WO2023109199A1 - 一种个体慢病演进风险可视化评估方法及系统 - Google Patents

一种个体慢病演进风险可视化评估方法及系统 Download PDF

Info

Publication number
WO2023109199A1
WO2023109199A1 PCT/CN2022/116969 CN2022116969W WO2023109199A1 WO 2023109199 A1 WO2023109199 A1 WO 2023109199A1 CN 2022116969 W CN2022116969 W CN 2022116969W WO 2023109199 A1 WO2023109199 A1 WO 2023109199A1
Authority
WO
WIPO (PCT)
Prior art keywords
risk
individual
chronic disease
patient
disease
Prior art date
Application number
PCT/CN2022/116969
Other languages
English (en)
French (fr)
Inventor
李劲松
蒋一帆
周天舒
田雨
Original Assignee
之江实验室
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 之江实验室 filed Critical 之江实验室
Priority to JP2023538042A priority Critical patent/JP7430295B2/ja
Publication of WO2023109199A1 publication Critical patent/WO2023109199A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis

Definitions

  • the invention belongs to the field of medical information technology, and in particular relates to a method and system for visually assessing the risk of individual chronic disease evolution.
  • Chronic disease is the abbreviation of chronic non-communicable disease. It is a general term for a class of diseases with insidious onset, long course, protracted illness, and lack of definite etiological evidence. It mainly includes diabetes, hypertension, cardiovascular disease, and chronic kidney disease. A major feature of chronic diseases is that there are many complications and complications, and the incidence rate is high. At least half of the patients with chronic diseases have complications or complications. Complex complications will affect the quality of life of patients, increase the cost of medical care for patients, and even directly increase mortality. Many complications can be prevented by intervention methods such as lifestyle changes and medication, but effective health education and effective self-management are required for patients.
  • the purpose of the present invention is to address the deficiencies of the prior art, to provide a method and system for visual assessment of individual chronic disease progression risk, to present personalized chronic disease progression risk assessment for doctors and patients, and to help doctors and patients make joint decisions.
  • the present invention discloses a method for visual assessment of individual chronic disease evolution risk, which includes the following steps:
  • (1) Construct an evidence-based semantic knowledge base of chronic disease risk, including: constructing disease entities of chronic diseases and their complications, comorbidities, and disease-related risk factor entities. For each disease entity, retrieve relevant research literature and extract Semantic knowledge of risk relationships, merging the relative risk ratios of different research documents for the same research purpose, using the combined relative risk ratios to represent the relationship between entities, and constructing an evidence-based semantic knowledge base for chronic disease risks;
  • Calculating and drawing the patient's individual chronic disease evolution risk path map including: based on the chronic disease risk evidence-based semantic knowledge base, the population exposure rate database and the patient's individual data standard model, constructing a representation of the individual chronic disease evolution
  • a weighted directed graph of risk The vertices of the directed graph are the risk factors and possible diseases that have occurred in the patient, corresponding to the entities in the chronic disease risk evidence-based semantic knowledge base.
  • the edge weights of the directed graph are two The relative risk ratio between entities, the edge direction is the direction of influence between entities; the overall relative risk ratio of each vertex is stored in the risk matrix; the patient's individual chronic disease is drawn according to the weighted directed graph and risk matrix of the individual patient Evolution risk roadmap;
  • step (1) the relevant research literature on chronic diseases and their complications and comorbidities is used as the knowledge source of the knowledge base, and the medical and health standard terminology is used as the data semantic identification to construct chronic diseases and their complications, combined Semantic knowledge of disease-risk relationships.
  • the data structure of the chronic disease risk evidence-based semantic knowledge base is designed as an RDF triplet conforming to the OWL language format specification, and each triplet includes two entities, representing the source entity A and the entity B that characterizes the result, and the relationship C between entities;
  • the entity class includes demographic information, inspection, drug, non-pharmaceutical intervention treatment, disease, symptom;
  • the class of the entity B is disease, and the entity A is the risk factor of the entity B, and the relationship C between the entities includes increasing risk and reducing risk, and is characterized by a relative risk ratio.
  • step (1) the relative risk ratios of different research documents for the same research purpose are combined, including:
  • Var i is the variance of the relative risk ratio of the samples contained in the i-th research document
  • RR i is the relative risk ratio of the i-th research document, is the average value of the relative risk ratio of each research literature
  • exp( ⁇ ) is an exponential function
  • ln( ⁇ ) is a logarithmic function
  • the random effect model DL method is used to correct the weight W i , and the correction formula is as follows:
  • W i ′ is the corrected weight value of the i-th research document.
  • the patient individual data standard model includes personal basic information, as well as demographic information, inspection and testing, drugs, non-pharmaceutical intervention methods, diseases, and symptoms.
  • the patient’s individual health data comes from: clinical electronic medical records of medical institutions, personal health files, health data collected by smart hardware and wearable devices, health questionnaire data, chronic disease management institutions, pension Agency management data.
  • step (4) the overall relative risk ratio of each possible disease is estimated by combining all the risk factors that have occurred in the individual patient, and the estimated overall relative risk ratio of each vertex is stored in the risk matrix , the estimation methods include:
  • b j is the regression coefficient of the jth edge
  • a is estimated by the prevalence rate of the population corresponding to the disease at vertex i
  • P is the population prevalence rate of the corresponding disease, which is derived from the population exposure rate database
  • the Rothman-Keller model estimates the overall relative risk ratio: there are n edges in the directed graph ending at vertex i, and the weights of the edges are r 1 , r 2 , ..., r j , ..., r n ;
  • P j is the population exposure rate corresponding to the risk factor of the jth side, which is derived from the population exposure rate database;
  • m is the number of elements of the collection E
  • E i is the i-th element value of the collection E
  • k is the number of elements of the collection F
  • F j is the j-th element value of the collection E
  • the calculation and drawing of the patient's individual chronic disease progression risk path map includes: defining the position and size of the canvas, determining the scale of the coordinate axes, selecting the chart type, and defining the shape and color of the nodes; the node types include population Statistical information, inspection and testing, drugs, non-pharmaceutical interventions, diseases, and symptoms are distinguished by different node shapes; for different disease types, different node colors are used to distinguish; the size of the disease node is the overall number of vertices stored by the risk matrix
  • the calculation of the relative risk ratio shows a linear relationship; the thickness of the connection between nodes is calculated through the weight of the edge in the directed graph, and the relationship is linear, and the risk is increased and the risk is reduced by different connection colors; the node with an in-degree of 0 Align on one side, and arrange the lower nodes in order to form the risk path hierarchy of chronic disease evolution.
  • Another aspect of the present invention discloses a visual assessment system for individual chronic disease evolution risk, which includes:
  • Patient individual data acquisition and conversion module According to the patient's identity, obtain patient individual health data from multiple data sources, match the acquired patient individual health data, extract, convert, and load into the standard model of patient individual data;
  • the visual assessment module of individual patient's chronic disease evolution risk including the following units:
  • Chronic disease risk evidence-based semantic knowledge base construction unit construct disease entities of chronic diseases and their complications, comorbidities, and disease-related risk factor entities. For each disease entity, retrieve relevant research literature and extract risk relationship semantic knowledge , merge the relative risk ratios of different research literatures for the same research purpose, use the combined relative risk ratios to represent the relationship between entities, and build an evidence-based semantic knowledge base for chronic disease risks;
  • Crowd exposure rate database construction unit construct a crowd exposure rate database for all entities in the chronic disease risk evidence-based semantic knowledge base, and the crowd exposure rate corresponding to the associated entity;
  • Calculation and drawing unit of individual patient chronic disease evolution risk path diagram based on the chronic disease risk evidence-based semantic knowledge base, population exposure rate database and patient individual data standard model, construct a weighted directed graph representing the individual chronic disease evolution risk, directed
  • the vertices of the graph are the patient’s risk factors and possible diseases, corresponding to the entities in the chronic disease risk evidence-based semantic knowledge base.
  • the edge weight of the directed graph is the relative risk ratio between the two entities, and the edge direction is the entity The direction of the influence between them; store the overall relative risk ratio of each vertex in the risk matrix; draw the patient's individual chronic disease evolution risk path diagram according to the weighted directed graph and risk matrix of the individual patient;
  • Interactive analysis module of patient individual chronic disease evolution risk if the individual patient data changes, update the standard model of patient individual data, draw a new patient individual chronic disease evolution risk path map, and visually display the disease risk change on the evolution risk path , to provide doctors and patients with a dynamic and interactive risk analysis of individual chronic disease progression.
  • the present invention obtains risk relationship knowledge from a large number of research documents, and analyzes the same The different research results of the research purpose are combined to systematically construct an evidence-based semantic knowledge base of chronic disease risk containing various complex relationships.
  • the present invention is based on the chronic disease risk evidence-based semantic knowledge base, combined with patient Individual data and population exposure rate data are used to calculate the individual patient's chronic disease evolution risk and draw a patient's individual chronic disease evolution risk path map, which not only considers individual patient differences, gives personalized assessment results, but also ensures that the assessment results are evidence-based.
  • the present invention provides visual and interactive patient information based on the drawing method of individual patient chronic disease evolution risk path map.
  • the individual chronic disease evolution risk analysis method makes the information expression more clear, intuitive, and vivid, allowing patients to fully understand their own chronic disease evolution risk and the expected effect of intervention and treatment measures, fully participate in clinical decision-making, and realize joint decision-making between doctors and patients, thereby helping patients to comply Enhancement of sex and healing effect.
  • Fig. 1 is a flow chart of the method for visual assessment of individual chronic disease evolution risk provided by the embodiment of the present invention
  • Fig. 2 is the flow chart of building the chronic disease risk evidence-based semantic knowledge base provided by the embodiment of the present invention
  • Fig. 3 is an example of a patient's individual chronic disease evolution risk path diagram provided by an embodiment of the present invention.
  • Fig. 4 is a structural diagram of a system for visually assessing individual chronic disease evolution risk provided by an embodiment of the present invention.
  • An embodiment of the present invention provides a method for visually assessing the risk of individual chronic disease evolution, as shown in Figure 1, including the following steps:
  • the disease-risk relationship based on medical evidence is the basis for evaluating the risk of chronic disease evolution, so it is necessary to first construct an evidence-based semantic knowledge base for chronic disease risk.
  • the research literature related to chronic diseases and their complications and comorbidities is used as the knowledge source of the knowledge base, and the medical and health standard terminology is used as the data semantic label to construct the risk relationship semantic knowledge of chronic diseases and their complications and comorbidities.
  • Medical health standard terminology can be constructed using SNOMED CT, ICD-10, UMLS, etc.
  • the data structure of the chronic disease risk evidence-based semantic knowledge base is designed as an RDF (Resource Description Framework) triplet conforming to the OWL (Web Ontology Language) language format specification, making the chronic disease risk knowledge a semantic structure suitable for computer reasoning.
  • Each triple includes two entities, entity A representing the source and entity B representing the result, and the relationship C between the two entities.
  • Entity classes include demographic information, inspection tests, drugs, non-pharmaceutical interventions, diseases, symptoms, and more.
  • Non-pharmaceutical interventions include behavior, exercise, diet, etc.
  • the chronic disease risk evidence-based semantic knowledge base constructed by the present invention is mainly to represent the disease risk relationship.
  • the category of the entity B in the triplet is a disease, and the medical and health standard term set is used as the unique semantic identifier.
  • Entity A is a risk factor for Entity B, and Entity A contains all the above entity classes.
  • the risk factors of a disease include not only entities such as demographic information, examinations, drugs, non-pharmaceutical interventions, symptoms, but also other diseases.
  • the relationship C between entities mainly includes increasing risk and reducing risk, which is characterized by relative risk ratio (risk ratio, RR).
  • the relative risk ratio is the ratio of the incidence in the exposed group to the incidence in the control group.
  • RR 1, indicating that entity A is not related to entity B; RR>1, indicating that the relationship between entity A and entity B is risk-increasing; RR ⁇ 1, indicating that the relationship between entity A and entity B is risk-reducing.
  • knowledge base entities including chronic diseases and their complications, disease entities of comorbidities, and demographic information related to these diseases, inspection tests, drugs, non- Drug intervention means of treatment, disease, symptoms and other risk factors entities.
  • For each disease entity retrieve relevant published research literature (keyword retrieval can be used), and build a chronic disease clinical literature database.
  • Using natural language processing technology to obtain the semantic knowledge of risk relationship in research literature and semantically map each risk factor extracted from research literature with the constructed risk factor entity, so as to construct the risk relationship triplet of each disease.
  • the chronic disease risk evidence-based semantic knowledge base is stored using neo4j.
  • the relative risk ratio between the same two entities may be different. Therefore, after extracting the risk relationship semantic knowledge of each research literature, and referring to the idea of Meta analysis, different studies of the same research purpose The results are analyzed and combined to obtain the final and most reliable relative risk ratio.
  • K is the total number of research documents
  • W i is the weight of the i-th research document
  • RR i is the effect size of the i-th research document, that is, the relative risk ratio
  • It is the average value of the relative risk ratio of each research literature.
  • Var i is the variance of the relative risk ratio of the samples included in the i-th research literature.
  • the test quantity Q obeys the chi-square distribution with K-1 degrees of freedom. Using the chi-square test, a chi-square value X 2 was obtained. If Q>X 2 (1- ⁇ ), it indicates that there is heterogeneity among the research documents, otherwise it indicates that there is no heterogeneity among the research documents, where ⁇ is the confidence level and can be set as 0.05.
  • the combined relative risk ratio RR com is calculated using the fixed effect model Mantel-Hasenzel method.
  • exp( ⁇ ) is an exponential function
  • ln( ⁇ ) is a logarithmic function
  • the random effect model DL method is used to correct the weight W i , and the correction formula is as follows:
  • W i ′ is the corrected weight value of the i-th research document.
  • the population exposure rate refers to the ratio of the exposed number of entities in the population to the total number of people. For example, if an entity is "smoking", the population exposure rate for this entity is the number of smokers in the population divided by the total number of people. For disease entities, the population exposure rate is the population prevalence rate.
  • the data source of the crowd exposure rate database can be publicly released statistical data, or the statistical data of a certain crowd database.
  • the crowd can be a national crowd, a local crowd, a hospital crowd, etc., which are not specifically limited in the present invention. For example, if the system is used in a hospital, the data source can be the statistical data of all patients in the hospital; if the system is used in a certain area, the data source can be the statistical data of all patients in the area.
  • Individual patient health data comes from: clinical electronic medical records of multiple medical institutions, personal health files, health data collected by smart hardware and wearable devices, health questionnaire data, management data of chronic disease management institutions, elderly care institutions, etc.
  • the standard model of individual patient data includes basic personal information, as well as demographic information, inspection and testing, drugs, non-drug intervention methods, diseases, symptoms and other information. Match individual patient health data from multiple data sources, extract, convert, and load to the standard model of individual patient data.
  • the terminology used in the standard model of patient individual data is consistent with the medical health standard terminology in the chronic disease risk evidence-based semantic knowledge base.
  • a weighted directed graph G(V,E) representing the individual chronic disease evolution risk is constructed.
  • the vertices in the vertex set V of the directed graph G are the risk factors and possible diseases that have occurred in patients, corresponding to the entities in the chronic disease risk evidence-based semantic knowledge base;
  • the edge weights in the edge set E of the directed graph G are The relative risk ratio between two entities, the direction of the edge is the direction of the influence between the entities.
  • the chronic disease risk evidence-based semantic knowledge base stores the risk relationship between a single risk factor and a disease. Combining all the risk factors that have occurred in an individual patient, the overall relative risk ratio of each possible disease can be estimated.
  • the estimated overall relative risk ratio of each vertex is stored in the risk matrix M.
  • the overall relative risk ratio can be estimated by three methods: logistic regression model, Rothman-Keller model and mixed model.
  • the logistic regression model estimates the overall relative risk ratio
  • b j is the regression coefficient of the jth edge, and the calculation formula is:
  • a is estimated based on the population prevalence rate of the disease corresponding to the vertex, and the calculation formula is:
  • P is the population exposure rate of the corresponding disease, that is, the population prevalence rate, which comes from the above-mentioned population exposure rate database.
  • edges in the directed graph G that end at vertex i, and the weights of the edges are r 1 , r 2 ,..., r j ,..., r n .
  • P j is the exposure rate of the risk factor corresponding to the jth edge in the population, which comes from the above-mentioned population exposure rate database.
  • m is the number of elements of the set E
  • E i is the i-th element value of the set E
  • k is the number of elements of the set F
  • F j is the j-th element value of the set E.
  • a mixed model is a model that combines the results of the logistic regression model and the Rothman-Keller model.
  • the results of the above two models were weighted and averaged to obtain the final overall relative risk ratio.
  • the weight can be selected artificially based on experience, which is not specifically limited in the present invention.
  • the weighted directed graph G and risk matrix M of individual patients are visualized using the drawing plug-in echarts.
  • the types of nodes include demographic information, inspection and testing, drugs, non-drug intervention methods, diseases, symptoms, etc., which are distinguished by different node shapes.
  • different node colors are used to distinguish them.
  • the node size of the disease is calculated by the overall relative risk ratio of the vertices stored in the risk matrix M, and it has a linear relationship. The larger the overall relative risk ratio, the larger the node.
  • the thickness of the connection between nodes is calculated by the weight of the edge in the weighted directed graph G, which shows a linear relationship, reflecting the relative risk between two nodes.
  • the connection between nodes is distinguished by different colors to increase risk and reduce risk risk. Align the nodes with an in-degree of 0 to the left, and arrange the lower nodes in order to form the chronic disease evolution risk path hierarchy.
  • Figure 3 is an example of a patient's individual chronic disease evolution risk path map.
  • the calculation and drawing of the above-mentioned individual chronic disease evolution risk path map of the patient is mainly based on the initial data of the patient.
  • an interactive analysis method for the evolution risk of the individual patient's chronic disease is provided to achieve a dynamic, interactive, Visual analysis.
  • the standard model of a patient’s individual data is set A, where the non-adjustable data is set B, including basic information such as demographics, and the adjustable data is set C, including the physical examination results in the inspection, such as Weight, including data on drugs, non-drug interventions, diseases, symptoms, etc.
  • the embodiment of the present invention also provides a visual assessment system for individual chronic disease evolution risk, as shown in Figure 4, the system includes:
  • Patient individual data acquisition and conversion module According to the patient's identity, obtain patient individual health data from multiple data sources, match the acquired patient individual health data, extract, convert, and load into the standard model of patient individual data; the realization of this module You can refer to the above step (3).
  • the visual assessment module for the risk of chronic disease evolution of individual patients includes the following units:
  • Chronic disease risk evidence-based semantic knowledge base construction unit construct disease entities of chronic diseases and their complications, comorbidities, and disease-related risk factor entities. For each disease entity, retrieve relevant research literature and extract risk relationship semantic knowledge , merge the relative risk ratios of different research literatures for the same research purpose, use the combined relative risk ratios to represent the relationship between entities, and build an evidence-based semantic knowledge base for chronic disease risks; the implementation of this unit can refer to the above steps (1 );
  • the population exposure rate database construction unit for all entities in the chronic disease risk evidence-based semantic knowledge base, the population exposure rate corresponding to the associated entity is constructed to construct the population exposure rate database; the realization of this unit can refer to the above step (2);
  • Calculation and drawing unit of individual patient chronic disease evolution risk path diagram based on the chronic disease risk evidence-based semantic knowledge base, population exposure rate database and patient individual data standard model, construct a weighted directed graph representing the individual chronic disease evolution risk, directed
  • the vertices of the graph are the patient’s risk factors and possible diseases, corresponding to the entities in the chronic disease risk evidence-based semantic knowledge base.
  • the edge weight of the directed graph is the relative risk ratio between the two entities, and the edge direction is the entity The direction of the influence between them; store the overall relative risk ratio of each vertex in the risk matrix; draw the individual patient's chronic disease evolution risk path diagram according to the weighted directed graph and risk matrix of the individual patient; the realization of this unit can refer to the above Step (4).
  • Interactive analysis module of patient individual chronic disease evolution risk if the individual patient data changes, update the standard model of patient individual data, draw a new patient individual chronic disease evolution risk path map, and visually display the disease risk change on the evolution risk path, providing doctors Provide dynamic and interactive risk analysis of individual patient chronic disease evolution with patients; the implementation of this module can refer to the above step (5).

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Pathology (AREA)
  • Artificial Intelligence (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

本发明公开了一种个体慢病演进风险可视化评估方法及系统，本发明从研究文献中获取风险关系知识，并对同一研究目的的不同研究结果进行合并，从而系统地构建包含各种复杂关系的慢病风险循证语义知识库；基于慢病风险循证语义知识库，结合患者个体数据和人群暴露率数据，计算患者个体慢病演进风险，绘制患者个体慢病演进风险路径图，既考虑患者个体差异，给出个性化评估结果，又确保评估结果有循证支持；本发明提供可视化、交互式的患者个体慢病演进风险分析方法，让信息表达更加清晰、直观、生动，让患者充分了解自身慢病演进风险、干预治疗措施的预期效果，充分参与临床决策，实现医患共同决策，提升患者依从性及治疗效果。

Description

一种个体慢病演进风险可视化评估方法及系统 技术领域
本发明属于医疗信息技术领域，尤其涉及一种个体慢病演进风险可视化评估方法及系统。
背景技术
慢病是慢性非传染性疾病的简称，是一类起病隐匿、病程长且病情迁延不愈、缺乏确切病因证据的疾病的总称，主要包括糖尿病、高血压、心血管疾病和慢性肾病等。慢病的一大特点是合并症、并发症多，且发病率高，有至少一半以上的慢病患者具有合并症或并发症。复杂的并发症会影响患者的生活质量、提高患者医疗成本，严重的甚至直接提高死亡率。许多并发症可以通过改变生活方式、用药等干预治疗方法进行预防，但是需要对患者进行有效的健康教育、患者进行有效的自我管理。
慢病患者去医院就诊时，医生会根据患者问诊情况和临床数据对患者进行诊断，评估患者慢病可能演进的并发症、合并症风险，并给出干预治疗方案，干预治疗方案包括用药治疗、生活方式干预、患者自我管理措施等。对于患者的慢病演进风险评估，医生往往借助临床指南或现有临床决策支持系统。对于患者的干预治疗方案，主要通过口头或医嘱文本传达。存在的不足在于：
1.现有的临床指南或临床决策支持系统往往来自于各个研究文献的循证证据。研究文献一般是针对单个疾病进行相关风险因素或干预治疗的研究并形成循证证据，比如研究吸烟这一风险因素对于增加患高血压风险的影响，或研究跑步这一干预治疗措施对于降低高血压风险的影响。现有的临床指南或临床决策支持系统会针对某疾病筛选、汇集相关的研究文献结论，以文本形式呈现。但是慢病患者存在个体差异，且各种慢病及其并发症、合并症种类繁多、关系复杂，现有方法主要针对单个疾病分别进行风险评估，没有系统全面地评估慢病及其多种并发症、合并症的复杂演进关系。
2.现有存在一些采用机器学习、数据挖掘算法对患者个体数据分析并预测慢病及其并发症、合并症风险的方法，但是这些方法也难以做到对几十种疾病复杂关系的精准考虑，且缺乏循证支持，难以真正应用到临床实际场景中。
3.慢病的特殊性在于许多干预治疗措施需要在院外、日常生活中进行，例如生活方式干预、运动干预、饮食干预等，这需要患者具备自我管理的意愿和能力。医生将评估结果和干预治疗措施通过口头或医嘱文本形式传达给患者，不够清晰、直观，会使得患者对于自身慢病演进风险理 解不全面、不深入，对于医生给出的干预治疗措施的预期效果不了解，从而导致依从性降低，治疗效果不好。
发明内容
本发明的目的在于针对现有技术的不足，提供一种个体慢病演进风险可视化评估方法及系统，为医生和患者呈现针对患者个性化的慢病演进风险评估，帮助医患共同决策。
本发明的目的是通过以下技术方案来实现的：
本发明一方面公开了一种个体慢病演进风险可视化评估方法，该方法包括以下步骤：
(1)构建慢病风险循证语义知识库，包括：构建慢病及其并发症、合并症的疾病实体，以及与疾病相关的风险因素实体，对于每个疾病实体，检索相关研究文献，提取风险关系语义知识，对同一研究目的的不同研究文献的相对风险比进行合并，使用合并后的相对风险比表征实体之间的关系，构建慢病风险循证语义知识库；
(2)针对慢病风险循证语义知识库中的所有实体，关联实体对应的人群暴露率，构建人群暴露率数据库；
(3)构建患者个体数据标准模型，从多个数据源获取患者个体健康数据，将获取的患者个体健康数据进行匹配，抽取、转换、加载到所述患者个体数据标准模型；
(4)计算与绘制患者个体慢病演进风险路径图，包括：基于所述慢病风险循证语义知识库、所述人群暴露率数据库和所述患者个体数据标准模型，构建表征个体慢病演进风险的带权有向图，有向图的顶点为患者已发生的风险因素以及可能发生的疾病，对应所述慢病风险循证语义知识库中的实体，有向图的边权重为两个实体之间的相对风险比，边方向为实体之间产生影响的方向；将各个顶点的总体相对风险比存储在风险矩阵中；根据患者个体的带权有向图和风险矩阵绘制患者个体慢病演进风险路径图；
(5)患者个体慢病演进风险交互式分析，包括：若患者个体数据变化，则更新患者个体数据标准模型，绘制新的患者个体慢病演进风险路径图，可视化地展示演进风险路径上疾病风险变化。
进一步地，步骤(1)中，以慢病及其并发症、合并症的相关研究文献作为知识库的知识来源，以医疗健康标准术语集作为数据语义标识，构建慢病及其并发症、合并症的风险关系语义知识。
进一步地，步骤(1)中，所述慢病风险循证语义知识库的数据结构设计为符合OWL语言格式规范的RDF三元组，每个三元组包括两个实体，表征来源的实体A和表征结果的实体B，以及实体之间的关系C；实体类包括人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状；所述实体B的类为疾病，所述实体A是所述实体B的风险因素，所述实体之间的关系C 包括增加风险和降低风险，使用相对风险比进行表征。
进一步地，步骤(1)中，所述对同一研究目的的不同研究文献的相对风险比进行合并，包括：
(1.1)使用Q检验法对不同研究文献的异质性进行识别，得到检验量Q，检验量Q服从自由度为K-1的卡方分布，K为研究文献总数；使用卡方检验获得卡方值X 2，若Q＞X 2(1-α)，α为置信度，则表明各研究文献间存在异质性，否则表明各研究文献间无异质性；
(1.2)若各研究文献间无异质性，使用固定效应模型Mantel-Hasenzel法计算合并后的相对风险比，否则使用随机效应模型D-L法计算合并后的相对风险比。
进一步地，所述检验量Q的计算公式如下：
Figure PCTCN2022116969-appb-000001
其中，
Figure PCTCN2022116969-appb-000002
为第i个研究文献的权重，Var i为第i个研究文献所包含样本的相对风险比的方差，RR i为第i个研究文献的相对风险比，
Figure PCTCN2022116969-appb-000003
为各研究文献的相对风险比的平均值；
若各研究文献间无异质性，使用固定效应模型Mantel-Hasenzel法计算合并后的相对风险比RR com的公式如下：
Figure PCTCN2022116969-appb-000004
其中，exp(·)为指数函数，ln(·)为对数函数；
若各研究文献间存在异质性，使用随机效应模型D-L法对权重W i进行校正，校正公式如下：
Figure PCTCN2022116969-appb-000005
其中，
Figure PCTCN2022116969-appb-000006
W i′为第i个研究文献校正后的权重值。
进一步地，步骤(3)中，所述患者个体数据标准模型包括个人基本信息，以及人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状。
进一步地，步骤(3)中，所述患者个体健康数据来源于：医疗机构的临床电子病历，个人健康档案，智能硬件、穿戴式设备采集的健康数据，健康问卷数据，慢病管理机构、养老机构的管理数据。
进一步地，步骤(4)中，结合患者个体已发生的所有风险因素，对每个可能发生的疾病的总体相对风险比进行估算，将估算得到的各个顶点的总体相对风险比存储在风险矩阵中，估算方法包括：
(4.1)logistic回归模型估算总体相对风险比：设有向图中以顶点i为终点的边有n条，边的权重分别为r 1，r 2，…，r j，…，r n，则顶点i的总体相对风险比RR f计算公式如下：
RR f＝a+b 1+b 2+…+b j+…+b n
Figure PCTCN2022116969-appb-000007
b j＝ln(r j)
其中，b j为第j条边的回归系数，a以顶点i对应疾病的人群患病率来估计，P为对应疾病的人群患病率，来源于所述人群暴露率数据库；
(4.2)Rothman-Keller模型估算总体相对风险比：设有向图中以顶点i为终点的边有n条，边的权重分别为r 1，r 2，…，r j，…，r n；
计算基准发病比例ρ：
Figure PCTCN2022116969-appb-000008
其中，P j为第j条边对应风险因素的人群暴露率，来源于所述人群暴露率数据库；
计算每个风险因素的风险分数S j：
S j＝ρ×r j
设集合E和集合F，对每个风险因素的风险分数S j，若S j≥1，则将S j插入集合E，若S j＜1，则将S j插入集合F，则顶点i的总体相对风险比RR f计算公式如下：
Figure PCTCN2022116969-appb-000009
其中，m为集合E的元素个数，E i为集合E的第i个元素值，k为集合F的元素个数，F j为集合E的第j个元素值；
(4.3)混合模型估算总体相对风险比：对logistic回归模型和Rothman-Keller模型的估算结果进行加权平均，作为最终的总体相对风险比。
进一步地，步骤(4)中，所述计算与绘制患者个体慢病演进风险路径图，包括：定义画布位 置和尺寸，确定坐标轴比例，选择图表类型，定义节点形状和颜色；节点类型包括人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状，用不同的节点形状区分；对于不同疾病类型，用不同的节点颜色区分；疾病的节点大小通过风险矩阵存储的顶点的总体相对风险比计算，呈线性关系；节点之间的连线粗细通过有向图中边的权重计算，呈线性关系，通过不同的连线颜色区分增加风险和降低风险；将入度为0的节点靠一侧对齐，依次排列下层节点，形成慢病演进风险路径层次。
本发明另一方面公开了一种个体慢病演进风险可视化评估系统，该系统包括：
(1)患者个体数据获取及转换模块：根据患者身份标识，从多个数据源获取患者个体健康数据，将获取的患者个体健康数据进行匹配，抽取、转换、加载到患者个体数据标准模型；
(2)患者个体慢病演进风险可视化评估模块，包括以下单元：
慢病风险循证语义知识库构建单元：构建慢病及其并发症、合并症的疾病实体，以及与疾病相关的风险因素实体，对于每个疾病实体，检索相关研究文献，提取风险关系语义知识，对同一研究目的的不同研究文献的相对风险比进行合并，使用合并后的相对风险比表征实体之间的关系，构建慢病风险循证语义知识库；
人群暴露率数据库构建单元：针对慢病风险循证语义知识库中的所有实体，关联实体对应的人群暴露率，构建人群暴露率数据库；
患者个体慢病演进风险路径图计算与绘制单元：基于慢病风险循证语义知识库、人群暴露率数据库和患者个体数据标准模型，构建表征个体慢病演进风险的带权有向图，有向图的顶点为患者已发生的风险因素以及可能发生的疾病，对应慢病风险循证语义知识库中的实体，有向图的边权重为两个实体之间的相对风险比，边方向为实体之间产生影响的方向；将各个顶点的总体相对风险比存储在风险矩阵中；根据患者个体的带权有向图和风险矩阵绘制患者个体慢病演进风险路径图；
(3)患者个体慢病演进风险交互式分析模块：若患者个体数据变化，则更新患者个体数据标准模型，绘制新的患者个体慢病演进风险路径图，可视化地展示演进风险路径上疾病风险变化，为医生和患者提供动态的、交互式的患者个体慢病演进风险分析。
本发明的有益效果是：
1.针对现有临床指南或临床决策支持系统没有对多种慢病及其并发症、合并症系统全面评估其复杂关系的问题，本发明从大量的研究文献中获取风险关系知识，并对同一研究目的的不同研究结果进行合并，从而系统地构建包含各种复杂关系的慢病风险循证语义知识库。
2.针对现有针对患者个体慢病风险预测方法存在的无法覆盖几十种疾病复杂关系、在临床实际应用中缺乏循证支持的问题，本发明基于慢病风险循证语义知识库，结合患者个体数据和人群暴露率数据，计算患者个体慢病演进风险，并绘制患者个体慢病演进风险路径图，既考虑患者个体差异，给出个性化评估结果，又确保评估结果有循证支持。
3.针对现有医患诊疗中的决策与交流存在的患者理解不全面、不深入、依从性低的问题，本发明基于患者个体慢病演进风险路径图绘制方法，提供可视化、交互式的患者个体慢病演进风险分析方法，让信息表达更加清晰、直观、生动，让患者充分了解自身慢病演进风险、干预治疗措施的预期效果，充分参与临床决策，实现医患共同决策，从而帮助患者依从性以及治疗效果的提升。
附图说明
图1为本发明实施例提供的个体慢病演进风险可视化评估方法流程图；
图2为本发明实施例提供的慢病风险循证语义知识库构建流程图；
图3为本发明实施例提供的患者个体慢病演进风险路径图示例；
图4为本发明实施例提供的个体慢病演进风险可视化评估系统结构图。
具体实施方式
为使本发明的上述目的、特征和优点能够更加明显易懂，下面结合附图对本发明的具体实施方式做详细的说明。
在下面的描述中阐述了很多具体细节以便于充分理解本发明，但是本发明还可以采用其他不同于在此描述的其它方式来实施，本领域技术人员可以在不违背本发明内涵的情况下做类似推广，因此本发明不受下面公开的具体实施例的限制。
本发明实施例提供一种个体慢病演进风险可视化评估方法，如图1所示，包括以下步骤：
(1)构建慢病风险循证语义知识库
基于医学证据的疾病风险关系是评估慢病演进风险的基础，因此需要首先构建慢病风险循证语义知识库。以慢病及其并发症、合并症相关的研究文献作为知识库的知识来源，以医疗健康标准术语集作为数据语义标识，构建慢病及其并发症、合并症的风险关系语义知识。
医疗健康标准术语集可以采用SNOMED CT、ICD-10、UMLS等构建。
慢病风险循证语义知识库的数据结构设计为符合OWL(Web Ontology Language)语言格式规范的RDF(Resource Description Framework)三元组，使得慢病风险知识变为适合于计算机推理的语义结构。每个三元组包括两个实体，表征来源的实体A和表征结果的实体B，以及两个 实体之间的关系C。实体类包括人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状等。非药物干预治疗手段包括行为、运动、饮食等。
本发明构建的慢病风险循证语义知识库主要是为了表征疾病风险关系，三元组中实体B的类为疾病，采用医疗健康标准术语集作为唯一语义标识。实体A是实体B的风险因素，实体A包含上述所有实体类。也就是说，疾病的风险因素既包括人口统计学信息、检查检验、药物、非药物干预治疗手段、症状等实体类，也包括其他疾病。实体之间的关系C主要包括增加风险和降低风险，使用相对风险比(risk ratio,RR)进行表征。相对风险比是指暴露组的发病率与对照组的发病率之比。RR＝1，说明实体A与实体B无关联；RR>1，说明实体A与实体B之间的关系为增加风险；RR<1，说明实体A与实体B之间的关系为降低风险。
如图2所示，根据临床指南与专家意见，构建知识库的实体，包括慢病及其并发症、合并症的疾病实体，以及与这些疾病相关的人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状这些风险因素实体。对每个疾病实体，检索公开发表的相关研究文献(可以采用关键词检索)，构建慢病临床文献库。使用自然语言处理技术获取研究文献中的风险关系语义知识，并将研究文献中提取的各个风险因素与已构建的风险因素实体进行语义映射，从而构建各个疾病的风险关系三元组。慢病风险循证语义知识库使用neo4j进行存储。
在不同的研究文献中，同样两个实体之间的相对风险比可能是不一样的，因此在提取出各研究文献的风险关系语义知识后，借鉴Meta分析的思想，对同一研究目的的不同研究结果进行分析与合并，从而获得最终的最为可信的相对风险比。
1、首先，使用Q检验法对不同研究文献的异质性进行识别，得到检验量Q，计算公式如下：
Figure PCTCN2022116969-appb-000010
其中，K为研究文献总数，W i为第i个研究文献的权重，RR i为第i个研究文献的效应量，也就是相对风险比值，
Figure PCTCN2022116969-appb-000011
为各研究文献的相对风险比值的平均值。
Figure PCTCN2022116969-appb-000012
其中，Var i为第i个研究文献所包含样本的相对风险比值的方差。
检验量Q服从自由度为K-1的卡方分布。使用卡方检验，获得卡方值X 2。若Q＞X 2(1-α)，则表明各研究文献间存在异质性，否则表明各研究文献间无异质性，其中，α为置信度，可设置为0.05。
2、然后，对多个研究文献的相对风险比值结果进行合并。
若各研究文献间无异质性，使用固定效应模型Mantel-Hasenzel法计算合并后的相对风险比值RR com。
Figure PCTCN2022116969-appb-000013
其中，exp(·)为指数函数，ln(·)为对数函数。
若各研究文献间存在异质性，使用随机效应模型D-L法对权重W i进行校正，校正公式如下：
Figure PCTCN2022116969-appb-000014
其中，
Figure PCTCN2022116969-appb-000015
W i′为第i个研究文献校正后的权重值。
(2)构建人群暴露率数据库
针对慢病风险循证语义知识库中的所有实体，关联实体对应的人群暴露率。人群暴露率是指人群中实体的暴露人数与总人数之比。例如，某实体为“吸烟”，则该实体的人群暴露率为人群中吸烟人数除以总人数。对于疾病实体，人群暴露率即为人群患病率。人群暴露率数据库的数据来源可以是公开发布的统计数据、自行对某人群数据库进行统计的数据，人群可以是全国人群、本地人群、某医院人群等，本发明不做具体限定。例如，本系统用于某医院，则数据来源可以是医院全体患者的统计数据；本系统用于某区域，则数据来源可以是区域全体患者的统计数据。
(3)构建患者个体数据标准模型
患者个体健康数据来源于：多医疗机构的临床电子病历，个人健康档案，智能硬件、穿戴式设备采集的健康数据，健康问卷数据，慢病管理机构、养老机构等管理数据等。患者个体数据标准模型包括个人基本信息，以及人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状等信息。将多个数据源中的患者个体健康数据进行匹配，抽取、转换、加载到患者个体数据标准模型。患者个体数据标准模型采用的术语与慢病风险循证语义知识库中的医疗健康标准术语集保持一致。
(4)患者个体慢病演进风险路径图计算与绘制
针对患者个体，基于慢病风险循证语义知识库、人群暴露率数据库和患者个体数据标准模型，构建表征个体慢病演进风险的带权有向图G(V,E)。有向图G的顶点集V中的顶点为患者已发生的风险因素以及可能发生的疾病，对应慢病风险循证语义知识库中的实体；有向图G的边集E 中的边权重为两个实体之间的相对风险比，边的方向为实体之间产生影响的方向。另外，慢病风险循证语义知识库中存储的是单个风险因素与疾病的风险关系，结合患者个体已发生的所有风险因素，可以对每个可能发生的疾病的总体相对风险比进行估算，将估算获得的各个顶点的总体相对风险比存储在风险矩阵M中。
其中，总体相对风险比可以用三种方法进行估算：logistic回归模型、Rothman-Keller模型和混合模型。
1、logistic回归模型估算总体相对风险比
设有向图G中以顶点i为终点的边有n条，边的权重分别为r 1，r 2，…，r j，…，r n。则顶点i的总体相对风险比RR f计算公式如下：
RR f＝a+b 1+b 2+…+b j+…+b n
其中，b j为第j条边的回归系数，计算公式为：
b j＝ln(r j)
a以该顶点对应疾病的人群患病率来估计，计算公式为：
Figure PCTCN2022116969-appb-000016
其中，P为对应疾病的人群暴露率，即人群患病率，来源于上述人群暴露率数据库。
2、Rothman-Keller模型估算总体相对风险比
设有向图G中以顶点i为终点的边有n条，边的权重分别为r 1，r 2，…，r j，…，r n。
首先，计算基准发病比例ρ：
Figure PCTCN2022116969-appb-000017
其中，P j为第j条边对应风险因素在人群中的暴露率，来源于上述人群暴露率数据库。
然后，计算每个风险因素的风险分数S j：
S j＝ρ×r j
最后，计算顶点i的总体相对风险比RR f：
设集合E和集合F，对每个风险因素的风险分数S j，若S j大于等于1，则将S j插入集合E，若S j小于1，则将S j插入集合F，则：
Figure PCTCN2022116969-appb-000018
其中，m为集合E的元素个数，E i为集合E的第i个元素值，k为集合F的元素个数，F j为集合E的第j个元素值。
3、混合模型估算总体相对风险比
混合模型即为logistic回归模型和Rothman-Keller模型结果相结合的模型。对上述两种模型的结果进行加权平均，得到最终的总体相对风险比。权重可以人为根据经验选择，本发明不做具体限定。
将患者个体的带权有向图G和风险矩阵M使用绘图插件echarts进行可视化绘图。定义画布的位置和尺寸，确定坐标轴的比例，选择图表类型，定义节点形状和颜色。节点的类型包括人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状等，用不同的节点形状区分。对于不同疾病类型，例如心血管系统疾病、泌尿系统疾病、内分泌系统疾病、其他疾病等，用不同的节点颜色区分。疾病的节点大小通过风险矩阵M存储的顶点的总体相对风险比值计算，呈线性关系，总体相对风险比值越大，节点越大。节点之间的连线粗细通过带权有向图G中边的权重计算，呈线性关系，反映两个节点之间的相对风险大小，节点之间的连线通过不同的颜色区分增加风险和降低风险。将入度为0的节点靠左对齐，依次排列下层的节点，形成慢病演进风险路径层次。图3为患者个体慢病演进风险路径图的一个示例。
(5)患者个体慢病演进风险交互式分析
上述患者个体慢病演进风险路径图计算与绘制，主要基于患者初始的数据进行计算和绘制。为了让患者更加生动、清晰地了解个体执行干预治疗手段后、或者个体数据变化后演进路径上各个疾病风险的变化，提供患者个体慢病演进风险交互式分析方法，实现动态的、可交互的、可视化的分析。设t时刻，某患者个体数据标准模型为集合A，其中，不可调节的数据为集合B，包括人口统计学等基础信息，可调节的数据为集合C，包括检查检验中的物理检查结果，如体重，还包括药物、非药物干预治疗手段、疾病、症状等数据。A＝B+C。此时刻，根据上述患者个体慢病演进风险路径图计算与绘制方法绘制出对应的路径图D。若t′时刻，患者动态调节了可调节的某些数据，可调节数据集C变为C′，则此时刻的患者个体数据标准模型为A′＝B+C′。此时刻，根据上述患者个体慢病演进风险路径图计算与绘制方法绘制出对应的路径图D′。由于患者调节了某些参数，可以可视化地看到t′时刻相比于t时刻的变化，从而生动、清晰、准确地了解执行干预治疗手段后、或者个体数据变化后演进路径上各个疾病风险的变化。也可以对比不同的慢病干预治疗手段对于慢病及其并发症、合并症的风险，从而选择对该患者最有效的管理方案。医生能够对患者进行有效的解释说明，患者能够更好理解，实现医患共同决策，提升患者依从性。
本发明实施例还提供一种个体慢病演进风险可视化评估系统，如图4所示，该系统包括：
患者个体数据获取及转换模块：根据患者身份标识，从多个数据源获取患者个体健康数据，将获取的患者个体健康数据进行匹配，抽取、转换、加载到患者个体数据标准模型；该模块的实现可以参考上述步骤(3)。
患者个体慢病演进风险可视化评估模块，包括以下单元：
慢病风险循证语义知识库构建单元：构建慢病及其并发症、合并症的疾病实体，以及与疾病相关的风险因素实体，对于每个疾病实体，检索相关研究文献，提取风险关系语义知识，对同一研究目的的不同研究文献的相对风险比进行合并，使用合并后的相对风险比表征实体之间的关系，构建慢病风险循证语义知识库；该单元的实现可以参考上述步骤(1)；
人群暴露率数据库构建单元：针对慢病风险循证语义知识库中的所有实体，关联实体对应的人群暴露率，构建人群暴露率数据库；该单元的实现可以参考上述步骤(2)；
患者个体慢病演进风险路径图计算与绘制单元：基于慢病风险循证语义知识库、人群暴露率数据库和患者个体数据标准模型，构建表征个体慢病演进风险的带权有向图，有向图的顶点为患者已发生的风险因素以及可能发生的疾病，对应慢病风险循证语义知识库中的实体，有向图的边权重为两个实体之间的相对风险比，边方向为实体之间产生影响的方向；将各个顶点的总体相对风险比存储在风险矩阵中；根据患者个体的带权有向图和风险矩阵绘制患者个体慢病演进风险路径图；该单元的实现可以参考上述步骤(4)。
患者个体慢病演进风险交互式分析模块：若患者个体数据变化，则更新患者个体数据标准模型，绘制新的患者个体慢病演进风险路径图，可视化地展示演进风险路径上疾病风险变化，为医生和患者提供动态的、交互式的患者个体慢病演进风险分析；该模块的实现可以参考上述步骤(5)。
以上所述仅是本发明的优选实施方式，虽然本发明已以较佳实施例披露如上，然而并非用以限定本发明。任何熟悉本领域的技术人员，在不脱离本发明技术方案范围情况下，都可利用上述揭示的方法和技术内容对本发明技术方案做出许多可能的变动和修饰，或修改为等同变化的等效实施例。因此，凡是未脱离本发明技术方案的内容，依据本发明的技术实质对以上实施例所做的任何的简单修改、等同变化及修饰，均仍属于本发明技术方案保护的范围内。

Claims (9)

  1. 一种个体慢病演进风险可视化评估方法，其特征在于，包括以下步骤：
    (1)构建慢病风险循证语义知识库，包括：构建慢病及其并发症、合并症的疾病实体，以及与疾病相关的风险因素实体，对于每个疾病实体，检索相关研究文献，提取风险关系语义知识，对同一研究目的的不同研究文献的相对风险比进行合并，使用合并后的相对风险比表征实体之间的关系，构建慢病风险循证语义知识库；
    (2)针对慢病风险循证语义知识库中的所有实体，关联实体对应的人群暴露率，构建人群暴露率数据库；
    (3)构建患者个体数据标准模型，从多个数据源获取患者个体健康数据，将获取的患者个体健康数据进行匹配，抽取、转换、加载到所述患者个体数据标准模型；
    (4)计算与绘制患者个体慢病演进风险路径图，包括：基于所述慢病风险循证语义知识库、所述人群暴露率数据库和所述患者个体数据标准模型，构建表征个体慢病演进风险的带权有向图，有向图的顶点为患者已发生的风险因素以及可能发生的疾病，对应所述慢病风险循证语义知识库中的实体，带权有向图的边权重为两个实体之间的相对风险比，边方向为实体之间产生影响的方向；结合患者个体已发生的所有风险因素，对每个可能发生的疾病的总体相对风险比进行估算，将估算得到的各个顶点的总体相对风险比存储在风险矩阵中；根据患者个体的带权有向图和风险矩阵绘制患者个体慢病演进风险路径图；总体相对风险比的估算方法包括：
    (4.1)logistic回归模型估算总体相对风险比：设带权有向图中以顶点i为终点的边有n条，边的权重分别为r 1，r 2，...，r j，...，r n，则顶点i的总体相对风险比RR f计算公式如下：
    RR f＝a+b 1+b 2+...+b j+...+b n
    Figure PCTCN2022116969-appb-100001
    b j＝ln(r j)
    其中，b j为第j条边的回归系数，a以顶点i对应疾病的人群患病率来估计，P为对应疾病的人群患病率，来源于所述人群暴露率数据库；
    (4.2)Rothman-Keller模型估算总体相对风险比：设带权有向图中以顶点i为终点的边有n条，边的权重分别为r 1，r 2，...，r j，...，r n；
    计算基准发病比例ρ：
    Figure PCTCN2022116969-appb-100002
    其中，P j为第j条边对应风险因素的人群暴露率，来源于所述人群暴露率数据库；
    计算每个风险因素的风险分数S j：
    S j＝ρ×r j
    设集合E和集合F，对每个风险因素的风险分数S j，若S j≥1，则将S j插入集合E，若S j＜1，则将S j插入集合F，则顶点i的总体相对风险比RR f计算公式如下：
    Figure PCTCN2022116969-appb-100003
    其中，m为集合E的元素个数，E i为集合E的第i个元素值，k为集合F的元素个数，F j为集合E的第j个元素值；
    (4.3)混合模型估算总体相对风险比：对logistic回归模型和Rothman-Keller模型的估算结果进行加权平均，作为最终的总体相对风险比；
    (5)患者个体慢病演进风险交互式分析，包括：若患者个体数据变化，则更新患者个体数据标准模型，绘制新的患者个体慢病演进风险路径图，可视化地展示演进风险路径上疾病风险变化。
  2. 根据权利要求1所述的一种个体慢病演进风险可视化评估方法，其特征在于，步骤(1)中，以慢病及其并发症、合并症的相关研究文献作为慢病风险循证语义知识库的知识来源，以医疗健康标准术语集作为数据语义标识，构建慢病及其并发症、合并症的风险关系语义知识。
  3. 根据权利要求1所述的一种个体慢病演进风险可视化评估方法，其特征在于，步骤(1)中，所述慢病风险循证语义知识库的数据结构设计为符合OWL语言格式规范的RDF三元组，每个三元组包括两个实体，表征来源的实体A和表征结果的实体B，以及实体之间的关系C；实体类包括人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状；所述实体B的类为疾病，所述实体A是所述实体B的风险因素，所述实体之间的关系C包括增加风险和降低风险，使用相对风险比进行表征。
  4. 根据权利要求1所述的一种个体慢病演进风险可视化评估方法，其特征在于，步骤(1)中，所述对同一研究目的的不同研究文献的相对风险比进行合并，包括：
    (1.1)使用Q检验法对不同研究文献的异质性进行识别，得到检验量Q，检验量Q服从自由度为K-1的卡方分布，K为研究文献总数；使用卡方检验获得卡方值X 2，若Q＞X 2(1-α)，α为 置信度，则表明各研究文献间存在异质性，否则表明各研究文献间无异质性；
    (1.2)若各研究文献间无异质性，使用固定效应模型Mantel-Hasenzel法计算合并后的相对风险比，否则使用随机效应模型D-L法计算合并后的相对风险比。
  5. 根据权利要求4所述的一种个体慢病演进风险可视化评估方法，其特征在于，所述检验量Q的计算公式如下：
    Figure PCTCN2022116969-appb-100004
    其中，
    Figure PCTCN2022116969-appb-100005
    为第i个研究文献的权重，Var i为第i个研究文献所包含样本的相对风险比的方差，RR i为第i个研究文献的相对风险比，
    Figure PCTCN2022116969-appb-100006
    为各研究文献的相对风险比的平均值；
    若各研究文献间无异质性，使用固定效应模型Mantel-Hasenzel法计算合并后的相对风险比RR com的公式如下：
    Figure PCTCN2022116969-appb-100007
    其中，exp(·)为指数函数，ln(·)为对数函数；
    若各研究文献间存在异质性，使用随机效应模型D-L法对权重W i进行校正，校正公式如下：
    Figure PCTCN2022116969-appb-100008
    其中，
    Figure PCTCN2022116969-appb-100009
    W i′为第i个研究文献校正后的权重值。
  6. 根据权利要求1所述的一种个体慢病演进风险可视化评估方法，其特征在于，步骤(3)中，所述患者个体数据标准模型包括个人基本信息，以及人口统计学信息、检查检验、药物、非药物干预治疗手段、疾病、症状。
  7. 根据权利要求1所述的一种个体慢病演进风险可视化评估方法，其特征在于，步骤(3)中，所述患者个体健康数据来源于：医疗机构的临床电子病历，个人健康档案，智能硬件、穿戴式设备采集的健康数据，健康问卷数据，慢病管理机构、养老机构的管理数据。
  8. 根据权利要求1所述的一种个体慢病演进风险可视化评估方法，其特征在于，步骤(4)中，所述计算与绘制患者个体慢病演进风险路径图，包括：定义画布位置和尺寸，确定坐标轴比例，选择图表类型，定义节点形状和颜色；节点类型包括人口统计学信息、检查检验、药物、非药物 干预治疗手段、疾病、症状，用不同的节点形状区分；对于不同疾病类型，用不同的节点颜色区分；疾病的节点大小通过风险矩阵存储的顶点的总体相对风险比计算，呈线性关系；节点之间的连线粗细通过带权有向图中边的权重计算，呈线性关系，通过不同的连线颜色区分增加风险和降低风险；将入度为0的节点靠一侧对齐，依次排列下层节点，形成慢病演进风险路径层次。
  9. 一种个体慢病演进风险可视化评估系统，其特征在于，包括：
    (1)患者个体数据获取及转换模块：根据患者身份标识，从多个数据源获取患者个体健康数据，将获取的患者个体健康数据进行匹配，抽取、转换、加载到患者个体数据标准模型；
    (2)患者个体慢病演进风险可视化评估模块，包括以下单元：
    慢病风险循证语义知识库构建单元：构建慢病及其并发症、合并症的疾病实体，以及与疾病相关的风险因素实体，对于每个疾病实体，检索相关研究文献，提取风险关系语义知识，对同一研究目的的不同研究文献的相对风险比进行合并，使用合并后的相对风险比表征实体之间的关系，构建慢病风险循证语义知识库；
    人群暴露率数据库构建单元：针对慢病风险循证语义知识库中的所有实体，关联实体对应的人群暴露率，构建人群暴露率数据库；
    患者个体慢病演进风险路径图计算与绘制单元：基于慢病风险循证语义知识库、人群暴露率数据库和患者个体数据标准模型，构建表征个体慢病演进风险的带权有向图，带权有向图的顶点为患者已发生的风险因素以及可能发生的疾病，对应慢病风险循证语义知识库中的实体，带权有向图的边权重为两个实体之间的相对风险比，边方向为实体之间产生影响的方向；结合患者个体已发生的所有风险因素，对每个可能发生的疾病的总体相对风险比进行估算，将估算得到的各个顶点的总体相对风险比存储在风险矩阵中；根据患者个体的带权有向图和风险矩阵绘制患者个体慢病演进风险路径图；总体相对风险比的估算方法包括：
    a.logistic回归模型估算总体相对风险比：设带权有向图中以顶点i为终点的边有n条，边的权重分别为r 1，r 2，...，r j，...，r n，则顶点i的总体相对风险比RR f计算公式如下：
    RR f＝a+b 1+b 2+...+b j+...+b n
    Figure PCTCN2022116969-appb-100010
    b j＝ln(r j)
    其中，b j为第j条边的回归系数，a以顶点i对应疾病的人群患病率来估计，P为对应疾病的人群患病率，来源于所述人群暴露率数据库；
    b.Rothman-Keller模型估算总体相对风险比：设带权有向图中以顶点i为终点的边有n条，边的权重分别为r 1，r 2，...，r j，...，r n；
    计算基准发病比例ρ：
    Figure PCTCN2022116969-appb-100011
    其中，P j为第j条边对应风险因素的人群暴露率，来源于所述人群暴露率数据库；
    计算每个风险因素的风险分数S j：
    S j＝ρ×r j
    设集合E和集合F，对每个风险因素的风险分数S j，若S j≥1，则将S j插入集合E，若S j＜1，则将S j插入集合F，则顶点i的总体相对风险比RR f计算公式如下：
    Figure PCTCN2022116969-appb-100012
    其中，m为集合E的元素个数，E i为集合E的第i个元素值，k为集合F的元素个数，F j为集合E的第j个元素值；
    c.混合模型估算总体相对风险比：对logistic回归模型和Rothman-Keller模型的估算结果进行加权平均，作为最终的总体相对风险比；
    (3)患者个体慢病演进风险交互式分析模块：若患者个体数据变化，则更新患者个体数据标准模型，绘制新的患者个体慢病演进风险路径图，可视化地展示演进风险路径上疾病风险变化。
PCT/CN2022/116969 2021-12-14 2022-09-05 一种个体慢病演进风险可视化评估方法及系统 WO2023109199A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2023538042A JP7430295B2 (ja) 2021-12-14 2022-09-05 個体慢性疾患進行リスク可視化評価方法及びシステム

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111523880.7A CN113921141B (zh) 2021-12-14 2021-12-14 一种个体慢病演进风险可视化评估方法及系统
CN202111523880.7 2021-12-14

Publications (1)

Publication Number Publication Date
WO2023109199A1 true WO2023109199A1 (zh) 2023-06-22

Family

ID=79249112

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/116969 WO2023109199A1 (zh) 2021-12-14 2022-09-05 一种个体慢病演进风险可视化评估方法及系统

Country Status (3)

Country Link
JP (1) JP7430295B2 (zh)
CN (1) CN113921141B (zh)
WO (1) WO2023109199A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116959715A (zh) * 2023-09-18 2023-10-27 之江实验室 一种基于时序演进过程解释的疾病预后预测系统
CN117198527A (zh) * 2023-08-24 2023-12-08 北京大学人民医院 一种亲缘造血干细胞移植术后的风险评估系统及方法
CN117912708A (zh) * 2024-03-14 2024-04-19 苏州市相城区疾病预防控制中心 公共卫生中慢性非传染性疾病防控方法、系统及存储介质
CN117936114A (zh) * 2024-03-13 2024-04-26 湖北福鑫科创信息技术有限公司 基于大语言模型的电子病历智能化分析与优化系统

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113921141B (zh) * 2021-12-14 2022-04-08 之江实验室 一种个体慢病演进风险可视化评估方法及系统
CN115036034B (zh) * 2022-08-11 2022-11-08 之江实验室 一种基于患者表征图的相似患者识别方法及系统

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108877943A (zh) * 2018-06-21 2018-11-23 天津医科大学 基于循证医学证据的ii型糖尿病风险评估模型
CN110718302A (zh) * 2019-10-22 2020-01-21 江苏健康无忧网络科技有限公司 一种基于大数据的糖尿病管理路径的系统
CN111507827A (zh) * 2020-04-20 2020-08-07 上海商涌网络科技有限公司 一种健康风险评估的方法、终端及计算机存储介质
CN112102937A (zh) * 2020-11-13 2020-12-18 之江实验室 一种慢性病辅助决策的患者数据可视化方法及系统
CN112336358A (zh) * 2020-04-30 2021-02-09 中山大学孙逸仙纪念医院 一种预测致密型乳腺的乳腺病灶恶性风险模型及其构建方法
CN112820415A (zh) * 2021-02-08 2021-05-18 郑州大学 一种基于gis的慢性病时空演化特征分析及环境健康风险监测系统及方法
US20210272696A1 (en) * 2020-03-02 2021-09-02 University Of Cincinnati System, method computer program product and apparatus for dynamic predictive monitoring in the critical health assessment and outcomes study (chaos)
US20210375472A1 (en) * 2020-06-01 2021-12-02 University Of Washington Methods and systems for decision support
CN113921141A (zh) * 2021-12-14 2022-01-11 之江实验室 一种个体慢病演进风险可视化评估方法及系统

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004530971A (ja) 2001-02-21 2004-10-07 デルファイ ヘルス システムズ，インコーポレイテッド 慢性疾患のアウトカム、教育、および連絡のためのシステム
EP2788925A4 (en) * 2011-12-09 2015-08-26 Vis Res Inst Tecnologias E Servicos Para Pesquisa Clinika S A SYSTEM AND METHOD FOR EVALUATING AND MARKETING CLINICAL RESEARCH CENTERS
CN103559397A (zh) * 2013-10-31 2014-02-05 合肥学院 一种应用于高血压慢性病的智慧服务系统及方法
CN108511070A (zh) * 2018-04-18 2018-09-07 郑州大学第附属医院 一种糖尿病患者评估及管理系统
JP7107375B2 (ja) 2018-08-31 2022-07-27 日本電信電話株式会社 状態遷移予測装置、予測モデル学習装置、方法およびプログラム
CN110277167A (zh) * 2019-05-31 2019-09-24 南京邮电大学 基于知识图谱的慢性非传染性疾病风险预测系统
CN111370127B (zh) * 2020-01-14 2022-06-10 之江实验室 一种基于知识图谱的跨科室慢性肾病早期诊断决策支持系统
CN112133445A (zh) * 2020-10-21 2020-12-25 万达信息股份有限公司 一种心血管疾病管理服务方法和系统
CN113626624B (zh) * 2021-10-12 2021-12-21 腾讯科技（深圳）有限公司 一种资源识别方法和相关装置
CN113643821B (zh) * 2021-10-13 2022-02-11 浙江大学 一种多中心知识图谱联合决策支持方法与系统

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108877943A (zh) * 2018-06-21 2018-11-23 天津医科大学 基于循证医学证据的ii型糖尿病风险评估模型
CN110718302A (zh) * 2019-10-22 2020-01-21 江苏健康无忧网络科技有限公司 一种基于大数据的糖尿病管理路径的系统
US20210272696A1 (en) * 2020-03-02 2021-09-02 University Of Cincinnati System, method computer program product and apparatus for dynamic predictive monitoring in the critical health assessment and outcomes study (chaos)
CN111507827A (zh) * 2020-04-20 2020-08-07 上海商涌网络科技有限公司 一种健康风险评估的方法、终端及计算机存储介质
CN112336358A (zh) * 2020-04-30 2021-02-09 中山大学孙逸仙纪念医院 一种预测致密型乳腺的乳腺病灶恶性风险模型及其构建方法
US20210375472A1 (en) * 2020-06-01 2021-12-02 University Of Washington Methods and systems for decision support
CN112102937A (zh) * 2020-11-13 2020-12-18 之江实验室 一种慢性病辅助决策的患者数据可视化方法及系统
CN112820415A (zh) * 2021-02-08 2021-05-18 郑州大学 一种基于gis的慢性病时空演化特征分析及环境健康风险监测系统及方法
CN113921141A (zh) * 2021-12-14 2022-01-11 之江实验室 一种个体慢病演进风险可视化评估方法及系统

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117198527A (zh) * 2023-08-24 2023-12-08 北京大学人民医院 一种亲缘造血干细胞移植术后的风险评估系统及方法
CN117198527B (zh) * 2023-08-24 2024-02-23 北京大学人民医院 一种亲缘造血干细胞移植术后的风险评估系统及方法
CN116959715A (zh) * 2023-09-18 2023-10-27 之江实验室 一种基于时序演进过程解释的疾病预后预测系统
CN116959715B (zh) * 2023-09-18 2024-01-09 之江实验室 一种基于时序演进过程解释的疾病预后预测系统
CN117936114A (zh) * 2024-03-13 2024-04-26 湖北福鑫科创信息技术有限公司 基于大语言模型的电子病历智能化分析与优化系统
CN117912708A (zh) * 2024-03-14 2024-04-19 苏州市相城区疾病预防控制中心 公共卫生中慢性非传染性疾病防控方法、系统及存储介质
CN117912708B (zh) * 2024-03-14 2024-05-17 苏州市相城区疾病预防控制中心 公共卫生中慢性非传染性疾病防控方法、系统及存储介质

Also Published As

Publication number Publication date
JP2023553504A (ja) 2023-12-21
CN113921141A (zh) 2022-01-11
CN113921141B (zh) 2022-04-08
JP7430295B2 (ja) 2024-02-09

Similar Documents

Publication Publication Date Title
WO2023109199A1 (zh) 一种个体慢病演进风险可视化评估方法及系统
Krumholz et al. Do non-clinical factors improve prediction of readmission risk? results from the Tele-HF study
Raghupathi et al. An overview of health analytics
Cole et al. Profiling risk factors for chronic uveitis in juvenile idiopathic arthritis: a new model for EHR-based research
Brennan et al. Observing health in everyday living: ODLs and the care-between-the-care
Shenas et al. Identifying high-cost patients using data mining techniques and a small set of non-trivial attributes
Parimbelli et al. A review of AI and Data Science support for cancer management
CN113345583A (zh) 一种构建全生命周期居民智慧健康档案的方法及系统
Hunter-Zinck et al. Predicting emergency department orders with multilabel machine learning techniques and simulating effects on length of stay
US11875884B2 (en) Expression of clinical logic with positive and negative explainability
Shaw et al. Timing of onset, burden, and postdischarge mortality of persistent critical illness in Scotland, 2005–2014: a retrospective, population-based, observational study
KR20200113954A (ko) 사용자 맞춤형 건강 정보 서비스 제공 시스템 및 그 방법
Nietert et al. Using a summary measure for multiple quality indicators in primary care: the Summary QUality InDex (SQUID)
Wongtangman et al. Development and validation of a machine learning ASA-score to identify candidates for comprehensive preoperative screening and risk stratification
RU2752792C1 (ru) Система для поддержки принятия врачебных решений
US20170186120A1 (en) Health Care Spend Analysis
US11238988B2 (en) Large scale identification and analysis of population health risks
CN114566284A (zh) 疾病预后风险预测模型训练方法、装置及电子设备
Nie et al. Forecasting medical state transition using machine learning methods
Gauthier et al. Challenges to building a platform for a breast cancer risk score
Alghamdi Health data warehouses: reviewing advanced solutions for medical knowledge discovery
US20240086771A1 (en) Machine learning to generate service recommendations
Velichkovska et al. A Survey of Bias in Healthcare: Pitfalls of Using Biased Datasets and Applications
Srinivasan Essays on Digital Health and Preventive Care Analytics
Saravi et al. Machine Learning in Apache Spark Environment for Diagnosis of Diabetes

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 2023538042

Country of ref document: JP

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22905954

Country of ref document: EP

Kind code of ref document: A1