Medical patient similarity clustering analysis method and system and application
Through multi-dimensional scaling algorithm and attraction and exclusion optimization algorithm, the clustering problem of high-dimensional and nonlinear medical data is solved, and more accurate similarity analysis of medical patients is achieved. It is suitable for large-scale medical data, improving the effect of personalized treatment and Quality of medical management.
Patent Information
- Application Number
- CN202410664608.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-27
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing technology cannot effectively identify and analyze high-dimensional and nonlinear distribution of medical data, resulting in insufficient accuracy in the clustering of similarity of medical patients, inaccurate identification of complex patterns and relationships, and cannot be applied to the analysis of large-scale medical data.
A multi-dimensional scaling algorithm is used to reduce the dimension of medical data, and a nonlinear relationship is dynamically explored in combination with an attraction and exclusion optimization algorithm, and the candidate solutions are iteratively updated to obtain similarity clustering results of medical patients.
It improves the accuracy of cluster analysis of medical patients, can more accurately identify pathological patterns and patient similarities, is suitable for large-scale medical data, supports personalized treatment and management, reduces treatment costs, and improves patients' quality of life.
Smart Images

Figure CN118919006A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of smart medical care, and in particular to a medical patient similarity clustering analysis method, system and application. Background Art
[0002] Medical group similarity analysis is of great significance in the field of modern medical health. Through similarity analysis of patient groups, patient groups with similar symptoms, physiological indicators or genetic backgrounds can be identified, which helps doctors diagnose diseases more accurately, especially when dealing with complex or rare diseases. In addition, after identifying similar patient groups through analytical means, personalized treatment plans can be formulated based on the response of these groups to treatment, thereby improving the quality of medical services while improving the level of personalized medicine.
[0003] The existing solutions for similarity analysis of medical patients mainly include cluster analysis methods: the common method is to use traditional clustering algorithms such as K-means and hierarchical clustering to divide patient groups. However, this method cannot effectively identify complex patterns and relationships and is not well suited for the analysis of high-dimensional, nonlinearly distributed medical data; principal component analysis: extracting the main components in medical data through dimensionality reduction to simplify the data structure, but this analysis method can only capture the linear relationship between data and is not accurate enough for the analysis of medical data with nonlinear relationships; manual classification method based on medical rules: relying on rules defined by experts or clinical pathways to divide patient groups. This method is highly dependent on the professional level of experts and will ignore unforeseen patterns in the data.
[0004] In other words, since the medical data of medical patients are nonlinear and high-dimensional, and the medical data have complex patterns and relationships, the current conventional clustering analysis methods cannot cluster similar medical patients based on medical data. Summary of the invention
[0005] The present application provides a medical patient similarity clustering analysis method, system and application, which effectively reduces the data dimension of multi-dimensional medical data through a multi-dimensional scaling algorithm, and uses an attraction-repulsion optimization algorithm to dynamically explore the nonlinear relationship between medical data, so as to accurately cluster similar medical patients through medical data.
[0006] To achieve the above objectives, in a first aspect, this solution provides a medical patient similarity clustering analysis method, comprising the following steps:
[0007] Acquire multi-dimensional medical data of multiple medical patients related to clustering targets;
[0008] constructing a multidimensional scaling matrix based on the multidimensional medical data of all medical patients, wherein the matrix elements of the multidimensional scaling matrix represent the position of each medical patient in a set low-dimensional space;
[0009] The position of each medical patient in the multidimensional scaling matrix is taken as a candidate solution, and the candidate solution is iteratively updated according to the attraction and repulsion optimization rule until the iteration condition is met, and the optimal solution among the candidate solutions of multiple iterations is taken as the medical patient similarity clustering result; if the comprehensive fitness index of the new candidate solution generated in each iteration is less than the comprehensive fitness index of the previous candidate solution, the new candidate solution is updated as the optimal solution, and the comprehensive fitness index is calculated by weighted calculation of the fitness value variance and the average fitness difference of the current candidate solution; the attraction and repulsion optimization rule is to iterate the current candidate solution according to the attraction and repulsion vector, the optimal solution fine-tuning vector and the local search vector to obtain a new candidate solution.
[0010] In the second aspect, the present solution provides a medical patient similarity clustering analysis system, including:
[0011] A data acquisition module, used for acquiring multi-dimensional medical data of multiple medical patients related to the clustering target;
[0012] A data dimension reduction module, used for constructing a multi-dimensional scaling matrix based on the multi-dimensional medical data of all medical patients, wherein the matrix elements of the multi-dimensional scaling matrix represent the position of each medical patient in a set low-dimensional space;
[0013] The clustering analysis module is used to take the position of each medical patient in the multi-dimensional scaling matrix as a candidate solution, iteratively update the candidate solution according to the attraction and repulsion optimization rule until the iteration condition is met, and take the optimal solution among the candidate solutions of multiple iterations as the medical patient similarity clustering result; wherein if the comprehensive fitness index of the new candidate solution generated in each iteration is less than the comprehensive fitness index of the previous candidate solution, the new candidate solution is updated as the optimal solution, and the comprehensive fitness index is obtained by weighted calculation of the fitness value variance and the average fitness difference of the current candidate solution; wherein the attraction and repulsion optimization rule is to iterate the current candidate solution according to the attraction and repulsion vector, the optimal solution fine-tuning vector and the local search vector to obtain a new candidate solution
[0014] In a third aspect, the present solution provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the above-mentioned medical patient similarity clustering analysis method.
[0015] In a fourth aspect, the present solution provides a readable storage medium, in which a computer program is stored. The computer program includes a program code for controlling a process to execute a process, and the process includes the above-mentioned medical patient similarity clustering analysis method.
[0016] Different from the prior art, the main contributions and innovations of the present invention are as follows:
[0017] 1. Adapted to solving the complex nonlinear relationships of high-dimensional medical data and improving the accuracy of cluster analysis of medical patients. This solution uses a multidimensional scaling algorithm to effectively reduce the data dimension of high-dimensional medical data and simplify the data structure to solve the difficulty of high-dimensional data processing; then the attraction and repulsion optimization algorithm is used to explore and utilize the nonlinear relationships in medical data through a dynamic optimization process, thereby effectively processing the nonlinearity and high dimensionality of medical data to improve more accurate medical patient similarity cluster analysis results, more accurately divide patient groups, identify true pathological patterns and patient similarities, and support personalized treatment and management.
[0018] 2. It has high computational flexibility and adaptability to provide a dynamic optimization strategy. The attraction-repulsion optimization algorithm provided by this solution provides a method for dynamically updating solutions. It can not only guide the search for similar medical patients based on fitness, but also take into account the interactions between medical patients, allowing the solution to be continuously updated in the search space based on the interactions, optimizing the diversity and quality of similar clustering results, and is more suitable for processing data sets of various sizes and types. It can be flexibly adjusted according to specific data characteristics and requirements, and is suitable for medical data sets of different types and sizes, especially in a changing clinical environment, and can better adapt to different data and requirements.
[0019] 3. Suitable for the analysis of large-scale medical data with complex data distribution. The present invention is applicable to large-scale data sets, can effectively process a large amount of patient data, can adapt to the complex distribution of medical data to discover potential similar medical patient groups, and support medical health management and research in the big data era;
[0020] Moreover, more accurate patient grouping can improve the effectiveness of chronic disease management, achieve early intervention, reduce treatment costs, and improve patients' quality of life.
[0021] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0023] Figure 1 It is a logical diagram of the training method of the medical news heat decay prediction model provided in this solution.
[0024] Figure 2It is a schematic diagram of the framework of the medical patient similarity clustering analysis system. Figure 3 It is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION
[0025] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of this specification. Instead, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.
[0026] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may be combined into a single step for description in other embodiments.
[0027] Embodiment 1
[0028] The embodiment of the present application provides a medical patient similarity clustering analysis method, which combines a multidimensional scaling algorithm and an attraction-repulsion optimization algorithm to cluster similar medical patient groups based on medical data. First, the scheme reduces the dimensionality of complex medical data through multidimensional scaling data, and then introduces a dynamically changing attraction-repulsion optimization algorithm to optimize the points representing each medical patient after dimensionality reduction, and simulates the attraction and repulsion relationship between medical patients to cluster similar medical patients and stay away from dissimilar medical patients, thereby achieving accurate clustering of similar medical patient groups.
[0029] like Figure 1 As shown, the medical patient similarity clustering analysis method provided by this scheme includes the following steps:
[0030] Acquire multi-dimensional medical data of multiple medical patients related to clustering targets;
[0031] constructing a multidimensional scaling matrix based on the multidimensional medical data of all medical patients, wherein the matrix elements of the multidimensional scaling matrix represent the position of each medical patient in a set low-dimensional space;
[0032] The position of each medical patient in the multidimensional scaling matrix is taken as a candidate solution, and the candidate solution is iteratively updated according to the attraction and repulsion optimization rule until the iteration condition is met, and the optimal solution among the candidate solutions of multiple iterations is taken as the medical patient similarity clustering result; if the comprehensive fitness index of the new candidate solution generated in each iteration is less than the comprehensive fitness index of the previous candidate solution, the new candidate solution is updated as the optimal solution, and the comprehensive fitness index is calculated by weighted calculation of the fitness value variance and the average fitness difference of the current candidate solution; the attraction and repulsion optimization rule is to iterate the current candidate solution according to the attraction and repulsion vector, the optimal solution fine-tuning vector and the local search vector to obtain a new candidate solution.
[0033] The medical patient similarity clustering analysis method provided in this solution utilizes multidimensional scaling dimensionality reduction technology and attraction-repulsion optimization algorithm to more accurately analyze and identify the similarities and differences between medical patients, thereby performing effective patient grouping, which helps doctors provide targeted treatment based on the characteristics of different patient groups.
[0034] Specifically, in the step of obtaining multi-dimensional medical data:
[0035] Since the medical patient similarity clustering analysis method provided by this scheme is applicable to the clustering of medical patient groups with various clustering targets, in other words, the medical patient similarity clustering analysis method provided by this scheme can be applied to clustering targets in many different medical clustering analysis scenarios. The medical data of medical patients that need to be collected for different clustering targets are different, so this scheme needs to collect multi-dimensional medical data related to the clustering target according to the clustering target.
[0036] The "clustering target" mentioned in this solution refers to the standard for clustering similar medical patients. For example, the clustering target can be: clustering medical patients with the same type of chronic diseases for chronic disease management. In this case, the multi-dimensional medical data includes multi-dimensional medical data related to the current chronic disease (such as diabetes / hypertension / cardiovascular disease), including but not limited to clinical medical indicators / long-term monitoring data (such as drug use records, physiological indicator monitoring (such as blood sugar monitoring records, blood pressure records) and disease progression record data (such as the time points of aggravation and relief of the disease and the effects of related treatment measures), etc.; another example of the clustering target can be: clustering medical patients with high-risk diseases for early disease screening. In this case, the multi-dimensional medical data includes medical data of different dimensions related to the current disease, including but not limited to personal information of medical patients (such as genetic information, environmental exposure, and living habits), early symptom data (such as symptoms that may not be obvious in the early stage), and historical health data (such as previous physical examination records and early health assessment results).
[0037] In other words, in the step of "obtaining multi-dimensional medical data of multiple medical patients related to clustering targets", the present solution selects a clustering target of a clustered medical patient group, and obtains relevant multi-dimensional medical data according to the clustering target. In some embodiments, the multi-dimensional medical data includes health records with detailed medical records, age, gender, family medical history, etc., clinical data with physical examination results, laboratory test data, imaging data, etc., and life habits with information such as the patient's eating habits, exercise frequency, sleep quality, etc.
[0038] In addition, since the multi-dimensional medical data acquired by this solution covers medical data of multiple different dimensions, it is necessary to perform data preprocessing on the multi-dimensional data, wherein the data preprocessing includes data cleaning, data feature processing, and data standardization processing. In some embodiments, data cleaning refers to supplementing missing values and removing abnormal values from multi-dimensional medical data to ensure the quality of multi-dimensional medical data; data feature processing refers to extracting key features from medical data based on clinical knowledge, such as extracting index features of patients from continuous biochemical indicators, extracting image features of patients from medical images, etc.; data standardization refers to normalizing or standardizing numerical medical data to ensure data comparability between medical data of different dimensions.
[0039] Specifically, in the step of constructing a multidimensional scaling matrix:
[0040] This solution uses a multidimensional scaling algorithm to reduce high-dimensional multidimensional medical data to low dimensions, simplifying the complexity of the data and making it easier to identify and analyze the structure and patterns between data in a lower-dimensional space.
[0041] In the step of "constructing a multidimensional scaling matrix based on the multidimensional medical data of all medical patients", an n-dimensional dissimilarity matrix is constructed according to the Euclidean distances between different medical patients, wherein the matrix elements of the dissimilarity matrix represent the Euclidean distances between every two medical patients; a doubly centered matrix is constructed based on the square of the dissimilarity matrix and the centered matrix; the doubly centered matrix is eigenvalue decomposed to obtain m eigenvalues and corresponding eigenvectors, and an m-dimensional multidimensional scaling matrix is constructed according to the m eigenvalues and the corresponding eigenvectors, wherein the matrix elements in the multidimensional scaling matrix represent the position of each medical patient in the m-dimensional space.
[0042] Specifically, the dissimilarity matrix is expressed as:
[0043]
[0044] Where D is the n*n dimensional dissimilarity matrix, d i,jrepresents the Euclidean distance between the i-th medical patient and the j-th medical patient, and the dissimilarity matrix is constructed by the Euclidean distances between n medical patients.
[0045] The construction of a dual-centered matrix based on the square of the dissimilar matrix and the centered matrix includes the following steps: constructing a centered matrix based on the unit matrix and the n*n full square matrix, taking the square of the dissimilar matrix and performing two centered matrix processes, and then taking half of the value to obtain a dual-centered matrix. The dual-centered matrix is expressed as:
[0046]
[0047]
[0048] Where B is the double-centered matrix, D 2 is the square of the dissimilarity matrix, C is the centering matrix, I is the identity matrix, and Jn is an n*n square matrix whose matrix elements are all 1.
[0049] The formula for the m-dimensional multidimensional scaling matrix constructed based on m eigenvalues and corresponding eigenvectors is as follows:
[0050] X=EΛ 1 / 2 ;
[0051]
[0052] Where X is an m-dimensional multidimensional scaling matrix, E is a matrix composed of m eigenvectors, Λ 1 / 2 is a diagonal matrix consisting of the square roots of m eigenvalues, where x nm Represents the position of the nth medical patient in the mth dimensional space.
[0053] It should be noted that the multidimensional scaling matrix constructed in this scheme is n*m-dimensional, and the high-dimensional dissimilarity data of the multidimensional scaling matrix is reduced to m-dimensional space to prepare for the subsequent attraction and repulsion optimization. Each matrix element in the multidimensional scaling matrix corresponds to a medical patient, and the position of each medical patient represents a "compressed" expression of its multidimensional medical data; then the attraction and repulsion optimization algorithm optimizes the multidimensional scaling matrix to simulate the attraction and repulsion between medical patients to adjust the position of each medical patient, thereby obtaining the optimal configuration of the medical patient positions, so as to cluster similar medical groups and stay away from dissimilar medical patients.
[0054] Specifically, in the process of iteratively updating the multidimensional scaling matrix according to the attraction and repulsion optimization rule:
[0055] This solution uses the position of each medical patient in the multidimensional scaling matrix as a candidate solution, and uses the attraction-repulsion optimization algorithm to optimize the solution of the multidimensional scaling matrix to cluster similar medical groups. The attraction-repulsion optimization algorithm mentioned in this solution is an efficient optimization framework that combines global search and local search, which is particularly suitable for solving complex optimization problems. The algorithm simulates the interaction between biological individuals in nature, that is, the attraction between similar individuals and the repulsion between different individuals, to find the optimal solution. The present invention utilizes the characteristics of the algorithm and applies it to the similarity evaluation of medical patient groups in medical data analysis.
[0056] In some embodiments, in order to increase the diversity of candidate solutions, it is necessary to iteratively update multiple groups of candidate solutions. At this time, the position of each medical patient in the multidimensional scaling matrix that has not been iteratively updated is set as the initial candidate solution, and the position of each medical patient in the initial candidate solution is randomly perturbed to obtain multiple perturbed candidate solutions, which are collectively referred to as the initial candidate solution and the perturbed candidate solution as the candidate solution.
[0057] Specifically, the perturbation candidate solution is obtained as follows:
[0058] x i ′=x i +rand☉(x max -x min );
[0059] where x i ′ is the position of the medical patient in the perturbation candidate solution, x i is the position of the medical patient in the initial candidate solution, xmax, xmin are the limits of the search space and can be set as needed, and rand is a random perturbation.
[0060] It should be noted that, in some embodiments, the initial candidate solution corresponding to the position of each medical patient in the multidimensional scaling matrix that has not been iteratively updated is not perturbed, and the candidate solution in this case refers to the initial candidate solution; while in other embodiments, the initial candidate solution corresponding to the position of each medical patient in the multidimensional scaling matrix that has not been iteratively updated is perturbed to obtain a perturbed candidate solution, and the candidate solution in this case refers to the initial candidate solution and the perturbed candidate solution.
[0061] In the process of iteratively updating the candidate solutions using the attraction and repulsion optimization rule, this scheme judges whether the candidate solution generated by the current iteration is a better solution based on the comprehensive fitness index of the new candidate solution generated in each iteration. Correspondingly, the comprehensive fitness index is calculated by weighting the fitness value variance and the average fitness difference of the current candidate solution. The specific formula is:
[0062] S = α·MeanDiff avg +β·Var(F);
[0063] Where S is the comprehensive adaptability index of the current candidate solution, MeanDiff avg is the average difference in fitness, Var(F) is the variance of fitness value, α and β are weight coefficients for adjusting the variance of fitness value and the average difference in fitness, which are used to adjust the relative importance of the variance of fitness value and the average difference in fitness when evaluating the comprehensive fitness index.
[0064] In an embodiment of the present solution, the present solution calculates the fitness value of each medical patient of the current candidate solution according to the designed fitness function, obtains the fitness value variance and the fitness average difference of the current candidate solution based on the fitness value, and calculates the comprehensive fitness index of the current candidate solution according to the fitness value variance and the fitness average difference. It should be noted that the optimization goal of iterative optimization is to minimize the distance between medical patients with similar fitness values. Therefore, the present solution uses a comprehensive fitness index to make medical patients with similar fitness values closer to each other in the low-dimensional space after multi-dimensional scaling, which helps to more effectively identify and manage medical patient groups.
[0065] The method for obtaining the average fitness difference is: for each medical patient, calculate the individual average difference between the fitness value of the medical patient and the fitness values of all other medical patients, and then take the average of the individual average differences as the average fitness difference. The specific calculation formula is as follows:
[0066]
[0067]
[0068] MeanDiff i is the average individual difference, f i is the fitness value of the i-th medical patient, f j is the fitness value of the jth medical patient, n is the number of all medical patients in the current candidate solution, MeanDiff avg is the average fitness difference.
[0069] The method for obtaining the variance of the fitness value is: obtain the average fitness value of all medical patients of the current candidate solution, and calculate the variance of the fitness value of the current candidate solution based on the average fitness value and the fitness value of each medical patient. The specific calculation formula is as follows:
[0070]
[0071] Where Var(F) is the variance of the fitness value, f i is the fitness value of the i-th medical patient, n is the number of all medical patients in the current candidate solution, is the average fitness value.
[0072] In addition, it should be noted that this solution designs different fitness functions according to the clustering target. The fitness function is a weighted function of the medical indicators related to the current clustering target, where the medical indicators are taken from at least one multidimensional medical data of the multidimensional medical data. In other words, the fitness function of this solution can be flexibly designed according to the needs of different multidimensional medical data characteristics and clustering targets, combined with factors such as disease characteristics, clinical outcomes, and treatment responses. The fitness function not only helps the algorithm understand the health status of each medical patient under the current clustering target, but also guides the attraction-repulsion optimization algorithm on how to significantly improve the analysis and understanding of the medical patient group by adjusting the position of the medical patient in the low-dimensional space.
[0073] Specifically, this scheme calculates the performance of each medical patient in the current candidate solution according to the fitness function to guide the attraction-repulsion optimization rule to better adjust the position of the medical patient. If the fitness value of the medical patient indicates that they are in good health or similar to some other medical patients, the positions of these medical patients will be attracted to each other under the attraction-repulsion optimization rule, and thus clustered together in a low-dimensional space, which helps to identify groups of medical patients with similar health characteristics or disease patterns; similarly, if the health conditions between medical patients vary greatly, these medical patients will repel each other under the attraction-repulsion optimization rule to clearly distinguish different medical patients with significantly different health conditions or disease states. By optimizing the positions of medical patients after attraction-repulsion, similar medical groups under the clustering target can be better clustered.
[0074] Taking the clustering goal of clustering medical patients with the same type of chronic diseases for chronic disease management as an example, the fitness function designed in this scheme is expressed as follows:
[0075] f i =w1×z1+w2×z2+...+w m × m ;
[0076] where f i is the fitness value of the i-th medical patient, |z m is the mth medical index after standardization, w m are the weights of the mth medical indicator. These weights reflect the importance of each indicator. The weight values can be obtained based on human experience or through design modeling training.
[0077] Specific medical indicators are selected from multidimensional medical data. For example, medical indicators can be selected as blood pressure, cholesterol level, glycosylated hemoglobin, etc. These medical indicators are key indicators of the current clustering target.
[0078] Regarding the attraction-repulsion optimization rule mentioned in this solution, the attraction-repulsion optimization rule mentioned in this solution is to iterate the current candidate solution according to the attraction-repulsion vector, the optimal solution fine-tuning vector and the local search vector to obtain a new candidate solution. That is to say, the position of each medical patient is updated at each iteration according to the dynamics of attraction and repulsion and the result of the local search operation. Specifically, the formula for each iteration update is as follows:
[0079] x i (t) = x i (t-1)+n i +b i +r i ;
[0080] where x i (t-1) is the position of the i-th medical patient in the candidate solution before the current iteration, x i (t) is the position of the i-th medical patient in the candidate solution after the current iteration, t is the number of iterations, n i is the attraction and repulsion vector corresponding to the i-th medical patient, b i is the optimal solution fine-tuning vector corresponding to the i-th medical patient, r i is the local search vector corresponding to the i-th medical patient.
[0081] About attraction and repulsion vector: The update of each medical patient's position is affected by the attraction and repulsion of other medical patients' positions. The attraction comes from the medical patients with higher fitness values, while the repulsion comes from the medical patients with lower fitness values. This mechanism ensures the diversity and exploration efficiency of the understanding space. This scheme defines the attraction and repulsion vector based on the weighted sum of the distance between the current medical patient and other medical patients and the interpolation of the fitness value. The specific formula is as follows:
[0082]
[0083]
[0084]
[0085]
[0086] where n i is the attraction and repulsion vector corresponding to the i-th medical patient, x i is the location of the i-th medical patient, x j is the position of the jth medical patient, n is the number of all medical patients in the current candidate solution, c is the custom step size, I() is the function that reflects the influence of distance on attraction or repulsion, d ij is the distance between the i-th medical patient and the j-th medical patient, d i,maxis the maximum distance between the ith medical patient and all medical patients, s(f i , f j ) is the direction function, f i is the fitness value of the i-th medical patient, f j is the fitness value of the jth medical patient, k is the number of neighbors considered in each iteration, k decreases as the number of iterations increases, t is the number of iterations, t max is the maximum number of iterations.
[0087] About the optimal solution fine-tuning vector: At each iteration, each medical patient in the candidate solution will be affected by the optimal solution. This scheme uses the optimal solution fine-tuning vector to represent the impact of the optimal solution on the candidate solution. The optimal solution fine-tuning vector not only promotes convergence to the global optimal solution, but also maintains the breadth of the search and prevents premature falling into the local optimum. The optimal solution fine-tuning vector is:
[0088]
[0089]
[0090] where x best is the optimal solution, x i is the i-th medical patient, a1 is a random number vector, c is the step size, m is the adjustment factor that controls the size of attraction, and the value of m is dynamically adjusted as the number of iterations increases to balance exploration and utilization. t is the number of iterations, t max is the maximum number of iterations, r1 is a random number in the range of [0,1], which is used to decide whether to introduce additional random perturbations; p1 is the probability threshold that controls the frequency of introducing random perturbations.
[0091] About local search vector: The local search operation of this solution is a key link, which is used to explore and utilize the solution space to find the optimal solution. These operations determine which search strategy to use based on the probability threshold and iteration progress, so that the algorithm maintains a balance between global exploration and local utilization. In other words, the local search vector is selected from the Brownian motion search vector, trigonometric function search vector, and random position search vector based on the probability value in the iteration process and the preset threshold. The specific selection formula is as follows;
[0092]
[0093] where r B is the Brownian motion search vector, r tri Search vector for trigonometric functions, r R is the random position search vector, t is the number of iterations, t max is the maximum number of iterations, r2 and r3 are randomly generated probability values, and p2 is the preset threshold.
[0094] The Brownian motion search vector is a local search that simulates the random walking behavior of small particles. It is usually used for extensive exploration in the search space. The step size is determined by a normal distribution, where the mean is 0 and the standard deviation is dynamically adjusted according to the number of iterations. The Brownian motion search vector provided by this solution is expressed as:
[0095]
[0096] Where u1 is a binary vector, f r1 is a small coefficient for adjusting the step size, t is the number of iterations, t max is the maximum number of iterations, x max refers to the upper limit of the search space, x min Refers to the lower bound of the search space.
[0097] The trigonometric function search vector simulates periodic changes through sine or cosine functions, which is suitable for fine search in a local area. The trigonometric function search vector provided by this solution is expressed as:
[0098]
[0099] x w =Select(X,p);
[0100] where f r2 is a small coefficient to adjust the step size, u2 is a binary vector used to randomly select which dimensions will be affected, a2 is a random number vector that increases the randomness and diversity of the search, r4 is a random number in the interval [0,1], r5 is a random number in the interval [0,1], and x w is the position selected by roulette, representing a potential optimal position in the current candidate solution, X is the position set of all medical patients in the current candidate solution, p is the selection probability array calculated according to the fitness value, the higher the fitness of the individual, the greater its value in the array p, and the higher the probability of being selected. The selection function Select is executed based on the cumulative probability distribution.
[0101] The random position search vector randomly selects a new position in the solution space to increase the diversity of solutions. The random position search vector of this scheme is expressed as:
[0102] r R =u3☉(2·a3-1)☉(x max -x min );
[0103] Where u3 is a binary vector, a3 is a random number vector in the interval [0, 1], and x max refers to the upper limit of the search space, x min Refers to the lower bound of the search space.
[0104] This scheme iteratively updates the candidate solution according to the above-mentioned attraction and repulsion optimization rule until the iteration condition is met. Its ultimate goal is to minimize the distance between medical patient groups with similar fitness values. This scheme provides two iteration conditions: 1. If the change value of the comprehensive fitness index of the candidate solution generated by the iteration is less than the convergence threshold, the optimization is considered to have converged and the iteration is stopped; 2. If the number of iterations reaches the set maximum value, the iteration is stopped. Correspondingly, in the medical patient similarity clustering results obtained by this scheme, medical patients with similar distances are clustered into similar medical patient groups.
[0105] The medical patient similarity clustering analysis method provided in this solution obtains the medical patient similarity clustering results by clustering medical patients based on multidimensional medical data. According to the position of each medical patient in the medical patient similarity clustering results, the medical patients can be divided into different medical patient groups, and each medical patient group has similar clinical characteristics and disease risks that meet the clustering goals. Clinically, relevant personnel can analyze the differences between different medical patient groups, such as lifestyles, treatment response physiological indicators, etc. for medical management, and can also provide data support for early disease identification and management, so that doctors can provide targeted treatment for different medical patient groups.
[0106] Embodiment 2
[0107] like Figure 2 As shown, referring to Example 1, this solution provides a medical patient similarity clustering analysis system, including the following functional modules:
[0108] A data acquisition module, used for acquiring multi-dimensional medical data of multiple medical patients related to the clustering target;
[0109] A data dimension reduction module, used for constructing a multi-dimensional scaling matrix based on the multi-dimensional medical data of all medical patients, wherein the matrix elements of the multi-dimensional scaling matrix represent the position of each medical patient in a set low-dimensional space;
[0110] The clustering analysis module is used to take the position of each medical patient in the multi-dimensional scaling matrix as a candidate solution, iteratively update the candidate solution according to the attraction and repulsion optimization rule until the iteration condition is met, and take the optimal solution among the candidate solutions of multiple iterations as the medical patient similarity clustering result; wherein if the comprehensive fitness index of the new candidate solution generated in each iteration is less than the comprehensive fitness index of the previous candidate solution, the new candidate solution is updated as the optimal solution, and the comprehensive fitness index is obtained by weighted calculation of the fitness value variance and the average fitness difference of the current candidate solution; wherein the attraction and repulsion optimization rule is to iterate the current candidate solution according to the attraction and repulsion vector, the optimal solution fine-tuning vector and the local search vector to obtain a new candidate solution.
[0111] The contents of the second embodiment that are the same as those of the first embodiment will not be described in detail here.
[0112] Embodiment 3
[0113] This embodiment also provides an electronic device, referring to Figure 3 , including a memory 404 and a processor 402, wherein the memory 404 stores a computer program, and the processor 402 is configured to run the computer program to execute the steps in any of the above-mentioned embodiments of the medical patient similarity clustering analysis method.
[0114] Specifically, the processor 402 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.
[0115] The memory 404 may include a large-capacity memory 404 for data or instructions. The memory 404 may be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 402.
[0116] The processor 402 reads and executes the computer program instructions stored in the memory 404 to implement any one of the medical patient similarity clustering analysis methods in the above embodiments.
[0117] Optionally, the electronic device may further include a transmission device 406 and an input / output device 408 , wherein the transmission device 406 is connected to the processor 402 , and the input / output device 408 is connected to the processor 402 .
[0118] The transmission device 406 may be used to receive or send data via a network. Specific examples of the above network may include a wired or wireless network provided by a communication provider of the electronic device.
[0119] The input / output device 408 is used to input or output information. In this embodiment, the input information may be multi-dimensional medical data of medical patients, and the output information may be a group of similar medical patients.
[0120] Optionally, in this embodiment, the processor 402 may be configured to perform the following steps through a computer program:
[0121] Acquire multi-dimensional medical data of multiple medical patients related to clustering targets;
[0122] constructing a multidimensional scaling matrix based on the multidimensional medical data of all medical patients, wherein the matrix elements of the multidimensional scaling matrix represent the position of each medical patient in a set low-dimensional space;
[0123] The position of each medical patient in the multidimensional scaling matrix is taken as a candidate solution, and the candidate solution is iteratively updated according to the attraction and repulsion optimization rule until the iteration condition is met, and the optimal solution among the candidate solutions of multiple iterations is taken as the medical patient similarity clustering result; if the comprehensive fitness index of the new candidate solution generated in each iteration is less than the comprehensive fitness index of the previous candidate solution, the new candidate solution is updated as the optimal solution, and the comprehensive fitness index is calculated by weighted calculation of the fitness value variance and the average fitness difference of the current candidate solution; the attraction and repulsion optimization rule is to iterate the current candidate solution according to the attraction and repulsion vector, the optimal solution fine-tuning vector and the local search vector to obtain a new candidate solution.
[0124] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.
[0125] In general, various embodiments may be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. Some aspects of the invention may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device, but the invention is not limited thereto. Although various aspects of the invention may be shown and described as block diagrams, flow charts, or using some other graphical representation, it should be understood that, as non-limiting examples, the boxes, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.
[0126] Embodiments of the present invention may be implemented by computer software that is executable by a data processor of a mobile device, such as in a processor entity, or by hardware, or by a combination of software and hardware. Computer software or programs (also referred to as program products) including software routines, applets and / or macros may be stored in any device-readable data storage medium, and they include program instructions for performing specific tasks. A computer program product may include one or more computer executable components configured to perform an embodiment when the program is run. One or more computer executable components may be at least one software code or a portion thereof. In addition, at this point, it should be noted that, for example, Figure 3Any block of the logic flow in the program may represent program steps, or interconnected logic circuits, blocks and functions, or a combination of program steps and logic circuits, blocks and functions. The software may be stored on physical media such as memory chips or storage blocks implemented within a processor, magnetic media such as hard disks or floppy disks, and optical media such as, for example, DVDs and data variants thereof, CDs, etc. Physical media are non-transitory media.
[0127] Those skilled in the art should understand that the technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0128] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A medical patient similarity clustering analysis method, characterized in that: The following steps are involved: Acquire multi-dimensional medical data of multiple medical patients related to clustering targets; constructing a multidimensional scaling matrix based on the multidimensional medical data of all medical patients, wherein the matrix elements of the multidimensional scaling matrix represent the position of each medical patient in a set low-dimensional space; The position of each medical patient in the multidimensional scaling matrix is taken as a candidate solution, and the candidate solution is iteratively updated according to the attraction and repulsion optimization rule until the iteration condition is met, and the optimal solution among the candidate solutions of multiple iterations is taken as the medical patient similarity clustering result; if the comprehensive fitness index of the new candidate solution generated in each iteration is less than the comprehensive fitness index of the previous candidate solution, the new candidate solution is updated as the optimal solution, and the comprehensive fitness index is calculated by weighting the fitness value variance and the average fitness difference of the current candidate solution; The attraction-repulsion optimization rule is to iterate the current candidate solution according to the attraction-repulsion vector, the optimal solution fine-tuning vector and the local search vector to obtain a new candidate solution.
2. The medical patient similarity clustering analysis method according to claim 1, characterized in that: An n-dimensional dissimilarity matrix is constructed according to the Euclidean distances between different medical patients, wherein the matrix elements of the dissimilarity matrix represent the Euclidean distances between every two medical patients; Construct a dual-centered matrix based on the square of the dissimilar matrix and the centered matrix; The doubly centered matrix is subjected to eigenvalue decomposition to obtain m eigenvalues and corresponding eigenvectors, and an m-dimensional multidimensional scaling matrix is constructed based on the m eigenvalues and the corresponding eigenvectors, wherein the matrix elements in the multidimensional scaling matrix represent the position of each medical patient in the m-dimensional space.
3. The medical patient similarity clustering analysis method according to claim 1, characterized in that: Different fitness functions are designed according to the clustering objectives. The fitness function is a weighted function of the medical indicators related to the current clustering objective, wherein the medical indicators are taken from at least one multidimensional medical data of the multidimensional medical data. The fitness value of each medical patient of the current candidate solution is calculated according to the designed fitness function, and the fitness value variance and the fitness average difference of the current candidate solution are calculated based on the fitness value, and the comprehensive fitness index of the current candidate solution is calculated according to the fitness value variance and the fitness average difference.
4. The medical patient similarity clustering analysis method according to claim 3, characterized in that: The method for obtaining the average fitness difference is: for each medical patient, calculate the individual average difference between the fitness value of the medical patient and the fitness values of all other medical patients, and then take the average of the individual average differences as the average fitness difference; the method for obtaining the fitness value variance is: obtain the fitness value average of the fitness values of all medical patients of the current candidate solution, and calculate the fitness value variance of the current candidate solution based on the fitness value average and the fitness value of each medical patient.
5. The medical patient similarity clustering analysis method according to claim 1, characterized in that: The attraction and repulsion vector is defined according to the weighted sum of the distance between the current medical patient and other medical patients and the interpolation of the fitness value: where n i is the attraction and repulsion vector corresponding to the i-th medical patient, x i is the location of the i-th medical patient, x j is the position of the jth medical patient, n is the number of all medical patients in the current candidate solution, c is the custom step size, I( ) is the function that reflects the influence of distance on attraction or repulsion, d i,j is the distance between the i-th medical patient and the j-th medical patient, d i,max is the maximum distance between the ith medical patient and all medical patients, s(f i , f j ) is the direction function, f i is the fitness value of the i-th medical patient, f j is the fitness value of the jth medical patient.
6. The medical patient similarity clustering analysis method according to claim 1, characterized in that: The optimal solution fine-tuning vector is: where x best is the optimal solution, x i is the i-th medical patient, a1 is a random number vector, c is the step size, m is the adjustment factor that controls the size of attraction, t is the number of iterations, t max is the maximum number of iterations, r1 is a random number in the range of [0, 1], and p1 is the probability threshold for controlling the frequency of introducing random disturbances.
7. The medical patient similarity clustering analysis method according to claim 1, characterized in that: According to the probability value in the iteration process and the preset threshold, a local search vector is selected from the Brownian motion search vector, the trigonometric function search vector, and the random position search vector. The selection formula is as follows; where r B is the Brownian motion search vector, r tri Search vector for trigonometric functions, r R is the random position search vector, t is the number of iterations, t max is the maximum number of iterations, r2 and r3 are randomly generated probability values, and p2 is the preset threshold.
8. A medical patient similarity clustering analysis system, characterized in that: include: A data acquisition module, used for acquiring multi-dimensional medical data of multiple medical patients related to the clustering target; A data dimension reduction module, used for constructing a multi-dimensional scaling matrix based on the multi-dimensional medical data of all medical patients, wherein the matrix elements of the multi-dimensional scaling matrix represent the position of each medical patient in a set low-dimensional space; A clustering analysis module is used to take the position of each medical patient in the multi-dimensional scaling matrix as a candidate solution, iteratively update the candidate solution according to the attraction and repulsion optimization rule until the iteration condition is met, and take the optimal solution among the candidate solutions of multiple iterations as the medical patient similarity clustering result; wherein if the comprehensive fitness index of the new candidate solution generated in each iteration is less than the comprehensive fitness index of the previous candidate solution, the new candidate solution is updated as the optimal solution, and the comprehensive fitness index is calculated by weighting the fitness value variance and the average fitness difference of the current candidate solution; The attraction-repulsion optimization rule is to iterate the current candidate solution according to the attraction-repulsion vector, the optimal solution fine-tuning vector and the local search vector to obtain a new candidate solution.
9. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to run the computer program to execute the medical patient similarity clustering analysis method according to any one of claims 1 to 7.
10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, wherein the computer program includes a program code for controlling a process to execute a process, wherein the process includes the medical patient similarity clustering analysis method according to any one of claims 1 to 7.
Citation Information
Cited By
Clinical medication data recording method and system
CN119580916A
Medical care data processing method and system and electronic equipment
CN119851846A
Training method of southern semen cuscutae grating cell intelligent detection model
CN119863791A
Medical image denoising system and method based on 3D chaos and attraction and rejection algorithm
CN120471795A
Medical image denoising system and method based on 3D chaos and attraction-repulsion algorithm
CN120471795B
