized and configured with high reuse.It combines the advantages and specialties of medicine,information science,computer science and other disciplines to form an organic whole,Lay the foundation for clinical scientific research in the hospital.The specific research contents are as follows:(1)A form area detection algorithm based on triangular positioning and a image correction algorithm is proposed.First,edge detection,image smoothing,and threshold binarization are performed on the form image,and the approximate area of the form area is determined according to the three position marker features on the form image.And then use the contour extraction algorithm to find the largest contour in the area,and extract the form area according to the coordinates of the four vertices of the fitting polygon of the largest contour.Considering the problem of image distortion caused by improper placement of the input image and wrinkles,affine transformation is performed to correct the image,and finally resizing is performed to obtain the segmentation result.(2)A Yolov5-TCT form recognition method based on form prior knowledge is proposed.First,the center position and content related information of the form checkboxes are recorded in the database,and the yolov5 target detection model with Transformer Encoder and CBAM(Convolutional Block Attetion Block)is used to recognize the form image processed by the above form area detection algorithm,get the coordinate position of the check box information filled in the form,and finally determine the specific content of the information according to the prior knowledge of the form.(3)According to the content and structure of clinical trial reports related to TAVR technology,EAV(entity attribute value)data model is used to study and design the database to realize the high reuse function of custom configuration of e CRF forms.(4)According to the current development environment of the hospital,we choose B/S architecture,use Java as the main development language,adopt the popular framework with Spring Boot + Vue + SQLServer + Redis,and carry out system development in strict accordance with the development specifications of relevant departments and information departments.This paper finally formed a complete set of clinical data collection systems for TAVR registration research.The implementation results in a third class hospital show that this system meets the needs of departmental for data acquisition,query and export.Combined with scanning input,follow-up reminder and other functions,it not only greatly improves the work efficiency of medical staff,but also leaves valuable data for future clinical,teaching and scientific research.