A disease state diagnosis method and system based on a three-dimensional fuzzy semantic model

Abstract

The application belongs to the field of medical information technology and artificial intelligence auxiliary diagnosis, and discloses a disease state diagnosis method and system based on a three-dimensional fuzzy semantic model. The method comprises the following steps: acquiring multi-source clinical data of a patient, wherein the data at least comprises clinical pathology data, biomarker data and functional evaluation data; pre-processing the data, and decomposing the data into three mutually independent orthogonal dimensions of clinical severity dimension, molecular etiology dimension and functional state dimension according to a preset semantic modeling rule; constructing a feature set for each of the three orthogonal dimensions and performing dimension reduction processing to obtain principal component coordinates of each dimension; constructing a three-dimensional fuzzy semantic model based on the principal component coordinates of each dimension, calculating the membership degree of the patient in each semantic state by using a medical semantic scoring rule and a membership function; determining the semantic state description and the confidence score on each dimension according to the membership degree, and outputting a disease state diagnosis result.

Description

Technical Field

[0001] This invention belongs to the field of medical information technology and artificial intelligence-assisted diagnosis, and in particular relates to a disease state diagnosis method and system based on multi-source heterogeneous data and a three-dimensional fuzzy semantic model. Background Technology

[0002] In the clinical diagnosis and prognostic assessment of various complex diseases, disease status is typically determined by multiple dimensions of information, including but not limited to pathophysiological indicators, molecular biological characteristics, and the patient's functional health status. However, existing technologies generally face the following common challenges: multi-source data often exist in isolated forms, lacking a unified expression and fusion framework; different assessment models are based on their own independent scoring systems, making it difficult to form structured correlations; traditional thresholding or stratification methods ignore the continuous and gradual changes in disease status, leading to significant uncertainty in assessment results for boundary cases; furthermore, most assessment results are output as numerical values ​​or discrete labels, lacking interpretable expressions oriented towards clinical semantics, limiting physicians' intuitive understanding and decision support. These problems are evident in various diseases such as tumors, chronic diseases, and neurological disorders, urgently requiring a universal technical framework capable of unified modeling and semantic output of multi-dimensional information.

[0003] Taking breast cancer as an example, with the development of precision medicine, clinical practice has gradually evolved from single-stage assessment to multi-factor comprehensive evaluation. Currently, widely used technical approaches mainly include staging and scoring methods based on clinicopathological characteristics and gene expression scoring methods based on molecular markers. For instance, prognostic index models constructed based on tumor size, lymph node metastasis, and histological grade can stratify patients' risks and guide treatment decisions to some extent. Simultaneously, recurrence scoring methods based on multi-gene expression profiles generate recurrence risk scores by quantitatively analyzing the expression levels of several key genes, which are used to assist in determining the benefit of chemotherapy and prognosis. These methods constitute the mainstream technical system for current breast cancer risk assessment and clinical decision support.

[0004] Furthermore, in practical applications, relevant studies typically use the aforementioned different types of assessment methods for different dimensions of judgment. For example, clinicopathological staging is mainly used to describe tumor burden and progression, while gene expression scoring focuses on revealing the molecular biological characteristics of the tumor. Existing literature indicates that these methods can provide certain prognostic reference value in specific populations or under specific conditions, but their predictive results vary across different patient groups, exhibiting a degree of heterogeneity. Moreover, in clinical practice, physicians often need to combine multiple independent indicators for comprehensive judgment, but these indicators lack a unified expression space and a consistent semantic interpretation framework.

[0005] However, existing technologies generally suffer from the following technical problems: First, existing methods mostly model based on a single or few dimensions, lacking a unified expression mechanism for multidimensional information such as clinical severity, molecular etiological characteristics, and patient functional status, making it difficult to form a structured correlation between different evaluation results; second, different assessment models usually adopt independent calculation and discrete output methods, lacking a unified fusion and collaboration mechanism, making the final conclusion rely on human experience integration, making it difficult to achieve standardization and automation; third, existing technologies mostly use rigid stratification or threshold division for classification, making it difficult to accurately express the transitional state at the boundary of different risk intervals, resulting in uncertainty in the diagnostic results of some complex cases; in addition, existing assessment results are mostly presented in numerical or category label form, lacking an interpretable expression method oriented towards clinical semantics, which is not conducive to doctors' understanding and decision support.

[0006] Therefore, under the existing technological framework, how to construct a diagnostic method capable of unified modeling of multi-source clinical data, achieving collaborative expression across multiple key dimensions, and reflecting disease status in a continuous semantic form has become a pressing technical problem to be solved in this field. These issues directly constrain the consistency, interpretability, and clinical application efficiency of breast cancer diagnostic results, and also provide clear directions for improvement in subsequent technical solutions. Summary of the Invention

[0007] To address the limitations of existing technologies, such as subjective assessment, data isolation, and binary classification, this invention provides a disease state modeling and diagnosis method based on a three-dimensional fuzzy semantic model.

[0008] This invention is implemented as follows: a disease state diagnosis method based on a three-dimensional fuzzy semantic model, taking breast cancer as an example, includes the following steps: Step 1: Obtain multi-source clinical data of the patient, which includes at least clinicopathological data, biomarker data, and functional assessment data.

[0009] Step 2: Preprocess the multi-source clinical data and deconstruct it into three independent orthogonal dimensions according to preset semantic modeling rules: clinical severity dimension, molecular etiology dimension, and functional state dimension.

[0010] Step 3: Construct corresponding feature sets for each of the three orthogonal dimensions, and perform dimensionality reduction processing on each feature set to obtain the principal component coordinates of each orthogonal dimension.

[0011] Step 4: Construct a three-dimensional fuzzy semantic model based on the principal component coordinates of each orthogonal dimension, and calculate the membership degree of the patient in each semantic state using the medical semantic scoring rules and membership functions corresponding to each orthogonal dimension.

[0012] Step 5: Determine the semantic state description and confidence score of the patient in each orthogonal dimension based on the membership degree of each semantic state, and output the disease state diagnosis result.

[0013] Furthermore, the preprocessing includes imputing missing data, converting categorical variables to numerical values, and standardizing feature data.

[0014] The feature set includes composite indicators constructed from raw data, which include at least a tumor burden indicator based on tumor size and lymph node involvement, a receptor composite indicator based on hormone receptor expression, and a psychosocial burden indicator based on symptom data.

[0015] Furthermore, the semantic scoring rule for the clinical severity dimension is as follows: multiply the tumor size by 0.2, and then add it to the lymph node stage value and histological grade value to obtain the clinical severity score.

[0016] Based on the clinical severity score, patients are mapped to at least one of an early low-risk semantic state, a locally late semantic state, or a progressive high-risk semantic state.

[0017] Furthermore, the semantic scoring rule for the molecular etiology dimension is as follows: a weighted combination of proliferation-related indicators and receptor-related indicators is performed, wherein Ki67 expression value and HER2 expression value are positively weighted, and ER expression value and PR expression value are negatively weighted to obtain the molecular etiology score.

[0018] The semantic scoring rule for the functional state dimension is as follows: take 100 points as the baseline value, subtract the product of the average total symptom burden and the scaling factor to obtain the functional state score.

[0019] Based on the molecular etiology score and functional status score, a semantic state description of the molecular etiology and a semantic state description of the functional status are generated respectively.

[0020] Another objective of this invention is to provide a disease state diagnosis system based on a three-dimensional fuzzy semantic model, comprising: The data acquisition module is used to acquire multi-source clinical data from patients.

[0021] The data preprocessing module is used to perform missing value processing, variable transformation, and standardization on the multi-source clinical data.

[0022] The orthogonal decomposition module is used to deconstruct the preprocessed data into three orthogonal dimensions: clinical severity, molecular etiology, and functional status.

[0023] The dimensionality reduction module is used to perform dimensionality reduction on the three orthogonal dimensions respectively to obtain the corresponding principal component coordinates.

[0024] The fuzzy semantic mapping module is used to generate semantic state descriptions and confidence scores based on the principal component coordinates of each orthogonal dimension, medical semantic scoring rules, and membership functions.

[0025] The results output module is used to output the disease status diagnosis results.

[0026] Furthermore, the fuzzy semantic mapping module uses triangular membership functions and trapezoidal membership functions to perform semantic mapping on the coordinates of continuous principal components.

[0027] Among them, the triangular membership function is used to determine the transition interval between adjacent semantic states, and the trapezoidal membership function is used to determine the stable interval of the boundary semantic state.

[0028] Furthermore, the result output module includes a report generation unit and a visualization display unit.

[0029] The report generation unit is used to integrate semantic state descriptions, confidence scores, and corresponding clinical features into a personalized semantic analysis report.

[0030] The visualization unit is used to display the spatial distribution of patients across the dimensions of clinical severity, molecular etiology, and functional status.

[0031] Furthermore, the three clinically orthogonal dimensions and their feature quantification methods are as follows: Clinical severity dimension: integrates tumor size, lymph node status, and TNM staging information; uses the Nottingham Prognostic Index (NPI) as the basis for feature weight allocation, and its quantification formula is as follows:

[0032]

[0033] in, The size of the tumor. For lymph node staging, For histological grading.

[0034] Molecular etiology dimension: Integrating hormone receptor (ER / PR), HER2 status, and Ki67 indicator to reflect molecular driving mechanisms and intrinsic subtypes; its characteristic weighted logical simulation of the Oncotype DX relapse score (RS) algorithm is approximated using the following formula:

[0035]

[0036] Functional status dimension: Patients' quality of life was assessed based on the Campbell Unmet Needs Scale and mapped to the Karnofsky Performance Status (KPS) score. The mapping formula is as follows:

[0037]

[0038] A three-dimensional fuzzy semantic model is used, employing customized triangle and trapezoidal member relationship functions.

[0039] Trigonometric functions are used for transitions between clearly defined intermediate states between adjacent categories.

[0040] Trapezoid functions are used to handle extreme state transitions where the number of members gradually increases or decreases.

[0041] The system calculates the attribution degree of all semantic states within each dimension, identifies the main state based on the maximum attribution degree, and records this value as a confidence score.

[0042] Another object of the present invention is to provide an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is invoked by the processor, the electronic device executes a disease state diagnosis method based on a three-dimensional fuzzy semantic model, and the electronic device is connected to or integrated with a disease state diagnosis system based on a three-dimensional fuzzy semantic model.

[0043] Furthermore, when the processor executes the computer program, it uses the semantic state corresponding to the maximum membership degree as the target output state and determines the corresponding maximum membership degree value as the confidence score.

[0044] When the confidence score is greater than 0.8, a clear classification label is output.

[0045] When the confidence score is between 0.5 and 0.8, a transition state flag is output.

[0046] When the confidence score is less than 0.5, a further evaluation prompt is output.

[0047] When the processor executes the computer program, it generates independent semantic results for the three orthogonal dimensions, and performs joint interpretation of the independent semantic results to form a comprehensive three-dimensional disease state assessment result oriented towards clinical severity, molecular etiology, and functional status.

[0048] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows: The technical solution of this invention does not simply employ fuzzy models or multi-source data fusion. Instead, it first orthogonally deconstructs multi-source clinical data into a three-dimensional semantic space of clinical severity—molecular etiology—functional state according to medical semantics. Then, it reduces the dimensionality of each data source and constructs a unified three-dimensional fuzzy semantic model. Finally, it describes the disease state and outputs confidence scores based on semantic membership. This technical approach realizes the transformation from data-driven classification to semantic space modeling + fuzzy reasoning, enabling data from different sources and scales to be expressed in a unified semantic coordinate system. This solves the problems of difficulty in uniformly interpreting multimodal data and insufficient model interpretability in existing technologies.

[0049] This invention transcends traditional unidimensional assessment frameworks, establishing a comprehensive framework that simultaneously assesses clinical severity, molecular etiology, and functional status, ensuring a complete capture of the disease landscape while minimizing information redundancy between dimensions. By introducing a fuzzy semantic mapping system, this invention effectively addresses the curse of dimensionality in complex medical data and adapts to the inherent uncertainties in clinical diagnosis. It transforms continuous coordinates into clinically interpretable descriptors and relevant confidence measures, preserving the continuous nature of the disease state.

[0050] This invention generates personalized semantic analysis reports that combine natural language descriptions, specific clinical data, and confidence indices, providing clinicians with a detailed understanding of each patient's unique disease characteristics and supporting precision medicine and personalized treatment plans. By introducing the Campbell's Unmet Needs Scale, this invention formally incorporates patient-reported functional status into disease assessment, addressing the shortcomings of traditional biomedical indicators that neglect patients' quality of life and promoting patient-centered precision medicine practices.

[0051] Compared with existing technologies (such as simple thresholding methods): Existing technologies typically output continuous risk scores or discrete classifications directly, requiring physicians to set thresholds themselves. This leads to the forced classification of borderline cases, resulting in inconsistent treatment decisions. This invention inputs the dimensionality-reduced coordinates into a membership function based on a medical prototype, outputting semantic labels with confidence levels. For example, in borderline breast cancer cases (NPI scores between 2.4 and 4.0), using traditional thresholding methods, approximately 30% of cases are forcibly classified into low-risk or high-risk groups. However, using the fuzzy semantic model of this invention, by calculating the membership degrees (e.g., 0.65 and 0.35) for "locally advanced" and "early low-risk" and outputting a "transitional state marker," it clearly indicates the uncertainty of the case to clinicians, avoiding the risk of misjudgment caused by forced binary classification. This continuous and interpretable semantic output is something that traditional scoring thresholding methods cannot achieve. This comparative effect is also applicable to the assessment of borderline cases for other diseases.

[0052] Traditional methods focus solely on clinical severity dimensions (such as NPI and TNM staging), while this invention demonstrates that molecular etiology and functional status are independent and important informational dimensions. EFA confirms that ignoring the latter two dimensions results in a loss of nearly 40% of disease-related information (the three dimensions collectively explain 82.7% of the variance, while the clinical dimension alone explains only 42%).

[0053] In existing technologies, PCA is only used for dimensionality reduction visualization, and fuzzy logic is only used for post-classification processing, lacking inherent coupling between the two. This invention creatively uses PCA principal component coordinates as input to medical scoring rules, and then outputs confidence semantics through membership functions, forming a three-level linkage architecture of "data compression → medical benchmark → fuzzy inference," rather than simple splicing. This invention performs exploratory factor analysis (EFA) on the original data, and the results are shown in Table 1: Factor 1 (clinical severity) has high loading variables of tumor size, lymph node metastasis, and TNM stage; Factor 2 (molecular etiology) has high loading variables of ER, PR, HER2, and Ki67; Factor 3 (functional status) has high loading variables of each item in the Campbell scale. The three factors cumulatively explain 82.7% of the total variance, and the covariance between factors is close to zero, proving that the division of the three dimensions has an inherent structure and is not arbitrary classification.

[0054] To verify the diagnostic consistency of this invention across the clinical severity dimension, this invention used data from 150 breast cancer patients as a basis, with clinical expert consensus as the gold standard. Three methods were employed for risk stratification (early stage low risk / locally intermediate risk / advanced stage high risk), and the adjusted Rand Index (ARI) was calculated relative to the gold standard. The results are as follows:

[0055] The above results demonstrate that the consistency of this invention in identifying clinical severity is significantly superior to traditional hard classification methods and fuzzy models using only the clinical dimension (p<0.01). Regarding the molecular etiology and functional status dimensions, since there is currently no universally accepted gold standard, this invention does not provide a comparison with the gold standard. However, this invention is the first to incorporate these two dimensions into a unified semantic framework, and its clinical effectiveness has been confirmed through qualitative assessment by clinical experts (92% of case semantic descriptions were consistent with expert intuition). Attached Figure Description

[0056] Figure 1 This is a flowchart of a disease state diagnosis method based on a three-dimensional fuzzy semantic model provided by an embodiment of the present invention.

[0057] Figure 2 This is a block diagram of a disease state diagnosis system based on a three-dimensional fuzzy semantic model provided in an embodiment of the present invention.

[0058] Figure 3This is a flowchart of disease state modeling for semantic medical perception provided by the present invention.

[0059] Figure 4 This invention provides a multi-dimensional spatial distribution and a relationship diagram between features and latent factors (including a three-dimensional view of PCA principal component distribution).

[0060] Figure 5 This invention provides a semantic report based on the patient's disease status.

[0061] Figure 6 This is a heatmap of potential factor loadings of clinicopathological features and psychosocial indicators in exploratory factor analysis provided by the present invention. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments.

[0063] like Figure 1 As shown in the figure, a disease state diagnosis method based on a three-dimensional fuzzy semantic model provided by an embodiment of the present invention includes: S101, acquire multi-source clinical data of the patient, wherein the multi-source clinical data includes at least clinicopathological data, biomarker data and functional assessment data.

[0064] S102, the multi-source clinical data is preprocessed and deconstructed into three independent orthogonal dimensions according to preset semantic modeling rules: clinical severity dimension, molecular etiology dimension, and functional state dimension.

[0065] S103: Construct corresponding feature sets for the three orthogonal dimensions respectively, and perform dimensionality reduction processing on each feature set to obtain the principal component coordinates of each orthogonal dimension.

[0066] S104. A three-dimensional fuzzy semantic model is constructed based on the principal component coordinates of each orthogonal dimension. The membership degree of the patient in each semantic state is calculated by using the medical semantic scoring rules and membership functions corresponding to each orthogonal dimension.

[0067] S105 determines the semantic state description and confidence score of the patient in each orthogonal dimension based on the membership degree of each semantic state, and outputs the disease state diagnosis result.

[0068] The preprocessing provided in this embodiment of the invention includes imputing missing data, converting categorical variables into numerical values, and standardizing feature data.

[0069] The feature set includes composite indicators constructed from raw data, which include at least a tumor burden indicator based on tumor size and lymph node involvement, a receptor composite indicator based on hormone receptor expression, and a psychosocial burden indicator based on symptom data.

[0070] The semantic scoring rule for the clinical severity dimension provided in this embodiment of the invention is as follows: multiply the tumor size by 0.2, and then add it to the lymph node stage value and histological grade value to obtain the clinical severity score.

[0071] Based on the clinical severity score, patients are mapped to at least one of an early low-risk semantic state, a locally late semantic state, or a progressive high-risk semantic state.

[0072] The semantic scoring rule for the molecular etiology dimension is as follows: a weighted combination of proliferation-related indicators and receptor-related indicators is performed, with Ki67 expression value and HER2 expression value being positively weighted and ER expression value and PR expression value being negatively weighted to obtain the molecular etiology score.

[0073] The semantic scoring rule for the functional state dimension is as follows: take 100 points as the baseline value, subtract the product of the average total symptom burden and the scaling factor to obtain the functional state score.

[0074] Based on the molecular etiology score and functional status score, a semantic state description of the molecular etiology and a semantic state description of the functional status are generated respectively.

[0075] like Figure 2 As shown in the figure, an embodiment of the present invention provides a disease state diagnosis system based on a three-dimensional fuzzy semantic model, comprising: The data acquisition module is used to acquire multi-source clinical data from patients.

[0076] The data preprocessing module is used to perform missing value processing, variable transformation, and standardization on the multi-source clinical data.

[0077] The orthogonal decomposition module is used to deconstruct the preprocessed data into three orthogonal dimensions: clinical severity, molecular etiology, and functional status.

[0078] The dimensionality reduction module is used to perform dimensionality reduction on the three orthogonal dimensions respectively to obtain the corresponding principal component coordinates.

[0079] The fuzzy semantic mapping module is used to generate semantic state descriptions and confidence scores based on the principal component coordinates of each orthogonal dimension, medical semantic scoring rules, and membership functions.

[0080] The results output module is used to output the disease status diagnosis results.

[0081] The fuzzy semantic mapping module provided in this embodiment of the invention uses triangular membership functions and trapezoidal membership functions to perform semantic mapping on the coordinates of continuous principal components.

[0082] Among them, the triangular membership function is used to determine the transition interval between adjacent semantic states, and the trapezoidal membership function is used to determine the stable interval of the boundary semantic state.

[0083] The result output module provided in this embodiment of the invention includes a report generation unit and a visualization display unit.

[0084] The report generation unit is used to integrate semantic state descriptions, confidence scores, and corresponding clinical features into a personalized semantic analysis report.

[0085] The visualization unit is used to display the spatial distribution of patients across the dimensions of clinical severity, molecular etiology, and functional status.

[0086] The three clinically orthogonal dimensions and their feature quantification methods provided in this embodiment of the invention are as follows: Clinical severity dimension: integrates tumor size, lymph node status, and TNM staging information; uses the Nottingham Prognostic Index (NPI) as the basis for feature weight allocation, and its quantification formula is as follows:

[0087]

[0088] in, The size of the tumor. For lymph node staging, For histological grading.

[0089] Molecular etiology dimension: Integrating hormone receptor (ER / PR), HER2 status, and Ki67 indicator to reflect molecular driving mechanisms and intrinsic subtypes; its characteristic weighted logical simulation of the Oncotype DX relapse score (RS) algorithm is approximated using the following formula:

[0090]

[0091] Functional status dimension: Patients' quality of life was assessed based on the Campbell Unmet Needs Scale and mapped to the Karnofsky Performance Status (KPS) score. The mapping formula is as follows:

[0092]

[0093] A three-dimensional fuzzy semantic model is used, employing customized triangle and trapezoidal member relationship functions.

[0094] Trigonometric functions are used for transitions between clearly defined intermediate states between adjacent categories.

[0095] Trapezoid functions are used to handle extreme state transitions where the number of members gradually increases or decreases.

[0096] The system calculates the attribution degree of all semantic states within each dimension, identifies the main state based on the maximum attribution degree, and records this value as a confidence score.

[0097] Another object of the present invention is to provide an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is invoked by the processor, the electronic device executes a disease state diagnosis method based on a three-dimensional fuzzy semantic model, and the electronic device is connected to or integrated with a disease state diagnosis system based on a three-dimensional fuzzy semantic model.

[0098] Furthermore, when the processor executes the computer program, it uses the semantic state corresponding to the maximum membership degree as the target output state and determines the corresponding maximum membership degree value as the confidence score.

[0099] When the confidence score is greater than 0.8, a clear classification label is output.

[0100] When the confidence score is between 0.5 and 0.8, a transition state flag is output.

[0101] When the confidence score is less than 0.5, a further evaluation prompt is output.

[0102] When the processor executes the computer program, it generates independent semantic results for the three orthogonal dimensions, and performs joint interpretation of the independent semantic results to form a comprehensive three-dimensional disease state assessment result oriented towards clinical severity, molecular etiology, and functional status.

[0103] The fuzzy semantic mapping module provided in this embodiment of the invention uses triangular membership functions and trapezoidal membership functions to perform semantic mapping on the coordinates of continuous principal components.

[0104] Among them, the triangular membership function is used to determine the transition interval between adjacent semantic states, and the trapezoidal membership function is used to determine the stable interval of the boundary semantic state.

[0105] The result output module provided in this embodiment of the invention includes a report generation unit and a visualization display unit.

[0106] The report generation unit is used to integrate semantic state descriptions, confidence scores, and corresponding clinical features into a personalized semantic analysis report.

[0107] The visualization unit is used to display the spatial distribution of patients across the dimensions of clinical severity, molecular etiology, and functional status.

[0108] A three-dimensional fuzzy semantic model is used, employing customized triangle and trapezoidal member relationship functions.

[0109] Trigonometric functions are used for transitions between clearly defined intermediate states between adjacent categories.

[0110] Trapezoid functions are used to handle extreme state transitions where the number of members gradually increases or decreases.

[0111] The system calculates the membership degree of all semantic states within each dimension, identifies the main state based on the maximum membership degree, and defines the membership function mathematically as follows: Let the continuous score of a certain dimension be x, and the corresponding three semantic states be S1, S2, and S3, with the center points of their score intervals being c1, c2, and c3, respectively.

[0112] Then the triangular membership function μ of x with respect to the intermediate state S2 S2 (x) is defined as follows: μ S2 (x) = max(0, min( , )) For boundary state S1, a trapezoidal membership function is used: μ S1 (x)=1, if x≤c1; μ S1 (x)= If c1 <x<c2; μ S1 (x)=0, if x≥c2.

[0113] For boundary state S3, a trapezoidal membership function is used: μ S3 (x)=1, if x≥c3; μ S3 (x)= If c2 <x<c3; μ S3 (x)=0, if x≤c2.

[0114] In the embodiments of this invention (breast cancer), under the clinical severity dimension, c1=1.2, c2=3.2, and c3=5.2, corresponding to the center points of "early stage, low risk," "locally advanced stage," and "high risk of progression," respectively; under the molecular etiology dimension, c1=5, c2=15, and c3=25; under the functional status dimension, c1=85, c2=65, and c3=45 (higher scores indicate better function). For other diseases, those skilled in the art can similarly determine the values ​​of c1, c2, and c3 based on the clinical distribution data of the disease. These values ​​are recorded as confidence scores.

[0115] The breast cancer disease state diagnosis method based on a three-dimensional fuzzy semantic model provided in this invention includes the following core processing layers: Multi-source data input layer: used to acquire multi-source clinical data of patients, including clinicopathological data, biomarker profiles, and patient-reported functional assessment scales.

[0116] Feature engineering and core modeling layer: Deconstruct multi-source data into three clinically orthogonal dimensions: clinical severity, molecular etiology, and functional status. Principal component analysis (PCA) is applied independently to each dimension for dimensionality reduction and principal components are extracted.

[0117] Semantic output layer: Construct a three-dimensional fuzzy semantic model, and through a customized medical rating scale and membership function, convert continuous PCA coordinates into a clinically interpretable semantic state description with confidence scores.

[0118] Furthermore, the three clinically orthogonal dimensions and their feature quantification methods are as follows: Clinical severity dimension: Integrates tumor size, lymph node status, and TNM staging information. The Nottingham Prognostic Index (NPI) is used as the basis for feature weight allocation, and its quantification formula is as follows:

[0119]

[0120] in, The size of the tumor. For lymph node staging, For histological grading.

[0121] The molecular etiology dimension provided in this invention integrates hormone receptor (ER / PR), HER2 status, and the Ki67 indicator to reflect molecular driving mechanisms and intrinsic subtypes. Its feature weighted logical simulation of the Oncotype DX relapse score (RS) algorithm uses the following formula for approximate evaluation:

[0122]

[0123] The functional status dimension provided in this embodiment of the invention is as follows: Based on the Campbell scale, the patient's quality of life is assessed and mapped to the Karnofsky Performance Status (KPS) score. The mapping formula is as follows:

[0124]

[0125] The three-dimensional fuzzy semantic model provided in this embodiment of the invention employs customized triangle and trapezoid member relation functions: Trigonometric functions are used for transitions between clearly defined intermediate states between adjacent categories.

[0126] Trapezoid functions are used to handle extreme state transitions where the number of members gradually increases or decreases.

[0127] The system calculates the attribution degree of all semantic states within each dimension, identifies the main state based on the maximum attribution degree, and records this value as a confidence score.

[0128] like Figure 3 The flowchart shown illustrates a disease state modeling process based on semantic medical perception. This invention provides a disease state diagnosis and assessment system based on a three-dimensional fuzzy semantic model, whose data processing flow is divided into five coherent levels: Step S1: The multi-source data input layer collects and prepares heterogeneous medical data of patients, specifically including: clinicopathological data (tumor size, lymph node size, lymph node metastasis); biomarker profiles (ER, PR, HER2, Ki67 expression); and functional assessment data based on the Campbell Unmet Needs Scale.

[0129] Step S2: Data preprocessing and feature engineering layers handle missing values ​​in the raw data (e.g., imputation using the median), perform numerical transformations on categorical variables, and standardize them using normalization (StandardScaler). Advanced feature engineering is used to create composite indicators, including: a tumor burden index based on tumor size and lymph node involvement, a hormone receptor composite score, and a psychosocial burden index combining pain and fatigue.

[0130] Step S3: Feature Engineering and Core Modeling Layer (Orthogonal Dimensionality Decomposition and PCA Dimensionality Reduction) To address feature redundancy in high-dimensional data, this system independently applies Principal Component Analysis (PCA) across three orthogonal dimensions (clinical severity, molecular etiology, and functional status). PCA effectively reduces the high-dimensional feature space, extracting principal components (PC1, PC2, PC3) as input coordinates for the subsequent fuzzy model. Exploratory Factor Analysis (EFA) validates that: ER positivity and PR positivity highly cluster in Factor 1 (corresponding to clinical features); tumor size and metastasis cluster in Factor 2 (corresponding to molecular etiology); and Campbell's score assessment indicators cluster in Factor 3 (corresponding to functional status).

[0131] Step S4: Semantic Output Layer (3D Fuzzy Semantic Mapping) This is the core step of the invention. A classic medical scoring prototype is introduced into each dimension, and the membership degree of each semantic state is calculated: Clinical severity mapping: Based on the Nottingham Prognostic Index (NPI) principle, tumor size was selected. Lymph node staging and histological grading Formula I is used to construct a spatial structure, which is then mapped to three semantic states: early low risk, local high risk, and progressive high risk.

[0132] Molecular etiology dimension mapping: Referring to the Oncotype DX relapse scoring logic, the positive weighted proliferation group (Ki67 and HER2) and the negative weighted estrogen group (ER / PR) are used to distinguish between hormone-sensitive type and HER2 dominant / triple-negative type using Formula II.

[0133] Functional status dimension mapping: The symptom data in the Campbell scale are mapped to the Karnofsky Functional Status Score (KPS) of 0-100 using Formula III, and the output is a gradual functional status from well-adapted to severe impairment.

[0134] The model employs custom triangular and trapezoidal membership functions to convert continuous principal component coordinates into confidence scores for the aforementioned semantic states. High confidence (>0.8) indicates clear classification, while medium confidence (0.5-0.8) indicates transitional states.

[0135] Step S5: Apply the interaction and output layer using a multi-view visualization system (such as...) Figure 4 The two-dimensional and three-dimensional scatter plots show the distribution of patients across the three core dimensions. Simultaneously, through template-based natural language generation technology, fuzzy logic results and specific clinical data are integrated to generate personalized analysis reports (e.g., output: localized early-stage tumor (2.1 cm, 0 lymph nodes) [confidence level: 0.85]).

[0136] Experimental verification results: This example is based on data from 150 patients, and the main results are as follows: PCA dimensionality reduction effect: PC1 explained 84.2% of the variance in the clinical dimension, 83.9% in the molecular dimension, and 13.8% in the functional dimension. The first two indicate that the selected features are highly cohesive, while the latter reflects the diversity of functional states but still captures unique patient-centric information.

[0137] Fuzzy classification confidence: 78% of cases had a confidence level >0.7 for the main semantic state across all three dimensions; 92% of cases had semantic classification results consistent with independent clinical expert assessments.

[0138] The visualization system effectively assists doctors in identifying disease patterns, validating the model's clinical applicability.

[0139] This invention performed exploratory factor analysis (EFA) on the original data, and the results are shown in Table 1: Factor 1 (clinical severity) had high loadings of tumor size, lymph node metastasis, and TNM stage; Factor 2 (molecular etiology) had high loadings of ER, PR, HER2, and Ki67; and Factor 3 (functional status) had high loadings of each item on the Campbell scale. The three factors cumulatively explained 82.7% of the total variance, and the covariance between factors was close to zero, demonstrating that the division of the three dimensions has an inherent structure and is not arbitrary.

[0140] The performance comparison of this invention with the two comparative methods is as follows:

[0141] To verify the technical effects of this invention, an experimental verification was conducted based on multi-source data from 150 breast cancer patients. The main results are as follows: 1. Dimensionality Reduction Effect of PCA: The explained variance of independent PCA in each dimension is as follows: 84.2% for clinical severity, 83.9% for molecular etiology, and 13.8% for functional status. This indicates that the characteristics within the clinical and molecular dimensions are highly correlated, while the variance distribution of the functional status dimension is more dispersed due to the inclusion of diverse psychosocial indicators, but it still captures unique patient-centric information.

[0142] 2. Fuzzy semantic classification confidence: Among 150 patients, 78% of the cases had a confidence level of over 0.7 (high confidence) in the main semantic state across all three dimensions, and the semantic classification results of 92% of the cases were consistent with the assessment of independent clinical experts, demonstrating the clinical accuracy and reliability of the model.

[0143] 3. Representative Case Report: Taking patient 5 as an example, the generated semantic report is as follows: Clinical severity dimension (explained variance 84.2%): early low-risk lesion (stage I, tumor 1.0 cm, 0 lymph node metastases) [confidence: 0.85].

[0144] Molecular etiology dimension (explained 83.9% of variance): HER2 dominant type (Luminal B type (HER2+), major marker HER2) [confidence level: 0.73].

[0145] Functional status dimension (explained variance 13.8%): moderate functional distress (severe unmet needs, Campbell total score 95.8) [confidence level: 0.66].

[0146] 4. Visualization System: The 3D visualization scatter plot intuitively displays the distribution of patients in three dimensions, allowing doctors to quickly identify clusters and outliers, thus assisting in clinical decision-making.

[0147] The above evidence demonstrates that the present invention can effectively integrate multi-source heterogeneous data to generate semantic descriptions of disease states with high interpretability and confidence, and has significant clinical practical value.

[0148] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Equivalent substitutions or conventional adjustments made by those skilled in the art based on the content of this invention all fall within the protection scope of this invention.

[0149] The following four embodiments, respectively addressing different clinical scenarios such as early low burden, local progression, boundary transition, and significant functional impairment, fully demonstrate the core steps of the method of this invention: multi-source data acquisition and preprocessing, deconstruction into three orthogonal dimensions and construction of composite indicators, fractal principal component analysis for dimensionality reduction, fuzzy semantic mapping based on medical scoring rules and membership functions, confidence determination, and result output. Each embodiment directly corresponds to the technical features described in claims 1-10.

[0150] Example 1: Status assessment of newly diagnosed patients with low workload in the early stage (Corresponding to claims 1-4, 6, 10-11) I. Multi-source data acquisition and preprocessing The following multi-source clinical data were obtained from a newly diagnosed early-stage breast cancer patient: Clinical pathology data: The tumor measured 1.2 cm, with no lymph node metastasis, and its histological grade was grade 1. Biomarker data: Estrogen receptor (ER) expression was 90%, progesterone receptor (PR) expression was 85%, human epidermal growth factor receptor 2 (HER2) expression was negative, and Ki67 proliferation index was 8%.

[0151] Functional assessment data: Symptoms collected based on Campbell's Unmet Needs Scale included mild fatigue, no pain, good sleep, and stable mood. The average total symptom burden was calculated to be 0.18 (out of 1.0).

[0152] The above data is preprocessed as follows: missing values ​​are filled with the median (there are no missing values ​​in this embodiment), categorical variables (such as HER2 positive or negative) are converted into numerical values, and all continuous features are standardized.

[0153] II. Deconstructing into three orthogonal dimensions and constructing a composite index As described in claim 2, the preprocessed data is deconstructed into three mutually independent orthogonal dimensions, and a composite index is constructed: Clinical severity dimension: Integrating tumor size, lymph node status (stage 0), and histological grade (grade 1), a tumor burden index is constructed, which is tumor size multiplied by the lymph node involvement adjustment factor; Molecular etiology dimension: Integrating ER, PR, HER2, and Ki67, a comprehensive receptor index is constructed, which is the arithmetic mean of the expression percentages of ER and PR.

[0154] Functional status dimension: Integrating the symptom items of the Campbell scale, a psychosocial burden index is constructed, which is the average total symptom burden.

[0155] III. Dimensionality Reduction through Fractal Principal Component Analysis Perform principal component analysis independently on the feature sets of the three dimensions: The first principal component score of the clinical severity dimension was negative (indicating below average), and this principal component explained 84.2% of the total variance of this dimension.

[0156] The first principal component score of the molecular etiology dimension was positive (indicating above-average), explaining 83.9% of the total variance of this dimension.

[0157] The first principal component score of the functional status dimension was negative (indicating a symptom burden below average), explaining 13.8% of the total variance for this dimension.

[0158] IV. Three-dimensional fuzzy semantic mapping According to claims 3, 4, and 6, medical semantic scoring rules and membership functions are applied to each dimension respectively.

[0159] (a) Mapping of clinical severity dimensions Using the Nottingham Prognostic Index logic described in claim 3, the calculation formula is: 0.2 multiplied by the tumor size (cm), plus the lymph node stage value, plus the histological grade value. The calculated Nottingham Prognostic Index for this patient is 1.24.

[0160] Based on this index, three semantic states are defined: early low risk (index ≤ 2.4), local late stage (index between 2.4 and 4.0), and high risk of progression (index ≥ 4.0). A trapezoidal membership function is used to handle the boundary states: the right shoulder of the early low risk interval is 2.4, the center of the local late stage interval is 3.2, and the left shoulder of the high risk of progression interval is 3.8.

[0161] The patient's index of 1.24 is less than 2.4, therefore the membership degree for early low-risk is 1.0, and the membership degree for other states is 0. The maximum membership degree of 1.0 is greater than 0.8, and the output classification is clearly labeled.

[0162] (II) Molecular Etiology Dimension Mapping According to the approximate relapse scoring logic described in claim 4: Ki67 and HER2 are positively weighted, and ER and PR are negatively weighted. The calculated approximate relapse score for this patient is 0.5.

[0163] Semantic states were defined as: hormone-sensitive (score ≤ 10), HER2 dominant / triple-negative driven (score > 20), with a triangular membership function used for the intermediate transition zone. This patient's score of 0.5 had a membership of 1.0 to the hormone-sensitive type, with a confidence level greater than 0.8.

[0164] (III) Functional State Dimension Mapping According to the functional state mapping formula of claim 4: using 100 as a baseline, subtract the product of the average total symptom burden and the scaling factor. Taking the scaling factor as 40, the patient's functional state score is 92.8.

[0165] Semantic states were defined as: well-adjusted (score ≥ 80), mild to moderate impairment (50 ≤ score < 80), and severe impairment (score < 50). The patient's score of 92.8 had a membership of 1.0 to well-adjusted, with a confidence level greater than 0.8.

[0166] V. Results Output According to claims 10-11, output three-dimensional independent semantic results and joint comprehensive evaluation: Clinical severity dimension: early stage, low risk (confidence 1.0, clearly categorized).

[0167] Molecular etiology dimension: hormone-sensitive type (confidence level 1.0).

[0168] Functional state dimension: well adapted (confidence level 1.0).

[0169] The system generated a personalized semantic analysis report: The patient is in an early low-burden state (tumor 1.2 cm, no lymph node metastasis, histological grade 1), molecular subtype is hormone-sensitive (high ER / PR expression, HER2 negative, low Ki67), and functional status is good. All three-dimensional semantic confidence levels are high, and routine follow-up and endocrine therapy evaluation are recommended.

[0170] This embodiment fully demonstrates all the steps described in claim 1 and specifically implements the technical features of claims 2-4, 6, and 10-11.

[0171] Example 2: Assessment of Patients with Locally Progressive Disease (Corresponding to claims 1-4, 6, 10-11) Patient data: Tumor size 3.8 cm, 3 lymph node metastases (stage 2), histological grade 3; ER expression 20%, PR expression 15%, HER2 2+ and positive fluorescence in situ hybridization, Ki67 45%; Campbell's scale showed moderate fatigue, anxiety, and pain scores 4 / 10, with a mean total symptom burden of 0.55.

[0172] Pretreatment and construction of composite indices: performed according to the method in Example 1. The tumor burden index was 5.7, the receptor composite index was 17.5, and the psychosocial burden index was 0.55.

[0173] Principal component analysis dimensionality reduction: The first principal component scores for the three dimensions are as follows: positive for the clinical dimension, negative for the molecular dimension, and positive for the functional dimension.

[0174] Fuzzy semantic mapping: Clinical severity: Nottingham Prognostic Index = 0.2 × 3.8 + 2 + 3 = 5.76. This value is greater than 4.2 (the right shoulder of the high-risk progression trapezoidal function), therefore the membership degree for high-risk progression is 1.0, with a confidence level greater than 0.8.

[0175] Molecular etiology: Approximate relapse score = 0.5 × 45 + 0.8 × 1 0.02×(20+15)=22.6. This value is greater than 18 (the left shoulder of the HER2 dominant trapezoidal function), the membership of HER2 dominant / triple negative driving type is 1.0, and the confidence level is greater than 0.8.

[0176] Functional Status: Functional Status Score = 100 (0.55 × 40) = 78. This value falls within the boundary between well-adapted (≥80) and mild to moderate impairment (50-80). Using a triangular membership function, the membership degree is approximately 0.93 for mild to moderate impairment and approximately 0.4 for well-adapted individuals. The maximum membership degree of 0.93 is greater than 0.8, indicating a high confidence level.

[0177] Results: Clinical severity: High risk of progression (confidence 1.0); Molecular etiology: HER2 dominant (confidence 1.0); Functional status: Mild to moderate impairment (confidence 0.93). Recommendation: Intensive adjuvant chemotherapy combined with anti-HER2 targeted therapy, along with symptom management.

[0178] This embodiment demonstrates the practical application of the scoring formulas in claims 3-4, as well as the confidence threshold determination in claim 10.

[0179] Example 3: Assessment of patients with borderline transitional disease (Corresponding to claims 1-4, 6, 10-11, the focus is on demonstrating the transitional state between confidence levels of 0.5 and 0.8) Patient data: Tumor size 2.3 cm, 1 lymph node metastasis (stage 1), histological grade 2; ER expression 50%, PR expression 40%, HER2 1+ (negative), Ki67 25%; Campbell's total symptom burden mean 0.32.

[0180] Principal component analysis dimensionality reduction: The first principal component scores of all three dimensions are close to zero, indicating that the patient is in the average or intermediate range in each dimension.

[0181] Fuzzy semantic mapping: Clinical severity: Nottingham Prognostic Index = 0.2 × 2.3 + 1 + 2 = 3.46. This value is in the central region of the membership function of the locally advanced triangle (center point 3.2). The calculated membership for locally advanced disease is approximately 0.84, which is greater than 0.8, still indicating high confidence, but it is close to the boundary.

[0182] Molecular etiology: Approximate relapse score = 0.5 × 25 + 0 0.02 × (50 + 40) = 10.7. The right shoulder point of the hormone-sensitive type is set to 10, and the transition region extends to 12. The calculated membership degree for the hormone-sensitive type is 0.65, and the membership degree for the intermediate transition state is 0.14. The maximum membership degree of 0.65 falls within the 0.5-0.8 range. According to claim 10, the transition state marker is output.

[0183] Functional Status: Functional Status Score = 100 (0.32×40)=87.2. This value is greater than 80, the membership degree for good adaptation is 1.0, and the confidence level is greater than 0.8.

[0184] Results: System report: Clinical severity leans towards locally advanced (confidence 0.84, high confidence); Molecular etiology is in the transition zone between hormone-sensitive and atypical-driven states (confidence 0.65, transitional state marker), supplementary gene expression profiling is recommended; Functional status is well-adapted (confidence 1.0). Overall assessment: This patient's molecular subtype is atypical, requiring further evaluation to guide the trade-off between endocrine therapy and chemotherapy.

[0185] This embodiment directly supports the technical feature of outputting a transition state marker when the confidence score is between 0.5 and 0.8, as described in claim 10.

[0186] Example 4: Patients with triple-negative tendencies and significant functional impairment (Corresponding to claims 1-11, fully demonstrating the 3D joint output and visualization) Patient data: Tumor size 2.9 cm, 2 lymph node metastases (stage 1), histological grade 3; ER expression 2%, PR expression 1%, HER2 negative, Ki67 70%; Campbell's total symptom burden mean 0.78 (severe pain, severe fatigue, depressed mood).

[0187] Fuzzy semantic mapping: Clinical severity: Nottingham Prognostic Index = 0.2 × 2.9 + 1 + 3 = 4.58. A value greater than 4.2 indicates a high risk of progression with a membership of 1.0 and a confidence level greater than 0.8.

[0188] Molecular etiology: Approximate relapse score = 0.5 × 70 0.02×(2+1)=34.94. This value is greater than 18, and the membership of the HER2 dominant / triple negative driving type is 1.0, with a confidence level greater than 0.8.

[0189] Functional Status: Functional Status Score = 100 (0.78 × 40) = 68.8. This value falls within the light to moderate obstacle range (50-80). Using the triangular membership function, the membership degree for light to moderate obstacles is calculated to be 0.63, and the membership degree for severe obstacles is 0.06. The maximum membership degree of 0.63 falls within the 0.5-0.8 range, and a transition state flag is output.

[0190] Results: System report: Clinical severity: High risk of progression (confidence 1.0); Molecular etiology: Triple-negative driven (confidence 1.0); Functional status: Mild to moderate impairment (confidence 0.63, transitional state). Recommendations: Intensive chemotherapy regimen, active supportive treatment, and regular psychological intervention.

[0191] At the same time, such as Figure 4As shown, the patient's three-dimensional principal component coordinates are projected onto a scatter plot in a three-dimensional semantic space, which intuitively shows that the patient belongs to a cluster region with high clinical burden, high invasive molecular phenotype, and moderate functional impairment, assisting doctors in quickly identifying disease patterns.

[0192] Figure 5 Semantic reporting based on the patient's disease status.

[0193] Figure 6 EFA results: In this invention, the principal component coordinates, medical semantic scoring rules, and membership functions form a rigid causal linkage. Their calculation order, input / output constraints, and diagnostic output rules are as follows: Step 1: Calculation of principal component coordinates (data-driven dimensionality reduction) Principal component analysis (PCA) is performed independently on each dimension to compress the high-dimensional feature set into one-dimensional principal component coordinates, denoted as PCd (where d represents the clinical, molecular, and functional dimensions). The final output is the biomarker with the highest weight within each dimension, and the dimension with the highest weight among all dimensions.

[0194] Constraints: No cross-dimensionality reduction occurs between features of different dimensions. Principal component coordinates are unique numerical abstractions of that dimension.

[0195] Step 2: Mapping semantic scoring rules (knowledge-driven benchmark) Multi-source data (clinical, molecular, and functional data) are used as input and fed into the corresponding medical semantic scoring rules for each dimension.

[0196] Clinical dimension:

[0197] Molecular dimension:

[0198] Functional dimensions:

[0199] Constraints: The output (Score) of the semantic scoring rule does not participate in any other calculations and does not accept any input other than principal component coordinates (or the original features reconstructed from them). This step achieves the transformation from "data abstraction" to a "clinical continuous scale".

[0200] Step 3: Calculation of membership function (fuzzification and confidence generation) The continuous score output by the semantic scoring rule is used as the sole independent variable and input into the predefined membership function.

[0201] For the case of three semantic states (S1, S2, S3), the intermediate states use a triangular membership function: μS2(x) = max(0, min( , )) The boundary states (S1, S3) use trapezoidal membership functions (see above for the specific formula).

[0202] Constraints: The membership function takes only the score as input and does not directly accept principal component coordinates or original features. Within the same dimension of the same patient, the sum of the membership degrees of all semantic states does not have to be 1 (non-mutually exclusive fuzzy states are allowed).

[0203] Step 4: Determining the diagnostic output (maximum membership principle + confidence threshold) For each dimension, select the semantic state with the highest membership degree as the dominant semantic state: Stated = arg(max μk(x)).

[0204] Record the maximum membership value as the confidence score: Confidenced = max μk(x).

[0205] Final output rules: If Confidence > 0.8: Output "Clear Classification Marker", the semantic state can be directly used for clinical decision-making.

[0206] If 0.5≤Confidence≤0.8: Output "Transitional State Flag", indicating to the doctor that the patient is in the boundary range and suggesting additional examinations or dynamic observation.

[0207] If Confidence < 0.5: Output "Further evaluation prompt", indicating that the model has high uncertainty and requires manual review or the addition of features.

[0208] This process is unidirectional, non-skipping, and irreversible. The output of each level strictly serves as the sole input to the next level, with no bypasses or feedback. This rigid linkage mechanism ensures the reproducibility and clinical interpretability of diagnostic results, unlike the loosely coupled approach in existing technologies where PCA is used only for visualization and fuzzy logic is used only for post-processing classification.

[0209] Simulation results:

[0210] It should be noted that embodiments of the present invention can be implemented using hardware, software, or a combination of both. The above descriptions are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A disease state diagnosis method based on a three-dimensional fuzzy semantic model, characterized in that, Includes the following steps: Step 1: Obtain multi-source clinical data of the patient, which includes at least clinicopathological data, biomarker data, and functional assessment data; Step 2: Preprocess the multi-source clinical data and deconstruct it into three independent orthogonal dimensions according to preset semantic modeling rules: clinical severity dimension, molecular etiology dimension, and functional state dimension. Step 3: Construct corresponding feature sets for each of the three orthogonal dimensions, and perform dimensionality reduction processing on each feature set to obtain the principal component coordinates of each orthogonal dimension; Step 4: Construct a three-dimensional fuzzy semantic model based on the principal component coordinates of each orthogonal dimension, and calculate the membership degree of the patient in each semantic state using the medical semantic scoring rules and membership functions corresponding to each orthogonal dimension. Step 5: Determine the semantic state description and confidence score of the patient in each orthogonal dimension based on the membership degree of each semantic state, and output the disease state diagnosis result.

2. The disease state diagnosis method based on a three-dimensional fuzzy semantic model according to claim 1, characterized in that, The preprocessing includes filling in missing data, converting categorical variables to numerical values, and standardizing feature data. The feature set includes composite indicators constructed from raw data, which include at least a tumor burden indicator based on tumor size and lymph node involvement, a receptor composite indicator based on hormone receptor expression, and a psychosocial burden indicator based on symptom data.

3. The disease state diagnosis method based on a three-dimensional fuzzy semantic model according to claim 1, characterized in that, The semantic scoring rule for the clinical severity dimension is as follows: multiply the tumor size by 0.2, and then add it to the lymph node stage value and histological grade value to obtain the clinical severity score; Based on the clinical severity score, patients are mapped to at least one of an early low-risk semantic state, a locally late semantic state, or a progressive high-risk semantic state.

4. The disease state diagnosis method based on a three-dimensional fuzzy semantic model according to claim 1, characterized in that, The semantic scoring rule for the molecular etiology dimension is as follows: a weighted combination of proliferation-related indicators and receptor-related indicators is performed, wherein Ki67 expression value and HER2 expression value are positively weighted, and ER expression value and PR expression value are negatively weighted to obtain the molecular etiology score; The semantic scoring rule for the functional state dimension is as follows: take 100 points as the baseline value, subtract the product of the average total symptom burden and the scaling factor to obtain the functional state score. Based on the molecular etiology score and functional status score, a semantic state description of the molecular etiology and a semantic state description of the functional status are generated respectively.

5. A disease state diagnosis system based on a three-dimensional fuzzy semantic model, based on the disease state diagnosis method based on a three-dimensional fuzzy semantic model according to any one of claims 1 to 4, characterized in that, include: The data acquisition module is used to acquire multi-source clinical data from patients; The data preprocessing module is used to perform missing value processing, variable transformation, and standardization on the multi-source clinical data; The orthogonal decomposition module is used to deconstruct the preprocessed data into three orthogonal dimensions: clinical severity, molecular etiology, and functional status. The dimension reduction module is used to perform dimension reduction processing on the three orthogonal dimensions respectively to obtain the corresponding principal component coordinates; The fuzzy semantic mapping module is used to generate semantic state descriptions and confidence scores based on the principal component coordinates of each orthogonal dimension, medical semantic scoring rules, and membership functions. The results output module is used to output the disease status diagnosis results.

6. The disease state diagnosis system based on a three-dimensional fuzzy semantic model according to claim 5, characterized in that, The fuzzy semantic mapping module uses triangular membership functions and trapezoidal membership functions to perform semantic mapping on the coordinates of continuous principal components; Among them, the triangular membership function is used to determine the transition interval between adjacent semantic states, and the trapezoidal membership function is used to determine the stable interval of the boundary semantic state.

7. The disease state diagnosis system based on a three-dimensional fuzzy semantic model according to claim 5, characterized in that, The result output module includes a report generation unit and a visualization display unit; The report generation unit is used to integrate semantic state descriptions, confidence scores, and corresponding clinical features into a personalized semantic analysis report. The visualization unit is used to display the spatial distribution of patients across the dimensions of clinical severity, molecular etiology, and functional status.

8. The disease state diagnosis system based on a three-dimensional fuzzy semantic model according to claim 5, characterized in that, The three clinically orthogonal dimensions and their feature quantification methods are as follows: Clinical severity dimension: integrates tumor size, lymph node status, and TNM staging information; uses the Nottingham Prognostic Index (NPI) as the basis for feature weight allocation, and its quantification formula is as follows: in, The size of the tumor. For lymph node staging, For histological grading; Molecular etiology dimension: Integrating hormone receptor (ER / PR), HER2 status, and Ki67 indicator to reflect molecular driving mechanisms and intrinsic subtypes; its characteristic weighted logical simulation of the Oncotype DX relapse score (RS) algorithm is approximated using the following formula: Functional status dimension: Patients' quality of life was assessed based on the Campbell Unmet Needs Scale and mapped to the Karnofsky Performance Status (KPS) score. The mapping formula is as follows: A three-dimensional fuzzy semantic model is used, employing customized triangle and trapezoidal member relationship functions; Trigonometric functions are used for transitions between clearly defined intermediate states between adjacent categories; Trapezoid functions are used to handle extreme state transitions where the number of members gradually increases or decreases; The system calculates the attribution degree of all semantic states within each dimension, identifies the main state based on the maximum attribution degree, and records this value as a confidence score.

9. An electronic device, characterized in that, The device includes a processor and a memory, wherein the memory stores a computer program, which, when invoked by the processor, causes the electronic device to execute the disease state diagnosis method based on a three-dimensional fuzzy semantic model as described in any one of claims 1 to 4, and the electronic device is connected to or integrated with the disease state diagnosis system based on a three-dimensional fuzzy semantic model as described in any one of claims 5 to 8.

10. The electronic device according to claim 9, characterized in that, When the processor executes the computer program, it uses the semantic state corresponding to the maximum membership degree as the target output state and determines the corresponding maximum membership degree value as the confidence score. When the confidence score is greater than 0.8, a clear classification label is output; When the confidence score is between 0.5 and 0.8, a transition state flag is output; When the confidence score is less than 0.5, a further evaluation prompt is output. When the processor executes the computer program, it generates independent semantic results for the three orthogonal dimensions, and performs joint interpretation of the independent semantic results to form a comprehensive three-dimensional disease state assessment result oriented towards clinical severity, molecular etiology and functional status.

Images