目的 在影像归档和通信系统(Picture Archiving and Communication System,PACS)数据库文件丢失或损坏后,实现影像资料和PDF报告关键信息的快速识别和重组,供患者回诊使用.方法 利用基于深度学习的光学字符识别技术和Pydicom技术分别读取PDF和DCOM文件中的基本信息,重新建立起患者、影像、报告三者之间的联系,并将关联数据写入数据库.结果 经抽样验证,该方法识别同类图像精度的准确度、精准度及召回率均为100%,综合指标F1值为1,在不同组别独立样本间的识别精度表现出一致性.平均每份报告识别时间约为0.14 s(t=-1.005,P=0.315),说明不同组别独立样本间的识别时间表现出一致性.结论 该方法的使用能有效缩短数据库故障后患者等待时长,能够在短时间内恢复医疗秩序,可用于PACS数据库数据丢失后的应急处置,也为PACS的数据整合提供依据,为医学影像数据恢复和数据整合提供一种新思路.
Research on Emergency Response and Recovery of PACS Database Data After Loss Based on OCR and Pydicom
Objective To realize rapid identification and reorganization of key information from image data and PDF reports for patient return visits after the loss or damage of picture archiving and communication system(PACS)database files.Methods Optical character recognition technology and Pydicom technology based on deep learning were used to read the basic information in PDF and DCOM files,so as to re-establish the connection between patient,image and report,and write the associated data to the database.Results After sampling verification,the accuracy,precision and recall rate of the method for identifying similar image accuracy were 100%,and the F1 value was 1,indicating that the recognition accuracy among different groups of independent samples showed consistency.The average recognition time per report was about 0.14 s(t=-1.005,P=0.315),indicating that the recognition time among independent samples of different groups showed consistency.Conclusion This method can effectively shorten the waiting time of patients after database failure and quickly restore medical order,which can be used for emergency treatment after data loss in PACS database,and also provides a basis for data integration of PACS,and provides a new idea for medical image data recovery and data integration.
optical character recognition(OCR)PACS dataemergency responsedeep learningDCOM information extractionimage text recognition
国家科技期刊平台
NSTL
万方数据
