Healthcare Big Data Lifecycle
In Asker's medical big data methodology, the life cycle of medical data is divided into four quadrants in the following figure:
The first quadrant : data collection In the data collection stage, hospitals usually use relational databases (such as Oracle, DB2, MySQL...), and the core requirement is to ensure the integrity and consistency of the data, and ensure that the data will not be affected by software and hardware. loss due to failure. In addition, with the introduction of Internet traffic, the online processing scale and efficiency of data has also become an important consideration. In order to improve the processing efficiency of online transactions, some hospitals have introduced distributed databases to meet the needs of high concurrent access under Internet traffic.
The second quadrant : data collation Each hospital has many different business systems, and these systems use relatively independent databases to store and process different business data. Usually, the relational data design capacity used by the system is limited, and historical data needs to be regularly cleaned to the central data warehouse to ensure fast and efficient online transaction processing. The central data warehouse is to store the full historical data of each independent system and collect the data of each system at the same time. Therefore, a distributed and scalable technical architecture will be used in the design. The entire cluster capacity and processing power can be seamlessly scaled.
The third quadrant : Data analysis users need to use data assets to create value. First, the full amount of historical data from various discrete systems can be correlated and queried, and data analysis tables of different dimensions can be constructed through batch processing to drive BI and report display. Then, based on the exploratory analysis of the full amount of data, correlation analysis can be performed on the data of each system, relying on advanced machine learning algorithms to discover new business rules, and use the data rules to influence and guide decision-making.
Fourth quadrant : Data decision-making In the third quadrant, the data cubes, dimension tables and hierarchical tables constructed through batch data processing need to be transferred to the relational database to drive report display and generate decision views. The data cube and dimension table database for decision support are also called data marts. BI applications and data-driven applications can directly obtain data from the data mart for business analysis, or perform secondary data collection on the basis of dimension tables to form Higher-level business aggregation.
Medical big data platform architecture and operation mode
The Asker medical big data platform solution adopts the hybrid design of traditional relational database and big data platform. The overall basic data platform consists of online transaction database, data integration and exchange, big data platform, data mart, exploration and analysis, and data-driven application. Six parts:
online transaction library
The online transaction database is the online production business system that the enterprise has built, such as the database used by the transaction system, website, ERP, warehouse management, production process management, etc., as well as the database used by the system to be built for Internet business expansion. The online transaction library mainly faces real-time transaction processing, and is currently mainly composed of commercial or open source relational databases. In the future, in order to meet the high concurrent business requirements under Internet traffic, distributed databases and cloud databases can be introduced as required.
Big data platform
The big data platform is composed of big data technology components, including Hadoop, Spark, Hive, Hbase, Kylin, etc., which can be tailored, customized and expanded according to requirements. The big data platform obtains data from the online transaction database, which is the full data set of the online transaction database. At the same time, the big data platform can also obtain data from third parties, such as importing credit data, public opinion data, etc., and conduct correlation analysis with online transaction data.
data mart
The data mart is the data that is aggregated by multi-dimensional cubes on top of the basic fact data stored in the big data platform. It is the result of batch processing by the big data platform fact table (Fact Table). BI statistical reports and related application data display efficiency.
data-driven applications
Data-driven applications include different business systems such as BI reports, corporate integrity query, corporate business analysis, brokerage practice quality assessment, knowledge base theme construction, etc. They are characterized by the need for a large amount of data to drive business presentation or decision-making. These applications do not directly connect to the big data platform, because the granularity of the fact data stored in the big data platform is too fine, and it needs to be aggregated in batches before it can be used. Therefore, the data-driven application layer is supported by data marts. For the new business system, it is necessary to cooperate with the corresponding dimension table, and the business system is driven by the pre-collected dimension statistical data.
Exploratory Analysis
Exploration and analysis is the soul of big data, and it is also an important direction to explore the value of data for business development. Exploratory Analysis Through machine learning algorithms, correlation analysis is performed on the feature set constructed from the full amount of data, so as to discover business laws, predict in advance, and improve service quality. Commonly used analysis algorithms include mature algorithms such as decision trees and random forests, and new algorithms suitable for the characteristics of enterprise data can also be developed based on existing algorithms.
The application prospect of medical big data
Provide reference for clinical diagnosis and treatment of common diseases
Using big data technology to mine and analyze massive medical data, it can provide a series of intelligent human-computer interaction, such as repeated inspection and inspection prompts, treatment safety warnings, drug allergy warnings, efficacy evaluation, intelligent analysis of diagnosis and treatment plans, and prediction of disease progression. It can provide scientific decision-making reference for clinicians, improve the level of clinical diagnosis and treatment, and form a clinical decision support system that "originates from the clinic and returns to the clinic". Through the analysis of big data in patient files, it is possible to determine which people are susceptible to a certain disease, so that they can receive preventive interventions as soon as possible. These methods can also help patients choose appropriate treatment options. In addition, the clinical decision support system can also relieve doctors from time-consuming simple consultation work and improve the efficiency of their diagnosis and treatment.
Provide the basis for the refined management of the hospital
The refined management of the hospital takes standardization as the premise, systemization as the guarantee, dataization as the standard, and informatization as the means. Competitiveness. Through the big data analysis platform, the hospital's outpatient volume, operation volume, number of admitted/discharged patients, bed utilization rate, bed turnover rate, equipment utilization rate, equipment depreciation rate, disease spectrum, patient distribution area, financial income and expenditure and other data are analyzed. Compare and analyze the current data with the same period data and previous data. Comparing and analyzing the data of local hospitals with similar conditions, find out the causes and gaps of continuously improving the quality of the hospital's economic operation, seize the weak links in their own work, and take practical improvement measures.
Provide a platform for personalized medicine
Individualized medicine is based on a large amount of information of each patient, through comprehensive analysis and mining the characteristics of each patient's pathology, physiology, etc., to further formulate unique and optimal treatment and prevention plans for each patient, and improve the effectiveness of treatment. targeted for optimal efficacy. Personalized medicine requires comprehensive analysis of all aspects of each patient's information, and requires methods and capabilities to process this "big data". The detailed examination information and diagnosis information of patients are analyzed to facilitate the formulation of individualized treatment strategies, so as to obtain better curative effect. The development of technology has made the amount of patient information routinely collected in modern medicine very huge, and the ability to analyze information has also been greatly improved, making individualized medicine possible.
Provide information for clinical research
The emergence of massive data has spawned a new scientific research model, that is, in the face of massive data, researchers only need to directly find or mine the required information, knowledge and wisdom from it, and even do not need to directly contact the object to be studied. In 2007, the late Turing Award winner Jim Gray described "The Fourth Paradigm" of data-intensive scientific research in his final speech, transforming big data research from the third paradigm ( Computational Science) as a research paradigm, arguing that the "Fourth Paradigm" may be the only systematic approach to some of the toughest global challenges we face. In the process of scientific research, the utilization, development and organization of big data can subvert many previous research results and bring unexpected benefits.
The construction goal of Asker medical big data integration platform
Under the background of the advent of the Internet era, Asker is committed to providing the best overall solution for the data platform for the transformation of the hospital's "big data +".