The following content is compiled from the final defense report of the students in the compulsory course "Big Data System Foundation" of the Big Data Capacity Improvement Project.
Our report is divided into the following five sections.
First of all, the first part is the project background and demand analysis. The background of our project is the digital medical scene. Digital medicine is a combination of information technology and medical knowledge. As a development trend of modern medicine, digital medicine is of great significance to the realization of precision medicine and efficient medicine. Suzhou Yiduo Cloud Health Co., Ltd., which we cooperate with, is an enterprise that provides smart medical care, Internet services and digital medical products to patients, doctors and medical institutions. He cooperated with Hengrui Medicine to carry out the patient follow-up project of the tumor product line, and accumulated millions of real patient data, covering the four drugs they developed. For such a digital medical problem, its routine requirements mainly include the daily management of patient data, the realization of some tracking and recording of data related to cancer patients, and the analysis requirements related to these recorded follow-up data. In response to these two needs, our team strives to establish a management and analysis system for patient data based on the follow-up data provided by Yiduoyun, focusing on the degree of adverse reactions after patients take medication and the relationship between their drug withdrawal and medication status. It is hoped that the obtained results can serve the research of drug side effects and further provide guidance for clinical medication.
Our specific content can be divided into the following three parts. The first part is the data management part and we need to realize the conversion of the data given to us by the enterprise to a specific data type, upload, parse and store these files, and then perform some user management tasks. data manipulation. The second part is digital display and data display. We need to realize the visual display of relevant fields in the uploaded data. The third part is data analysis, which uses different machine learning models to analyze their correlation and rank them for adverse reactions and field and drug application status data.
The second part is the system design architecture. First of all, let me introduce our webpage part. The technical station of the webpage part is mainly composed of three parts: HTML, CSS and JavaScript. The first part of HTML defines the structure of our web page, CSS defines the style of the web page, and JavaScript provides some dynamic interactivity for the web page. Our visualization process mainly includes the following four steps. The first step is to clean and analyze, then we do some processing and supplementation, merge the data according to the requirements of the enterprise, and then obtain some geographic data through web crawlers, then we choose the appropriate visualization form, and use what we need data into the corresponding format. Then use Echarts for data visualization and embed it into our network homepage for interactive observation.
In the second part, we adopted the DWF data management system. Our design goal is mainly to facilitate follow-up personnel to conduct modular, structured and standardized follow-up, simplify the follow-up process and facilitate answers. Our design logic is to consider that we have obtained three drug data, and the questions they need to ask are different. We divide them into three corresponding systems, and each system is divided into two pages, mainly including Patient information query page and patient information editing page. On the query page, we need to retrieve the text information of each field, and then on the patient information editing page, we need to register the patient's basic information, medication status, and his adverse reactions by module.
Finally there is our analysis section architecture. Because we got serious data confusion in some Excel tables, we used some manual work to remove duplicate rows and data conversion. Due to the many symptoms and high dimensions of adverse reactions involved, we classified them according to the body system to which the adverse symptoms belong according to some standards of the U.S. Department of Public Health and Human Services, and then conducted some data screening, and removed some of them because of emergency, etc. The data for reasons of discontinuation were finally analyzed with some machine learning models.
The following shows what we have done. The first is the webpage part. Our code structure is as follows. HTML files are stored in the root directory, and other files are stored in corresponding folders. Our webpage is divided into the following four parts, the first part is our homepage company introduction, and then the visualization of follow-up information and contact page. The bottom three pictures are screenshots of our webpage, and a visualization case is listed here. Because the table contains some province information of patients, we will do some formatting and then visualize it dynamically. Using Echarts tabular memory visualization, the color shades of different provinces represent the number of patients in that province.
Next is the part of the DWF data management system. We focused on implementing the two parts of existing data import and new form management, cleaning the data provided by the enterprise, removing some irrelevant information, and Vacancy was filled. In the part of adding and editing data forms, which is the part on the right, we have also made some adjustments to the logic of the questionnaire, which can facilitate us to get some more useful information.
Next is the Analysis Results section. We used a variety of statistical learning models to analyze the relationship between independent adverse reactions and adverse reactions classified by body system and drug withdrawal, and obtained the conclusions shown in the figure above. We performed a similar analysis on the Aitan and Erica data.
Finally, our data system display. Let's show our webpage first. We can access it through the local port. This is our main page, which contains some drug introductions and our division of labor. On the enterprise information interface, we displayed some information about Suzhou Yiduoyun Company, followed by our contact information page. And then there's our focus, which is our follow-up data visualization. We visualize various data through various tables.
The second part is DWF display, here we take Aitan visual data management as an example. First of all, we can perform some more precise queries on the query page. Here, some targets are screened out using the range of death time as an example. Similar queries can also be made for information such as cancer. On the edit page, we also implement some logic management we need according to the original plan. The delete function is not demonstrated, but it can be done. Editing and adding are similar, and both can implement similar logic.
Editor: Wen Jing
Proofreading: Gong Li
