CN102156812A - Hospital decision-making aiding method based on symptom similarity analysis - Google Patents

Abstract

The medical-seeking assistant decision-making method based on symptom similarity analysis of the invention belongs to the field of information analysis and assistant decision-making. First, establish the disease database, symptom database, department database and their associations, establish a patient model for the symptoms entered by the user, perform similarity calculations with the disease model, perform reasoning and sort suspected diseases, and search for corresponding suspected diseases in the department model. The department of the disease, as a reference for consultation. The beneficial effect of the present invention is that it can weaken the wrong interference result when the symptoms mistaken by the user are taken as inference conditions, and according to the wrong symptom conditions, combined with the existing correlation between symptoms, the correct symptoms can be involved Reasoning operations, so as to achieve the correctness of reasoning and anti-interference, and use the mapping relationship between diseases and departments to give a list of departments that should be treated. It can be used in electronic triage or health consultation system, and can also be used in medical training or clinical reference.

Description

Medical aid decision-making method based on symptom similarity analysis

technical field

The invention relates to an auxiliary decision-making method for medical treatment, belongs to the field of information analysis and auxiliary decision-making, and can be used in a triage system.

Background technique

Five major crises in modern medical diagnosis: (1) The continuous increase of medical expenses exceeds the capacity of individuals and society; (2) The explosive growth and confusion of knowledge, young doctors cannot quickly master it, even experienced doctors It is difficult to master the knowledge of diagnosis and treatment across departments; (3) The geographical distribution of medical experts is uneven, and most areas lack high-level doctors; (4) The increasing number of people who need high-level medical services has a strong conflict with the high-quality services that may be provided. (5) Despite the continuous use of high-tech diagnostic techniques, the rate of misdiagnosis and mistreatment remains high. Solving the five major crises in modern medical diagnosis, especially how to effectively reduce medical diagnosis costs and misdiagnosis and mistreatment rates, should be the leading direction of research and development of information technology in the field of medical diagnosis, which is also the original intention of the present invention.

In addition, triage is the primary link that patients experience when they seek medical treatment in the hospital. Only by correct triage and diverting patients to appropriate departments can patients receive appropriate treatment. , It also causes unnecessary waste of medical resources. In reality, which department a patient goes to see a doctor (or which specialized hospital to choose) depends mainly on which department the disease causing the patient's discomfort belongs to. Therefore, judging the suspected disease is the key to triage. At this stage, hospital triage is completely realized manually. On the one hand, it is the patient's self-judgment, and on the other hand, it is through the manual consultation at the hospital triage desk. The correct triage work requires professional doctors to perform triage based on the patient's uncomfortable symptoms and medical experience, and according to the symptoms described by the patient object. factor interference.

In recent years, with the development of information technology, some new methods and new products for auxiliary diagnosis have emerged. Symptoms can help determine what kind of disease you are suffering from. In these systems, a series of symptom sets are usually defined in advance for users to choose to appear on their own. Symptoms, in order to improve the accuracy of diagnosis, the increase of symptom items and expressions usually leads to a huge database, difficulty in thesaurus classification, or cumbersome operations, resulting in unpleasant user experience; therefore, an auxiliary diagnosis system that allows patients to actively input symptoms has emerged However, the symptoms described by patients often deviate from the symptoms in the database due to lack of medical knowledge or the existence of language habits and diverse expressions.

Contents of the invention

The purpose of the present invention is to provide a kind of medical aided decision-making method based on symptom similarity analysis, with fast and accurate solution public (for the crowd of lack of medical knowledge or language habit and expression diversification) because wrong or inaccurate description symptom causes The triage results are wrong. Through the mapping of symptoms to diseases, and the mapping of diseases to departments, the function of assisting triage and medical treatment decision-making is realized.

The medical aid decision-making method based on symptom similarity analysis of the present invention comprises the following steps:

Step 1: According to the canonical names of the symptoms in the medical knowledge base, supplement the symptom entries that have the upper and lower, homonym and other correlations with these basic symptoms, take the basic symptom entries and the supplementary symptom entries as the constituent elements of the symptom library, and establish and Save the association relationship between every two symptoms in the symptom database; the association relationship between each two symptoms includes the upper and lower, and the same position association; wherein the upper and lower relationship includes the affiliation between symptoms, and the same position Relationships include synonymous, near-synonymous and similar relationships;

Step 2: The set of symptom items related to each disease constitutes a disease model (that is, the set of symptoms exhibited by each disease, and the set elements are the symptoms that may occur when suffering from the disease), and each disease model is used as a constituent element of the disease library ; For example, the model of suspected disease A is α=(a ₁ , a ₂ ,..., a _k ,..., a _m ), including m symptoms;

Step 3: Construct each disease model as a disease vector, the construction method is:

其中是该疾病模型中症状a_k的权值，其取值表示症状a_k与该疾病之间的关联程度，权值越大则关联程度越高；是根据临床经验和专家数据预先设置的；Constitute the disease vector with the weights of all the symptoms contained in the disease model

in is the weight of symptom a _k in the disease model, and its value indicates the degree of association between symptom a _k and the disease. The greater the weight, the higher the degree of association; it is preset based on clinical experience and expert data;

Step 4: The collection of diseases that can be treated by each department of the hospital constitutes a department model (that is, the collection of diseases that each department is suitable for receiving, and the collection elements are the collection of diseases that can be diagnosed in this department), and the models of each department As a constituent element of the department library; for example, the model of department C is γ=(c ₁ , c ₂ ,...,c _p ,...,c _q ), which contains q diseases in total;

其中

表示科室模型C中的疾病c_p的权值，对其赋值采用如下两种方法之一：Step 5: Construct the department vector with the weights of all diseases contained in the department model C

in

Represents the weight of the disease c _p in the department model C, and one of the following two methods is used for its assignment:

(1) Assign values according to the suitability of various diseases that the department can treat, and the more diseases the department treats, the greater the weight;

(2) Each disease in the department model C adopts the same weight, that is, it does not distinguish the degree of indications of diseases that can be treated by the department;

Step 6: Construct a patient model from the collection of all the symptoms provided by the current patient; that is, the patient model B is β=(b ₁ , b ₂ ,..., b _j ,..., b _n ), containing n symptoms in total;

Step 7: Construct the current patient model as a patient vector; the construction method is:

Construct the patient vector with the weights of all the symptoms contained in the current patient model B in Represents the weight of symptom b _j in symptom model B, and one of the following two methods is used for its assignment:

(1) Assign a value according to the order in which the user enters the symptom or the main degree of subjective feeling among all the symptoms. The more important the symptom or the higher the input order, the greater the weight;

(2) Each symptom in patient model B adopts the same weight, that is, no distinction is made between the input order and main degree of symptoms;

Step 8: Use one of the following two methods to obtain the probability that the patient has a suspected disease:

与步骤3构建的各疾病向量之间的角度，向量之间角度越小则认为患者患有该疾病的可能性越大；(1) When the patient vector and the disease vector need to have the same dimension, that is, the patient model and the disease model need to have the same number of symptoms, and the same symptoms should be located in the same position in the two models, that is, a _k with the same subscript and b _j should have the same meaning in case the patient vector is computed

The angle between each disease vector constructed in step 3, the smaller the angle between the vectors, the greater the possibility that the patient has the disease;

(2) Calculate the similarity between the current patient model B and each disease model. The greater the similarity, the greater the possibility that the patient has the disease. The similarity between disease model A and patient model B is as follows formula to calculate:

$\mathrm{<span\; class="notranslate">\; Sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; kj</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; \{</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; kj</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; jk</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>}^{\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$

是该疾病模型中症状a_k的权值，T_kj表示患者模型B中的症状b_j与待比对的疾病模型A中的症状a_k之间的距离，该距离依据步骤1建立的症状库所规定的症状之间的关联关系获得；in,

is the weight of the symptom a _k in the disease model, T _kj represents the distance between the symptom b _j in the patient model B and the symptom a _k in the disease model A to be compared, and the distance is based on the symptom library established in step 1 The association between the prescribed symptoms is obtained;

表示疾病模型A中症状a_k和a_j之间的距离，

表示患者模型B中症状b_j和b_k之间的距离。这两个距离也依据步骤1建立的症状库所规定的症状之间的关联关系获得；

Indicates the distance between symptoms a _k and a _j in disease model A,

Denotes the distance between symptoms b _j and b _k in patient model B. These two distances are also obtained according to the relationship between the symptoms specified in the symptom library established in step 1;

Preferably, before step 8 is carried out, it also includes coarse screening of the disease model in advance, and then performing step 8 on the suspected disease set obtained by the coarse screening, that is, calculating the disease of each disease in the current patient model B and the suspected disease set obtained by the coarse screening similarity between models.

Step 9: According to the possibility of the patient suffering from each disease given in step 8, sort from high to low, and use the top H diseases with the highest possibility as the final auxiliary diagnosis result. Preferably, H is less than or equal to 3;

Step 10: For the H diseases determined in step 9, search for the department corresponding to each disease one by one in the department database, that is, search for the department with the disease in the department database and give the result; as a preference, when step 5 adopts In the method (1), it also includes sorting the found H department results according to the weight value of the disease in the corresponding department vector, and sorting them into a department list from high to low.

Compared with the prior art, the beneficial effect of the present invention is that it can weaken the wrong interference results when taking the symptoms that the user mistakenly thinks as the reasoning conditions, and according to the wrong symptom conditions, combined with the existing correlation between symptoms, Let the correct symptoms participate in the reasoning operation, so as to achieve the correctness and anti-interference of reasoning, and use the mapping relationship between diseases and departments to give a list of departments that should be treated. The method can be used in electronic triage or health consultation system, and can also be used in medical training or clinical reference.

Description of drawings

Fig. 1 is a schematic diagram of the inference of disease-symptom relationship in the present invention;

Figure 2 is a schematic diagram of the spatial distribution of symptoms.

Detailed ways

The content of the present invention will be explained below in conjunction with the accompanying drawings.

In this technical solution, the triage function is realized through the mapping of symptoms to diseases and the mapping of diseases to hospitals/departments. First, construct the disease database and symptom database. The disease database and the symptom database are collections of disease and symptom entries respectively, and the association between them is established and related description information is included. There is a many-to-many relationship between symptoms and diseases. Each symptom corresponds to multiple diseases, and each disease corresponds to multiple symptoms. Different combinations of symptoms may be caused by different diseases; each disease will show one or several symptoms . Diseases and departments can have a many-to-one relationship or a many-to-many relationship. Each disease corresponds to one or more departments, and the same department can diagnose and treat multiple diseases. As shown in Figure 1. Of course, diseases and departments can also be designed as a many-to-many relationship. For example, there are multiple options for receiving departments for a certain disease, such as the Department of Traditional Chinese Medicine and the Department of Infectious Diseases. For example, for hematuria symptoms, you can go to the Department of Nephrology or Urology. Surgery visit.

Symptoms are the main feature of a suspected disease. In this technical solution, the patient is modeled according to the symptom information of each patient. Modeling is to record and organize the symptom information input by the patient, and use it to infer the disease. All symptom collections provided by each user (patient) constitute a patient model, and the symptom information is composed of the canonical names of symptoms in the medical knowledge base, and each disease-related symptom collection constitutes a disease model. The disease library consists of multiple disease models.

Suspected diseases are judged by calculating the probability that the patient model falls within the disease model.

Let E denote a specific symptom, H _j denote a disease,

According to Bayes formula:

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>}$

The meaning of using the Bayes conditional probability formula is:

P(H _j |E) When a patient has symptom E, the probability of suffering from disease H _j .

P(E|H _j ) is the probability that the patient suffers from disease H _j and has symptom E.

P(H _j ) is the prior probability of the patient suffering from disease H _j . The prior probability P(H _j ) is not easy to determine. Even if medical experts have spent a lot of effort to determine the prior probability, the prior probability P(H _j ) of suffering from the disease at different times and in different places is still different. The usual practice is to follow the usual situation without distinguishing special cases. Condition.

The premise of using the Bayes formula is that the antecedent can be correctly input, that is, the user must input the correct symptom and completely match the symptom entry in the symptom database. However, in reality, a patient may make a wrong judgment because of his own symptoms, resulting in erroneous item input, thus leading to wrong conclusions; because the present invention is aimed at the self-symptom description of the public, because most users lack medical professional knowledge and cannot accurately identify symptoms, it is very easy to cause deviations in symptom description, for example: vomiting and Nausea confusion, symptoms of dizziness and fatigue caused by psychological panic about suspected diseases. Based on the analysis of the above situation, we cannot simply copy Bayes' calculation formula, and we need to have certain identification or processing ability to exclude interfering symptoms in auxiliary diagnosis.

In the preferred solution of the present invention, after the formula calculated by Bayes, the suspected diseases with a probability higher than 0.2 are recalculated and sorted according to step 6, and the space vector model of the input antecedent deviation can be adapted.

We regard the collection of symptoms exhibited by each disease (the collection elements are the symptoms that may occur when suffering from the disease) as a vector, and the collection of symptoms entered by the user is also regarded as a vector, and each vector element is a symptom. Symptoms There is a relationship between upper and lower, same position and so on. The hyponym relationship refers to the subordination relationship between symptoms, and the homonym relationship includes synonymous, near-synonymous, similar, and easily mixed relationships.

Therefore, in the present invention, firstly, a symptom bank with associations such as hyponym and homonym will be established according to the basic symptom information. The basic symptom information is composed of the canonical name of the symptom database in the medical knowledge base, as shown in FIG. 2 . The established symptom library is a collection of subjective abnormal sensations caused by a series of abnormal changes in the body's function, metabolism, and morphological structure during the disease process, such as vomiting, pain, and chills. The same discomfort symptom may have different expressions when subjectively described, such as pain, pain, pain. There will be deviations or even errors in the subjective description of similar symptoms, such as nausea and vomiting, redness and cyanosis, that is, the homonym relationship.

According to the method described in the present invention, a symptom library and a disease library are established, and a disease model and a patient model are established;

The disease model of suspected disease A is α=(a ₁ , a ₂ ,..., a _k ,..., a _m ), including m symptoms;

Patient model B is β=(b ₁ , b ₂ ,..., b _j ,..., b _n ), which contains n symptoms in total; use one of the following two methods described in step 8 to obtain the patient's suspected disease possibility:

与步骤3构建的各疾病向量之间的角度，向量之间角度越小则认为患者患有该疾病的可能性越大；特别说明的是，在此情况下，待进行角度计算的，患者向量和疾病向量需具有同样的维度，即患者模型和疾病模型需具有同样的症状个数，并且相同症状应在这两个模型中位于相同的位置，即相同下标的a_k和b_j应具有相同的含义；(1) Calculate the patient vector

The angle between the disease vectors and the disease vectors constructed in step 3, the smaller the angle between the vectors, the greater the possibility of the patient suffering from the disease; in particular, in this case, the patient vector to be calculated for the angle and the disease vector must have the same dimension, that is, the patient model and the disease model must have the same number of symptoms, and the same symptoms should be located in the same position in the two models, that is, a _k and b _j with the same subscript should have the same the meaning of

未有的症状表述为0，该患者的所有症状的权重构成一个在向量空间的坐标

计算患者向量与各个疾病向量的角度，角度越小则认为该患者患有该疾病的可能性越大。This is because in the vector space of the disease model and the patient model, the position of the disease a _k that needs to be compared and the symptoms of the patient is expressed as

Unexisting symptoms are expressed as 0, and the weights of all symptoms of the patient form a coordinate in the vector space

Calculate the angle between the patient vector and each disease vector, the smaller the angle, the greater the possibility that the patient suffers from the disease.

However, when calculating the cosine of the angle between two vectors, the dimensions of the two vectors must be the same, otherwise it cannot be calculated. The cosine of the angle between the vectors can be obtained by the following method. During the calculation, we compare the inner product and the vector The calculation of the length has been improved (see step 8 for the second method below);

(2) Calculate the similarity between the current patient model B and each disease model. The greater the similarity, the greater the possibility that the patient has the disease. The similarity between disease model A and patient model B is as follows formula to calculate:

$\mathrm{<span\; class="notranslate">\; Sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; kj</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; \{</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; kj</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; jk</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>}^{\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$

是该疾病模型中症状a_k的权值，T_kj表示两个症状a_k和b_j之间的距离，在步骤1给定症状库中每两个症状之间的关联关系的基础上，由于不同症状的相似性，必须给出患者模型B中的症状b_j与待比对的疾病模型A中的症状a_k之间的距离，该距离依据步骤1建立的症状库所规定的症状之间的关联关系获得。例如：完全相同则距离为1，从属关系则距离为0.5，兄弟关系则距离为0.25，同义关系则距离为1，相似关系则距离为0.6，距离的取值是据经验来设置的，在使用过程中可以逐步调整。in,

is the weight of symptom a _k in the disease model, and T _kj represents the distance between two symptoms a _k and b _j , based on the correlation between every two symptoms in the given symptom database in step 1, because For the similarity of different symptoms, the distance between the symptom b _j in the patient model B and the symptom a _k in the disease model A to be compared must be given. relationship obtained. For example: the distance is 1 for the exact sameness, 0.5 for the subordinate relationship, 0.25 for the sibling relationship, 1 for the synonymous relationship, and 0.6 for the similar relationship. The value of the distance is set according to experience. It can be adjusted step by step during use.

表示疑似疾病模型A中症状a_k和a_j之间的距离，

表示患者模型B中症状b_j和b_k之间的距离。这两个距离也依据步骤1建立的症状库所规定的症状之间的关联关系获得。

和

的定值原则为：完全相同则距离为1，从属关系则距离为0.5，兄弟关系则距离为0.25，同义关系则距离为1，相似关系则距离为0.6，距离取值也是根据经验来设置的，在使用过程中可以逐步调整。

Indicates the distance between symptoms a _k and a _j in suspected disease model A,

Denotes the distance between symptoms b _j and b _k in patient model B. These two distances are also obtained according to the correlation between the symptoms specified in the symptom library established in step 1.

and

The principle of setting the value is: if they are identical, the distance is 1, for the subordinate relationship, the distance is 0.5, for the sibling relationship, the distance is 0.25, for the synonymous relationship, the distance is 1, for the similar relationship, the distance is 0.6, and the distance value is also set according to experience Yes, it can be adjusted step by step during use.

According to the possibility of the patient suffering from each disease given in step 8, the patients are sorted from high to low, and the disease with a higher possibility is taken as the final auxiliary diagnosis result.

Preferably, step 8 further includes: before performing step 8, it also includes performing rough screening on the disease model in advance, and then performing step 8 on the suspected disease set obtained by the rough screening, that is, calculating the current patient model B and the suspected disease set obtained by the rough screening The similarity between the disease models of each disease in .

The method for performing rough screening on disease models to obtain suspected disease collections includes the following steps:

Step a: Calculate the patient’s disease probability for each disease in the disease library; the method is that the patient’s probability of suffering from disease H _i is: P(H _i |β)=[P(H _i |b ₁ )+ P(H _i |b ₂ )+…+P(H _i |b _j )+…+P(H _i |b _n )]/n;

in:

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; D.</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}$

Where b _j represents a specific symptom, H _i represents a specific disease, P(H _i |b _j ) is the probability of suffering from disease H _i when the patient has symptom b _j ; P(b _j |H _i ) is the probability of the patient suffering from disease H _i The probability of symptom b _j appearing at the time; P(H _i ) is the prior probability of the patient suffering from disease H _i , usually a preset value; D is the number of disease models in the disease database;

Step b: select the disease models whose probability obtained in step a is higher than the preset threshold as the suspected disease set, preferably, the preset threshold is 0.2.

In this embodiment, the symptoms input by the user are used to establish the user's patient model, and after the Bayes probability calculation is performed first, the similarity calculation is performed on the suspected diseases with a probability greater than the threshold value and the patient model (the input symptom set) respectively, and the inference is performed. Sort the suspected diseases, list the suspected diseases according to the possibility, and find the departments with the corresponding suspected diseases in the department model, as a reference for seeing a doctor.

The specific description above further elaborates the purpose, technical solution and beneficial effect of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not used to limit the protection of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (9)

1. A medical aided decision-making method based on symptom similarity analysis, characterized in that it comprises the following steps: Step 1: According to the canonical names of the symptoms in the medical knowledge base, supplement the symptom entries that have the upper and lower, homonym and other correlations with these basic symptoms, take the basic symptom entries and the supplementary symptom entries as the constituent elements of the symptom library, and establish and Save the association relationship between every two symptoms in the symptom database; Step 2: Construct a disease model α=(a ₁ , a ₂ ,..., a _k ,..., a _m ) with a collection of symptom items related to each disease, and each disease model contains m symptoms; each disease model as a constituent element of the disease bank; Step 3: Construct each disease model as a disease vector, the construction method is: Constitute the disease vector with the weights of all the symptoms contained in the disease model

in

is the weight of symptom a _k in the disease model, and its value indicates the degree of association between symptom a _k and the disease. The greater the weight, the higher the degree of association; it is preset based on clinical experience and expert data; Step 4: The set of diseases that can be treated by each department of the hospital constitutes a department model γ=(c ₁ , c ₂ , ..., c _p , ..., c _q ), and each department model contains q kinds of diseases; Use the models of each department as the constituent elements of the department database; Step 5: Construct the department vector with the weights of all diseases contained in the department model C

in

Represents the weight of the disease c _p in the department model C, and one of the following two methods is used for its assignment: (1) Assign values according to the suitability of various diseases that the department can treat, and the more diseases the department treats, the greater the weight; (2) Each disease in the department model C adopts the same weight, that is, it does not distinguish the degree of indications of diseases that can be treated by the department; Step 6: Construct a patient model from the collection of all the symptoms provided by the current patient; that is, the patient model B is β=(b ₁ , b ₂ ,..., b _j ,..., b _n ), containing n symptoms in total; Step 7: Construct the current patient model as a patient vector; the construction method is: Construct the patient vector with the weights of all the symptoms contained in the current patient model B

in Represents the weight of symptom b _j in symptom model B, and one of the following two methods is used for its assignment: (1) Assign a value according to the order in which the user enters the symptom or the main degree of subjective feeling among all the symptoms. The more important the symptom or the higher the input order, the greater the weight; (2) Each symptom in patient model B adopts the same weight, that is, no distinction is made between the input order and main degree of symptoms; Step 8: Use one of the following two methods to obtain the probability that the patient has a suspected disease: (1) When the patient vector and the disease vector need to have the same dimension, that is, the patient model and the disease model need to have the same number of symptoms, and the same symptoms should be located in the same position in the two models, that is, a _k with the same subscript and b _j should have the same meaning in case the patient vector is computed

The angle between each disease vector constructed in step 3, the smaller the angle between the vectors, the greater the possibility that the patient has the disease; (2) Calculate the similarity between the current patient model B and each disease model. The greater the similarity, the greater the possibility that the patient has the disease. The similarity between disease model A and patient model B is as follows formula to calculate: $\mathrm{<span\; class="notranslate">\; Sim</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; kj</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; \{</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; kj</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; T</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; jk</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>}^{\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$ in,

is the weight of the symptom a _k in the disease model, T _kj represents the distance between the symptom b _j in the patient model B and the symptom a _k in the disease model A to be compared, and the distance is based on the symptom library established in step 1 The association between the prescribed symptoms is obtained; Indicates the distance between symptoms a _k and a _j in disease model A,

Indicates the distance between symptoms b _j and b _k in patient model B; these two distances are also obtained according to the relationship between symptoms specified in the symptom library established in step 1; Step 9: According to the possibility of the patient suffering from each disease given in step 8, sort from high to low, and use the top H diseases with the highest possibility as the final auxiliary diagnosis result; Step 10: For the H diseases determined in step 9, search for the department corresponding to each disease in the department database one by one, that is, search for the department with the disease in the department database and give the result. 2. A kind of auxiliary decision-making method for seeking medical treatment based on symptom similarity analysis according to claim 1, characterized in that, the association relationship between every two symptoms described in step 1 comprises upper and lower, and same position association; wherein said upper and lower The homonym relationship includes the affiliation relationship between symptoms, and the homonymy relationship includes synonymous, near-synonymous and similar relations. 3. A kind of medical aid decision-making method based on symptom similarity analysis according to claim 1, it is characterized in that, in step 8, the symptom b _j in the patient model B and the symptom a _k in the disease model A to be compared The value of the distance between two symptoms is as follows: if b _j and a _k are identical, the distance T _kj between them is 1; if _{b j and a k} are in a subordinate relationship, the distance is 0.5; The distance is 0.25; if b _j and a _k are synonymous, the distance is 1; if b _j and a _k are similar, the distance is 0.6; if b _j and a _k are not related, the distance is 0. 4. A kind of medical aid decision-making method based on symptom similarity analysis according to claim 1, it is characterized in that, in step 8,

and The principle of value selection is: if two symptoms are exactly the same, the distance is 1; for the affiliation relationship, the distance is 0.5; for the brother relationship, the distance is 0.25; for the synonymous relationship, the distance is 1; for the similar relationship, the distance is 0.6; for no relationship, the distance is 0 . 5. A method for assisting decision-making based on symptom similarity analysis according to claim 1, characterized in that, before step 8 is carried out, it also includes performing rough screening on the disease model in advance, and then re-screening the suspected disease set obtained by the rough screening. Go to step 8, that is, calculate the similarity between the current patient model B and the disease models of each disease in the suspected disease set obtained by coarse screening. 6. A kind of medical aid decision-making method based on symptom similarity analysis according to claim 5, it is characterized in that, the method for carrying out coarse screening to disease model to obtain suspected disease collection comprises the following steps: Step a: Calculate the patient’s disease probability for each disease in the disease library; the method is that the patient’s probability of suffering from disease H _i is: P(H _i |β)=[P(H _i |b ₁ )+ P(H _i |b ₂ )+…+P(H _i |b _j )+…+P(H _i |b _n )]/n; in: $\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; D.</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}$ Where b _j represents a specific symptom, H _i represents a specific disease, P(H _i /b _j ) is the probability of suffering from disease H _i when the patient has symptom b _j ; P(b _j |H _i ) is the patient suffering from disease H _i The probability of symptom b _j appearing at the time; P(H _i ) is the prior probability of the patient suffering from disease H _i , usually a preset value; D is the number of disease models in the disease database; Step b: Select the disease models whose probability obtained in step a is higher than the preset threshold as the suspected disease set. 7. A medical aid decision-making method based on symptom similarity analysis according to claim 6, characterized in that the preset threshold in step b is 0.2. 8 . A medical aid decision-making method based on symptom similarity analysis according to claim 1 , wherein in step 9, H is less than or equal to 3. 9. A kind of medical aid decision-making method based on symptom similarity analysis according to claim 1, it is characterized in that, when step 5 adopts described method (1), step 10 also comprises according to the weight of disease in corresponding department vector The value size sorts the found H department results, and arranges them into a department list from high to low.

Images