Research On The Key Techniques And System Of Visual Patient System

As the rapid development of medical digitalization and widely utilization of medical information system, amount of diverse medical data is growing dramatically. How to improve the usage rate and efficiency of medical information under the background of big data has become a significant research orientation in medical information application field. Even though current medical information system, such as HIS, PACS, EMR, etc., provides unified retrieval function for different medical data, problems, like low efficiency of data usage, complex of operation and lack of intelligence, still remain. These problems may cause omitting and misunderstanding of information, which have become bottleneck and obstacles for physicians to obtain information fast and efficiently. To tackle those problems and face the challenges directly, this thesis introduces a multidimensional visual representative method for medical information, i.e., Visual Patient Representative Method. This method displays medical information in multidimensional visual ways, including time dimension and spatial dimension, to vividly illustrate health condition of patient, which assists physicians to obtain medical information with high efficiency. Furthermore, it bridges intellectual gap between medical information and nonprofessionals.The major innovative research points in this thesis are stated as follows:1. Designed a visual representative medical information processing system, which is based on anatomical structure and medical terminology classification standards, i.e., Visual Patient System(VPS). VPS helps users learn historical health status without reading clinical reports. Major functions of VPS:(1) Representation of clinical and health status based on anatomical structure;(2) Quantitative representation of lesion trend;(3) Retrieval interfaces of original clinical data, including DICOM images.2. Utilizing natural language processing technology, designed a semantic comprehension algorithm with a combination of rule-based methods and statisticsbased methods for RIS radiology reports. This algorithm can extract critical information about clinical findings and description from free text radiology reports, and complete data processing, transmission and storing in data format of key-value pairs. This algorithm is the backbone of VPS. It implements intelligent analysis of radiology reports and technically supports clinical assessment.3. Designed a medical data acquisition, processing and storing method, which is based on Hadoop distributed processing structure and is applied to implementing VPS. Established Visual Patient Information Processing Unit System(VPIS) with a combination of Hadoop batch processing architecture and Storm stream processing architecture. VPIS realizes parallel processing of various clinical data, which come from disparate sources, with high efficiency. VPIS guarantees the expandability and timeliness of VPS.4. Studied methods of system integration of VPS and PACS/RIS, evaluated clinical probation of VPS. According to practice of physicians, the time-consuming of reviewing historical clinical information through VPS decreases over 50%, compared with one by using PACS/RIS system. Furthermore, accuracy of information extracting remains high. It proves the superiority and feasibility of VPS in aspects such as intelligent extraction and representation of patient historical records, content-based data display and processing. VPS offers an innovative technical proposal to physicians for high efficiency of reviewing and analyzing patient historical records.