Search

CN-121996694-A - Multi-symptom cooperative disease diagnosis system based on fuzzy reasoning

CN121996694ACN 121996694 ACN121996694 ACN 121996694ACN-121996694-A

Abstract

The invention provides a multi-symptom collaborative disease diagnosis system based on fuzzy reasoning, and relates to the technical field of disease diagnosis systems. The system comprises a symptom input module, a three-level fuzzy rule base module, an inference calculation module, a diagnosis result output module and a patient information database module. According to the invention, fuzzy symptoms such as intermittent abdominal pain are converted into quantized values in a 0-1 interval through a three-level semantic mapping mechanism, so that the precision is obviously improved compared with that of the traditional method. And the T-norm operator is adopted to dynamically process symptom association, the product operator for strong association combination and the minimum operator for weak association are adopted, the accuracy of association calculation is improved, and the missed diagnosis rate is reduced.

Inventors

  • WU WENJIAN
  • YIN LINGLING

Assignees

  • 徐州医科大学附属医院

Dates

Publication Date
20260508
Application Date
20251222

Claims (10)

  1. 1. The system is characterized by comprising a symptom input module, a three-level fuzzy rule base module, an inference calculation module, a diagnosis result output module and a patient information database module; The symptom input module is used for receiving symptom information of a patient, wherein the symptom information comprises fuzzy symptoms and clear symptoms, and transmitting the symptom information to the reasoning calculation module in a structured data format; The patient information database module is used for storing age, past medical history and family medical history information of a patient, and a distributed storage architecture is adopted to provide a real-time dynamic correction data calling interface for the reasoning calculation module; the three-level fuzzy rule base module comprises a first-level rule base, a second-level rule base and a third-level rule base, wherein the first-level rule base is internally provided with membership function models of more than 120 common symptoms and is used for converting input symptoms into membership values in a range of 0-1, the second-level rule base comprises more than 200 groups of multi-symptom association rules, the multi-symptom association degree is fused through a T-norm operator, the coupling weight of different symptom combinations is calculated, and the three-level rule base is preset with dynamic correction factor matrixes of 8 age segments and 32 common medical histories and is used for adjusting an inference threshold according to age and medical history information of patients; The reasoning calculation module is used for receiving the symptom information transmitted by the symptom input module, calling the three-level fuzzy rule base module through an API interface, converting symptoms into membership through the first-level rule base, fusing the multi-symptom association degree according to the second-level rule base through a T-norm operator to obtain an association value of a multi-symptom combination, finally combining the three-level rule base, adjusting a reasoning threshold value according to the age and medical history information of the patient provided by the patient information database module by using a dynamic correction factor, comparing the association value of the multi-symptom combination with the adjusted reasoning threshold value to obtain a diagnosis result, and transmitting the diagnosis result to the diagnosis result output module in a JSON format; The diagnosis result output module is used for receiving the diagnosis result transmitted by the reasoning calculation module, and the diagnosis result is presented in a graphic-text combination mode through a high-definition display screen, wherein the diagnosis result comprises a text diagnosis conclusion, a probability value and a symptom associated thermodynamic diagram.
  2. 2. The system for diagnosing a multi-symptom collaborative disease based on fuzzy inference according to claim 1, wherein the specific implementation of the distributed storage architecture comprises the following: 1) ‌ data partition design ‌ ‌ Hash partitioning ‌, namely uniformly distributing data to different nodes through a hash function, so as to avoid the hot spot problem; ‌ scope partition ‌, partitioning data according to the field value scope, and adapting to scope query; 2) ‌ data replication and redundancy ‌ ‌ Three copies of redundancy ‌ are that each data block stores 3 copies, 2 copies are in the same cabinet, and 1 copy is in different cabinets, so that high availability is ensured; ‌ Raft protocol ‌, ensuring data consistency and realizing automatic fault transfer; 3) ‌ real-time update interface ‌ ‌ RESTfulAPI ‌ providing a dynamic call interface to support real-time data update; 4) ‌ privacy protection ‌ ‌ Field level encryption ‌. Sensitive fields are stored by AES-256 encryption; ‌ desensitization processing ‌. Partial fields are displayed through the desensitization view, so that plaintext leakage is avoided.
  3. 3. The system for diagnosing a multi-symptom collaborative disease based on fuzzy reasoning of claim 1, wherein the membership function model construction is a core step of fuzzy logic application, comprising the following implementation steps: 1) ‌ determines the universe ‌ of explicitly entered variables; 2) ‌ selecting a method ‌, namely selecting a statistical method or an assignment method according to the data property; 3) ‌ parameterizing ‌, adjusting the function parameters to match the actual distribution; 4) ‌ verifies ‌ that membership functions are tested for rationality by actual data.
  4. 4. The system for diagnosis of multiple symptom-coordinated diseases based on fuzzy inference as set forth in claim 1, wherein the transmission of symptom information in the structured data format is achieved by ‌ 1) Converting unstructured natural language into structured data is a key step for realizing intelligent diagnosis, and the process mainly depends on natural language processing technology and a predefined medical ontology library; 2) ‌ symptom entity identification ‌. The system automatically identifies and extracts key symptom entities from the text; 3) ‌ attribute and relationship extraction ‌. For the identified symptoms, the symptom input module further extracts the attributes to construct a multi-dimensional symptom representation; 4) ‌ intensity/degree ‌. Subjective description is quantized into numerical values by using ‌ membership function model; 5) ‌ Property modification ‌. Categorical mapping to standard terms in a predefined medical thesaurus; 6) ‌ form structured data object ‌ -after processing, each symptom is converted into a standardized JSON object, ensuring the machine-readability and computability of the data.
  5. 5. The system for diagnosing the multi-symptom synergy disease based on fuzzy reasoning of claim 1, wherein the symptom input module adopts a three-level universality mapping mechanism for semantic analysis of fuzzy symptoms, and covers more than 90% of common fuzzy description scenes: The first-level mapping is to classify the fuzzy descriptors according to semantic types and correspond to general feature dimensions, frequency classes correspond to "attack frequency" dimensions, intensity classes correspond to "symptom intensity" dimensions, duration classes correspond to "duration" dimensions, and degree classes correspond to "influence degree" dimensions; The second-level mapping is to establish 3-5-level quantification standard for each fuzzy descriptor, wherein the frequency class is defined as that the frequency of attacks is lower than 30% of the clinical mean value of the symptom in unit time, the frequency class is defined as that the frequency of attacks is 30% -70% of the mean value, the frequency is defined as that the frequency of attacks is higher than 70% of the mean value, the intensity class is defined as that the frequency of attacks is 2-3 minutes, the intensity class is 4-6 minutes and the intensity class is 7-10 minutes, and the duration class and the degree class refer to the same class logic classification; Establishing a general association model between the grading value and the characteristic interval of the first-level rule base, directly mapping the frequency class grading value to the characteristic basic score of the 'attack frequency', mapping the intensity class to the characteristic basic score of the 'symptom intensity', and supporting mapping adaptation of the cross-symptom type; The fuzzy semantic recognition of text input is realized through a pre-trained medical semantic model, the model is a deep learning model based on a transducer architecture, the model comprises an input layer, a 6-layer encoder module and an output layer, the training corpus comprises more than 5 ten thousand fuzzy symptom descriptions of 12 departments related to internal medicine, surgery and endocrinology, the labeling data of more than 1200 common fuzzy descriptors are included, and the model is trained through the following steps: 1) Cleaning the original corpus, removing repeated and invalid data, and labeling semantic types and corresponding feature dimensions of the fuzzy word; 2) Converting the text into Word vectors by using a Word2Vec tool, inputting a model for pre-training, setting the learning rate to be 5e-5, and training for 50 rounds; 3) Designing a multi-label classification task aiming at the cross semantic description, and outputting a semantic analysis result through a Softmax function; The model performance index is that the recognition accuracy of core words of frequency class and intensity class is not lower than 95%, the resolution accuracy of cross semantic description is not lower than 90%, and the single text resolution response time is not higher than 0.3 seconds.
  6. 6. The system for collaborative multiple symptom diagnosis based on fuzzy inference of claim 1, wherein the patient information database module stores medical history information including disease name, age of diagnosis, current control status, fuzzy correction logic for supporting a three-level rule base: When a patient has a single basic disease, basic correction factors corresponding to the disease are called from a three-level rule base, and a calculation formula is as follows: single disease correction factor formula: ; Wherein F represents a single disease modifying factor; the method comprises the steps of representing basic correction factors corresponding to the diseases, representing the diagnosis-confirmed years correction factors by K, and calculating comprehensive correction factors by adopting a fuzzy superposition algorithm when a patient has various basic diseases, wherein the formula is as follows: And (3) synthesizing a correction factor formula: ; Wherein m is the number of underlying diseases; Representing the integrated correction factor; 、 ... single disease modifying factors respectively representing the 1 st to m th diseases; 、 ... The weights of the 1 st to the m th diseases are respectively expressed, the high association degree is 0.6-0.8, the medium association degree is 0.4-0.6 according to the association degree of the diseases and the diseases to be diagnosed, the comprehensive correction factor result is controlled within the range of 0.1-0.5, and the total correction factor is taken to be 0.5 when the upper limit is exceeded.
  7. 7. The system for diagnosing the multi-symptom collaborative disease based on fuzzy reasoning of claim 1, wherein corresponding membership calculation standards are preset for different symptoms in the first-level rule base, and membership of all symptoms is obtained through the following unified calculation process: 1) Determining a core feature dimension of symptoms, each symptom comprising 2-3 core features, each feature being assigned a weight according to clinical relevance; 2) Setting a quantization grading standard for each core feature, dividing the feature value into 3-5 intervals, wherein each interval corresponds to a basic score in the range of 0-1; 3) Carrying out fine correction on the characteristic value by a linear interpolation method, and adopting a formula if the characteristic value is in the critical value of two classification intervals: Correction score formula: ; Wherein S represents the corrected score of the feature; V represents the actual value of the feature; An upper limit value representing a low range; a lower limit value representing a high section; basic scores representing high intervals; 4) And calculating the symptom membership degree, wherein the formula is as follows: Membership formula: ; wherein n is the characteristic quantity, and the value of 2 or 3;M represents the membership degree of symptoms; 、 ... correction scores respectively representing the 1 st to nth features; 、 ... weights respectively representing the 1 st to nth features; The result remains two bits after the decimal point and is controlled within the 0-1 interval.
  8. 8. The system for diagnosing the multi-symptom synergy disease based on fuzzy reasoning of claim 1, wherein the coupling weight calculation of the multi-symptom combination in the secondary rule base adopts fuzzy association reasoning logic, and the fuzzy association strength is calculated through a T-norm operator: The operator selection, namely automatically matching a T-norm operator according to the symptom combination type, adopting a product operator for strong correlation symptom combination, adopting a minimum operator for weak correlation symptom combination, and calculating the product operator as follows: ; Wherein R represents fuzzy association strength; Representing the membership of symptom a; representing membership of symptom B, C representing weight coefficient, and minimum operator calculated as follows: ; Wherein R represents fuzzy association strength; Representing the membership of symptom a; Representing membership degree of symptom B, C representing weight coefficient, min () representing minimum function; The weight coefficient is determined that based on the symptom co-occurrence frequency distribution of more than 10 ten thousand cases, the co-occurrence frequency is not lower than 35 percent of the combination weight coefficient C is 1.2-1.3, the co-occurrence frequency is 15-35 percent of C is 1.0-1.2, and the co-occurrence frequency is less than 15 percent of C and is 0.8-1.0.
  9. 9. The system for diagnosing the multi-symptom collaborative disease based on fuzzy reasoning of claim 1, wherein the dynamic correction factor matrix of the three-level rule base is constructed based on fuzzy logic, and a trapezoid membership function is adopted to divide a multi-dimensional fuzzy set, so that the system is suitable for common disease diagnosis scenes of various departments: The dimension of age is divided into a subsection every 10 years into children, teenagers, young, middle-aged and old, and the old respectively have corresponding basic correction factors of 1.1, 1.05, 1.0, 1.1, 1.15, 1.2 and 1.3; The medical history dimension is that the 32 common medical histories are divided into four fuzzy sets of low risk, medium and high risk according to the disease risk level, wherein the fuzzy sets respectively correspond to membership degrees of 0.5, 0.65, 0.8 and 0.95, and basic correction factors are 0.05-0.1, 0.1-0.15, 0.15-0.2 and 0.2-0.25 in sequence; The correction rule is to fuse multidimensional correction factors by a weighted summation algorithm, the weight of the age correction factors is 0.3, the weight of the comprehensive correction factors of medical history is 0.7, and finally the dynamic correction factors are obtained The method comprises the following steps: in the formula, As a result of the age correction factor, For the integrated correction factor of the medical history, The result was controlled to be in the range of 1.0 to 1.5.
  10. 10. The system for diagnosing the multi-symptom synergy disease based on fuzzy reasoning of claim 1, wherein the probability value presented by the diagnosis result output module is confidence level of the fuzzy reasoning, and the fuzzy correlation strength calculated by the T-norm operator corresponding to different color depths in the dynamic symptom correlation map is obtained through fuzzy comparison of the multi-symptom correlation value and the dynamic threshold.

Description

Multi-symptom cooperative disease diagnosis system based on fuzzy reasoning Technical Field The invention relates to the technical field of disease diagnosis systems, in particular to a multi-symptom collaborative disease diagnosis system based on fuzzy reasoning. Background In the field of modern medical diagnosis, a computer-aided diagnosis system has become an important auxiliary tool for clinical diagnosis and treatment, but the prior art has significant limitations in processing complex clinical scenes, and mainly shows the following four dimensions: 1. Insufficient quantification accuracy of the fuzzy symptom The existing system mostly adopts simple binary classification (such as existence/nonexistence) or single dimension scoring (such as 1-10 score scale) for treating fuzzy symptoms such as intermittent abdominal pain, moderate hypodynamia and the like, and lacks a systematic semantic analysis mechanism. For example, a certain mainstream diagnostic system maps the intermittence to a fixed value of 0.5 uniformly, does not distinguish the difference between 3 times of daily attacks and 5 times of daily attacks, and the judgment of the central weakness is only subjectively scored by doctors, so that the judgment errors of different operators can reach more than 40%. The processing mode leads to larger deviation between the quantification result of the fuzzy symptom and the clinical actual result, and the accuracy of the follow-up reasoning is directly affected. 2. Multi-symptom association analysis mechanism rigidification When the traditional system analyzes the multi-symptom relationship, a linear weighting algorithm is generally adopted, and the nonlinear synergistic effect among symptoms cannot be reflected. For example, in diagnosing diabetes-related abdominal pain, the combined effects of dyslipidemia and abdominal pain are not simply a superposition of the individual effects, and clinical data shows that when empty abdominal blood glucose is greater than 7.0mmol/L and the daily abdominal pain onset is greater than 4 times, the correlation intensity is 3.2 times that of a single symptom, whereas existing systems can only calculate a correlation of 1.5 times. This defect results in insufficient diagnostic sensitivity for complex co-morbid scenes, with a missed diagnosis rate as high as 28%. 3. Loss of personalized correction mechanism The diagnostic threshold value in the prior art is a fixed value, and individual differences such as the age of a patient, basic diseases and the like are not considered. For example, when 65 years old diabetics and 30 years old healthy people have the same symptoms of blood sugar of 8.0mmol/L and abdominal pain, clinical significance of the patients is significantly different, and risk factors of the aged patients are 2.1 times that of young patients, but the existing system can give the same diagnosis results. In addition, only qualitative judgment is made on the influence of basic diseases, and the lack of quantitative correction models leads to 35% reduction in the diagnosis accuracy of chronic patients. 4. The result has poor interpretation Most systems only output disease names and matching probabilities, and no reasoning basis is shown. The clinician cannot know why the diagnosis is made and which symptoms contribute most, resulting in low confidence in the diagnosis. Some investigation shows that the adoption rate of doctors to the system diagnosis results is only 41% due to the lack of the visual correlation map. Meanwhile, the generation logic of the probability value is opaque, for example, the calculation process of the matching probability of 72% cannot be traced, and the interpretation requirements of medical decision are difficult to meet. The above drawbacks together lead to limited application of existing diagnostic systems in complex clinical settings, especially in basic medical institutions with practical use rates of less than 30%. Therefore, developing a diagnostic system that can precisely quantify fuzzy symptoms, dynamically analyze multi-symptom association, realize personalized reasoning and can interpret results is a technical problem to be solved. Disclosure of Invention Technical problem to be solved Aiming at the defects of the prior art, the invention provides a multi-symptom collaborative disease diagnosis system based on fuzzy reasoning, which solves the defects and the shortcomings in the prior art. Technical proposal The system comprises a symptom input module, a three-level fuzzy rule base module, an inference calculation module, a diagnosis result output module and a patient information database module; The symptom input module is used for receiving symptom information of a patient, wherein the symptom information comprises fuzzy symptoms and clear symptoms, and transmitting the symptom information to the reasoning calculation module in a structured data format; The patient information database module is used for storing age, past