Diagnosis and treatment information classification method of emergency treatment information management system

Abstract

The invention relates to the technical field of patient medical record data, in particular to a diagnosis and treatment information classification method of an emergency treatment information management system. According to the method, a reference word vector sequence is obtained, the similarity between a word vector sequence of each patient every day and the reference word vector sequence is analyzed, and sample points in a coordinate system are formed in combination with the structured degree of a diagnosis and treatment process of each patient every day; performing mean shift clustering on the sample points to obtain a shift cluster under each iteration; according to the distribution of relative distances between sample points corresponding to the word vector sequence of each patient every day and central points in the drift clusters at corresponding same positions under different iterations, obtaining the diagnosis and treatment process consistency of the corresponding sample points; according to the time characteristics of the sample points in the different optimization drift clusters, the optimal process time range of each optimization drift cluster is obtained; and managing the diagnosis and treatment process information. According to the invention, by obtaining the optimal process range, the diagnosis and treatment process information management effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of patient medical record data, and in particular to a diagnosis and treatment information classification method for an emergency information management system. Background Art

[0002] The increasing incidence of chronic diseases and limited medical resources have led to frequent outpatient clinic overcrowding and overload. Furthermore, most patients are unclear about the scope of care within different departments, requiring them to spend significant time consulting with medical staff. The imbalance between the growing demand for emergency care and the limited human resources is particularly pronounced. Emergency medical resources and information management are severely lagging, significantly impacting the healthcare system's ability to treat critically ill patients.

[0003] In the existing technology, similar diagnosis and treatment process information between different patients is clustered through mean shift clustering, and the corresponding similar diagnosis and treatment process information is managed; however, as patients visit and undergo examinations, considering the individual differences and diagnostic needs of patients, there are different symptom changes caused by changes in the patient's condition, and the updating and association of diagnosis and treatment information will be more complicated, resulting in deviations between different stages of the diagnosis process and the overall patient process structure, and poor information management effect. Summary of the Invention

[0004] In order to solve the technical problem that individual differences and diagnostic needs of patients may lead to deviations between different stages of the diagnostic process and the overall patient process structure, resulting in poor information management effects, the purpose of the present invention is to provide a method for classifying diagnosis and treatment information in an emergency information management system. The technical solutions adopted are as follows: The present invention proposes a method for classifying diagnosis and treatment information of an emergency information management system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: Obtain each patient's daily diagnosis and treatment process information and obtain the corresponding word vector sequence; Based on the correlation between the corresponding word vector sequences of different patients, a benchmark word vector sequence is obtained to obtain the degree of structured diagnosis and treatment process for each patient on a daily basis; based on the similarity between each patient's daily word vector sequence and the benchmark word vector sequence and the degree of structured diagnosis and treatment process, sample points in the coordinate system are constructed; Clustering the sample points based on their position distribution in the coordinate system to obtain multiple drift clustering clusters; obtaining the consistency of the diagnosis and treatment process for the corresponding sample points based on the distribution of the relative distances between the sample points and the center points in different drift clustering clusters corresponding to each patient's daily word vector sequence; adjusting the drift clustering clusters based on the consistency of the diagnosis and treatment process to obtain optimized drift clustering clusters; obtaining the optimal process time range for each optimized drift clustering cluster based on the time characteristics of the sample points in different optimized drift clustering clusters; The diagnosis and treatment process information is managed according to the optimal process time range.

[0005] Furthermore, the method for obtaining the benchmark word vector sequence includes: Obtain the DTW distance of the corresponding word vector sequences between different patients as the difference score between the corresponding word vector sequences; Get the mean difference scores between all word vector sequences, select the word vector sequence with a difference score greater than the mean difference score, and use the corresponding word vector sequence as the target word vector sequence; Normalize the target word vector sequence and obtain the mean of all target word vector sequences after normalization as the benchmark word vector sequence.

[0006] Furthermore, the method for obtaining the degree of structure of the diagnosis and treatment process includes: For any patient, the degree of structured diagnosis and treatment process for each patient is obtained based on the DTW distance between the daily word vector sequence and the benchmark word vector sequence, as well as the difference in sequence length. The DTW distance is negatively correlated with the degree of structured diagnosis and treatment process, while the difference in sequence length is positively correlated with the degree of structured diagnosis and treatment process.

[0007] Furthermore, the method for obtaining the drift cluster center includes: Based on the position distribution of sample points in the coordinate system, mean shift clustering is performed on all sample points to obtain multiple drift clustering clusters.

[0008] Furthermore, the method for obtaining consistency of the diagnosis and treatment process includes: Obtain the relative distance between the sample point corresponding to the word vector sequence of each patient every day and the center point in each drift cluster as the relative distance of the center point in each drift cluster relative to the sample point; Calculate the cumulative sum of the differences in the corresponding relative distances between the center points of all adjacent drift clusters as the first difference coefficient; According to the relative distance of the center point of the final drift cluster relative to the sample point and the first difference coefficient, the consistency of the diagnosis and treatment process of the corresponding sample point is obtained. The first difference coefficient and the relative distance of the center point of the final drift cluster relative to the sample point are both negatively correlated with the consistency of the diagnosis and treatment process.

[0009] Furthermore, the method for obtaining the optimized drift cluster is as follows: For any drift cluster, sample points corresponding to the diagnosis and treatment process consistency greater than the preset consistency threshold are screened out to form the corresponding optimized drift cluster.

[0010] Furthermore, the method for obtaining the optimal process time range includes: Obtain the time range corresponding to the sample points in each optimized drift clustering cluster; for optimized drift clustering clusters with overlapping time ranges, calculate the ratio of the overlapping time range to the time range of each optimized drift clustering cluster as a reference ratio; The difference between the time overlap range and the time range corresponding to the optimized drift clustering cluster with the smallest reference ratio is obtained as the new time range of the corresponding optimized drift clustering cluster, and the time range of all optimized drift clustering clusters is obtained as the optimal process-based time range.

[0011] Furthermore, the method for acquiring sample points in the coordinate system includes: The similarity between each patient's daily word vector sequence and the benchmark word vector sequence is used as the horizontal axis of the coordinate system, and the degree of structuring of the diagnosis and treatment process of each patient's daily word vector sequence is used as the vertical axis of the coordinate system to form the corresponding sample points.

[0012] Furthermore, the method for obtaining the word vector sequence includes: Perform word vectorization on each patient’s daily diagnosis and treatment process information to obtain a daily word vector sequence.

[0013] Furthermore, the similarity is cosine similarity.

[0014] The present invention has the following beneficial effects: The present invention obtains a benchmark word vector sequence based on the correlation between corresponding word vector sequences of different patients, obtains the degree of structured diagnosis and treatment process of each patient every day, and more accurately reflects the degree of similarity of each patient's corresponding diagnosis and treatment information to the overall process characteristics by obtaining the benchmark word vector sequence of the overall process change level; according to the similarity between each patient's daily word vector sequence and the benchmark word vector sequence and the degree of structured diagnosis and treatment process, sample points in the coordinate system are constructed to more intuitively show the similarities and differences between patients; based on the position distribution of sample points in the coordinate system, mean shift clustering is performed on the sample points to obtain the drift clustering cluster under each iteration, which can divide patients with similar characteristics into the same cluster, thereby revealing the potential differences between patients. In connection; according to the distribution of relative distances between the sample points corresponding to the word vector sequence of each patient every day and the center points in the drift clustering clusters at the same position under different iterations, the consistency of the diagnosis and treatment process of the corresponding sample points is obtained, and the degree of association between the sample points and each drift clustering cluster can be reflected by analyzing the relative distance distribution, and the consistency and stability of the diagnosis and treatment process are evaluated; according to the consistency of the diagnosis and treatment process, the drift clustering cluster under the final iteration is adjusted to obtain an optimized drift clustering cluster, ensuring that the patients in the clustering cluster have higher similarity and consistency, and avoiding the influence of individual differences among patients; according to the time characteristics of the sample points in different optimized drift clustering clusters, the optimal process time range of each optimized drift clustering cluster is obtained; and the diagnosis and treatment process information is managed. The present invention improves the effect of diagnosis and treatment process information management by obtaining the optimal process range. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A flowchart of an implementation method of a diagnosis and treatment information classification method of an emergency information management system provided by one embodiment of the present invention; Figure 2 A flowchart of a method for obtaining a benchmark word vector sequence provided by one embodiment of the present invention; Figure 3 A flowchart of a method for obtaining consistency in diagnosis and treatment processes provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation method, structure, features and effects of the diagnosis and treatment information classification method of the emergency information management system proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics of one or more embodiments may be combined in any suitable form.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The following describes in detail a specific scheme of a diagnosis and treatment information classification method for an emergency information management system provided by the present invention with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a flowchart of an implementation method of a diagnosis and treatment information classification method of an emergency information management system provided by an embodiment of the present invention, the method specifically includes: Step S1: Obtain the daily diagnosis and treatment process information data of each patient and obtain the corresponding word vector sequence.

[0021] In an embodiment of the present invention, an analysis is performed on emergency patients. Emergency patients are generally divided into many different types. For patients who need hospitalization for treatment, there is a certain degree of hospitalization and surgical process processing. Since the information data of emergency patients involves complex operations, postoperative monitoring and multidisciplinary collaboration, it is usually more detailed and complex than other departments. The integrated management of emergency patient treatment information has the characteristics of close correlation between internal information of treatment, surgery and other stages. Therefore, it is necessary to record and analyze the daily treatment process information. The treatment process information includes patient basic information, medication, diagnosis results, treatment plan and other information data, obtain each patient's daily treatment process information, and obtain the corresponding word vector sequence.

[0022] Medical treatment process information is a type of text data that contains a large number of medical terms and descriptions. By converting the text data in the medical treatment process information into word vectors, it can be converted into a numerical form that the model can process, which helps capture the association between words.

[0023] It should be noted that, in one embodiment of the present invention, the method for obtaining a word vector sequence is to vectorize the daily diagnosis and treatment process information of each patient to obtain a corresponding word vectorized sequence. The specific word vectorization is a technical means well known to those skilled in the art and will not be described in detail here.

[0024] Step S2: Based on the correlation between the corresponding word vector sequences of different patients, a benchmark word vector sequence is obtained, and the process structuring degree of each patient's daily word vector sequence is obtained; based on the similarity and process structuring degree between each patient's daily word vector sequence and the benchmark word vector sequence, sample points in the coordinate system are constructed.

[0025] The diagnosis and treatment process of patients often has a certain structure and regularity, that is, the distribution structure of diagnosis and treatment information over multiple days is similar, but the diagnosis and treatment cycle and content of different patients are not exactly the same. Therefore, by analyzing the correlation between word vector sequences of different patients, similar symptoms that may exist between different patients are shown, and the diagnosis and treatment information of different patients is structured. According to the correlation between word vector sequences of different patients, the benchmark word vector sequence is obtained, and the process structuring degree of each patient's daily word vector sequence is obtained.

[0026] Preferably, in one embodiment of the present invention, the method for obtaining the benchmark word vector sequence can be found in Figure 2 , which shows a flow chart of a method for obtaining a baseline word vector sequence, including: Step S201: Obtain the DTW distances of the corresponding word vector sequences between different patients as the difference scores between the corresponding word vector sequences.

[0027] Since different patients have different conditions and treatment processes, there are differences in the word vector sequences. The DTW algorithm is used to calculate the differences between word vectors, reflecting the similarity between the two patients in vocabulary usage, condition description or treatment process. The larger the DTW distance, the greater the difference, and the more different the medical record information.

[0028] It should be noted that the specific DTW algorithm is a technical means well known to those skilled in the art and is not limited or elaborated herein.

[0029] Step S202: Obtain the mean difference scores between all word vector sequences, select the word vector sequences whose difference scores between the word vector sequences are greater than the mean difference scores, and use the corresponding word vector sequences as the target word vector sequences.

[0030] The difference scores between all word vector sequences are quantified by taking the average value to reflect the average level of change in the overall word vector sequence; the word vector sequences with difference scores greater than the average difference scores are selected to screen out those that are more significant than the average, more valuable, and have significant changes.

[0031] Step S203: Normalize the target word vector sequence and obtain the mean of all target word vector sequences after normalization as the benchmark word vector sequence.

[0032] The target word vector sequence reflects the text content that has changed significantly in the dataset. By calculating the mean of the normalized target word vector sequence, we obtain a benchmark word vector sequence that reflects the overall level of information change and is used to compare the changes in other word vector sequences.

[0033] Preferably, in one embodiment of the present invention, the method for obtaining the degree of structure of the diagnosis and treatment process includes: For any patient, the degree of structured diagnosis and treatment process for each patient is obtained based on the DTW distance between the daily word vector sequence and the benchmark word vector sequence, as well as the difference in sequence length. The difference in DTW distance and sequence length is negatively correlated with the degree of structured diagnosis and treatment process.

[0034] Among them, the negative correlation indicates that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. That is, the larger the DTW distance, the greater the difference in sequence length, the more inconsistent the content and length of the two word vector sequences, and the less structured the diagnosis and treatment process.

[0035] In one embodiment of the present invention, the formula for the degree of structuring of the diagnosis and treatment process is expressed as: ; in, Indicates the Patient No. The degree of structure of the daily diagnosis and treatment process; Indicates the Patient No. word vector sequence of the day; Represents a sequence of baseline word vectors; Indicates the Patient No. The difference between the word vector sequence of the day and the benchmark word vector sequence distance; Indicates the Patient No. The length of the word vector sequence of the day; Indicates the The length of the baseline word vector sequence for each patient; represents an exponential function with a natural constant as its base; Indicates taking the absolute value.

[0036] In the formula of the degree of structuring of diagnosis and treatment process, the exponential function with natural constant as the base is used to convert Perform negative correlation mapping, Indicates calculation of Patient No. The difference between the word vector sequence of the day and the benchmark word vector sequence Distance: The larger the DTW distance, the greater the difference between the corresponding word vector sequence and the benchmark word vector sequence, the more inconsistent the corresponding diagnosis and treatment process is, and the less structured the diagnosis and treatment process is. right Add 0.01 to avoid the formula being 0, which makes the formula meaningless. Indicates calculation of Patient No. The difference in length between the word vector sequence of the day and the benchmark word vector sequence, the greater the difference, the greater the difference in content, the more different the diagnosis and treatment processes, and the less structured the diagnosis and treatment processes.

[0037] The degree of process structuring reflects the standardization and consistency of the patient's diagnosis and treatment process. By comparing the patient's daily word vector sequence with the benchmark word vector sequence, the similarity between each patient's daily diagnosis and treatment process information and the overall information change level can be quantified; by combining word vector similarity and process structuring as dimensions to constitute the coordinate system, it is helpful to analyze the regularity and differences of patient data, and more intuitively understand the distribution of diagnosis and treatment information of different patients, and can comprehensively evaluate the patient's health status, diagnosis and treatment effects, and potential health risks; based on the similarity between each patient's daily word vector sequence and the benchmark word vector sequence and the degree of diagnosis and treatment process structuring, sample points in the coordinate system are constructed.

[0038] Preferably, in one embodiment of the present invention, the method for acquiring sample points in the coordinate system includes: The similarity between each patient's daily word vector sequence and the benchmark word vector sequence is used as the horizontal axis of the coordinate system, and the degree of structuring of the diagnosis and treatment process of each patient's daily word vector sequence is used as the vertical axis of the coordinate system to form the corresponding sample points.

[0039] It should be noted that, in one embodiment of the present invention, the similarity is cosine similarity; the specific cosine similarity is a technical means well known to those skilled in the art and will not be described in detail here.

[0040] Step S3: Perform mean shift clustering on the sample points based on their position distribution in the coordinate system to obtain drift clustering clusters under each iteration; obtain the consistency of the diagnosis and treatment process for the corresponding sample points based on the distribution of the relative distances between the sample points corresponding to each patient's daily word vector sequence and the center points in the drift clustering clusters at the same position under different iterations.

[0041] The location distribution of the sample points exhibits a certain degree of clustering, likely due to the sample points sharing similar characteristics or attributes of the diagnosis and treatment process. Clustering groups these sample points into a single cluster, reflecting the inherent structure of the diagnosis and treatment process information. Mean shift clustering is performed on the sample points based on their location distribution in the coordinate system, yielding drift clusters for each iteration.

[0042] The mean shift clustering algorithm continuously iteratively updates the position of the cluster midpoint, causing the cluster midpoint to move toward areas with higher density, eventually clustering into multiple clusters and identifying clustering patterns in data sets that change over time. The specific mean shift clustering algorithm is a technical means well known to those skilled in the art and will not be described in detail here.

[0043] Because drift clusters under different iterations represent treatment processes with similar characteristics that change over time, analyzing the relative distance distribution can reflect the degree of association between sample points and each drift cluster, and evaluate the consistency and stability of the treatment process. Based on the distribution of the relative distances between the sample points corresponding to each patient's daily word vector sequence and the center points within the drift clusters at the same location under different iterations, the consistency of the treatment process for the corresponding sample points is obtained.

[0044] Preferably, in one embodiment of the present invention, the method for obtaining the consistency of diagnosis and treatment process can be found in Figure 3 , which shows a flow chart of a method for obtaining consistency in diagnosis and treatment processes, including: Step S301: For each sample point corresponding to the word vector sequence of each patient every day, obtain the relative distance between the center point in the drift cluster at the same position and the sample point in each iteration.

[0045] By calculating the relative distance, the similarity of the corresponding diagnosis and treatment process information between the sample point and the center point in the drift cluster at the same position in each iteration is expressed, which is conducive to observing the process changes of the sample point.

[0046] It should be noted that, in one embodiment of the present invention, the relative distance is obtained by calculating the Euclidean distance; in other embodiments of the present invention, the relative distance can also be obtained by using existing distance calculation methods such as the Manhattan distance; the specific Euclidean distance and Manhattan distance are technical means well known to those skilled in the art and will not be described in detail here.

[0047] Step S302: Calculate the cumulative sum of the differences in the relative distances between the center points of the drift clusters at the same position in all adjacent iterations as the first difference coefficient.

[0048] By analyzing the changes in the relative distances between the center points of the drift clusters at the same position and the sample points under different adjacent iterations, the severity of the changes between clusters can be reflected. That is, the cumulative sum of the differences in the corresponding relative distances between the center points of the drift clusters at the same position under all adjacent iterations is calculated. The larger the cumulative sum, the greater the severity of the change, the smaller the possibility of similar processes, and the larger the first difference coefficient.

[0049] Step S303: Based on the relative distance of the center point of the drift clustering cluster at the same position corresponding to the sample point and the first difference coefficient in the final iteration, the consistency of the diagnosis and treatment process corresponding to the sample point is obtained. The first difference coefficient and the relative distance of the center point of the drift clustering cluster at the same position corresponding to the sample point in the final iteration are both negatively correlated with the consistency of the diagnosis and treatment process.

[0050] Among them, the negative correlation indicates that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases, that is, the smaller the first difference coefficient, the smaller the change in the drift cluster center corresponding to the same position in adjacent iterations, and the greater the consistency of the diagnosis and treatment process; the smaller the relative distance, the closer the sample point is to the center point of the drift cluster cluster corresponding to the same position in the final iteration, the greater the structural similarity between patients, and the greater the consistency of the diagnosis and treatment process.

[0051] In one embodiment of the present invention, the formula for obtaining the consistency of diagnosis and treatment process is expressed as: ; in, Indicates the Patient No. The consistency of the diagnosis and treatment process of the word vector sequence corresponding to the sample point of the day; Indicates the Patient No. The word vector sequence of the day corresponds to the sample point and the The relative distance between the center points of the drift clusters at the same position under the final iteration; Indicates the Patient No. The word vector sequence of the day corresponds to the sample point and the The relative distance between the center points of the drift cluster at the same position under the iteration; Indicates the Patient No. The word vector sequence of the day corresponds to the sample point and the The relative distance between the center points of the drift cluster at the same position under the iteration; Indicates the number of drift clustering iterations; Represents the normalization function.

[0052] In the formula for consistency of diagnosis and treatment process, Indicates calculation of Patient No. The difference in relative distance between the sample point corresponding to the word vector sequence of the day and the center point in the drift cluster at the same position in the two adjacent iterations, Indicates the calculation of the center points of the drift clusters at the same position relative to the first Patient No. The cumulative sum of the differences in the relative distances of the word vector sequences corresponding to the sample points of the day is the first difference coefficient. The larger the cumulative sum of the differences, the larger the first difference coefficient, the greater the difference between the relative distances, the greater the distance fluctuation, and the more inconsistent the process. In the final iteration, the center point of the drift cluster at the same position relative to the first Patient No. The greater the relative distance between the word vector sequence of the day and the corresponding sample points, the more inconsistent the diagnosis and treatment information is, and the less consistent the diagnosis and treatment process is.

[0053] Step S4: Adjust the drift clustering clusters under the final iteration according to the consistency of diagnosis and treatment process to obtain the optimized drift clustering clusters; obtain the optimal process time range of each optimized drift clustering cluster according to the time characteristics of the sample points in different optimized drift clustering clusters.

[0054] Due to data noise, outliers, or special circumstances of specific patients, the initial clustering results may include some sample points that are inconsistent with or deviate significantly from the overall clustering trend. Greater consistency in the diagnosis and treatment process ensures that doctors follow unified standards and processes when recording and analyzing patient data, helping to improve the accuracy and readability of medical records, thereby providing patients with more accurate and effective diagnosis and treatment. Therefore, by analyzing the consistency of the diagnosis and treatment process, we can remove points that do not conform to the overall clustering trend, thereby improving the quality and accuracy of the clustering results. Therefore, based on the consistency of the diagnosis and treatment process, we adjust the drift clustering clusters in the final iteration to obtain the optimized drift clustering clusters.

[0055] Preferably, in one embodiment of the present invention, the method for obtaining drift clusters is optimized: For any drift clustering cluster in the final iteration, the sample points corresponding to the diagnosis and treatment process consistency greater than the preset consistency threshold are screened out to form the corresponding optimized drift clustering cluster.

[0056] It should be noted that, in one embodiment of the present invention, the size of the preset consistency threshold is 0.85; in one embodiment of the present invention, the size of the preset consistency threshold can be specifically set according to specific circumstances, and is not limited or elaborated herein.

[0057] The time characteristics of sample points within the optimized drift clustering cluster can capture the dynamic changes and trends of information, reflect the corresponding time changes of patients' consultation time and treatment cycle, and help to divide the time range of different stages of the diagnostic process; therefore, according to the time characteristics of sample points within different optimized drift clustering clusters, the optimal process time range of each optimized drift clustering cluster is obtained.

[0058] Preferably, in one embodiment of the present invention, the method for obtaining the optimal process-based time range includes: Obtain the time range corresponding to the sample points in each optimized drift clustering cluster; for optimized drift clustering clusters with overlapping time ranges, calculate the ratio of the overlapping time range to the time range of each optimized drift clustering cluster as a reference ratio; The difference between the time overlap range and the time range corresponding to the optimized drift clustering cluster with the smallest reference ratio is obtained as the new time range of the corresponding optimized drift clustering cluster, and the time range of all optimized drift clustering clusters is obtained as the optimal process-based time range.

[0059] It should be noted that, in one embodiment of the present invention, the time range refers to the time range between the earliest day and the latest day corresponding to the sample points in the optimized drift clustering cluster. For optimized drift clustering clusters where there is no time range overlap and the reference ratio is larger, the corresponding time range remains unchanged.

[0060] Step S5: Manage the diagnosis and treatment process information according to the optimal process time range.

[0061] Determine the optimal process time range to reflect the changing patterns of different drift clusters in the time dimension, ensure the temporal rationality of different stages of the diagnosis and treatment process, and package and store the corresponding diagnosis and treatment process information according to the optimal process time range, which will help to retrieve and query diagnosis and treatment information and improve the management and utilization efficiency of diagnosis and treatment information.

[0062] In summary, the present invention obtains a benchmark word vector sequence, analyzes the similarity between each patient's daily word vector sequence and the benchmark word vector sequence, and combines the degree of structured diagnosis and treatment process of each patient every day to form sample points in the coordinate system; performs mean shift clustering on the sample points to obtain drift clustering clusters under each iteration; obtains the diagnosis and treatment process consistency of the corresponding sample points based on the distribution of relative distances between the sample points corresponding to each patient's daily word vector sequence and the center points in the drift clustering clusters at the same position under different iterations; obtains optimized drift clustering clusters, and obtains the optimal process time range of each optimized drift clustering cluster based on the time characteristics of the sample points in different optimized drift clustering clusters; and manages the diagnosis and treatment process information. The present invention improves the effect of diagnosis and treatment process information management by obtaining the optimal process range.

[0063] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0064] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for classifying medical information in an emergency information management system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the following steps are implemented: Obtain each patient's daily diagnosis and treatment process information and obtain the corresponding word vector sequence; Based on the correlation between the corresponding word vector sequences of different patients, a baseline word vector sequence is obtained to obtain the degree of structured diagnosis and treatment process for each patient on a daily basis; Based on the similarity between each patient's daily word vector sequence and the benchmark word vector sequence and the degree of structured diagnosis and treatment process, sample points in the coordinate system are formed; Based on the position distribution of sample points in the coordinate system, mean shift clustering is performed on the sample points to obtain drift clustering clusters under each iteration; based on the distribution of the relative distances between the sample points corresponding to each patient's daily word vector sequence and the center points in the drift clustering clusters at the same position under different iterations, the consistency of the diagnosis and treatment process of the corresponding sample points is obtained; Adjusting the drift clustering clusters under the final iteration according to the consistency of the diagnosis and treatment process to obtain the optimized drift clustering clusters; obtaining the optimal process time range of each optimized drift clustering cluster according to the time characteristics of the sample points in different optimized drift clustering clusters; The diagnosis and treatment process information is managed according to the optimal process time range.

2. The method for classifying medical information of an emergency information management system according to claim 1, characterized in that: The method for obtaining the benchmark word vector sequence includes: Obtain the DTW distance of the corresponding word vector sequences between different patients as the difference score between the corresponding word vector sequences; Get the mean difference scores between all word vector sequences, select the word vector sequence with a difference score greater than the mean difference score, and use the corresponding word vector sequence as the target word vector sequence; Normalize the target word vector sequence and obtain the mean of all target word vector sequences after normalization as the benchmark word vector sequence.

3. The method for classifying medical information of an emergency information management system according to claim 1, characterized in that: Methods for obtaining the structured degree of the diagnosis and treatment process include: For any patient, the degree of structured diagnosis and treatment process for each patient is obtained based on the DTW distance between the daily word vector sequence and the benchmark word vector sequence, as well as the difference in sequence length. The DTW distance is negatively correlated with the degree of structured diagnosis and treatment process, while the difference in sequence length is positively correlated with the degree of structured diagnosis and treatment process.

4. The method for classifying medical information of an emergency information management system according to claim 1, characterized in that: The method for obtaining the consistency of the diagnosis and treatment process includes: For each patient's daily word vector sequence corresponding to the sample point, obtain the relative distance between the center point of the drift cluster and the sample point at the same position in each iteration; Calculate the cumulative sum of the differences in the corresponding relative distances between the center points of the drift clusters at the same position in all adjacent iterations as the first difference coefficient; Based on the relative distance and the first difference coefficient of the center point of the drift cluster at the same position in the final iteration relative to the sample point, the consistency of the diagnosis and treatment process of the corresponding sample point was obtained. The first difference coefficient and the relative distance of the center point of the drift cluster at the same position in the final iteration relative to the sample point were negatively correlated with the consistency of the diagnosis and treatment process.

5. The method for classifying medical information in an emergency information management system according to claim 1, characterized in that: The method for obtaining the optimized drift clustering cluster is as follows: For any drift clustering cluster in the final iteration, the sample points corresponding to the diagnosis and treatment process consistency greater than the preset consistency threshold are screened out to form the corresponding optimized drift clustering cluster.

6. The method for classifying medical information in an emergency information management system according to claim 1, characterized in that: The method for obtaining the optimal process time range includes: Obtain the time range corresponding to the sample points in each optimized drift clustering cluster; for optimized drift clustering clusters with overlapping time ranges, calculate the ratio of the overlapping time range to the time range of each optimized drift clustering cluster as a reference ratio; The difference between the time overlap range and the time range corresponding to the optimized drift clustering cluster with the smallest reference ratio is obtained as the new time range of the corresponding optimized drift clustering cluster, and the time range of all optimized drift clustering clusters is obtained as the optimal process-based time range.

7. The method for classifying medical information in an emergency information management system according to claim 1, characterized in that: The method for obtaining sample points in the coordinate system includes: The similarity between each patient's daily word vector sequence and the benchmark word vector sequence is used as the horizontal axis of the coordinate system, and the degree of structuring of the diagnosis and treatment process of each patient's daily word vector sequence is used as the vertical axis of the coordinate system to form the corresponding sample points.

8. The method for classifying medical information in an emergency information management system according to claim 1, characterized in that: The method for obtaining the word vector sequence includes: Perform word vectorization on each patient’s daily diagnosis and treatment process information to obtain a daily word vector sequence.

9. The method for classifying medical information in an emergency information management system according to claim 1, characterized in that: The similarity is cosine similarity.

10. The method for classifying medical information in an emergency information management system according to claim 5, characterized in that: The preset consistency threshold is 0.85.

Images