An electronic health record management system

Abstract

The present invention relates to the technical field of data management, and specifically provides an electronic health record management system, which includes: an image data analysis module, a medical record information classification module, an input error detection module, and a data access management module. In the present invention, by identifying and annotating the organs and tissue types in medical images, combining the shooting date and device type of the images, an image feature index is created, enhancing the retrieval efficiency of image data. Combining the automatic analysis and classification of medical record texts improves the efficiency and accuracy of medical record management. Through the analysis and cross-validation of data consistency, data input errors are effectively detected and corrected. The monitoring of user input behavior and the recognition of behavior patterns strengthen the security management of data. By analyzing and comparing user behavior data, dynamic adjustment of data access permissions is achieved, ensuring the security and legality of data access, improving the security of patient information, and enhancing the quality and effectiveness of medical services.

Description

Technical Field

[0001] The present invention relates to the technical field of data management, and particularly to an electronic health record management system. Background Art

[0002] The technical field of data management covers multiple aspects of data collection, storage, maintenance, processing, and use, focusing on efficiently and securely processing data, ensuring data quality and integrity, and meeting compliance and accessibility requirements. It utilizes database management, data cleaning, data integration, data protection, and data analysis, combined with machine learning algorithms to improve the data processing process and the accuracy of data analysis, enhance the automation and intelligence level of data management, achieve data-driven insights and decisions, and is applied to multiple aspects such as financial services, healthcare, public services, and education.

[0003] Among them, the electronic health record management system focuses on managing and storing patient health information, including collecting the patient's medical history, medication information, treatment process, treatment outcomes, and various key health indicators, and ensuring the security and privacy of patient information, enabling healthcare providers to effectively track patient conditions, provide real-time and accurate access to medical records, achieve accurate diagnosis and personalized treatment plans, promote the optimal allocation of medical resources, support the exchange and sharing of medical information by medical institutions on the premise of ensuring data security, reduce medical errors, enhance patient satisfaction, and the overall effectiveness of medical services.

[0004] Traditional electronic health record management systems lack the ability of effective integration and automated processing, have poor data management levels when dealing with complex and diverse medical data, are unable to accurately identify and process key information in images, resulting in low efficiency of medical record management and inaccurate information extraction, are insufficient in data consistency detection and correction of input errors, lack effective error prevention and correction measures, leading to the occurrence of medical errors, affecting the treatment effect of patients, limiting the performance of the medical system when dealing with large-scale data, and affecting the quality and speed of medical decision-making. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose an electronic health record management system.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions. An electronic health record management system includes:

[0007] Based on the patient's medical image information, the image data analysis module uses edge detection to identify the organs and tissue types in the image, combines the shooting date and shooting device type of the image, matches annotation information and retrieval tags for the medical image, and generates an image feature index;

[0008] Based on the image feature index, the medical record information classification module analyzes the text in the user's medical record, identifies various keywords in the medical record, and automatically marks and classifies the medical record information according to the symptom characteristics and diagnosis results of multiple patients to generate a medical record classification result;

[0009] The input error detection module uses the medical record classification result to detect data input errors by analyzing the data consistency of multiple data sources in the target medical record, combining the cross-validation of image annotation information and the integrity check of data, and corrects the input data of the medical record to generate verified medical record data;

[0010] Based on the verified medical record data, the data access management module monitors the user's input behavior, identifies the behavior pattern of the target user, compares it with the behavior pattern of the user's login identity, evaluates the legality of the access request, and adjusts the data access permission to obtain access control parameters.

[0011] As a further solution of the present invention, the steps for obtaining the organ and tissue types in the image are specifically as follows:

[0012] Based on the patient's medical image information, analyze the input image data through the formula:

[0013] ;

[0014] Calculate the gradient magnitude of the image to obtain the pixel edge intensity value;

[0015] Wherein, is the gradient magnitude of the image, represents the partial derivative of the gray value , is the gray value of the image, is the horizontal coordinate of the image, is the vertical coordinate of the image;

[0016] Based on the pixel edge intensity value, mark the edge information in the image by adjusting the contrast of the image to obtain an edge information extraction record;

[0017] Based on the edge information extraction record, identify various organs and tissues in the image, including the heart and lungs, by analyzing the combined characteristics of the edge information to generate organ and tissue types.

[0018] As a further solution of the present invention, the steps for obtaining the image feature index are specifically as follows:

[0019] According to the organ and tissue types, match annotation information for multiple identified tissues and organs, including the heart and lungs, to generate an annotation information matching record;

[0020] Match records based on the annotation information, combine the shooting date and shooting device type of the medical image, and use the formula:

[0021] ;

[0022] Calculate the index values of multiple images to obtain an image index list;

[0023] Among them, is the image feature index value, represents the th element in the feature vector, is the weight coefficient corresponding to , is the feature vector the number of elements in, is the index of the element in the feature vector;

[0024] According to the image index list, store the medical image in the database, associate the image index value with the storage path, and generate an image feature index.

[0025] As a further solution of the present invention, the steps for obtaining multiple keywords in the medical record are specifically as follows:

[0026] Based on the image feature index, preprocess the medical record text, including word segmentation, removal of stop words, calculate the frequency of each word appearing in multiple documents, and count the number of documents containing the target word to obtain a text word data set;

[0027] Using the text word data set, use the formula:

[0028] ;

[0029] Calculate the text importance index of multiple words in the document;

[0030] Among them, represents the text importance index of the word in the document , is the word in the document the number of occurrences in, represents any word in the document the number of occurrences in, is the index of the target vocabulary, is any vocabulary in the document, is the index of the target document, is the total number of documents in the document set, contains the word the number of documents;

[0031] According to the text importance index, by comparing the importance indexes of multiple words, various keywords in the medical record are identified.

[0032] As a further solution of the present invention, the steps for obtaining the medical record classification result are specifically as follows:

[0033] Based on the various keywords in the medical record, by analyzing the medical record document and expert evaluation, evaluate the importance of multiple keywords in multiple classifications, match weights for multiple medical record classifications and keywords, and establish a keyword weight list;

[0034] Based on the keyword weight list, through the formula:

[0035] ;

[0036] Calculate the scores of the target document in multiple categories to obtain the document category information;

[0037] Wherein, is the keyword set in the document , is the weight of the keyword for the classification , is the text importance index of the keyword in the document , is the document classified into the category, is used to determine the value of the variable that makes the given function reach the maximum value, is the index of the medical record classification, is the target document, is the document in the keyword;

[0038] Based on the document category information, classify and file the medical records of multiple patients to generate a medical record classification result.

[0039] As a further solution of the present invention, the steps for obtaining the verified medical record data are specifically as follows:

[0040] Based on the medical record classification result, analyze the information of multiple data sources in the target medical record, including analyzing the consistency between the diagnosis result and the treatment information, identifying inconsistencies and contradictions, and generating a document error identification result;

[0041] Based on the document error identification result, perform cross-verification on the image data and the text data of the associated medical record, including comparing the consistency of the pathological features in the text and the type information of the organs and tissues recorded in the image annotation, and generating a cross-verification result;

[0042] Based on the cross-validation results and combined with data integrity checks, through the formula:

[0043] ;

[0044] calculate the document accuracy;

[0045] wherein, is the document accuracy, is the weight coefficient of the cross-validation score, is the weight coefficient of the consistency analysis score, is the weight coefficient of the data integrity score, is the cross-validation score of the document information, is the consistency analysis score of the document information, is the integrity score of a single data field, is the total number of data fields, is the index of the data field;

[0046] Based on the document accuracy, detect data input errors and correct them, including correcting inconsistent information and filling in missing information, to generate verified medical record data.

[0047] As a further solution of the present invention, the step of obtaining the behavior pattern of the target user is specifically:

[0048] Based on the verified medical record data, monitor the user's input behavior, including collecting the user's keyboard typing speed and mouse movement pattern data to obtain user behavior data;

[0049] According to the user behavior data, through the formula:

[0050] ;

[0051] calculate the eigenvalue of the user behavior;

[0052] wherein, is the variability score of the target user behavior pattern, is the speed of the th keyboard keystroke, is the average value of the keyboard typing speed, is the speed of the th mouse movement, is the average value of the mouse movement speed, is the number of measurements of the keyboard keystrokes, is the number of measurements of the mouse movement, is the iteration index of the keyboard keystrokes, is the iteration index of the mouse movement;

[0053] Based on the eigenvalue of the user behavior, evaluate the consistency and regularity of the user behavior, analyze the stability of the user behavior, and identify the behavior pattern of the target user.

[0054] As a further solution of the present invention, the step of obtaining the access control parameter is specifically as follows:

[0055] Based on the behavior pattern of the target user, extract the behavior pattern information associated with the user login identity, including the keyboard tapping pattern and the mouse operation habit, and obtain the behavior feature data of the login identity.

[0056] According to the behavior feature data of the login identity, use the formula:

[0057] ;

[0058] Calculate the consistency between the current access request behavior and the standard behavior pattern of the user to obtain the behavior consistency score.

[0059] Wherein, represents the consistency between the current user behavior and the historical behavior pattern, represents the behavior eigenvalue of the th event in the user standard behavior pattern, represents the behavior eigenvalue of the th event in the current access request, represents the total number of behavior events recorded within a given monitoring period, is the index of the behavior feature event;

[0060] According to the behavior consistency score, evaluate the legitimacy of the access request in real time, and adjust the data access rights of the operating user, including restricting access and authentication, to obtain the access control parameter.

[0061] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0062] In the present invention, by identifying and labeling the organs and tissue types in the medical images, combining the shooting date and device type of the images, creating an image feature index, the retrieval efficiency of the image data is enhanced. Combining the automatic analysis and classification of the medical record text, the efficiency and accuracy of the medical record management are improved. Through the analysis and cross-validation of data consistency, data input errors are effectively detected and corrected. The monitoring of the user input behavior and the recognition of the behavior pattern strengthen the security management of the data. By analyzing and comparing the user behavior data, the dynamic adjustment of the data access rights is realized, ensuring the security and legitimacy of the data access, improving the security of the patient information, and strengthening the quality and effect of the medical service. Brief Description of the Drawings

[0063] Figure 1It is the system flowchart of the present invention;

[0064] Figure 2 It is the flowchart for identifying the types of organs and tissues in the image of the present invention;

[0065] Figure 3 It is the flowchart for obtaining the image feature index of the present invention;

[0066] Figure 4 It is the flowchart for extracting multiple keywords in the medical record of the present invention;

[0067] Figure 5 It is the flowchart for obtaining the classification result of the medical record of the present invention;

[0068] Figure 6 It is the flowchart for obtaining the verified medical record data of the present invention;

[0069] Figure 7 It is the flowchart for analyzing the behavior pattern of the target user of the present invention;

[0070] Figure 8 It is the flowchart for adjusting the access control parameters of the present invention. Detailed implementation manners

[0071] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0072] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more unless otherwise specifically defined.

[0073] Please refer to Figure 1 , an electronic health record management system includes:

[0074] Based on the patient's medical image information, the image data analysis module uses edge detection to identify the types of organs and tissues in the image, combines the shooting date and shooting device type of the image, matches annotation information and retrieval tags for the medical image, and generates an image feature index;

[0075] The medical record information classification module analyzes the text in the user's medical record based on image feature indexing, identifies various keywords in the medical record, and automatically marks and classifies the medical record information according to the symptom characteristics and diagnosis results of multiple patients to generate a medical record classification result;

[0076] The input error detection module uses the medical record classification result to detect data input errors by analyzing the data consistency of multiple data sources in the target medical record, combining cross-validation of image annotation information and data integrity checking, and corrects the input data of the medical record to generate verified medical record data;

[0077] The data access management module, based on the verified medical record data, monitors the user's input behavior, identifies the behavior pattern of the target user, compares it with the behavior pattern of the user's login identity, evaluates the legality of the access request, and adjusts the data access permission to obtain access control parameters.

[0078] The image feature indexing specifically refers to organ and tissue type identification, date tags, and device tags. The medical record classification result includes symptom marking information, diagnosis classification results, and treatment record classification information. The verified medical record data specifically refers to corrected diagnosis data, updated treatment information, and corrected symptom records. The access control parameters specifically refer to user behavior scores, permission adjustment levels, and legality verification results.

[0079] Please refer to Figure 2 , and the specific steps for obtaining the organ and tissue types in the image are as follows:

[0080] Based on the patient's medical image information, analyze the input image data through the formula:

[0081] ;

[0082] Calculate the gradient magnitude of the image to obtain the pixel edge intensity value;

[0083] Among them, is the gradient magnitude of the image, represents the partial derivative of the gray value , is the gray value of the image, is the horizontal coordinate of the image, is the vertical coordinate of the image;

[0084] Formula:

[0085] ;

[0086] Parameter meaning and acquisition method:

[0087] : The grayscale value of the image, obtained by converting the original color image to a grayscale image;

[0088] and : Respectively, the grayscale gradients of the image in the horizontal and vertical directions, calculated from the differences in grayscale values of adjacent pixels in the image. Among them, , ;

[0089] and : The pixel coordinates of the image, directly obtaining the position of each pixel through the pixel traversal method in image processing software or programming libraries;

[0090] Calculation example:

[0091] Set the position of the target pixel in the grayscale image to (x,y) = (100,100), and the grayscale value , and the grayscale values of adjacent pixels are respectively , , , ;

[0092] Calculate the gradient:

[0093] ;

[0094] ;

[0095] Calculate the gradient magnitude :

[0096] ;

[0097] The calculation results show that the edge intensity at the position (100,100) of the image is 11.18. The value reflects the edge sharpness of the image content at this position. The calculation process is used to identify and analyze the edge information in the image, providing a data basis for the edge enhancement and feature extraction steps.

[0098] Based on the pixel edge intensity value, by adjusting the contrast of the image, mark the edge information in the image to obtain the edge information extraction record;

[0099] The local histogram equalization method is adopted. This method involves the redistribution of the gray values in the area around each pixel point to enhance the local contrast of the image. During implementation, a 3x3 neighborhood window is selected, the pixel gray histogram within the neighborhood is calculated, and the gray value of the target pixel is remapped. This process is iteratively performed for the entire image. The adjusted image shows the edges of the organs. The Canny edge detection algorithm is used for image edge detection. Potential edges are determined using double thresholds, and strong and weak edges are determined by setting high and low thresholds. Strong edges are used as the final edges, and weak edges are retained when connected to strong edges. The obtained edge information extraction record includes edge intensity and edge position. The target data is used for subsequent organ and tissue type recognition.

[0100] Based on the edge information extraction record, by analyzing the combined features of the edge information, multiple organs and tissues in the image, including the heart and lungs, are recognized to generate organ and tissue types.

[0101] The support vector machine is used for the classification of organs and tissues in the image. The feature vectors of the organs are extracted through image preprocessing steps. The features include edge density, edge direction, and edge intensity. The target features are input into a pre-trained SVM model. The model training is based on a batch of labeled medical image data to distinguish different organ and tissue types such as the heart and lungs. Through the classification decision of the SVM model, various organ and tissue types in the image are recognized. Each image region is labeled as a specific organ type according to the output of the SVM. This process ensures the accuracy of image recognition and generates organ and tissue type labels.

[0102] Please refer to Figure 3 , and the specific steps for obtaining the image feature index are as follows:

[0103] According to the organ and tissue types, annotation information is matched for multiple recognized tissues and organs, including the heart and lungs, to generate an annotation information matching record.

[0104] The annotation propagation algorithm is applied. Based on the existing annotation templates, newly recognized organs and tissues are automatically matched to the corresponding annotation categories. The steps include loading the feature vectors of the organ and tissue recognition results, such as edge intensity and shape descriptors, and comparing the target features with the templates in the annotation database. The feature similarity is calculated for each matching process, and the template with the highest similarity is selected for annotation to generate a specific annotation information matching record. The record lists the annotation information corresponding to each organ and tissue and records the confidence level of the match, providing a basis for subsequent review or adjustment.

[0105] Based on the annotation information matching record, combined with the shooting date and shooting device type of the medical image, through the formula:

[0106] ;

[0107] Calculate the index values of multiple images to obtain an image index list;

[0108] Among them, is the image feature index value, indicating the th element in the feature vector, including the encoding of organ and tissue types, the encoding of the shooting date, and the encoding of the device type, is the weight coefficient corresponding to , is the number of elements in the feature vector ; is the index of the element in the feature vector;

[0109] Formula:

[0110] ;

[0111] Parameter meaning and acquisition method:

[0112] : The image feature index value is used to identify and retrieve images;

[0113] : The numerical encoding of the th feature, including organ or tissue type, shooting date, and device type;

[0114] : The weight of the th feature;

[0115] : The total number of elements in the feature vector;

[0116] Calculation example:

[0117] Set the target image as an ultrasonic medical image of the heart, and the shooting date is January 1, 2023, , , , , , , , calculate :

[0118] ;

[0119] ;

[0120] The calculation result is the index value of the target image, ensuring that the image can be effectively identified and retrieved through the index value, and optimizing the retrieval efficiency.

[0121] Store the medical images into the database according to the image index list, associate the image index values with the storage paths, and generate image feature indexes;

[0122] Use the insert and update operations of the SQL database to associate the feature index values of each medical image, such as organ type codes, shooting date codes, and device type codes, with the image files. By writing SQL commands, batch insert the target information into the medical image database. During the process, the storage path of each image file is also recorded in the database to ensure that any specific medical image can be quickly retrieved through the index value. After the operation, generate an operation report, including the number of successfully stored images and any errors or exceptions for subsequent maintenance or investigation. The process ensures data integrity and traceability, providing strong support for the management and use of medical images.

[0123] Please refer to Figure 4 , the steps for obtaining multiple keywords in the medical record are specifically as follows:

[0124] Based on the image feature index, preprocess the medical record text, including word segmentation, stop word removal, calculate the occurrence frequency of each word in multiple documents, and count the number of documents containing the target word to obtain a text word dataset;

[0125] Apply word segmentation and stop word removal operations in natural language processing technology to ensure that the text data is prepared for subsequent analysis. Use the NLTK library in the Python programming language, which provides a wide range of text processing functions. In the word segmentation step, the NLTK tokenizer decomposes the text string into individual lexical elements. Using the English stop word list provided by NLTK, remove common but low-information words from the tokenization results. Use the TF-IDF algorithm to calculate the occurrence frequency of each word in multiple documents and the total number of documents containing the word. During the process, the TF-IDF values of the words are calculated and stored to obtain a text word dataset. The dataset details each word and its statistical information in different documents.

[0126] Using the text word dataset, use the formula:

[0127] ;

[0128] Calculate the text importance index of multiple words in the document;

[0129] Among them, represents the text importance index of word in document , is the number of occurrences of word in document , represents any word The number of occurrences in the document is the index of the target vocabulary, is any vocabulary in the document, used to count the number of occurrences of all vocabularies in the document is the index of the target document, is the total number of documents in the document set, is the number of documents containing the word ; Formula: ;

[0130] Parameter meaning and acquisition method:

[0131] : is the number of occurrences of the word

[0132] in the document

[0133] obtained by directly counting the word frequency of the document ;

[0134] : is the total number of occurrences of all words in the document obtained by accumulating the occurrence frequencies of all words in the document;

[0135]

[0136] : is the total number of documents in the document set, obtained from the document management system database;

[0137] : is the number of documents containing the word

[0138] obtained by querying the database;

[0138] Calculation example:

[0139]

[0140]

[0141]

[0142] ;

[0141] ;

[0142] The calculation result indicates the word in the document The importance of the text quantifies the uniqueness and information content of the word in this document compared to other documents.

[0143] According to the text importance index, by comparing the importance indexes of multiple words, various keywords in the medical record are identified;

[0144] Determine the importance of each word in the text based on its TF-IDF value. During the process, determine the TF-IDF threshold for each word. The threshold is determined based on statistical analysis and is usually set as the average value of the TF-IDF values of all words. If the TF-IDF value of a word is higher than this threshold, it is considered a keyword. This is executed through the Scikit-learn library in Python. After calculating the TF-IDF value of each word, select the word with the highest TF-IDF value as the keyword. The target keyword represents the main content and theme of the document. This process ensures the effective identification of important words related to specific diseases, treatments, or medical conditions from a large number of medical record texts.

[0145] Please refer to Figure 5 , and the specific steps for obtaining the medical record classification results are as follows:

[0146] Based on various keywords in the medical record, by analyzing the medical record document and expert evaluation, evaluate the importance of multiple keywords in multiple classifications, match weights for multiple medical record classifications and keywords, and establish a keyword weight list;

[0147] Use a combination of automated text analysis techniques and manual evaluation, including text classification algorithms based on machine learning, such as support vector machines and decision trees. The target algorithm can automatically identify the correlation between keywords and categories when processing text data classification. During the process, extract keywords from the medical record document and use the keyword importance feedback provided by the expert group to adjust the initial weight of each keyword. According to the degree of association between the keyword and the medical record classification, use the algorithm model to adjust the keyword weight to ensure that the weight reflects the actual importance of the keyword in the medical record classification. The weight adjustment process includes multiple iterations, and each iteration adjusts the keyword weight based on expert feedback and algorithm output to generate a keyword weight list. The list records the weights of each keyword corresponding to different medical record classifications.

[0148] Based on the keyword weight list, through the formula:

[0149] ;

[0150] Calculate the scores of the target document in multiple categories to obtain the document category information;

[0151] Among them, is the set of keywords in the document , is the keyword For classification The weight of is the text importance index of the keyword in the document ; is the category to which the document is classified; is used to determine the value of the variable that maximizes a given function; is the index for medical record classification; is the target document; is the document ; the keyword in

[0152] Formula:

[0153] ;

[0154] Parameter meaning and acquisition method:

[0155] : is the weight of the keyword in the category , reflecting the importance of the word in a specific classification;

[0156] : is the text importance index of the keyword in the document , obtained by performing text analysis and word frequency statistics on the document;

[0157] : is the set of all keywords in the document ;

[0158] : is all possible medical record classifications, including cardiology, neurology;

[0159] Calculation example:

[0160] Suppose there are three classifications , representing cardiology, representing neurology, representing digestive diseases, and the set of keywords in the document is , being treatment, being pain, being heart, and suppose , , , ;

[0161] Calculate the score for each category:

[0162] Cardiology Score:

[0163] ;

[0164] Neurology Score:

[0165] ;

[0166] Digestive system diseases Score:

[0167] ;

[0168] By quantifying the relevance of each classification to the document, the classification result of the target document is obtained. The process is used for document classification in automated document management.

[0169] Based on the document category information, the medical records of multiple patients are classified and archived to generate the medical record classification result;

[0170] A multi-class logistic regression model based on keyword weights is used to handle multi-label classification problems, which is applicable to medical document classification. In the process, a classification score is calculated for each medical record document. The score is based on the weighted sum of all keywords in the document, and the weights are derived from a previously established keyword weight list. By setting a classification threshold, when the score of each medical record document is higher than a certain threshold, the document is classified into the corresponding medical record category. The classification process is automatic. By programming, the classification label of each patient's medical record is updated in the database. After classification, the system automatically stores the classification result and generates a medical record classification report. The result includes the number of medical records in each category and the evaluation result of classification accuracy, providing a detailed overview of the medical record classification status for medical institutions and facilitating further medical decision-making and medical record management.

[0171] Please refer to Figure 6 , the specific steps for obtaining the verified medical record data are as follows:

[0172] Based on the medical record classification result, the information of multiple data sources in the target medical record is analyzed, including analyzing the consistency between the diagnosis result and the treatment information, identifying inconsistencies and contradictions, and generating the document error identification result;

[0173] Use data consistency analysis tools, such as data comparison algorithms, to ensure the matching of diagnosis and treatment information, including extracting diagnosis and treatment records from the medical record database, executing through programming scripts in a Python environment, checking the matching of each diagnosis record with relevant treatment suggestions, the analysis tool examines the correlation between the two types of information, and identifies mismatched or contradictory data points. If it is found that the diagnosis result is inconsistent with the treatment suggestion, record the target information as a potential error, analyze to ensure a high degree of consistency between the diagnosis data and the treatment suggestion, generate a document error recognition result, and list in detail all identified inconsistencies and contradictions.

[0174] Based on the document error recognition result, perform cross-verification on the image data and the text data of the associated medical records, including comparing the consistency of the pathological features in the text and the type information of the organs and tissues recorded in the image annotations, and generating a cross-verification result;

[0175] Use image processing algorithms and text comparison algorithms to extract medical images related to the medical records and their annotation information from the database, and at the same time obtain the pathological feature descriptions in the medical record text. Analyze the types of organs and tissues in the image through image processing software, compare the target information with the pathological features in the text, the comparison algorithm analyzes the consistency of the two data sources, identifies information points that do not conform or are incorrect, the steps ensure the matching degree of the image data and the text data, enhance the accuracy and reliability of the medical record data, generate a cross-verification result, and record all discovered data inconsistencies.

[0176] Based on the cross-verification result, combined with data integrity checking, through the formula:

[0177] ;

[0178] Calculate the document accuracy;

[0179] Among them, is the document accuracy, is the weight coefficient of the cross-verification score, is the weight coefficient of the consistency analysis score, is the weight coefficient of the data integrity score, is the cross-verification score of the document information, is the consistency analysis score of the document information, is the integrity score of a single data field, is the total number of data fields, is the index of the data field;

[0180] Formula:

[0181] ;

[0182] Parameter meaning and acquisition method:

[0183] : is the cross - validation score, obtained by comparing the diagnostic information in the medical record with the image annotation information;

[0184] : is the consistency analysis score, calculated by comparing the match degree between the diagnostic results and treatment measures in the medical record;

[0185] : is the integrity score of a single data field, obtained by checking whether each field in the medical record is filled in completely, including patient name, gender, age information, symptom description, diagnostic results;

[0186] : is the total number of data fields, reflecting the total number of fields that need to be filled in the medical record;

[0187] , , : is the weight coefficient of the corresponding parameter;

[0188] Calculation example:

[0189] Set , , , , , , ;

[0190] Calculate :

[0191] ;

[0192] ;

[0193] ;

[0194] ;

[0195] The result 52.9 reflects the comprehensive accuracy and integrity of the medical record data. The value is used to judge whether the medical record data meets the preset accuracy and integrity standards, and further make decisions on the processing and use of the medical record data.

[0196] Based on the document accuracy, detect data input errors and correct them, including correcting inconsistent information and filling in missing information, to generate verified medical record data;

[0197] Use data correction tools, such as the update function of the database management system, combined with custom data correction scripts, including automatic correction of inconsistent and missing information. The script is executed in the database. For each identified error or missing point, the medical record is updated to reflect more accurate information. For complex contradictions that cannot be automatically resolved, they are marked and a report is generated for manual review by medical staff. After the operation is completed, the system generates verified medical record data, including corrected and filled information, ensuring the integrity and accuracy of the medical record data.

[0198] Please refer to Figure 7 , and the steps for obtaining the behavior pattern of the target user are specifically as follows:

[0199] Based on the verified medical record data, monitor the user's input behavior, including collecting data on the user's keyboard typing speed and mouse movement pattern, to obtain user behavior data;

[0200] Record and analyze the user's keyboard typing speed and mouse movement pattern in the electronic health record system, and collect input data in real time. The keyboard typing speed is obtained by calculating the average time between each keyboard keystroke, and the mouse movement speed is measured by tracking the number of pixels the mouse moves per second. The collected data is encrypted and sent to the central server for analysis to obtain a dataset including each user's behavior characteristics. The target dataset includes key parameters such as timestamps, typing speeds, and movement speeds, providing basic data for behavior analysis.

[0201] According to the user behavior data, through the formula:

[0202] ;

[0203] Calculate the characteristic value of the user behavior;

[0204] Among them, is the variability score of the target user behavior pattern, is the speed of the th keyboard keystroke, is the average value of the keyboard typing speed, is the speed of the th mouse movement, is the average value of the mouse movement speed, is the number of measurements of keyboard keystrokes, is the number of measurements of mouse movements, is the iteration index of keyboard keystrokes, is the iteration index of mouse movements;

[0205] Formula:

[0206] ;

[0207] Parameter meanings and acquisition methods:

[0208] : The speed of the th keyboard keystroke, obtained by monitoring the interval time of each keystroke in real time with monitoring software;

[0209] : The average value of the keyboard keystroke speed;

[0210] : The speed of the th mouse movement, obtained by monitoring the rate of each mouse movement in real time with monitoring software;

[0211] : The average value of the mouse movement speed;

[0212] : The number of measurements of keyboard keystrokes, that is, the total number of keystrokes recorded within the monitoring time window;

[0213] : The number of measurements of mouse movements, that is, the total number of movements recorded within the monitoring time window;

[0214] , : Iteration indicators, representing the number indicators of keyboard keystrokes and mouse movements respectively;

[0215] Calculation example:

[0216] Set , , the keyboard keystroke speed is milliseconds per keystroke, the mouse movement speed is pixels per second, milliseconds per keystroke, pixels per second, calculate the variability:

[0217] For keyboard keystrokes, calculate the square of the difference for each item:

[0218] ;

[0219] ;

[0220] ;

[0221] ;

[0222] ;

[0223] ;

[0224] ;

[0225] ;

[0226] ;

[0227] ;

[0228] For mouse movement, calculate the square of the difference for each item:

[0229] ;

[0230] ;

[0231] ;

[0232] ;

[0233] ;

[0234] ;

[0235] ;

[0236] ;

[0237] ;

[0238] ;

[0239] ;

[0240] ;

[0241] Calculate the standard deviation:

[0242] ;

[0243] ;

[0244] ;

[0245] The value obtained is 3.34. The value indicates the deviation of the consistency and regularity of the user's behavior. The lower the value, the more consistent the user's behavior. The calculation process is used to evaluate and identify the behavior pattern of the target user, providing support for security monitoring.

[0246] Evaluate the consistency and regularity of the user's behavior based on the characteristic value of the user's behavior, analyze the stability of the user's behavior, and identify the behavior pattern of the target user;

[0247] Use statistical analysis methods and machine learning models to evaluate the consistency and regularity of user behavior, identify user behavior patterns, use time series analysis and clustering algorithms such as K-means. Time series analysis is applied to data on keyboard typing speed and mouse movement speed to identify any periodic changes or abnormal patterns in the behavior data. The time series data is input into the clustering algorithm to classify user behavior and identify typical behavior patterns or abnormal behaviors. The analysis results help system administrators and security teams understand typical user behavior and its changes, and evaluate the stability of behavior. The behavior pattern of each user is recorded in detail and compared with the user's historical behavior data to determine any significant behavior changes or inconsistencies.

[0248] Please refer to Figure 8 , and the specific steps for obtaining access control parameters are as follows:

[0249] Based on the behavior pattern of the target user, extract behavior pattern information associated with the user's login identity, including keyboard typing patterns and mouse operation habits, to obtain login identity behavior characteristic data;

[0250] Through the behavior monitoring software deployed on the user terminal, record the keystroke interval, keystroke force, mouse movement speed, and acceleration parameters. Each time the user logs in during the process, the software is automatically activated to seamlessly record behavior data. The target data is transmitted to the central analysis server through an encrypted channel. The data processing script running on the server is written in the Python language, uses the Pandas library for data cleaning and feature extraction, and performs mathematical operations through the NumPy library to accurately calculate the statistical characteristics of the behavior pattern. The obtained behavior characteristic data includes the average keyboard keystroke interval of each user, the average value and standard deviation of the mouse movement speed, etc. The target data provides a basis for identifying the user's login identity.

[0251] According to the login identity behavior characteristic data, use the formula:

[0252] ;

[0253] Calculate the consistency between the current access request behavior and the user's standard behavior pattern to obtain a behavior consistency score;

[0254] Among them, represents the consistency between the user's current behavior and the historical behavior pattern, represents the behavior characteristic value of the th event in the user's standard behavior pattern, represents the behavior characteristic value of the th event in the current access request, represents the total number of behavior events recorded within a given monitoring period, is the index of the behavioral characteristic event;

[0255] Formula:

[0256] ;

[0257] Parameter meaning and acquisition method:

[0258] : The th behavioral characteristic value of the user in the standard behavior pattern, obtained through the user behavior monitoring system, reflecting the user's typical operation speed;

[0259] : The th behavioral characteristic value in the current access request, also obtained through real-time monitoring, reflecting the user's current operation speed;

[0260] : The number of behavioral events, obtained by analyzing the user activity data within the target time window;

[0261] Calculation example:

[0262] Set the keyboard tapping speed of the target operating user in multiple interfaces to , and the keyboard tapping speed of the standard behavior pattern to , , calculate the behavior consistency:

[0263] Calculate :

[0264] ;

[0265] Calculate and the sum of squares:

[0266] ;

[0267] ;

[0268] Calculate :

[0269] ;

[0270] ;

[0271] The calculation result 0.99971 indicates that the user's current behavior is very similar to the historical standard pattern, reflecting that the user's current access request is legal. The calculation process is used to verify the user's identity and adjust the access permissions.

[0272] Evaluate the legitimacy of access requests in real time according to the behavior consistency score, and adjust the data access permissions of the operating user, including access restriction and authentication, to obtain access control parameters;

[0273] Use an anomaly detection model in machine learning, including a classification algorithm based on support vector machines, to train a model from historical behavior data, define the boundary of normal behavior. When a user performs a login operation, capture the behavior data in real time and quickly perform a consistency score through this model. If the user's current behavior has a high consistency with the pattern in the training set, their behavior consistency score will be higher than the preset threshold, and the user will have normal access. If the consistency score is lower than the threshold, the system will trigger security response measures, such as sending a warning, requiring secondary authentication, or restricting access. This evaluation process runs in the form of a microservice in the server background, ensuring the system's response speed and the security of data processing. The result of the process is to generate access control parameters. The target parameters record the evaluation results and measures taken for each access request, providing real-time data support for the system's security monitoring.

[0274] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An electronic health record management system, characterized in that: The system comprises: The image data analysis module uses edge detection based on the patient's medical image information to identify the organ and tissue types in the image. It combines the image's shooting date and the type of shooting device to match the annotation information and retrieval tags for the medical image and generate an image feature index. The medical record information classification module analyzes the text in the user's medical record based on the image feature index, identifies multiple keywords in the medical record, and automatically marks and classifies the medical record information according to the symptom characteristics and diagnosis results of multiple patients to generate a medical record classification result; The input error detection module uses the medical record classification results to analyze the data consistency of information from multiple data sources in the target medical record, combines cross-validation of image annotation information and data integrity check, detects data input errors, and corrects the input data of the medical record to generate verified medical record data; The data access management module monitors the user's input behavior based on the verified medical record data, identifies the behavior pattern of the target user, and compares it with the behavior pattern of the user's login identity, evaluates the legitimacy of the access request, and adjusts the data access rights to obtain access control parameters; The steps of obtaining the image feature index are specifically as follows: matching annotation information for a plurality of identified tissues and organs, including a heart and a lung, according to the organ and tissue types, and generating an annotation information matching record; Based on the matching records of the annotation information, combined with the shooting date and shooting device type of the medical image, the formula is: ； Calculate the index values ​​of multiple images and obtain the image index list; in, is the image feature index value, represents the first elements, is corresponding to The weight coefficient of is the eigenvector The number of elements in , is the index of the element in the feature vector; According to the image index list, the medical image is stored in a database, and the image index value and the storage path are associated to generate an image feature index; The steps for obtaining the verified medical record data are specifically as follows: Based on the medical record classification results, analyzing information from multiple data sources in the target medical record, including analyzing the consistency of the diagnosis results and treatment information, identifying inconsistencies and contradictions, and generating document error identification results; Based on the document error recognition results, cross-validating the image data with the text data of the associated medical records, including comparing the consistency of the pathological features in the text with the type information of the organs and tissues recorded in the image annotations, and generating a cross-validation result; Based on the cross-validation results, combined with data integrity check, the formula: ； Calculate document accuracy; in, is the document accuracy, is the weight coefficient of the cross-validation score, is the weight coefficient of the consistency analysis score, is the weight coefficient of the data integrity score, is the cross-validation score of the document information, is the consistency analysis score of the document information, is the completeness score of a single data field, is the total number of data fields, is the index of the data field; Based on the accuracy of the document, detect and correct data entry errors, including correcting inconsistent information, filling in missing information, and generating verified medical record data; The steps for obtaining the access control parameters are specifically as follows: Based on the behavior pattern of the target user, extract the behavior pattern information associated with the user login identity, including keyboard tapping pattern and mouse operation habits, and obtain the login identity behavior feature data; According to the login identity behavior characteristic data, use the formula: ； Calculate the consistency between the current access request behavior and the user's standard behavior pattern to obtain a behavior consistency score; in, Indicates the consistency of the user's current behavior with the historical behavior pattern. Represents the user's standard behavior pattern The behavior characteristic value of the event, Represents the first The behavior characteristic value of the event, represents the total number of behavioral events recorded in a given monitoring period, is the index of the behavior feature event; According to the behavior consistency score, the legitimacy of the access request is evaluated in real time, and the data access rights of the operating user are adjusted, including restricting access and identity authentication, to obtain access control parameters.

2. The electronic health record management system according to claim 1, characterized in that: The steps for acquiring the organ and tissue types in the image are specifically as follows: Based on the patient's medical image information, the input image data is analyzed through the formula: ； Calculate the gradient amplitude of the image to obtain the pixel edge intensity value; in, is the gradient amplitude of the image, Represents the gray value The partial derivative of is the gray value of the image, is the horizontal coordinate of the image, is the vertical coordinate of the image; Based on the pixel edge intensity value, the contrast of the image is adjusted to mark the edge information in the image, thereby obtaining an edge information extraction record; Based on the edge information extraction record, multiple organs and tissues in the image, including the heart and lungs, are identified by analyzing the combined features of the edge information, and organ and tissue types are generated.

3. The electronic health record management system according to claim 1, characterized in that: The steps for obtaining the multiple keywords in the medical record are specifically as follows: Based on the image feature index, preprocess the medical record text, including word segmentation, removing stop words, calculating the frequency of occurrence of each word in multiple documents, and counting the number of documents containing the target word to obtain a text word data set; Using the text word dataset, use the formula: ； Calculate the text importance index of multiple words in a document; in, Expressive words In the documentation The text importance index in , Words In the documentation The number of occurrences in Represents any word In the documentation The number of occurrences in is the index of the target vocabulary, is any word in the document, is the index of the target document, is the total number of documents in the document set, Contains words The number of documents; According to the text importance index, multiple keywords in the medical record are identified by comparing the importance indexes of multiple words.

4. The electronic health record management system according to claim 3, characterized in that: The steps for obtaining the medical record classification results are specifically as follows: Based on the multiple keywords in the medical records, by analyzing the medical record documents and expert evaluation, evaluating the importance of multiple keywords in multiple categories, matching weights for multiple medical record categories and keywords, and establishing a keyword weight list; Based on the keyword weight list, by formula: ； Calculate the scores of the target document in multiple categories to obtain document category information; in, For Documentation The keyword set in For keywords For classification The weight of For keywords In the documentation The text importance index in , For Documentation Categorized into categories, It is used to determine the value of the variable that makes the given function reach the maximum value. C is the index of the medical record classification. For the target document, For Documentation Keywords in; Based on the document category information, the medical records of multiple patients are classified and archived to generate medical record classification results.

5. The electronic health record management system according to claim 1, characterized in that: The steps for acquiring the behavior pattern of the target user are specifically as follows: Based on the verified medical record data, monitoring the user's input behavior, including collecting the user's keyboard typing speed and mouse movement pattern data to obtain user behavior data; According to the user behavior data, through the formula: ； Calculate the characteristic value of user behavior; in, is the variability score of the target user's behavior pattern, It is The speed of each keyboard stroke, is the average keyboard typing speed, It is The speed of mouse movement. is the average mouse movement speed, is the measured number of keyboard strokes, is the number of measurements of mouse movement, is the iteration index of keyboard strokes, is the iteration index of the mouse movement; Based on the characteristic values ​​of the user behaviors, the consistency and regularity of the user behaviors are evaluated, the stability of the user behaviors is analyzed, and the behavior patterns of the target users are identified.

Images