See the medical health big data (on)
Author: Zhang Jiwu March 2017 published in "World of Medical Devices" special issue (as cited article, please indicate the source, thank you.)
"Big data" is so hot, there are "the wealth of the data to get the wealth", and even the "data is the world".
The era of big data has come. Everyone in the vertical field is talking about big data, as if big data is being raised, and any problems are solved. Artificial intelligence, cognitive computing, various intractable diseases, health prediction, scientific management, finance, business Problems in other fields are solved in the face of big data applications.
Big data is in full swing, and medical health big data is the focus of people's livelihood development. It is favored by the state and favored by industry, academia and research doctors. This year, the National Development and Reform Commission set up a “Medical Big Data Application Technology Engineering Laboratory”, which is ready to be established. The number of reporting units has reached 18, which basically includes the hospitals of research institutes of various key institutions in the field in China, and the leading figures are mostly Academicians take the lead, showing the importance of medical big data in China.
At the same time, domestic conferences on medical and health big data are intensive. In November and December 2017 alone, in Beijing, in Guangzhou, and in Shanghai, there were many sessions of medical big data, all related to health, medical care and health. Big data is an important topic at domestic and international conferences. In August 2017, the “Medical Health Big Data Alliance” led by the Health Planning Commission and China Electronics was established. And each listed company has linked its concept to big data in order to obtain sufficient market value recognition.
In particular, investment institutions, fund managers, and listed companies are all very interested. The most frequent question is "Can you get big data?", "Is there an information technology vendor that says that you can get big data, is it worth investing?", "Is there a team that says that there is a business model for operating big data, is it worth it? "Investment", "We are doing research on artificial intelligence, can automatically diagnose medical images, whether you are willing to invest", "We have established a third-party image hosting center, we will have big data."
Here I want to clarify a simple logic. When a system or a way that has a great impact on society exists, its return may be very large, and its investment cycle will be very long. For example, facebook changes the relationship of human social communication. However, it is still in the investment period until now, and only institutions that are optimistic about its long-term development will continue to invest. At the same time, the ability to create such a major social impact constitutes an ecological circle. This ecosystem is either top-down, perfect top-level design, and promoted; or spontaneously produced, bloody, such as Alibaba. Winning and forming an ecological environment. Regardless of the path, the formation of an ecological environment is a need for massive funds, and the entrepreneurial sacrifice that continues to run at a speed of sprinting can be achieved. Therefore, big data brings huge business opportunities, but it must have huge and long-term investment.
Big data, Baidu Encyclopedia gives the definition of "big data", which refers to a collection of data that cannot be captured, managed, and processed by conventional software tools within a certain time frame. It requires a new processing model to make stronger decisions. Power, insight, and process optimization capabilities, massive, high growth rates, and diverse information assets.” IBM proposes 5V features of big data: Volume, Velocity, Variety, Value Value density), Veracity.
The strategic significance of big data technology is not to master huge data information, but to professionalize these meaningful data. Some companies in this field have introduced different concepts. For example, IBM has launched the concept of “cognitive computing” and actively developed the commercial application “Watson”. Google developed AlphaGo with Big Data. In March 2017, it launched a century-long battle with the famous Korean chess player Li Shishi, and won the world's hot debate on artificial intelligence with machine wins. (The predictions of optimism and pessimism for the future of mankind are raging).
1.1.1 Impact of Big Data (Social Change)
- Big data and social costs (medical fee control)
- Big data and social quality (diagnosis rules, medical quality control, auxiliary diagnosis)
- Big data and precision medicine
- Big data and health management
- Big data and statistics, alarms, warnings, predictions
- Big data and insurance
- Big data data acquisition, data modeling
- Big data and internet of things
- Artificial intelligence, machine learning, cognitive science
- Big data and pension
- Big data application example
- Big data system design experience
- Big data and wearables
- Big data and mobile healthcare
- Privacy security
- Social management (data ownership, use rights, etc.)
1.2.1 Clinical treatment
a) Comparative effect study
b) Clinical decision support system
c) Transparency of medical data.
d) Remote patient monitoring
e) Advanced analysis of patient profiles
1.2.2 Payment and Pricing
a) Payment fraud and insurance payment automatic inspection system
b) Pricing plans based on medical economics and therapeutic effects studies
1.2.3 Drug development
a) predictive modeling
b) Improve the design of clinical trials using statistical tools and algorithms
c) Analysis of data from clinical trials
d) personalized medication
e) Analysis of disease patterns
1.2.4 New business model
a) Aggregation and synchronization of patient clinical data and insurance claims data sets
b) Online community platform
1.2.5 Public health
a) The use of big data can promote public health surveillance and response.
(2) Key to engineering technology for big data applications
The status quo of big data is reminiscent of the story of mice giving bells to cats: the mice meet to discuss ways to deal with cats, and propose a good way to lick a bell on the cat's neck. When the cat approaches, the mouse can listen. Go to the ringtone and you can run away. Every mouse agrees with this suggestion, but an old mouse stands up and says: Who is going to put a bell on the cat?
The status quo of big data is very similar to this fable. Everyone is talking about the revolutionary changes brought about by big data. The question is "Who is going to hang the bell and how to hang the bell?"
I once participated in a national medical big data application competition. The contestants and review experts all lamented that there was a lack of data and the data quality was relatively poor.
The application and research of big data is multi-layered. The first is data acquisition, data modeling, and then data processing, analysis, knowledge acquisition, cognition, and application.
As the hottest example of the advancement of artificial intelligence based on big data applications, Go game under AlphaGo is actually very tricky. Big data engineering techniques include data acquisition, data modeling, data processing and analysis, acquiring knowledge, building awareness, and presenting them in a friendly manner. At present, in the big data implementation technology, the so-called data mining and knowledge learning algorithms such as DeepLearning have matured, and many published scientific research papers. The most challenging thing is how to abstract the data in the display life into a model, or make the intrinsic relevance (knowledge) of the data easier to detect. Go, only black and white two colors, 19 lines across 19 lines, 361 intersections, seem to change infinitely, in fact, for the computer is the easiest to simplify the model, only the calculation speed and calculation and Memory (memory) is a challenge, but these are not a problem for today's computing technology. Therefore, the success of AlphaGo just shows that the bottleneck of today's big data application engineering technology lies in data acquisition and data modeling.
For the development of big data applications in China, the current strategic considerations for breakthrough development are:
2.1 Solve the bottleneck problem of data acquisition and data modeling
The main problem of big data research and application in China is not the hardware environment construction, but the corresponding system platform and methodology research. Including, data acquisition key technologies, data collection and interconnection standards establishment and promotion; data quality, including data models, management of heterogeneous data, correlation between data, time distribution of data; data mining methodology, clinical Feature parameter extraction of data; data application, clinical data mining methodology applied to clinically assisted diagnosis of CDSS mode; precision medical research. The focus is on:
2.1.2 Collection of medical big data
A large amount of data can analyze the correlation of disease, symptoms and laboratory data, helping clinical researchers to establish predictive models for some typical diseases. In the hospital's diagnosis and treatment process, for the specific application of each department, long-term clinical monitoring parameters related to specific diseases have been accumulated, and a large amount of data has been accumulated with the operation process of the hospital.
At the same time, with the development of mobile Internet technology and wearable medical devices and technologies, user vital signs obtained through various wearable devices provide great convenience for the acquisition of user health data.
On the one hand, the health data can be analyzed to obtain the user's health information to guide the living habits such as exercise and diet; on the other hand, the combination with medical data can improve the scientific and diagnostic accuracy of the user's disease diagnosis.
2.1.3 Analysis of medical big data
In the traditional medical industry, the hospital information system completed the process control and data accumulation within the hospital. The medical industry has long encountered the challenge of massive data and unstructured data. In recent years, many countries are actively promoting the development of medical informatization, which has enabled many medical institutions to have funds for big data analysis. Medical data is the data generated by medical personnel during the patient's diagnosis and treatment process, including the patient's basic condition, behavior data, medical treatment data, management data, inspection data, electronic medical records, and so on. In modern hospitals, the above data is stored in various information systems of hospitals, which is the basis of medical big data analysis.
Medical health data is a continuous and high-growth complex data, and the information value is rich and diverse. The effective storage, processing, query and analysis of medical health data, tapping its potential value, and discovering medical knowledge will deeply affect human health. Level and treatment. On the basis of traditional medical statistical methods, the emergence of new models and technologies provides new ideas for acquiring new knowledge from data.
Medical Health Data Mining Common methods for health information data analysis include classification, regression analysis, clustering, association rules, feature analysis, change, and bias analysis. Different types of patients are used to reason and judge different types of physiological data and health perception data. The big data analysis technology achieves the purpose of serving clinical treatment, predicting the incidence of diseases, and tracking the patient's condition.
2.1.4 Application of medical big data
The traditional data application of the medical industry has important reference value. It must be clear that the development of big data is based on its own technical foundation and data accumulation. New information analysis technologies and communication technologies have brought new ideas to traditional medical network applications and data analysis.
Based on the collection and analysis of the user's medical data and health monitoring data, it is possible to predict and monitor the user's physical condition, and even to determine which type of disease the user is susceptible to. Improve the user's health status and reduce the risk of the user's disease. Accurate analysis of large data sets including patient vital data, cost data, and efficacy data can help doctors determine the most effective and cost-effective treatments in the clinic. Medical care systems will have the potential to reduce over-treatment, such as treatments that avoid side effects greater than efficacy.
2.2 Critical and Critical Departments may be the first to break through in big data acquisition and application
It is unrealistic to cover all the diseases from the beginning, and the workload is far beyond the scope that can be completed in the initial stage of a key laboratory. Selecting special diseases (such as cardiovascular disease, echinococcosis) can focus on the application and engineering practice of big data and establish a high-quality data platform. Choose some typical departments, such as emergency department and intensive care unit. This is the emergency department that needs to be improved under the new medical environment in China. It is the entrance to various common diseases and frequently-occurring diseases (most cardiovascular and cerebral infarction patients enter the hospital. One link is the emergency department), and it is also the most life-saving department. It is also the most complete department of all kinds of examinations (various life parameters, various testing equipment, image information, etc.), and the special nature of emergency department and intensive care department. Make the recorded data have time information, continuous records, and the treatment results have a certain clear assessment in the short term, all of which makes the data of these departments have a large amount of data, and various clinical data can establish a good correspondence. Relationship, with a certain amount of time continuous information, and may be verified by the results of the diagnosis. The data quality is very good.
First of all, the critically ill department is one of the departments with the most abundant vital signs and the largest amount of data. Taking the intensive care unit as an example, the patient's electrophysiological data (EC, EEG), blood pressure, blood sugar, blood routine, and even patients Image data (CT, MRI, etc.) will be continuously detected, only electrophysiological information, each patient produces hundreds of thousands of data per day;
Thirdly, the data of the critically ill departments have greater complexity. Given the typical cases of the critically ill departments and the occurrence of complications, the acquired disease data has a complex combination of disease states;
Fourth, the data of the critically ill department has a stronger time value, and the data obtained by half of the departments lacks time significance. It takes a decade to record the chronic disease, and it takes ten years to complete the large-scale data recording. tracking record). In fact, a person's information over time is more valuable than statistically significant (evidence-based medicine) information. In April 2015, Nature issued "Personalized medicine: Time for one-person trials", emphasizing The time series information of an individual is more meaningful for personalized medicine than the statistical information of a large number of people. Long-term tracking, such as the 10-year chronic disease tracking research, is of great significance, and it is also costly and achievable. It takes a long time to accumulate enough data to form big data, conduct big data research, and extract rules. In critically ill departments, it records exactly a large amount of data in a short period of time, along the time axis, so it is possible to test various big data research algorithms and results and explicit results more quickly.
A good example of this is the recent success of the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) program at the Massachusetts Institute of Technology (MIT). They use the emergency department as the entry point, and have accumulated 60,000 cases for a long time. The big data research is the rare medical big data with scale and quality in the world. Based on this, they published many important articles in important journals such as Science, Nature, and Lancet.