Intelligent management method for scrotum eczema inquiry record data

Abstract

The invention provides an intelligent management method for scrotal eczema inquiry record data, and relates to the technical field of digital medical treatment. Automatic word segmentation and vectorization are performed through a natural language processing technology, so that the processing efficiency and accuracy of text information are remarkably improved; an isolated forest algorithm is adopted to accurately recognize abnormal feature vectors and calculate abnormal scores, and accurate recognition of key abnormal features is ensured; a K-means clustering algorithm is applied to optimize data classification, and a better clustering result is provided; classification indexes are generated through weighted calculation and standardized processing, and scientificity and accuracy of scrotum eczema diagnosis are improved; the dynamic adjustment of the screening threshold of the importance score is beneficial for improving the diagnosis precision, supporting personalized treatment and optimizing medical resource allocation, so that the efficiency of scrotum eczema data management and the diagnosis accuracy are greatly improved, and the method has important significance for medical informatization development.

Description

Technical Field

[0001] The invention relates to the technical field of digital medicine, and in particular to an intelligent management method for scrotal eczema consultation record data. Background Art

[0002] With the development of medical information technology, more and more medical institutions have begun to adopt electronic medical record systems to record patients' medical history, symptom descriptions and examination results. However, traditional electronic medical record systems mainly rely on manual input and query, which is inefficient and prone to errors. In addition, the consultation record data for scrotal eczema is often multidimensional, involving multiple aspects such as the patient's medical history, symptom descriptions and examination results. These data usually exist in text form. How to effectively process and manage these multidimensional text information has become an important challenge in the current field of medical information technology;

[0003] In the prior art, the publication number is CN113592345A, and the name is a medical triage method, system, computer equipment and storage medium based on a clustering model, including: determining the patient's medical needs according to the received category selection command; analyzing the received patient's main complaint to obtain the patient's triage result; obtaining the recommended doctor group according to the triage result and the doctor's expertise index for different types of diseases; real-time monitoring and analysis of message events in the doctor's consultation to obtain the busyness and fatigue of the recommended doctor group; according to the medical needs, configuring the weight factor and weight coefficient of the recommended weight, calculating the recommended weight of the recommended doctor group, and recommending the doctor with the highest weight to the patient. It can accurately equip patients with doctors, reduce patients' waiting time, improve the quality of consultation, and also improve the workload of doctors, improve doctors' work efficiency, and achieve a win-win effect of improving the quality of Internet medical services and improving patient satisfaction.

[0004] The existing technology has the following deficiencies when processing scrotal eczema consultation record data. First, the manual input and query method is not only inefficient, but also prone to missing key information, affecting the accuracy of the diagnosis. Secondly, traditional text processing methods are difficult to extract useful feature information when faced with a large amount of unstructured data, making it impossible to perform effective data analysis and classification. In addition, the existing methods lack intelligent means to identify and process abnormal data, making it difficult to quickly and accurately identify abnormal features that have an important impact on diagnosis;

[0005] The above information disclosed in the above Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a method for intelligent management of scrotal eczema consultation record data to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for intelligent management of scrotal eczema consultation record data, the specific steps comprising:

[0009] Step S1: Collect multi-dimensional text information from the patient's medical consultation on the mobile phone, the multi-dimensional text information includes medical history, symptom description and examination results;

[0010] Step S2: Use natural language processing technology to perform word segmentation on the multi-dimensional text information, apply a language model to convert the word segmentation into a numerical vector, and combine these numerical vectors to generate a multi-dimensional feature vector;

[0011] Step S3: Identify abnormal feature vectors in the multidimensional feature vector by using the isolation forest algorithm to form an abnormal feature vector set, and calculate the abnormal score SP of each abnormal feature vector;

[0012] Step S4: Obtain an abnormal feature vector set, calculate the importance score of each abnormal feature vector, and set a screening threshold for the importance score. Combine the abnormal score SP and the screening threshold to select the key abnormal feature vector with the most significant contribution from the scored abnormal feature vector set;

[0013] Step S5: using the K-means clustering algorithm to classify the patient data using the key abnormal feature vectors selected in step S4 to obtain optimized clustering results; storing the clustering results and the key abnormal feature vectors in a database, and providing a system interface to support data query and analysis;

[0014] Step S6: Obtain the clustering results obtained by the K-means clustering algorithm, use the importance score calculated in step S4 to perform weighted calculation on the key abnormal feature vector of each cluster in the clustering results, and obtain a weighted comprehensive value of each cluster;

[0015] Step S7: Collect clinical data scores of key abnormal feature vectors, combine the clinical data scores with the weighted comprehensive value of each cluster for analysis, and generate a classification index for scrotal eczema. At the same time, the classification index can dynamically adjust the screening threshold of the importance score.

[0016] Further explanation: the classification index of scrotal eczema is defined as SCEI, and the calculation formula is as follows:

[0017]

[0018] Among them, w j′is the weight coefficient of the j′th cluster; is an exponential function that adjusts the influence of the normalized weighted comprehensive value, η1 is a positive number, 0.01≤η1≤0.11; β′ is the weight coefficient of clinical data, and the value range of β′ is between 0 and 1; CD(X u ) is the key abnormal feature vector X u Clinical data scores;

[0019] Define CD(X u ) is calculated as:

[0020]

[0021] Among them, D p is the raw score of the pth clinical data item; W p is the weight of the pth clinical data item; L is the total number of clinical data items; assign a weight W to each data item p , the sum of weights is 1.

[0022] Further explanation: Dynamically adjust the screening threshold of importance score, including:

[0023] Set the valid value range of SCEI to (0,1);

[0024] When the SCEI value is closer to 0, it means that the patient's clinical and classification characteristics are less obvious;

[0025] When the SCEI value is closer to 1, it means that the patient's clinical characteristics and classification characteristics are more obvious;

[0026] The adjustment formula for the screening threshold T is:

[0027] T′=T×(1-SCEI)

[0028] Where T is the pre-set importance score screening threshold; T′ is the dynamically adjusted screening threshold;

[0029] Define the following interaction rules:

[0030] When the SCEI value is in the interval (0,0.25]:

[0031] T′=T×(1-0.25)

[0032] Among all feature vectors, select the top 25% as effective features;

[0033] When the SCEI value is in the interval (0.25,0.5]:

[0034] T′=T×(1-0.5)

[0035] Among all the key anomaly feature vectors, the top 55% are selected as effective features;

[0036] When SCEI is in the interval (0.5,0.75]:

[0037] T′=T×(1-0.75)

[0038] All key abnormal feature vectors are taken as effective features;

[0039] When the SCEI value is in the interval (0.75,1):

[0040] T′=T×(1-0.85).

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows: automatic word segmentation and vectorization through natural language processing technology significantly improve the processing efficiency and accuracy of text information; the isolation forest algorithm is used to accurately identify abnormal feature vectors and calculate abnormal scores to ensure accurate identification of key abnormal features; the K-means clustering algorithm is used to optimize data classification and provide better clustering results; the classification index is generated through weighted calculation and standardized processing, which improves the scientificity and accuracy of scrotal eczema diagnosis; dynamic adjustment of the screening threshold of the importance score helps to improve diagnostic accuracy, support personalized treatment, and optimize the allocation of medical resources, thereby greatly improving the efficiency of scrotal eczema data management and diagnostic accuracy, which is of great significance to the development of medical informatization. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0044] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0045] Embodiment 1:

[0046] See also Figure 1 , the present invention provides a technical solution:

[0047] An intelligent management method for scrotal eczema consultation record data;

[0048] Step S1: Collect multi-dimensional text information from the patient's medical consultation on the mobile phone, the multi-dimensional text information includes medical history, symptom description and examination results;

[0049] Step S2: Clean the collected multi-dimensional text information data, remove invalid and duplicate records, use natural language processing technology to perform word segmentation on the multi-dimensional text information, apply the language model to convert the word segmentation into numerical vectors, and combine these numerical vectors to generate a multi-dimensional feature vector;

[0050] Step S3: Identify abnormal feature vectors in the multidimensional feature vector by using the isolation forest algorithm to form an abnormal feature vector set, and calculate the abnormal score SP of each abnormal feature vector;

[0051] Step S4: Obtain an abnormal feature vector set, calculate the importance score of each abnormal feature vector, and set a screening threshold for the importance score. Combine the abnormal score SP and the screening threshold to select the key abnormal feature vector with the most significant contribution from the scored abnormal feature vector set;

[0052] Step S5: using the K-means clustering algorithm to classify the patient data using the key abnormal feature vectors selected in step S4 to obtain optimized clustering results; storing the clustering results and the key abnormal feature vectors in a database, and providing a system interface to support data query and analysis;

[0053] Step S6: Obtain the clustering results obtained by the K-means clustering algorithm, use the importance score calculated in step S4 to perform weighted calculation on the key abnormal feature vector of each cluster in the clustering results, and obtain a weighted comprehensive value of each cluster;

[0054] Step S7: Collect clinical data scores of key abnormal feature vectors, combine the clinical data scores with the weighted comprehensive value of each cluster for analysis, and generate a classification index for scrotal eczema. At the same time, the classification index can dynamically adjust the screening threshold of the importance score.

[0055] Further explanation: The multi-dimensional text information collected from the patient's medical consultation on the mobile phone includes:

[0056] Data collection process:

[0057] 1.1) User login or registration: User authentication through mobile phone number or other identity authentication methods;

[0058] 1.2) Information filling: Users fill in information such as medical history, symptom description and examination results according to the prompts in the application;

[0059] 1.3) Data transmission: After filling in the form, the data will be transmitted to the server through a secure encrypted channel to ensure patient privacy;

[0060] Data Storage:

[0061] Design database structure, including storage fields for multi-dimensional text information;

[0062] Data storage: Store the collected medical information in the database to ensure the integrity and consistency of the data;

[0063] Remove invalid records: Write a script to check if there are empty records or format errors in the database and delete them;

[0064] Remove duplicate records: Using the patient ID and timestamp key fields, write an algorithm to detect and merge duplicate records to ensure data uniqueness;

[0065] Use Jieba word segmentation tool to segment medical history, symptom description and examination result fields; for example, the input text "patient has persistent itching and rash" becomes "patient / has / persistent / itching / and / rash" after word segmentation;

[0066] Use the Word2Vec or BERT language model to convert the segmented text into a numerical vector; define the conversion formula as:

[0067] V i =Word2Vec(w i )

[0068] Among them, V i For vocabulary w i The numerical vector representation of , and i is the position index of the current word in the text sequence; Word2Vec is a word vector generation model; Word2Vec can map words to a high-dimensional continuous vector space;

[0069] Combine numerical vectors of different dimensions to generate multidimensional feature vectors; set the feature vectors of medical history, symptom description and examination results to be V h,i ,V s,i ,V r,i , then the multidimensional feature vector formula is as follows:

[0070] V total =[V h,i ,V s,i,V r,i ]

[0071] Among them, V total is a set of multidimensional feature vectors.

[0072] The above multidimensional feature vector [V h,i ,V s,i ,V r,i The feature vectors of different dimensions in ] are obtained through standardization processing to convert them to the same magnitude for analysis; the standardization formula is:

[0073]

[0074] Among them, V is the original eigenvector value, μ′ is the mean of the eigenvector, σ is the standard deviation of the eigenvector, V standard is the normalized feature vector.

[0075] Further explanation, for the construction of the isolation forest algorithm model:

[0076] 2.1) Set the algorithm parameters: the number of trees (n_estimators) and the number of subsamples (max_samples); set the number of trees to M and the number of subsamples to automatic selection;

[0077] 2.2) Model training: Use the standardized multidimensional feature vector set V total Perform model training; the specific steps are as follows:

[0078] Initialize the isolation forest model and set the algorithm parameters to:

[0079] The number of trees n_estimators = 100;

[0080] number of subsamples max_samples = 'auto';

[0081] Set random_state=42;

[0082] The multidimensional feature vector set V total Input the isolation forest model for training;

[0083] 2.3) Anomaly detection: Use the trained model to predict the multi-dimensional feature vector, and calculate the anomaly score SP of each feature vector through the decision_function of the isolation forest algorithm model; the specific steps are as follows:

[0084] The standardized multidimensional feature vector set V total Input the trained isolation forest model to get the predicted anomaly score, where -1 indicates anomaly and 1 indicates normality;

[0085] The specific steps are as follows:

[0086] The calculation formula of abnormal score SP is as follows:

[0087]

[0088] Where SP(i) is the eigenvector X i The anomaly score of X i ∈{V h,i ,V s,i ,V r,i};

[0089] M is the total number of trees in the isolation forest; path_length(X i ,T j ) is the eigenvector X i In the tree j The path length in , j represents the index of the tree in the isolation forest;

[0090] The constant C(M) is:

[0091]

[0092] The harmonic number H(M) is calculated as:

[0093] H(M)=ln(M)+γ

[0094] Among them, γ is the Euler-Marshall constant, which is 0.57721;

[0095] 2.4) Obtain the abnormal scores of all feature vectors; extract the feature vectors with a prediction result of -1 to form an abnormal feature vector set;

[0096] The anomaly score SP of each abnormal feature vector is integrated and expressed as an anomaly score set:

[0097] {SP(V h,i′ ), SP(V s,i′ ), SP(V r,i′ )|i′=1,2,…,m}

[0098] Where i′ represents the index of the i′th abnormal feature vector, and m represents the total number of abnormal feature vectors;

[0099] The abnormal feature vector set is extracted as {V h,i′ ,V s,i′ ,V r,i′ |i′=1,2,…,m}.

[0100] Further explanation: obtaining the extracted abnormal feature vector set {V h,i′,V s,i′ ,V r，i′ |i′=1,2,…,m};

[0101] Select the feature importance algorithm in random forest to calculate the importance score of each abnormal feature vector;

[0102] The steps to build a random forest model include:

[0103] 3.1) Set the model parameters of the random forest model; the number of trees and the maximum depth; set the number of trees to 100, the maximum depth to None; random state random_state = 42;

[0104] 3.2) Using the abnormal feature vector set {V h,i′ ,V s,i′ ,V r,i′ ∣i′=1,2,…,m} for model training;

[0105] 3.3) Importance score calculation: The importance score IS of each feature vector is obtained through the trained random forest model. The specific formula is:

[0106]

[0107] Among them, IS(X i′ ) is the abnormal feature vector X i′ The importance score of X i′ ∈{V h,i′ ,V s,i′ ,V r,i′}; n is the number of trees in the random forest; I(t,X i′ ) is the eigenvector X i′ Contribution in tree t;

[0108] 3.4) Obtain the importance scores of all abnormal feature vectors and generate an importance score set:

[0109] {IS(V h,i′ ),IS(V s,i′ ),IS(V r,i′ )|i′=1,2,...,m}

[0110] 3.5) Setting the screening threshold of importance score to T, the threshold set is an empirical value or a median value obtained through statistical analysis;

[0111] T=median({IS(V h,i′ ),IS(V s,i′ ),IS(V r,i ′)})

[0112] Mark the abnormal feature vectors with importance scores higher than the screening threshold T as preliminary key abnormal feature vectors;

[0113] 3.6) Combine the anomaly score SP and the importance score IS to calculate the comprehensive score. The formula is:

[0114] CS(X i′ ) = α × SP (X i′ )+β×IS(X i′ )

[0115] Among them, α and β are positive weight factors, and the value range of α and β is between 0 and 1, and Adjust α and β according to actual conditions;

[0116] In this embodiment, α and β are 0.3 and 0.7 respectively. At this time, the weight of the abnormality score SP is less than the weight of the importance score IS. The application scenarios are as follows:

[0117] When analyzing the patient's medical records, there are noise and irrelevant personal habit data, environmental factors and other data. In this case, relying on importance scores to screen out the key features that really affect scrotal eczema can improve the effectiveness of data management and analysis;

[0118] SP(X i′ ) is the abnormal feature vector X i′ Abnormal score; IS(X i′ ) is the abnormal feature vector X i′ Importance rating;

[0119] 3.7) Calculate the comprehensive scores of all abnormal feature vectors and generate a comprehensive score set:

[0120] {CS(V h,i ′),CS(V s,i′ ),CS(V r,i ′)|i′=1,2,...,m}

[0121] 3.8) Based on experimental demonstration or expert group systematic analysis, determine the critical value as T CS , according to the comprehensive score set, select the vectors with comprehensive scores higher than the critical value T from the preliminary key abnormal feature vectors CS The abnormal feature vector of is used as the final key abnormal feature vector, and the key abnormal feature vector set {X u ∣CS(X i′ )>T CS ,X i′ ∈{V h,i′ ,V s,i′ ,V r,i′}};

[0122] T CS =median({CS(V h,i′ ),CS(V s,i′ ),CS(V r,i′ )})

[0123] Among them, X u is the u-th key abnormal feature vector; together with SP(i) and IS(X i′ ) is calculated in the same way, and the importance score of the key abnormal feature vector is recorded as IS(X u ); the abnormal score of the key abnormal feature vector is recorded as SP(X u ).

[0124] Further explanation: K-means clustering algorithm is used for classification, including:

[0125] 4.1) Determine the optimal K′ value by the elbow method (ElbowMethod) or the average silhouette coefficient (SilhouetteScore); In this embodiment, K′=3 is determined by the elbow method;

[0126] 4.2) Initialize K-means algorithm:

[0127] Set the number of clusters K′ to 3; set the maximum number of iterations to 300; set the random seed to 42;

[0128] 4.3) Perform K-means clustering:

[0129] Input key abnormal feature vector set {X u ∣CS(X i′ )>T CS ,X i′ ∈{V h,i′ ,V s,i′ ,V r,i′}};

[0130] Run the K-means algorithm to obtain the cluster label Cu of each key abnormal feature vector; the cluster label Cu is obtained as follows:

[0131] 5.1) Initialize cluster centers: Randomly select K′ initial cluster centers μ j′ ;

[0132] 5.2) Assign cluster labels:

[0133] For each key anomaly feature vector X u , calculate its relationship with all cluster centers μ j′ distance;

[0134] X uAssign to the cluster corresponding to the nearest cluster center; thus, each X u There is a cluster label Cu, indicating the cluster to which it belongs;

[0135] 5.3) Update cluster center:

[0136] Calculate the new center of each cluster, which is the average of all eigenvectors in the cluster;

[0137] 5.4) Iteration:

[0138] Repeat steps 5.2) and 5.3) until the cluster center no longer changes significantly or the preset number of iterations is reached;

[0139] Finally, Cu represents X u Assign labels to the clusters they belong to; this process gradually approaches the optimal clustering result through continuous reallocation and updating.

[0140] The mathematical formula for clustering using the K-means algorithm is:

[0141]

[0142] Among them, X u is the u-th key abnormal feature vector, u is the index of the key abnormal feature vector; μ j′ is the j′th cluster center, and j′ is the index of the cluster center; K′ is the number of clusters, and U is the total number of key abnormal feature vectors.

[0143] Get the clustering results, including the cluster label to which each key anomaly feature vector belongs;

[0144] 4.4) Define the table structure, including the key anomaly feature vector ID, key anomaly feature vector content, comprehensive score CS, and clustering label Cu fields;

[0145] Store all key abnormal feature vectors and their related information in the database;

[0146] Ensure that the database record contains the feature vector ID, detailed content, comprehensive score and cluster label to facilitate subsequent query and analysis.

[0147] 4.5) Configure database index:

[0148] Create indexes for key fields including feature vector ID and cluster labels to improve query efficiency;

[0149] Provide RESTful API interface to support data query and analysis;

[0150] Design backend service interfaces to support the query of key abnormal feature vectors;

[0151] Ensure that the interface can query based on cluster labels, comprehensive scores, and other conditions.

[0152] To further illustrate, the acquisition of the weighted comprehensive value of each cluster specifically includes:

[0153] 6.1) Initialize variables:

[0154] Define the weighted comprehensive value set {WCV j′ |j′=1,2,...,K′}, where WCV j′ represents the weighted comprehensive value of the j′th cluster;

[0155] Initialize the weighted comprehensive value of each cluster to 0: WCV j′ = 0, where j′ is the cluster index, ranging from 1 to K′;

[0156] 6.2) Get the clustering labels of key anomaly feature vectors:

[0157] For each key anomaly feature vector X u , get its corresponding cluster label Cu;

[0158] {C u |u=1,2,...,U}, where U is the total number of key anomaly feature vectors;

[0159] 6.3) Calculate the weighted comprehensive value of each cluster:

[0160] For each cluster, calculate the weighted comprehensive value of all key abnormal feature vectors in the cluster to obtain the weighted comprehensive value WCV of each cluster j′ :

[0161] This formula can be expressed using summation notation as follows:

[0162]

[0163] Among them, δ(C u ,j′) is the indicator function, when the key abnormal feature vector X u When it belongs to the j′th cluster, δ(C u , j′)=1, otherwise 0; IS(X u ) is the key abnormal feature vector X u The importance score of SP(X u ) is the key abnormal feature vector X u Abnormal score of

[0164] A technical effect of traversing all feature vectors and calculating the weighted comprehensive value of the feature vectors belonging to the j′th cluster is achieved;

[0165] 6.4) Calculate and output the final weighted comprehensive value of each cluster:

[0166] The weighted comprehensive value WCV of each cluster is calculated j′ Normalize and get NWCV j′ . In order to provide a more intuitive evaluation:

[0167]

[0168] in, is the sum of all cluster weighted composite values;

[0169] 6.5) The final weighted comprehensive value set is expressed as:

[0170] {NWCV j′ |j′=1,2,...,K′}

[0171] Among them, NWCV j′ It is the weighted comprehensive value of the j′th cluster after normalization.

[0172] Further explanation: the classification index of scrotal eczema is defined as SCEI, and the calculation formula is as follows:

[0173]

[0174] Among them, w j′ is the weight coefficient of the j′th cluster, wj ′ The value is 0.25; is an exponential function that adjusts the influence of the normalized weighted comprehensive value. η1 is a positive number, 0.01≤η1≤0.11; it prevents the denominator from being zero; β′ is the weight coefficient of clinical data, and the value range of β′ is between 0 and 1; it is used to adjust the influence of clinical data on the classification index; CD(X u ) is the key abnormal feature vector X u Clinical data score; in this embodiment, η1 = 0.01; β′ = 0.5;

[0175] Define CD(X u ) is calculated as:

[0176]

[0177] Among them, D p is the raw score of the pth clinical data item; W p is the weight of the pth clinical data item; L is the total number of clinical data items; assign a weight W to each data item p , the sum of weights is 1;

[0178] The clinical data items are defined as age, disease course, and symptom severity, and the scoring is set as follows:

[0179] Let age be D1:

[0180] 0-10 years old: D1=2;

[0181] 11-20 years old: D1=4;

[0182] 21-40 years old: D1=6;

[0183] 41-60 years old: D1=8;

[0184] 61 years and above: D1=10

[0185] The duration of the disease is recorded as D2:

[0186] Day 1-3: D2 = 2;

[0187] 4-7 days: D2 = 4;

[0188] 8-14 days: D2 = 6;

[0189] 15-30 days: D2 = 8;

[0190] 31 days and above: D2=10;

[0191] The severity of the symptoms was recorded as D3:

[0192] 20% of the scrotal surface area is red, swollen, and itchy to a degree within 25% of the patient's tolerance limit: D3 = 2;

[0193] 20% to 45% of the scrotal surface area is red and swollen, and the degree of itching is between 25% and 60% of the patient's tolerance limit: D3 = 5;

[0194] Redness and swelling occur within 45% of the scrotal surface, and the redness and swelling are accompanied by ulceration and itching that exceeds 60% of the patient's tolerance limit: D3 = 8;

[0195] More than 45% of the scrotal surface area is red and swollen, accompanied by ulceration and itching that exceeds 60% of the patient's tolerance limit: D3 = 10;

[0196] In this embodiment, the weights W corresponding to age, course of disease, and severity of symptoms are set according to professional medical opinions or statistical analysis. p They are 0.3, 0.4, and 0.3 respectively:

[0197] Calculate CD(X u ) is:

[0198]

[0199] Further explanation: Dynamically adjust the screening threshold of importance score, including:

[0200] Set the valid value range of SCEI to (0,1);

[0201] When the SCEI value is closer to 0, the patient's clinical and classification characteristics are less obvious. The reasons include the following:

[0202] Mild illness: Patients only have mild symptoms, such as mild redness, swelling and slight itching.

[0203] Unclear features: The patient's clinical data items (such as age, disease course, and symptom severity) scored low, and different types of scrotal eczema could not be clearly distinguished.

[0204] Low priority treatment: In this case, the patient does not need emergency treatment and routine care and observation are sufficient.

[0205] When the SCEI value is closer to 1, the patient's clinical characteristics and classification characteristics are more obvious. The reasons include the following:

[0206] Severe illness: Patients have severe symptoms such as severe redness, swelling, severe itching, ulceration or infection.

[0207] Obvious characteristics: The patients' clinical data items (such as age, disease course, and symptom severity) scored high, which can clearly distinguish different types of scrotal eczema.

[0208] High priority for treatment: In this case, the patient requires urgent treatment and intervention.

[0209] The adjustment formula for the screening threshold T is:

[0210] T′=T×(1-SCEI)

[0211] Where T is the pre-set importance score screening threshold; T′ is the dynamically adjusted screening threshold;

[0212] Define the following interaction rules:

[0213] When the SCEI value is in the interval (0,0.25]:

[0214] T′=T×(1-0.25)

[0215] Among all feature vectors, select the top 25% as effective features;

[0216] When the SCEI value is in the interval (0.25,0.5]:

[0217] T′=T×(1-0.5)

[0218] Among all the key anomaly feature vectors, the top 55% are selected as effective features;

[0219] When SCEI is in the interval (0.5,0.75]:

[0220] T′=T×(1-0.75)

[0221] All key abnormal feature vectors are taken as effective features;

[0222] When the SCEI value is in the interval (0.75,1):

[0223] T′=T×(1-0.85).

[0224] The beneficial effects of the above-mentioned dynamic adjustment of the screening threshold T are:

[0225] Improve diagnostic accuracy: By dynamically adjusting the screening threshold according to actual clinical data, scrotal eczema can be more accurately identified and classified to avoid misdiagnosis and missed diagnosis.

[0226] Personalized treatment: The clinical manifestations of different patients may vary. Dynamic adjustment of thresholds can better reflect individual differences and thus support the formulation of personalized treatment plans.

[0227] Flexible response to changes in condition: The condition may change in a short period of time. Dynamic adjustment of thresholds can reflect these changes in a timely manner, making classification and diagnosis more flexible and real-time.

[0228] Optimize resource allocation: Adjusting the screening threshold according to changes in importance scores can optimize the allocation of medical resources and devote more attention and resources to cases that need priority treatment.

[0229] Improve model adaptability: Dynamically adjusting the threshold can make the classification model more adaptable, better cope with different data distributions and changes in the external environment, and improve the robustness of the model.

[0230] Data-driven decision support: Through dynamic adjustments driven by clinical data, the decision-making process becomes more scientific and accurate, improving the quality and efficiency of overall medical services.

[0231] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0232] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0233] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0234] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for intelligent management of scrotal eczema consultation record data, characterized in that: The specific steps include: Step S1: Collect multi-dimensional text information from the patient's medical consultation on the mobile phone, the multi-dimensional text information includes medical history, symptom description and examination results; Step S2: Use natural language processing technology to perform word segmentation on the multi-dimensional text information, apply a language model to convert the word segmentation into a numerical vector, and combine these numerical vectors to generate a multi-dimensional feature vector; Step S3: Identify abnormal feature vectors in the multidimensional feature vector by using the isolation forest algorithm to form an abnormal feature vector set, and calculate the abnormal score SP of each abnormal feature vector; Step S4: Obtain an abnormal feature vector set, calculate the importance score of each abnormal feature vector, and set a screening threshold for the importance score. Combine the abnormal score SP and the screening threshold to select the key abnormal feature vector with the most significant contribution from the scored abnormal feature vector set; Step S5: using the K-means clustering algorithm to classify the patient data using the key abnormal feature vectors selected in step S4 to obtain optimized clustering results; storing the clustering results and the key abnormal feature vectors in a database, and providing a system interface to support data query and analysis; Step S6: Obtain the clustering results obtained by the K-means clustering algorithm, use the importance score calculated in step S4 to perform weighted calculation on the key abnormal feature vector of each cluster in the clustering results, and obtain a weighted comprehensive value of each cluster; Step S7: Collect clinical data scores of key abnormal feature vectors, combine the clinical data scores with the weighted comprehensive value of each cluster for analysis, and generate a classification index for scrotal eczema. At the same time, the classification index can dynamically adjust the screening threshold of the importance score.

2. The intelligent management method for scrotal eczema consultation record data according to claim 1, characterized in that: Use Jieba word segmentation tool to segment medical history, symptom description and examination result fields; Use the Word2Vec language model to convert the segmented text into a numerical vector; define the conversion formula as: V i =Word2Vec(w i ) Among them, V i For vocabulary w i The numerical vector representation of , and i is the position index of the current word in the text sequence; Word2Vec is a word vector generation model; Combine numerical vectors of different dimensions to generate multidimensional feature vectors; set the feature vectors of medical history, symptom description and examination results to be V h,i ,V s,i ,V r,i , then the multidimensional feature vector formula is as follows: V total =[V h,i ,V s,i ,V r,i ] Among them, V total is a set of multidimensional feature vectors.

3. The intelligent management method for scrotal eczema consultation record data according to claim 2, characterized in that: For the construction of the isolation forest algorithm model: 2.1) Set the algorithm parameters: the number of trees and the number of subsamples; set the number of trees to M and the number of subsamples to automatic selection; 2.2) Model training: Use the standardized multidimensional feature vector set V total Conduct model training; 2.3) Anomaly detection: Use the trained model to predict the multi-dimensional feature vector, and calculate the anomaly score SP of each feature vector through the decision_function of the isolation forest algorithm model; The abnormality score of -1 indicates abnormality and 1 indicates normality; 2.4) Obtain the abnormal scores of all feature vectors; extract the feature vectors with a prediction result of -1 to form an abnormal feature vector set; The anomaly score SP of each abnormal feature vector is integrated and expressed as an anomaly score set: {SP(V h,i′ ),SP(V s,i′ ),SP(V r,i′ )|i′=1,2,…,m} Where i′ represents the index of the i′th abnormal feature vector, and m represents the total number of abnormal feature vectors; The abnormal feature vector set is extracted as {V h,i′ ,V s,i′ ,V r,i′ |i′=1,2,…,m}.

4. The intelligent management method for scrotal eczema consultation record data according to claim 3, characterized in that: Get the extracted abnormal feature vector set {V h,i′ ,V s,i′ ,V r,i′ |i′=1,2,…,m}; Select the feature importance algorithm in random forest to calculate the importance score of each abnormal feature vector; 3.1) Set the model parameters of the random forest model; 3.2) Using the abnormal feature vector set {V h,i′ ,V s,i′ ,V r,i′ ∣i′=1,2,…,m} for model training; 3.3) Obtain the importance score IS of each feature vector through the trained random forest model. The specific formula is: Among them, IS(X i′ ) is the abnormal feature vector X i′ The importance score of X i′ ∈{V h,i′ ,V s,i′ ,V r,i′ }; n is the number of trees in the random forest; I(t,X i′ ) is the eigenvector X i′ The contribution in tree t, where t represents the index of the decision tree in the random forest model; 3.4) Obtain the importance scores of all abnormal feature vectors and generate an importance score set: {IS(V h,i′ ),IS(V s,i′ ),IS(V r,i′ )∣i′=1,2,...,m} 3.5) Setting the screening threshold of importance score to T, the threshold set is an empirical value or a median value obtained through statistical analysis; T=median({IS(V h,i′ ),IS(V s,i′ ),IS(V r,i′ )}) Mark the abnormal feature vectors with importance scores higher than the screening threshold T as preliminary key abnormal feature vectors; 3.6) Combine the anomaly score SP and the importance score IS to calculate the anomaly feature vector X i′ The comprehensive score CS(X i′ ), the formula is: CS(X i′ )=α×SP(X i′ )+β×IS(X i′ ) Among them, CS(X i′ ) is the abnormal feature vector X i′ The comprehensive score of ; α and β are positive weight factors, and the value range of α and β is between 0 and 1, and SP(X i′ ) is the abnormal feature vector X i′ Abnormal score; IS(X i′ ) is the abnormal feature vector X i′ Importance rating; 3.7) Calculate the comprehensive scores of all abnormal feature vectors and generate a comprehensive score set: {CS(V h,i′ ),CS(V s,i′ ),CS(V r,i′ )∣i′=1,2,...,m} 3.8) Based on experimental demonstration or expert group systematic analysis, determine the critical value as T CS , according to the comprehensive score set, select the vectors with comprehensive scores higher than the critical value T from the preliminary key abnormal feature vectors CS The abnormal feature vector of is used as the final key abnormal feature vector; and a key abnormal feature vector set {X u ∣CS(X i′ )>T CS ,X i′ ∈{V h,i′ ,V s,i′ ,V r,i′ }}; T CS =median({CS(V h,i′ ),CS(V s,i′ ),CS(V r,i′ )}) Among them, X u is the u-th key abnormal feature vector; together with SP(i) and IS(X i′ ) is calculated in the same way, and the importance score of the key abnormal feature vector is recorded as IS(X u ); the abnormal score of the key abnormal feature vector is recorded as SP(X u ).

5. The intelligent management method for scrotal eczema consultation record data according to claim 4, characterized in that: The K-means clustering algorithm is used for classification, including: 4.1) Determine the optimal K′ value by the elbow method or the average silhouette coefficient; 4.2) Initialize K-means algorithm: 4.3) Perform K-means clustering: Input key abnormal feature vector set {X u ∣CS(X i′ )>T CS ,X i′ ∈{V h,i′ ,V s,i′ ,V r,i′ }}; Run the K-means algorithm to obtain the cluster label C of each key abnormal feature vector u ; The mathematical formula for clustering using the K-means algorithm is: Among them, X u is the u-th key abnormal feature vector, u is the index of the key abnormal feature vector; μ j′ is the j′th cluster center, j′ is the index of the cluster center; K′ is the number of clusters, and U is the total number of key abnormal feature vectors.

6. The intelligent management method for scrotal eczema consultation record data according to claim 5, characterized in that: The weighted comprehensive value of each cluster is obtained, including: Define the weighted comprehensive value set {WCV j′ |j′=1,2,...,K′}, where WCV j′ represents the weighted comprehensive value of the j′th cluster; Initialize the weighted comprehensive value of each cluster to 0. For each key abnormal feature vector X u , get its corresponding cluster label C u ; For each cluster, calculate the weighted comprehensive value of all key abnormal feature vectors in the cluster to obtain the weighted comprehensive value WCV of each cluster j′ : The weighted comprehensive value WCV of each cluster is calculated j′ Normalize and get NWCV j′ .

7. The intelligent management method for scrotal eczema consultation record data according to claim 6, characterized in that: The classification index of scrotal eczema is defined as SCEI, and the calculation formula is as follows: Among them, w j′ is the weight coefficient of the j′th cluster; is an exponential function that adjusts the influence of the normalized weighted comprehensive value, η1 is a positive number, 0.01≤η1≤0.11; β′ is the weight coefficient of clinical data, and the value range of β′ is between 0 and 1; CD(X u ) is the key abnormal feature vector X u Clinical data scores; Define CD(X u ) is calculated as: Among them, D p is the raw score of the pth clinical data item; W p is the weight of the pth clinical data item; L is the total number of clinical data items; assign a weight W to each data item p , the sum of weights is 1.

8. The intelligent management method for scrotal eczema consultation record data according to claim 7, characterized in that: Dynamically adjust the screening thresholds for importance scores, including: Set the valid value range of SCEI to (0,1); When the SCEI value is closer to 0, it means that the clinical and classification characteristics of the patient are less obvious; When the SCEI value is closer to 1, it means that the patient's clinical characteristics and classification characteristics are more obvious; The adjustment formula for the screening threshold T is: T′=T×(1-SCEI) Where T is the pre-set importance score screening threshold; T′ is the dynamically adjusted screening threshold; the following interaction rules are defined: When the SCEI value is in the interval (0,0.25]: T′=T×(1-0.25) When the SCEI value is in the interval (0.25,0.5]: T′=T×(1-0.5) When SCEI is in the interval (0.5,0.75]: T′=T×(1-0.75) When the SCEI value is in the interval (0.75,1): T′=T×(1-0.85).

Images