A medical record management method and system

Abstract

The present invention discloses a medical record management method and system, which relates to the technical field of big data analysis. By generating a comprehensive set of feature vectors Fp, through comparing the features with historical cases, the similarity XS between the historical cases and the patient is obtained, the degree of feature aggregation of the historical cases is calculated to obtain the feature concentration D, and the feature concentration correlation data set R is generated by integrating the feature concentration D. Based on this, a tree-like association graph is established, which helps doctors quickly identify the most relevant historical cases when managing and retrieving patient records, optimizing the management efficiency of medical records. Especially in solving the deficiencies of low retrieval efficiency, loose data organization, and inability to effectively associate patients with historical cases in existing medical record management, it shows significant advantages. Doctors can quickly locate historical cases that highly match the patient's features, greatly shortening the retrieval time and improving the accuracy of personalized medical decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and particularly to a medical record management method and system. Background Art

[0002] The medical informatization system is one of the important infrastructures in the current medical industry, which aims to improve the management and utilization efficiency of medical resources through information technology means. The medical record management system is a crucial branch that is responsible for recording and managing patients' medical record information. With the continuous accumulation of medical data, simple storage and retrieval can no longer meet the needs of modern medicine, especially the need for refined management of patients' individual characteristics is becoming increasingly important.

[0003] However, in the current medical record management system, most patients' records exist in a linear data storage manner, lacking multi-dimensional classification and intelligent analysis tools for different patient characteristics. Existing systems often can only screen data according to simple single dimensions, such as sorting by disease type or medical record time. For situations involving cross-analysis of multiple dimensions, such as considering age, gender, and disease duration simultaneously, the processing capabilities of existing systems are limited. When doctors conduct research on group diseases, they often need to rely on manual or external tools for reprocessing of data, which not only consumes a large amount of time but also easily leads to data loss or processing errors. In addition, for some patient groups with special disease processes, it is impossible to quickly locate similar case files, which limits the reference of historical cases by the medical team during decision-making. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a medical record management method and system, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A medical record management method and system, including the following steps:

[0006] S1. By establishing a communication channel between the hospital registration system and the electronic medical record system, extract patient information, including personal information, diagnosis results, examination reports, treatment records, and patient complication records, to form a feature vector set Fp;

[0007] S2. Based on the feature vector set Fp, extract the patient types, and then compare them with the case characteristics under the same patient types in historical cases to obtain the similarity XS between the historical cases and the feature vector set Fp, and form a case similarity matrix S;

[0008] S3. Classify patients according to the obtained case similarity matrix S, establish the association relationship between the patients and the historical cases in the case similarity matrix S, and generate the classification result set C;

[0009] S4. Analyze the association between the patients and the historical cases according to the classification result set C and the similarity matrix S. By calculating all the historical cases in the classification, obtain the feature concentration degree D, and then integrate the feature concentration degree D to form the feature concentration degree association data set R;

[0010] S5. Sort according to the feature concentration degree association data set R, and obtain the sorted feature concentration degree association data set R to establish a tree-like association graph.

[0011] Preferably, the S1 includes S11;

[0012] S11. Extract patient information, including personal information, diagnosis results, inspection reports, treatment records, and patient complication records, by establishing a communication channel between the hospital registration system and the electronic case system;

[0013] Among them, the personal information includes age A, gender G, and medical history MH; the diagnosis results include disease diagnosis DIA, blood glucose volatility BGV, and blood pressure BP; the treatment records include medication frequency MED and treatment cycle TE; the patient complication records include the number of complications CN and the duration of illness D;

[0014] Among them, the medical history MH specifically represents the number of years from the onset of the patient's illness to the present; the disease diagnosis DIA specifically represents the standardized value of the disease diagnosis; the duration of illness D specifically represents the duration of the complication;

[0015] Synchronously perform dimensionless processing on the personal information, diagnosis results, inspection reports, treatment records, and patient complication records and then integrate them to form the feature vector set Fp = {A, G, MH, DIA, BGV, BP, MED, TE, CN, D}.

[0016] Preferably, the S2 includes S21 and S22;

[0017] S21. Extract the patient type information based on the feature vector set Fp, including the disease diagnosis DIA, blood glucose volatility BGV, and the number of complications CN, to form the patient type T of the current patient, and then match the patient type T with the case type information in the historical case library. Among them, the matching is to input the patient type T as a matching item into the historical case library for retrieval to obtain the historical case samples including the patient type T, and form the matching historical case feature set Fhist;

[0018] Among them, the historical case feature set Fhist is specifically a two-dimensional matrix, including a number of historical case samples. The specific format of the historical case feature set Fhist is as follows:

[0019]

[0020] In the formula, m represents the total number of historical case samples.

[0021] Preferably, in S22, calculate the similarity XS between each case sample and the disease type T of the current patient in the feature vector set Fp according to the obtained historical case feature set Fhist, and form a case similarity matrix S = {XS i,j} with the current disease type T and a number of historical case samples in the historical case feature set Fhist;

[0022] The similarity XS is obtained through the following calculation formula:

[0023]

[0024] In the formula, XS i,j represents the similarity between the i-th patient and the j-th historical case sample in the historical case feature set Fhist, Fp i,k represents the k-th feature parameter of the i-th patient, Fhist j，k represents the k-th feature parameter of the j-th historical case sample in the historical case feature set Fhist, m represents the total number of features, specifically representing the length of the feature vector set Fp, and α k represents the weight coefficient of the k-th feature parameter.

[0025] Preferably, the S3 includes S31 and S32;

[0026] S31. Classify the patients according to the obtained case similarity matrix S. Specifically, match the similarity XS between the patient i and the historical case sample j in the case similarity matrix S with the upper threshold SZ and the lower threshold XZ of the preset similarity evaluation to obtain the association relationship between the patient i and the historical case sample j;

[0027] The association relationship is specifically obtained through the following matching method:

[0028] When the similarity XS i,j ≥ the upper threshold SZ, obtain the association relationship between the patient i and the historical case sample j as a first-level association relationship, and mark the association level GJ between the patient i and the historical case sample j as 1;

[0029] When the lower threshold XZ ≤ the similarity XS i,jWhen it is less than the upper threshold SZ, the association relationship between patient i and historical case sample j is a secondary association relationship, and the association level GJ between patient i and historical case sample j is marked as 2;

[0030] When the similarity XS i,j When it is less than the lower threshold XZ, the association relationship between patient i and historical case sample j is a tertiary association relationship, and the association level GJ between patient i and historical case sample j is marked as 3.

[0031] Preferably, in S32, the historical case samples with the association level GJ = 3 in the historical case feature set Fhist are removed, the historical case samples with the association levels GJ = 1 and GJ = 2 are retained, and then the historical case feature set Fhist is sorted according to the association level GJ to obtain the historical case feature set Fhist in the form of a list. The historical case feature set Fhist is cut with the historical case sample with the association level GJ = 1 that appears last in the historical case feature set Fhist, and divided into the historical case feature set Fhist1 with the association level GJ = 1 and the historical case feature set Fhist2 with the association level GJ = 2, forming the classification result set C = {Fhist1, Fhist2}.

[0032] Preferably, in S4, according to the classification result set C and the similarity matrix S, the association between the patient and the historical case is analyzed. By calculating all the historical cases in the classification, the feature concentration D = {D1, D2} is obtained. Specifically, by comparing the feature vector set Fp of the current patient with the classification result set C, the feature concentration D1 of the primary association historical case feature set Fhist1 and the feature concentration D2 of the secondary association historical case feature set Fhist2 are obtained;

[0033] Among them, the calculated feature concentration D1 is associated with the subscript of the primary association historical case feature set Fhist1, and the calculated feature concentration D2 is associated with the subscript of the primary association historical case feature set Fhist2, forming the feature concentration association data set R = {R1, R2}, where R1 represents the first feature concentration association data set, specifically R1 = {D1 ∪ Fhist1}; R2 represents the second feature concentration association data set, specifically R2 = {D2 ∪ Fhist2}.

[0034] Preferably, the feature concentration D1 of the primary association historical case feature set Fhist1 and the feature concentration D2 of the secondary association historical case feature set Fhist2 are obtained through the following calculation formulas:

[0035]

[0036] Wherein, n1 and n2 respectively represent the total number of historical case samples in the first-level associated historical case feature set Fhist1 and the second-level associated historical case feature set Fhist2, j represents the j-th historical case sample, m represents the total number of features, Fhist1 j,w and Fhist2 j,w respectively represent the w-th feature value of the j-th historical case sample in the first-level associated historical case feature set Fhist1 and the second-level associated historical case feature set Fhist2, Fp w represents the w-th feature parameter in the feature vector set Fp.

[0037] Preferably, S5, sorting according to the feature concentration association data set R, including using bubble sort to sort based on the feature concentrations D1 and D2 in the feature concentration association data set R, obtaining the sorted feature concentration association data set R, and establishing a tree-like association graph, including establishing the associated nodes of the root node according to the sorted feature concentration association data set R, where the root node is the feature vector set Fp of the patient, and the associated nodes include the first associated node and the second associated node;

[0038] The first associated node is specifically the first feature concentration association data set R1 in the feature concentration association data set R; the second associated node is specifically the second feature concentration association data set R2 in the feature concentration association data set R.

[0039] A medical record management method and system.

[0040] The present invention provides a medical record management method and system, having the following beneficial effects:

[0041] (1) By generating a comprehensive feature vector set Fp. By comparing with the features of historical cases, the similarity XS between the historical cases and the patient is obtained, the feature aggregation degree of the historical cases is calculated, the feature concentration D is obtained, and the feature concentration association data set R is generated by integrating the feature concentration D. And based on this, a tree-like association graph is established, thereby helping doctors to quickly identify the most relevant historical cases when managing and retrieving patient files, optimizing the management efficiency of medical records, especially in solving the deficiencies of low retrieval efficiency, loose data organization and inability to effectively associate patients with historical cases in the existing medical record management. It shows significant advantages. Doctors can quickly locate historical cases that highly match the patient's features, greatly shortening the retrieval time and improving the accuracy of personalized medical decisions. At the same time, the establishment of the tree-like association graph shows the historical cases associated with the patient at different levels in a hierarchical structure, providing an intuitive and orderly file management view for doctors, and further enhancing the utilization efficiency and pertinence of historical cases.

[0042] (2) By accurately classifying the case similarity matrix S according to the similarity level, the association relationship between patients and historical case samples can be divided into different association levels. The classified historical cases are further sorted according to the association level to generate a classification result set C, and the historical case set is segmented based on the association level to form a clear hierarchical structure, which effectively solves the problem that the traditional medical archive management cannot effectively distinguish the association strength between cases and patients, resulting in excessive data redundancy and low retrieval efficiency. Through the similarity level classification and the setting of the association level GJ, the system can accurately eliminate irrelevant historical cases and ensure that only the cases with the most reference value for patient diagnosis and treatment are retained, thereby significantly improving the utilization and accuracy of the data. At the same time, the hierarchical sorting and cutting mechanism further simplifies the review and management process of historical cases, allowing doctors to quickly locate the most relevant reference files in complex historical cases, optimizing the support for medical decision-making.

[0043] (3) By analyzing the classification result set C and the similarity matrix S, the feature concentration D of the historical cases is further calculated, and the feature concentration is associated with the corresponding historical case feature set to generate a feature concentration association data set R. On this basis, after sorting the feature concentration by bubble sorting, the system constructs a tree-like association graph based on the patient feature vector set Fp, in which the sorted historical case samples are presented in a hierarchical manner, with the first-level associated historical case samples close to the patient node, and the second-level associated historical case samples are arranged in sequence according to the similarity. This hierarchical structure can intuitively reflect the similarity between different historical cases and patient characteristics, and has shown significant advantages in solving the problem that traditional medical archive management cannot effectively visualize the correlation between historical cases and patients, resulting in complicated information and difficulty in quickly distinguishing highly correlated cases. By constructing a tree-like association graph, the system can help doctors intuitively and quickly view historical cases that are highly correlated with patient characteristics based on the sorting results of the feature concentration D, so that doctors can not only see the quantitative results of the similarity, but also directly obtain specific case characteristics, significantly improving the accuracy and efficiency of personalized medical decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a schematic diagram of the steps of a medical archive management method of the present invention;

[0045] Figure 2 The figure is a schematic diagram of a medical record management system according to the present invention. DETAILED DESCRIPTION

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0047] Embodiment 1

[0048] The present invention provides a medical record management method. Please refer to Figure 1 , which includes the following steps:

[0049] S1. By establishing a communication channel between the hospital registration system and the electronic case system, extract patient information, including personal information, diagnosis results, examination reports, treatment records, and patient complication records, to form a feature vector set Fp.

[0050] S2. Based on the feature vector set Fp, extract the patient type, and then compare it with the case features under the same patient type in the historical cases to obtain the similarity XS between the historical cases and the feature vector set Fp, and form a case similarity matrix S.

[0051] S3. Classify the patients according to the obtained case similarity matrix S, and establish an association relationship between the patients and the historical cases in the case similarity matrix S to generate a classification result set C.

[0052] S4. Analyze the association between the patients and the historical cases according to the classification result set C and the similarity matrix S. By calculating all the historical cases in the classification, obtain the feature concentration D, and then integrate the feature concentration D to form a feature concentration association data set R.

[0053] S5. Sort according to the feature concentration association data set R, and obtain the sorted feature concentration association data set R to establish a tree-like association graph.

[0054] In this embodiment, by establishing a communication channel between the hospital registration system and the electronic medical record system, personal information, diagnosis and examination information, treatment records, and patient complications of the patient are extracted to generate a comprehensive set of feature vectors Fp. By comparing the features with historical cases, the similarity XS between the historical cases and the patient is obtained, and further a case similarity matrix S is generated, providing accurate similarity level classification for the patient and correlation analysis of historical cases, and generating a classification result set C. Then, based on the classification result set C and the similarity matrix S, the degree of feature aggregation of historical cases is calculated to obtain the feature concentration D, and a feature concentration correlation data set R is generated by integrating the feature concentration D. By sorting the feature concentration correlation data set R and establishing a tree-like correlation graph based on this, doctors can quickly identify the most relevant historical cases when managing and retrieving patient files, optimizing the management efficiency of medical files. Especially in solving the deficiencies of low retrieval efficiency, loose data organization, and inability to effectively associate patients with historical cases in existing medical file management, it shows significant advantages. Through the calculation and analysis of the feature vector set Fp and the feature concentration D, doctors can quickly locate historical cases that highly match the patient's features, greatly shortening the retrieval time and improving the accuracy of personalized medical decisions. At the same time, the establishment of the tree-like correlation graph shows the historical cases associated with the patient at different levels in a hierarchical structure, providing an intuitive and orderly file management view for doctors, further enhancing the utilization efficiency and pertinence of historical cases.

[0055] Embodiment 2

[0056] This embodiment is an explanatory note based on Embodiment 1. Please refer to Figure 1 , specifically: S1 includes S11;

[0057] S11. By establishing a communication channel between the hospital registration system and the electronic medical record system, patient information is extracted, including personal information, diagnosis results, examination reports, treatment records, and patient complication records;

[0058] Among them, personal information includes age A, gender G, and medical history MH; diagnosis results include disease diagnosis DIA, blood glucose volatility BGV, and blood pressure BP; treatment records include medication frequency MED and treatment cycle TE; patient complication records include the number of complications CN and the duration of illness D;

[0059] Among them, the medical history MH specifically represents the number of years from the onset of the patient's illness to the present; the disease diagnosis DIA specifically represents the standardized diagnosis value of the disease, which can be classified according to the severity of the disease, such as 0 indicating no disease, 1 indicating mild, 2 indicating moderate, and 3 indicating severe; the duration of illness D specifically represents the duration of the complication;

[0060] Synchronously, after dimensionless processing of personal information, diagnosis results, examination reports, treatment records, and patient complication records, they are integrated to form a feature vector set Fp = {A, G, MH, DIA, BGV, BP, MED, TE, CN, D}.

[0061] S2 includes S21 and S22;

[0062] S21: Extract patient type information based on the feature vector set Fp, including disease diagnosis DIA, blood glucose volatility BGV, and the number of complications CN, to form the patient type T of the current patient. Then, match the patient type T with the case type information in the historical case database. Specifically, the matching is performed by inputting the patient type T as a matching item into the historical case database for retrieval to obtain historical case samples including the patient type T, and forming a matching historical case feature set Fhist;

[0063] Among them, the historical case feature set Fhist is specifically a two-dimensional matrix, including several historical case samples. The specific format of the historical case feature set Fhist is as follows:

[0064]

[0065] In the formula, m represents the total number of historical case samples.

[0066] S22: Calculate the similarity XS between each case sample and the patient type T of the current patient in the feature vector set Fp according to the obtained historical case feature set Fhist, and form a case similarity matrix S = {XS i,j};

[0067] The similarity XS is obtained through the following calculation formula:

[0068]

[0069] In the formula, XS i,j represents the similarity between the i-th patient and the j-th historical case sample in the historical case feature set Fhist, Fp i,k represents the k-th feature parameter of the i-th patient, Fhist j，k represents the k-th feature parameter of the j-th historical case sample in the historical case feature set Fhist, m represents the total number of features, specifically representing the length of the feature vector set Fp, and α k represents the weight coefficient of the k-th feature parameter.

[0070] In this embodiment, by extracting the patient's personal information, diagnosis results, examination reports, treatment records, and patient complication records, and performing dimensionless processing on all the data, they are integrated into a feature vector set Fp. On this basis, the patient's disease type T is extracted and generated, and by matching the disease type with the historical case feature set Fhist in the historical case database, accurate and efficient historical case retrieval is achieved. By calculating the similarity XS between each historical case sample and the patient's disease type, a case similarity matrix S is generated, optimizing the formulation process of personalized treatment plans. In addition, more accurate historical case stratification can be achieved, significantly improving the management efficiency of medical records and promoting the precise development of personalized diagnosis and treatment.

[0071] Embodiment 3

[0072] This embodiment is an explanatory description carried out in Embodiment 1. Please refer to Figure 1 , specifically: S3 includes S31 and S32;

[0073] S31. Classify the patients according to the obtained case similarity matrix S. Specifically, by matching the similarity XS between patient i and historical case sample j in the case similarity matrix S with the upper threshold SZ and lower threshold XZ of the preset similarity evaluation, the association relationship between patient i and historical case sample j is obtained;

[0074] The association relationship is specifically obtained through the following matching method:

[0075] When the similarity XS i,j ≥ upper threshold SZ, the association relationship between patient i and historical case sample j is obtained as a first-level association relationship, and the association level GJ between patient i and historical case sample j is marked as 1;

[0076] When the lower threshold XZ ≤ similarity XS i,j < upper threshold SZ, the association relationship between patient i and historical case sample j is obtained as a second-level association relationship, and the association level GJ between patient i and historical case sample j is marked as 2;

[0077] When the similarity XS i,j < lower threshold XZ, the association relationship between patient i and historical case sample j is obtained as a third-level association relationship, and the association level GJ between patient i and historical case sample j is marked as 3.

[0078] S32. Exclude the historical case samples in the historical case feature set Fhist with an association level GJ = 3, retain the historical case samples with an association level GJ = 1 and an association level GJ = 2, then sort the historical case feature set Fhist according to the association level GJ to obtain the historical case feature set Fhist in list form, and use the last-occurring historical case sample with an association level GJ = 1 in the historical case feature set Fhist to cut the historical case feature set Fhist into the historical case feature set Fhist1 with an association level GJ = 1 and the historical case feature set Fhist2 with an association level GJ = 2, forming the classification result set C = {Fhist1, Fhist2}.

[0079] In this embodiment, by performing precise similarity level classification on the case similarity matrix S, the association relationship between the patient and the historical case samples can be divided into different association levels, and each historical case sample can be classified according to the upper threshold SZ and the lower threshold XZ of the similarity evaluation. Through a clear grading strategy, the system classifies highly relevant historical cases as a first-level association relationship, less relevant cases as a second-level association relationship, and excludes case samples with weak relevance to the patient, optimizing the screening process of historical cases. Subsequently, the system further sorts the classified historical cases according to the association level to generate the classification result set C, and divides the historical case set based on the association level to form a clear hierarchical structure, effectively solving the problem in traditional medical record management that the association strength between cases and patients cannot be effectively distinguished, resulting in overly redundant data and low retrieval efficiency. Through the similarity level classification and the setting of the association level GJ, the system can accurately exclude unimportant historical cases, ensuring that only the cases most valuable for patient diagnosis and treatment are retained, thus significantly improving the utilization rate and accuracy of the data. At the same time, the hierarchical sorting and cutting mechanism further simplifies the process of consulting and managing historical cases, enabling doctors to quickly locate the most relevant reference files among the numerous historical cases, optimizing the support for medical decision-making.

[0080] Example 4

[0081] This embodiment is an explanatory description based on Embodiment 1. Please refer to Figure 1 , specifically: Among them, S4. Analyze the association between the patient and the historical cases according to the classification result set C and the similarity matrix S. By calculating all the historical cases in the classification, obtain the feature concentration degree D = {D1, D2}. Specifically, by comparing the feature vector set Fp of the current patient with the classification result set C, obtain the feature concentration degree D1 of the first-level associated historical case feature set Fhist1 and the feature concentration degree D2 of the second-level associated historical case feature set Fhist2;

[0082] Among them, the calculated feature concentration D1 is subscript-associated with the first-level associated historical case feature set Fhist1, and the calculated feature concentration D2 is subscript-associated with the first-level associated historical case feature set Fhist2 to form a feature concentration association data set R = {R1, R2}, where R1 represents the first feature concentration association data set, specifically R1 = {D1 ∪ Fhist1}; R2 represents the second feature concentration association data set, specifically R2 = {D2 ∪ Fhist2}.

[0083] The feature concentration D1 of the first-level associated historical case feature set Fhist1 and the feature concentration D2 of the second-level associated historical case feature set Fhist2 are obtained through the following calculation formulas:

[0084]

[0085] In the formula, n1 and n2 respectively represent the total number of historical case samples in the first-level associated historical case feature set Fhist1 and the second-level associated historical case feature set Fhist2, j represents the jth historical case sample, m represents the total number of features, Fhist1 j,w and Fhist2 j,w respectively represent the wth feature value of the jth historical case sample in the first-level associated historical case feature set Fhist1 and the second-level associated historical case feature set Fhist2, and Fp w represents the wth feature parameter in the feature vector set Fp.

[0086] Among them, S5. Sort according to the feature concentration association data set R, including using bubble sort to sort based on the feature concentrations D1 and D2 in the feature concentration association data set R, and obtain the sorted feature concentration association data set R. Establish a tree-like association graph, including establishing associated nodes of the root node according to the sorted feature concentration association data set R, where the root node is the feature vector set Fp of the patient, and the associated nodes include the first associated node and the second associated node;

[0087] The first associated node is specifically the first feature concentration association data set R1 in the feature concentration association data set R; the second associated node is specifically the second feature concentration association data set R2 in the feature concentration association data set R;

[0088] The first associated node represents the historical case samples in the sorted first-level associated historical case set Fhist1. According to the sorting result of the bubble-sorted feature concentration D1 j of the feature concentration D1 jHistorical cases with lower severity levels are placed higher in the tree, indicating a higher similarity to the patient's characteristics; each first associated node contains two key pieces of information: the feature concentration D1 j and the corresponding historical case feature vector Fhist1 j ;

[0089] The second associated node represents the historical case samples in the sorted set of second-level associated historical cases Fhist2. According to the sorting result of the feature concentration D2 j after bubble sorting, samples with a smaller feature concentration D2 j are closer to the first-level associated node, and samples with a larger feature concentration D2 j are placed farther away; each second associated node also contains the feature concentration D2 j and the corresponding historical case feature vector Fhist2 j ;

[0090] In this embodiment, by analyzing the classification result set C and the similarity matrix S, the feature concentration D of historical cases is further calculated, and the feature concentration is associated with the corresponding historical case feature set to generate the feature concentration association data set R. On this basis, after sorting the feature concentration by bubble sorting, the system constructs a tree-like association graph based on the patient feature vector set Fp, where the sorted historical case samples are presented in a hierarchical manner, with the first-level associated historical case samples close to the patient node, and the second-level associated historical case samples arranged in sequence according to the similarity. This hierarchical structure can intuitively reflect the similarity degree between different historical cases and the patient's characteristics, and shows significant advantages in solving the problem in traditional medical record management that the correlation between historical cases and patients cannot be effectively visualized, resulting in complex information and difficulty in quickly distinguishing highly associated cases. By constructing the tree-like association graph, the system can help doctors intuitively and quickly view the historical cases highly related to the patient's characteristics according to the sorting result of the feature concentration D, enabling doctors to not only see the quantitative results of the similarity, but also directly obtain the specific case characteristics, significantly improving the accuracy and efficiency of personalized medical decision-making.

[0091] Embodiment 5 <{

[0092] A medical record management system, please refer to Figure 2 , specifically: including a vector acquisition module, a similarity analysis module, a classification association module, a concentration evaluation module, and a visualization module;

[0093] The vector acquisition unit extracts the patient's personal information, diagnosis and examination information, treatment records, and patient progress and complications by establishing a communication channel between the hospital registration system and the electronic case system to form the feature vector set Fp;

[0094] The similarity analysis module extracts the patient type based on the feature vector set Fp, and then compares it with the case features of the same patient type in the historical cases to obtain the similarity XS between the historical cases and the feature vector set Fp, and forms a case similarity matrix S;

[0095] The classification association module classifies the patients according to the obtained case similarity matrix S, and establishes the association relationship between the patients and the historical cases in the case similarity matrix S, and generates a classification result set C;

[0096] The concentration evaluation module analyzes the association of the patients with the historical cases according to the classification result set C and the similarity matrix S. By calculating all the historical cases in the classification, the feature concentration D is obtained, and then the feature concentration D is integrated to form a feature concentration association data set R;

[0097] The visualization module sorts according to the feature concentration association data set R, and obtains the sorted feature concentration association data set R to establish a tree-like association graph.

[0098] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A medical record management method, characterized in that: It includes the following steps: S1. By establishing a communication channel between the hospital registration system and the electronic medical record system, extract patient information, including personal information, diagnosis results, examination reports, treatment records, and patient complication records, to form a feature vector set Fp; S2. Based on the feature vector set Fp, extract the patient type, and then compare it with the case features under the same patient type in the historical cases to obtain the similarity XS between the historical cases and the feature vector set Fp, and form a case similarity matrix S; S3. Classify the patients according to the obtained case similarity matrix S, and establish the association relationship between the patients and the historical cases in the case similarity matrix S to generate a classification result set C; S4. Analyze the association relationship between the patients and the historical cases according to the classification result set C and the similarity matrix S. By calculating all the historical cases in the classification, obtain the feature concentration D, and then integrate the feature concentration D, to form a feature concentration association data set R; Among them, S4. Analyze the association relationship between the patients and the historical cases according to the classification result set C and the similarity matrix S. By calculating all the historical cases in the classification, obtain the feature concentration D = {D1, D2}. Specifically, by comparing the feature vector set Fp of the current patient with the classification result set C, obtain the feature concentration D1 of the first-level associated historical case feature set Fhist1 and the feature concentration D2 of the second-level associated historical case feature set Fhist2; Among them, associate the calculated feature concentration D1 with the first-level associated historical case feature set Fhist1 by subscript, and associate the calculated feature concentration D2 with the first-level associated historical case feature set Fhist2 by subscript to form a feature concentration association data set R = {R1, R2}, where R1 represents the first feature concentration association data set, specifically R1 = {D1 ∪ Fhist1}; R2 represents the second feature concentration association data set, specifically R2 = {D2 ∪ Fhist2}; The feature concentration D1 of the first-level associated historical case feature set Fhist1 and the feature concentration D2 of the second-level associated historical case feature set Fhist2 are obtained through the following calculation formulas: Wherein, n1 and n2 respectively represent the total number of historical case samples in the first-level associated historical case feature set Fhist1 and the second-level associated historical case feature set Fhist2, j represents the j-th historical case sample, m represents the total number of features, Fhist1 j,w and Fhist2 j,w respectively represent the w-th feature value of the j-th historical case sample in the first-level associated historical case feature set Fhist1 and the second-level associated historical case feature set Fhist2, Fp w represents the w-th feature parameter in the feature vector set Fp; S5. Sort according to the feature concentration association data set R, and obtain the sorted feature concentration association data set R to establish a tree-like association graph.

2. The medical record management method according to claim 1, characterized in that: The S1 includes S11; S11. By establishing a communication channel between the hospital registration system and the electronic medical record system, extract patient information, including personal information, diagnosis results, examination reports, treatment records, and patient complication records; Among them, the personal information includes age A, gender G, and medical history MH; the diagnosis results include disease diagnosis DIA, blood glucose volatility BGV, and blood pressure BP; the treatment records include medication frequency MED and treatment cycle TE; the patient complication records include the number of complications CN and the duration of illness D; Among them, the medical history MH specifically represents the number of years from the onset of the patient's illness to the present; the disease diagnosis DIA specifically represents the standardized diagnosis value of the disease; the duration of illness D specifically represents the duration of the complication; Synchronously, after dimensionless processing of personal information, diagnosis results, examination reports, treatment records, and patient complication records, they are integrated to form a feature vector set Fp = {A, G, MH, DIA, BGV, BP, MED, TE, CN, D}.

3. A medical record management method according to claim 1, characterized in that: The S2 includes S21 and S22; S21. Extract patient type information based on the feature vector set Fp, including disease diagnosis DIA, blood glucose volatility BGV, and the number of complications CN, to form the patient type T of the current patient. Then, match the patient type T with the case type information in the historical case database. Specifically, the patient type T is used as a matching item to retrieve in the historical case database to obtain historical case samples including the patient type T, and a matching historical case feature collection Fhist is formed; Among them, the historical case feature collection Fhist is specifically a two-dimensional matrix, including several historical case samples. The specific format of the historical case feature collection Fhist is as follows: In the formula, m represents the total number of historical case samples.

4. A medical record management method according to claim 3, characterized in that: S22. Calculate the similarity XS between each case sample and the disease type T of the current patient in the feature vector set Fp according to the obtained historical case feature set Fhist, and form a case similarity matrix S = {XS i,j} with the current disease type T and several historical case samples in the historical case feature set Fhist; The similarity XS is obtained through the following calculation formula: where XS i,j represents the similarity between the i-th patient and the j-th historical case sample in the historical case feature set Fhist, and Fp i,k represents the k-th feature parameter of the i-th patient, and Fhist j，k represents the k-th feature parameter of the j-th historical case sample in the historical case feature set Fhist, m represents the total number of features, specifically representing the length of the feature vector set Fp, and α k represents the weight coefficient of the k-th feature parameter.

5. A medical record management method according to claim 4, characterized in that: The S3 includes S31 and S32; S31. Classify the patients according to the obtained case similarity matrix S. Specifically, match the similarity XS between patient i and historical case sample j in the case similarity matrix S with the upper threshold SZ and lower threshold XZ of the preset similarity evaluation to obtain the association relationship between patient i and historical case sample j; The association relationship is specifically obtained through the following matching method: When the similarity XS i,j ≥ the upper limit threshold SZ, obtain the association relationship between patient i and historical case sample j as a first-level association relationship, and mark the association level GJ between patient i and historical case sample j as 1; When the lower threshold XZ ≤ similarity XS i,j <upper threshold SZ, obtain the association relationship between patient i and historical case sample j as a secondary association relationship, and mark the association level GJ = 2 between patient i and historical case sample j; When the similarity XS i,j is less than the lower threshold XZ, obtain the association relationship between patient i and historical case sample j as a third-level association relationship, and mark the association level GJ between patient i and historical case sample j as 3.

6. The medical record management method according to claim 5, wherein: S32. Eliminate the historical case samples with an association level GJ = 3 in the historical case feature collection Fhist, retain the historical case samples with an association level GJ = 1 and an association level GJ = 2, then sort the historical case feature collection Fhist according to the association level GJ to obtain the historical case feature collection Fhist in list form, and cut the historical case feature collection Fhist with the last historical case sample with an association level GJ = 1 in the historical case feature collection Fhist into a historical case feature collection Fhist1 with an association level GJ = 1 and a historical case feature collection Fhist2 with an association level GJ = 2, forming a classification result set C = {Fhist1, Fhist2}.

7. A medical record management method according to claim 1, characterized in that: Among them, S5. Sort according to the feature concentration association data set R, including using bubble sort to sort based on the feature concentrations D1 and D2 in the feature concentration association data set R, and obtain the sorted feature concentration association data set R to establish a tree-like association graph, including establishing associated nodes of the root node according to the sorted feature concentration association data set R, where the root node is the feature vector set Fp of the patient, and the associated nodes include the first associated node and the second associated node; The first associated node is specifically the first feature concentration association data set R1 in the feature concentration association data set R; the second associated node is specifically the second feature concentration association data set R2 in the feature concentration association data set R.

8. A medical record management system, which is applied to the medical record management method described in any one of claims 1 to 7, and is characterized in that: It includes a vector acquisition module, a similarity analysis module, a classification association module, a concentration evaluation module, and a visualization module; The vector acquisition unit extracts patients' personal information, diagnosis and examination information, treatment records, and patient progress and complications by establishing a communication channel between the hospital registration system and the electronic medical record system, to form a feature vector set Fp; The similarity analysis module extracts the patient types based on the feature vector set Fp, and then compares them with the case features under the same patient types in the historical cases to obtain the similarity XS between the historical cases and the feature vector set Fp, and forms a case similarity matrix S; The classification association module classifies the patients according to the obtained case similarity matrix S into similarity levels, and establishes the association relationship between the patients and the historical cases in the case similarity matrix S to generate a classification result set C; The concentration evaluation module analyzes the association relationship between the patients and the historical cases according to the classification result set C and the similarity matrix S, calculates all the historical cases in the classification to obtain the feature concentration D, and then integrates the feature concentration D, to form a feature concentration association data set R; The visualization module sorts according to the feature concentration association data set R, and obtains the sorted feature concentration association data set R to establish a tree-like association graph.

Images