Data processing method for full-life-cycle resident health archive

Abstract

The invention discloses a data processing method for a full-life-cycle resident health record. The method comprises the following steps: S1, integrating data from various devices and systems through a standardized interface; s2, preprocessing the collected health data, and storing the health data in a distributed database and a cloud storage system; s3, performing consistency processing on data from different sources by using data cleaning, duplicate removal and standardization technologies; s4, performing deep analysis on the health data by applying a long-short-term memory network; s5, carrying out data encryption through AES and RSA technologies, and guaranteeing the privacy of the user through role-based access control and a differential privacy technology; and S6, integrating the mobile health application and the intelligent wearable device into a health file management system, and supporting real-time health monitoring and remote medical services. The system has the advantages of comprehensive coverage, real-time processing, efficient storage, intelligent analysis, data security and privacy protection, cross-platform sharing and remote medical integration.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for processing data of resident health records throughout their life cycle. Background Art

[0002] With the rapid development of information technology and medical technology, the management of residents' health records has gradually attracted widespread attention. The life cycle resident health record system aims to comprehensively record and manage residents' health data from birth to old age, and provide support for personalized health management and medical services. However, the existing health record management system still has many deficiencies in data collection, storage, processing, analysis and security protection, and it is difficult to meet the growing medical and health management needs.

[0003] The existing technology has obvious defects in the following aspects:

[0004] 1. Insufficient data collection and integration: Existing health record systems often rely on data collection from a single source and lack the ability to integrate multi-source data. Specifically, there is a lack of effective standardized interfaces between data sources such as medical institutions, wearable devices, and personal health records, which makes it difficult to seamlessly integrate data and affects the comprehensiveness and consistency of data.

[0005] 2. Inefficient data storage and management: Traditional health data storage methods mostly use centralized storage, which lacks the ability to efficiently manage and expand large-scale data. When faced with massive amounts of health data, the existing system has low storage efficiency, insufficient data backup and disaster recovery capabilities, and it is difficult to ensure high availability and reliability of data.

[0006] 3. Limited data processing and analysis capabilities: The existing health record system has limited processing and analysis capabilities, mainly relying on simple statistics and rule algorithms, and it is difficult to process and analyze massive amounts of health data in real time. Lacking the support of advanced data processing frameworks and machine learning algorithms, it is unable to provide accurate health trend predictions, disease risk assessments, and personalized health recommendations.

[0007] 4. Weak data security and privacy protection: Existing health data management systems have significant deficiencies in data security and privacy protection. Traditional encryption and access control measures are often not sound enough to effectively prevent data leakage and unauthorized access. There is a lack of comprehensive security protection mechanisms based on differential privacy and blockchain technology, making it difficult to fully comply with the requirements of relevant laws and regulations.

[0008] 5. Poor data sharing and interoperability: Poor data sharing and interoperability between different medical systems and institutions result in the inability of medical data to circulate effectively between systems, hindering the development of collaborative medical services. The existing system lacks the ability to use interoperability standards and blockchain technology to achieve data sharing, which affects the efficiency of medical resource utilization.

[0009] Therefore, how to provide a data processing method for residents’ health records throughout their life cycle is an urgent problem that technicians in this field need to solve. Summary of the invention

[0010] One purpose of the present invention is to propose a data processing method for residents' health records throughout their life cycle, aiming to solve the deficiencies in data collection, storage, processing, analysis, security and sharing in the prior art. The method comprehensively utilizes a variety of advanced technologies such as multi-source data integration, distributed storage and cloud storage, big data processing framework, deep learning algorithm, data encryption and access control, differential privacy protection, Internet of Things technology, blockchain technology and natural language processing, and describes in detail the entire process from data collection to the generation of personalized health recommendations. The method has the advantages of comprehensive coverage, real-time processing, efficient storage, intelligent analysis, data security and privacy protection, cross-platform sharing and telemedicine integration, which significantly improves the intelligence and reliability of residents' health record management.

[0011] A data processing method for a resident health record throughout the life cycle according to an embodiment of the present invention is characterized by comprising the following steps:

[0012] S1. Integrate data from multiple devices and systems through standardized interfaces, including wearable devices, medical institutions and personal health records, to collect multi-dimensional health data of residents in real time;

[0013] S2, pre-processing the collected health data and storing it in a distributed database and cloud storage system;

[0014] S3. Use data cleaning, deduplication and standardization technologies to process data from different sources in a consistent manner, and use ETL processes to achieve data integration;

[0015] S4. Apply long short-term memory networks to conduct in-depth analysis of health data to generate health trend predictions, disease risk assessments, and personalized health recommendations;

[0016] S5, data encryption through AES and RSA technology, and user privacy protection through role-based access control and differential privacy technology;

[0017] S6. Integrate mobile health applications and smart wearable devices into the health record management system to support real-time health monitoring and telemedicine services, achieve data synchronization and interoperability between devices through the Internet of Things technology, and generate personalized health management plans based on the analysis results.

[0018] Optionally, the S1 specifically includes:

[0019] S11, receiving raw data from wearable devices, medical institutions and personal health records through a multi-protocol interface module;

[0020] S12. Standardize the various data formats collected:

[0021]

[0022] T si =T i +ΔT;

[0023] Among them, S(D i ) represents the standardized data item, D i represents the collected original data item, μ represents the mean of the data item, σ represents the standard deviation of the data item, T si Represents the standardized timestamp, T i represents the original timestamp, and ΔT represents the correction value of the time zone difference or time offset;

[0024] S13, achieving real-time data collection, by setting the data collection module to synchronize data at a fixed frequency f, where the value of f is determined according to the data type and device characteristics;

[0025] S14, data verification and quality control, verify the validity of the received data, eliminate errors and abnormal data, and use the verification algorithm V(d i ) to verify each data item, where d i Represents the i-th data item to be verified:

[0026] Calculate the expected value E(d i ) and standard deviation σ(d i ):

[0027]

[0028]

[0029] The validity range of the data item is L,U, where:

[0030] L=E(d i )-k·σ(d i );

[0031] U=E(d i )+k·σ(d i );

[0032] k is a control parameter, which is determined according to actual needs and has a value range of 2 to 3;

[0033] Verify data item d i Is it within the validity range?

[0034]

[0035] Among them, V(d i ) has an output of 1 for valid data and 0 for invalid data;

[0036] S15. Use time series analysis techniques to preprocess continuous health data. t}, where d t Represents the data value at time t, and calculates the moving average MA of the time series t :

[0037]

[0038] Where n represents the size of the time window;

[0039] Calculate the difference Δd of the time series t ;

[0040] Δd t =d t -d t-1 ;

[0041] Apply exponential smoothing to the time series after difference and calculate the smoothed data value S t :

[0042] S t =αd t +(1-α)S t-1 ;

[0043] Among them, α is the smoothing coefficient, and its value range is 0<α≤1;

[0044] Apply the autoregressive integrated moving average model for time series modeling:

[0045] ARIMA(p,d,q);

[0046] Where p is the order of the autoregressive term, d is the number of differences, and q is the order of the moving average term:

[0047]

[0048] Among them, φ i represents the autoregressive coefficient, θ j represents the moving average coefficient, ∈ t represents the white noise term.

[0049] Optionally, the S2 specifically includes:

[0050] S21, convert the format of the collected health data and perform denoising processing, and use the Gaussian filtering algorithm to denoise the data. The original data sequence is {d i}, the filtered data sequence is {d′ i}

[0051]

[0052] Among them, σ is the standard deviation, k is the window size of the filter;

[0053] S22: The preprocessed data is partitioned according to the data type and timestamp and stored in a distributed database, with the partition key P k A combination of data type T and timestamp t:

[0054] P k =(T,t);

[0055] S23. Use the cloud storage system to regularly back up the data in the distributed database and realize data disaster recovery in different locations. The number of copies of the backup data is n. b , the backup frequency is f b ;

[0056] S24. Establish multidimensional indexes for stored data to speed up data query, and use B-tree and R-tree structures to index structured data and geospatial data respectively:

[0057] I k =(T,t);

[0058] Among them, T represents the data type, t represents the timestamp, and the R-tree index is used for fast query of spatial data. The bounding box of spatial data is B(x min ,y min ,x max ,y max ), where x and y represent geographic coordinates;

[0059] S25. Use LZ77 compression algorithm to compress the stored data:

[0060] B′ i =LZ77(B i );

[0061] Among them, LZ77 represents the application of data compression algorithm, and the original data block is {Bi}, the compressed data block is {B′ i};

[0062] S26. Use a role-based access control model to set access control policies for stored data:

[0063]

[0064] Among them, user u can access the corresponding data only when he is granted access rights in the permission p corresponding to his role r. The role set is R, the permission set is P, and the user set is U;

[0065] S27. Monitor the health status of the distributed database and cloud storage system, use the heartbeat mechanism to detect the status of the storage node, and the heartbeat interval is Δt. If the heartbeat signal is not received for n consecutive times, the node is determined to be failed and fault recovery is performed.

[0066] Optionally, the S3 specifically includes:

[0067] S31. Clean the collected health data and use the median filter algorithm to remove outliers, noise and outliers. The original data set is D = {d 1 ,d 2 ,…,d n}, the cleaned data set is D ′ , where d i For the i-th data item:

[0068] d′ i =median{d i-k ,…,d i ,…,d i+k};

[0069] Where k is the filter window size, d′ i is the data item after cleaning;

[0070] S32. De-duplicate the data to avoid duplicate records. The data set after de-duplication is D ″ , use the hash algorithm to calculate the unique identifier h(d i ), if the hash value of a data item already exists in the data set, the data item is considered a duplicate and removed:

[0071] h(d i )=hash(d i );

[0072] Among them, h(d i ) is the data item d i The hash value of

[0073] S33, standardize the data from different sources, the original data item is D i , the standardized data item is S(D i ):

[0074]

[0075] Among them, μ is the mean of the data item, σ is the standard deviation of the data item;

[0076] S34. Calibrate the timestamps of data from different sources. The original timestamp is T i , the calibrated timestamp is T si :

[0077] T si =T i +ΔT;

[0078] Wherein, ΔT is the correction value of the time offset;

[0079] S35. Use the ETL process to integrate the cleaned, deduplicated and standardized data. The extracted data set is E, the converted data set is T, and the loaded data set is L. The ETL process is expressed as:

[0080] E→T→L;

[0081] Among them, E represents the data extracted from the source system, T represents the transformed and standardized data, and L represents the data loaded into the target database;

[0082] S36. Perform consistency verification on the integrated data using a data consistency verification algorithm C(D), where D is the data set to be verified:

[0083]

[0084] Among them, the output of C(D) is 1, which means that the data consistency verification passes, and the output is 0, which means that the data consistency verification fails.

[0085] Optionally, the S4 specifically includes:

[0086] S41. Extract key features from the preprocessed health data. The data set is D = {d 1 ,d 2 ,…,d n}, the feature set is F = {f 1 ,f 2 ,…,f m}, where each feature f i By function F i (d) Generate:

[0087] fi =F i (d);

[0088] Where d represents the data item, F i represents the feature extraction function;

[0089] S42, the extracted features are standardized, and the standardized feature set is S(F):

[0090]

[0091] Among them, μ f The feature f i The mean value, σ f The feature f i The standard deviation of

[0092] S43, the standardized feature set and the corresponding label Y = {y 1 ,y 2 ,…,y n}, construct the training data set T = {(S(f 1 ),y 1 ),(S(f 2 ),y 2 ),…,(S(f n ),y n )}, where y i is the data item d i Labels;

[0093] S44. Use the long short-term memory network to train the training data set. The input sequence is X t ={x t1 ,x t2 ,…,x tn}, the output is the prediction result

[0094] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0095] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0096] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0097] C t =f t ·C t-1 +i t tanh(W C ·[h t-1 ,x t ]+b C );

[0098] h t =o t tanh(C t );

[0099] Among them, i t is the input gate, f t For the forget gate, t is the output gate, C t is the unit state, h t is the hidden state, W and b represent the weight matrix and bias vector respectively, σ is the sigmoid activation function, tanh is the hyperbolic tangent activation function, x t is the current input data, h t-1 is the hidden state of the previous time step;

[0100] S45. Use the validation data set to validate the trained model. The validation data set is V = {(S(f 1 ),y 1 ),(S(f 2 ),y 2 ),…,(S(f n ),y n )}, the error between the predicted result and the true label is calculated by the loss function L:

[0101]

[0102] in, is the model prediction result, y i is the true label, and the loss function L is minimized by adjusting the model parameters W and b;

[0103] S46, use the trained model to predict the health trend of the newly input health data, the new input data is X′={x′ 1 ,x′ 2 ,…,x′ t}, the model output prediction result is

[0104]

[0105] Among them, X′ t represents the new input sequence, Indicates predicted health trends;

[0106] S47. Generate personalized health recommendations based on prediction results, and analyze predicted health trends Provide specific health management suggestions and intervention measures. The suggestion set is G = {g 1 ,g 2 ,…,g n}, where g i According to the prediction results Generated suggestions.

[0107] Optionally, the S5 specifically includes:

[0108] S51, use the AES encryption standard to encrypt the health data, the original data set is D = {d 1 ,d 2 ,…,d n}, the encrypted data set is E(D):

[0109] E(d i )=AES(d i ,K);

[0110] Among them, d i is the i-th data item, K is the encryption key, E(d i ) is the encrypted data item;

[0111] S52. Decrypt the encrypted data using AES decryption technology:

[0112] D ′ (d i )=AES -1 (E(d i ),K);

[0113] Among them, AES -1 is the AES decryption algorithm, E(d i ) is the encrypted data item, K is the decryption key, D ′ (d i ) is the decrypted data item;

[0114] S53. Set access control policies for the stored data, with the role set as R, the permission set as P, and the user set as U:

[0115]

[0116] Among them, user u can access the corresponding data only when he is granted access rights in the permission p corresponding to his role r;

[0117] S54. Perform differential privacy processing on the data. The original data set is D, and the data set after differential privacy processing is D ′ , add noise ∈ to protect privacy, the processing process is:

[0118] D ′ =D+∈;

[0119] Among them, the noise ∈ follows the Laplace distribution:

[0120]

[0121] Among them, Δf is the sensitivity, ∈ is the privacy parameter;

[0122] S55, record the user's access operation to the data, the log record is L = {l 1 ,l 2 ,…,l n}, each log record contains the access user u, access time t, access data item d and operation type o:

[0123] l i =(u,t,d,o);

[0124] Among them, l i is the i-th log record.

[0125] Optionally, the S6 specifically includes:

[0126] S61. Integrate mobile health applications and smart wearable devices into the health record management system, collect data through standardized interfaces, and the collected health data is D m ={d m1 ,d m2 ,…,d mn} and D w ={d w1 ,d w2 ,…,d wn}, where D m represents the data collected by the mobile application, D w Represents the data collected by smart wearable devices;

[0127] S62. Use IoT technology to synchronize data between devices, including data transmission and timestamp alignment:

[0128] D s =D m ∪D w ;

[0129] Timestamp alignment:

[0130] T si =T i +ΔT;

[0131] Among them, T si is the timestamp after alignment, T i is the original timestamp, ΔT is the time offset;

[0132] S63, the integrated data set D s Analyze and realize real-time monitoring of residents’ health status. The health monitoring index is H = {h 1 ,h 2 ,…,h k}, through the monitoring algorithm M(H,D s )Calculate health status:

[0133] H i =M(H,D s );

[0134] Among them, H i is the calculated i-th health state;

[0135] S64, transmit health data to telemedicine platform to support doctors to make remote diagnosis and suggestions;

[0136] S65, generating a personalized health management plan P based on the health monitoring index H and the diagnosis result R of the telemedicine platform;

[0137] S66. Push the personalized health management plan to the user through the mobile application. Each health suggestion corresponds to a management plan and the health management plan is adjusted according to user feedback.

[0138] The beneficial effects of the present invention are:

[0139] (1) The present invention integrates data from wearable devices, medical institutions and personal health records through a standardized interface, collects multi-dimensional health data in real time, achieves seamless integration of multi-source data, ensures the comprehensiveness and consistency of data, and effectively solves the problem of insufficient data collection and integration in the prior art.

[0140] (2) The present invention utilizes a distributed database and a cloud storage system for data storage, and combines data preprocessing, denoising, partitioned storage, backup and disaster recovery, data compression and other technologies to significantly improve the efficiency and reliability of data storage and management, ensure the high availability and scalability of data, and overcome the limitations of traditional storage methods.

[0141] (3) The present invention adopts a big data processing framework and deep learning algorithms, such as long short-term memory network (LSTM), to perform intelligent analysis and real-time processing of health data, provide health trend prediction, disease risk assessment and personalized health advice, greatly improve data processing and analysis capabilities, and solve the problem of limited analysis capabilities of existing systems.

[0142] (4) The present invention ensures data security and user privacy protection through the Advanced Encryption Standard (AES) and Role-Based Access Control (RBAC) model, combined with differential privacy protection measures, effectively preventing data leakage and unauthorized access, complying with relevant laws and regulations, and enhancing user trust.

[0143] (5) The present invention integrates mobile health applications and smart wearable devices into the health record management system, supports real-time health monitoring and telemedicine services, generates personalized health management plans based on health monitoring indicators and the diagnosis results of the telemedicine platform and pushes them to users, provides convenient personalized health management services, and expands the accessibility and convenience of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0144] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0145] Figure 1 This is a schematic diagram of the structure of a multi-source data integration module of a data processing method for resident health records throughout the life cycle proposed by the present invention;

[0146] Figure 2 A schematic diagram of the structure of a data storage and management module of a data processing method for a resident health record throughout the life cycle proposed by the present invention;

[0147] Figure 3 A schematic diagram of the structure of an intelligent analysis module of a data processing method for a resident health record throughout the life cycle proposed by the present invention;

[0148] Figure 4 This is a structural schematic diagram of the data security and privacy protection module of the data processing method for residents’ health records throughout the life cycle proposed by the present invention. DETAILED DESCRIPTION

[0149] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0150] refer to Figure 1-Figure 4 , a data processing method for residents' health records throughout their life cycle, characterized by comprising the following steps:

[0151] S1. Integrate data from multiple devices and systems through standardized interfaces, including wearable devices, medical institutions and personal health records, to collect multi-dimensional health data of residents in real time;

[0152] In this implementation, S1 specifically includes:

[0153] S11, receiving raw data from wearable devices, medical institutions and personal health records through a multi-protocol interface module;

[0154] S12. Standardize the various data formats collected:

[0155]

[0156] T si =T i +ΔT;

[0157] Among them, S(D i ) represents the standardized data item, D i represents the collected original data item, μ represents the mean of the data item, σ represents the standard deviation of the data item, T si Represents the standardized timestamp, T i represents the original timestamp, and ΔT represents the correction value of the time zone difference or time offset;

[0158] S13, achieving real-time data collection, by setting the data collection module to synchronize data at a fixed frequency f, where the value of f is determined according to the data type and device characteristics;

[0159] S14, data verification and quality control, verify the validity of the received data, eliminate errors and abnormal data, and use the verification algorithm V(d i ) to verify each data item, where d i Represents the i-th data item to be verified:

[0160] Calculate the expected value E(d i ) and standard deviation σ(d i ):

[0161]

[0162]

[0163] The validity range of the data item is L,U, where:

[0164] L=E(d i )-k·σ(d i );

[0165] U=E(d i )+k·σ(d i );

[0166] k is a control parameter, which is determined according to actual needs and has a value range of 2 to 3;

[0167] Verify data item d i Is it within the validity range?

[0168]

[0169] Among them, V(d i ) has an output of 1 for valid data and 0 for invalid data;

[0170] S15. Use time series analysis techniques to preprocess continuous health data. t}, where d t Represents the data value at time t, and calculates the moving average MA of the time series t :

[0171]

[0172] Where n represents the size of the time window;

[0173] Calculate the difference Δd of the time series t ;

[0174] Δd t =d t -d t-1 ;

[0175] Apply exponential smoothing to the time series after difference and calculate the smoothed data value S t :

[0176] S t =αd t +(1-α)S t-1 ;

[0177] Among them, α is the smoothing coefficient, and its value range is 0<α≤1;

[0178] Apply the autoregressive integrated moving average model for time series modeling:

[0179] ARIMA(p,d,q);

[0180] Where p is the order of the autoregressive term, d is the number of differences, and q is the order of the moving average term:

[0181]

[0182] Among them, φ i represents the autoregressive coefficient, θ j represents the moving average coefficient, ∈ t represents the white noise term.

[0183] S2, pre-processing the collected health data and storing it in a distributed database and cloud storage system;

[0184] In this implementation, S2 specifically includes:

[0185] S21, convert the format of the collected health data and perform denoising processing, and use the Gaussian filtering algorithm to denoise the data. The original data sequence is {d i}, the filtered data sequence is {d′ i}

[0186]

[0187] Among them, σ is the standard deviation, k is the window size of the filter;

[0188] S22: The preprocessed data is partitioned according to the data type and timestamp and stored in a distributed database, with the partition key P k A combination of data type T and timestamp t:

[0189] P k =(T,t);

[0190] S23. Use the cloud storage system to regularly back up the data in the distributed database and realize data disaster recovery in different locations. The number of copies of the backup data is n. b , the backup frequency is f b ;

[0191] S24. Establish multidimensional indexes for stored data to speed up data query, and use B-tree and R-tree structures to index structured data and geospatial data respectively:

[0192] I k =(T,t);

[0193] Among them, T represents the data type, t represents the timestamp, and the R-tree index is used for fast query of spatial data. The bounding box of spatial data is B(x min ,y min ,x max ,y max ), where x and y represent geographic coordinates;

[0194] S25. Use LZ77 compression algorithm to compress the stored data:

[0195] B′ i =LZ77(B i );

[0196] Among them, LZ77 represents the application of data compression algorithm, and the original data block is {B i}, the compressed data block is {B′i};

[0197] S26. Use a role-based access control model to set access control policies for stored data:

[0198]

[0199] Among them, user u can access the corresponding data only when he is granted access rights in the permission p corresponding to his role r. The role set is R, the permission set is P, and the user set is U;

[0200] S27. Monitor the health status of the distributed database and cloud storage system, use the heartbeat mechanism to detect the status of the storage node, and the heartbeat interval is Δt. If the heartbeat signal is not received for n consecutive times, the node is determined to be failed and fault recovery is performed.

[0201] S3. Use data cleaning, deduplication and standardization technologies to process data from different sources in a consistent manner, and use ETL processes to achieve data integration;

[0202] In this implementation, S3 specifically includes:

[0203] S31. Clean the collected health data and use the median filter algorithm to remove outliers, noise and outliers. The original data set is D = {d 1 ,d 2 ,…,d n}, the cleaned data set is D′, where d i For the i-th data item:

[0204] d′ i =median{d i-k ,…,d i ,…,d i+k};

[0205] Among them, k is the filter window size, d′ i is the data item after cleaning;

[0206] S32, deduplicate the data to avoid duplicate records. The deduplicated data set is D″. The hash algorithm is used to calculate the unique identifier h(d i ), if the hash value of a data item already exists in the data set, the data item is considered a duplicate and removed:

[0207] h(d i )=hash(d i );

[0208] Among them, h(d i ) is the data item d iThe hash value of

[0209] S33, standardize the data from different sources, the original data item is D i , the standardized data item is S(D i ):

[0210]

[0211] Among them, μ is the mean of the data item, σ is the standard deviation of the data item;

[0212] S34. Calibrate the timestamps of data from different sources. The original timestamp is T i , the calibrated timestamp is T si :

[0213] T si =T i +ΔT;

[0214] Wherein, ΔT is the correction value of the time offset;

[0215] S35. Use the ETL process to integrate the cleaned, deduplicated and standardized data. The extracted data set is E, the converted data set is T, and the loaded data set is L. The ETL process is expressed as:

[0216] E→T→L;

[0217] Among them, E represents the data extracted from the source system, T represents the transformed and standardized data, and L represents the data loaded into the target database;

[0218] S36. Perform consistency verification on the integrated data using a data consistency verification algorithm C(D), where D is the data set to be verified:

[0219]

[0220] Among them, the output of C(D) is 1, which means that the data consistency verification passes, and the output is 0, which means that the data consistency verification fails.

[0221] S4. Apply long short-term memory networks to conduct in-depth analysis of health data to generate health trend predictions, disease risk assessments, and personalized health recommendations;

[0222] In this implementation, S4 specifically includes:

[0223] S41. Extract key features from the preprocessed health data. The data set is D = {d 1 ,d 2 ,…,d n}, the feature set is F = {f 1 ,f2 ,…,f m}, where each feature f i By function F i (d) Generate:

[0224] f i =F i (d);

[0225] Where d represents the data item, F i represents the feature extraction function;

[0226] S42, the extracted features are standardized, and the standardized feature set is S(F):

[0227]

[0228] Among them, μ f The feature f i The mean value, σ f The feature f i The standard deviation of

[0229] S43, the standardized feature set and the corresponding label Y = {y 1 ,y 2 ,…,y n}, construct the training data set T = {(S(f 1 ),y 1 ),(S(f 2 ),y 2 ),…,(S(f n ),y n )}, where y i is the data item d i Labels;

[0230] S44. Use the long short-term memory network to train the training data set. The input sequence is X t ={x t1 ,x t2 ,…,x tn}, the output is the prediction result

[0231] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0232] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0233] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0234] C t =f t ·C t-1 +i t tanh(W C ·[h t-1 ,x t ]+b C );

[0235] h t =o t tanh(C t );

[0236] Among them, i t is the input gate, f t For the forget gate, t is the output gate, C t is the unit state, h t is the hidden state, W and b represent the weight matrix and bias vector respectively, σ is the sigmoid activation function, tanh is the hyperbolic tangent activation function, x t is the current input data, h t-1 is the hidden state of the previous time step;

[0237] S45. Use the validation data set to validate the trained model. The validation data set is V = {(S(f 1 ),y 1 ),(S(f 2 ),y 2 ),…,(S(f n ),y n )}, the error between the predicted result and the true label is calculated by the loss function L:

[0238]

[0239] in, is the model prediction result, y i is the true label, and the loss function L is minimized by adjusting the model parameters W and b;

[0240] S46, use the trained model to predict the health trend of the newly input health data, the new input data is X′={x′ 1 ,x′ 2 ,…,x′ t}, the model output prediction result is

[0241]

[0242] Among them, X′ t represents the new input sequence, Indicates predicted health trends;

[0243] S47. Generate personalized health recommendations based on prediction results, and analyze predicted health trends Provide specific health management suggestions and intervention measures. The suggestion set is G = {g 1 ,g 2 ,…,g n}, where g i According to the prediction results Generated suggestions.

[0244] S5, data encryption through AES and RSA technology, and user privacy protection through role-based access control and differential privacy technology;

[0245] In this implementation, S5 specifically includes:

[0246] S51, use the AES encryption standard to encrypt the health data, the original data set is D = {d 1 ,d 2 ,…,d n}, the encrypted data set is E(D):

[0247] E(d i )=AES(d i ,K);

[0248] Among them, d i is the i-th data item, K is the encryption key, E(d i ) is the encrypted data item;

[0249] S52. Decrypt the encrypted data using AES decryption technology:

[0250] D ′ (d i )=AES -1 (E(d i ),K);

[0251] Among them, AES -1 is the AES decryption algorithm, E(d i ) is the encrypted data item, K is the decryption key, D ′ (d i ) is the decrypted data item;

[0252] S53. Set access control policies for the stored data, with the role set as R, the permission set as P, and the user set as U:

[0253]

[0254] Among them, user u can access the corresponding data only if he is granted access rights in the permission p corresponding to his role r;

[0255] S54. Perform differential privacy processing on the data. The original data set is D, and the data set after differential privacy processing is D ′ , add noise ∈ to protect privacy, the processing process is:

[0256] D ′ =D+∈;

[0257] Among them, the noise ∈ follows the Laplace distribution:

[0258]

[0259] Among them, Δf is the sensitivity, ∈ is the privacy parameter;

[0260] S55, record the user's access operation to the data, the log record is L = {l 1 ,l 2 ,…,l n}, each log record contains the access user u, access time t, access data item d and operation type o:

[0261] l i =(u,t,d,o);

[0262] Among them, l i is the i-th log record.

[0263] S6. Integrate mobile health applications and smart wearable devices into the health record management system to support real-time health monitoring and telemedicine services, achieve data synchronization and interoperability between devices through the Internet of Things technology, and generate personalized health management plans based on the analysis results.

[0264] In this implementation, S6 specifically includes:

[0265] S61. Integrate mobile health applications and smart wearable devices into the health record management system, collect data through standardized interfaces, and the collected health data is D m ={d m1 ,d m2 ,…,d mn} and D w ={d w1 ,d w2 ,…,d wn}, where D m represents the data collected by mobile applications, D w Represents the data collected by smart wearable devices;

[0266] S62. Use IoT technology to synchronize data between devices, including data transmission and timestamp alignment:

[0267] D s =D m ∪D w ;

[0268] Timestamp alignment:

[0269] T si =T i +ΔT;

[0270] Among them, T si is the timestamp after alignment, T i is the original timestamp, ΔT is the time offset;

[0271] S63, the integrated data set D s Analyze and realize real-time monitoring of residents’ health status. The health monitoring index is H = {h 1 ,h 2 ,…,h k}, through the monitoring algorithm M(H,D s )Calculate health status:

[0272] H i =M(H,D s );

[0273] Among them, H i is the calculated i-th health state;

[0274] S64, transmit health data to telemedicine platform to support doctors to make remote diagnosis and suggestions;

[0275] S65, generating a personalized health management plan P based on the health monitoring index H and the diagnosis result R of the telemedicine platform;

[0276] S66. Push the personalized health management plan to the user through the mobile application. Each health suggestion corresponds to a management plan and the health management plan is adjusted according to user feedback.

[0277] Embodiment 1:

[0278] In order to verify the feasibility of the present invention, the present invention is applied to the resident health record management system of a large medical institution. The medical institution includes multiple hospitals, community clinics and rehabilitation centers, involving health data of hundreds of thousands of residents, with various data types and huge quantities. Through practical application, we demonstrate the effects of the present invention in multi-source data integration, data storage and management, intelligent analysis, data security and privacy protection, cross-platform data sharing and telemedicine integration.

[0279] In practical applications, data from different devices and systems, including wearable devices, medical institutions, and personal health records, are first integrated through standardized interfaces. After the data is collected in real time, data synchronization between devices is achieved through IoT technology to ensure data consistency and accuracy. For example, a hospital's electronic medical record system (EMR) integrates data from smart bracelets, blood pressure monitors and other devices worn by residents to collect multi-dimensional health data in real time, including heart rate, blood pressure, blood sugar, body temperature, number of steps, etc.

[0280] In terms of data storage and management, we use distributed databases and cloud storage systems, combined with data preprocessing, denoising, partitioned storage, backup and disaster recovery, data compression and other technologies to achieve efficient data management. Through data compression technology, such as the LZ77 algorithm, the original data is compressed and stored, saving a lot of storage space, and distributed storage achieves high data availability and scalability. To ensure data security, we use the AES encryption standard to encrypt all health data and adopt a role-based access control (RBAC) model to ensure that only authorized users can access the corresponding data.

[0281] In terms of data processing and analysis, we apply big data processing frameworks and deep learning algorithms, such as long short-term memory networks (LSTM), to intelligently analyze health data. By processing and analyzing the collected health data in real time, we generate health trend predictions, disease risk assessments, and personalized health recommendations. For example, by analyzing a resident's blood pressure and heart rate data for a month, the LSTM model predicts that the resident is at risk of hypertension in the next two weeks and generates personalized health management recommendations, including diet adjustments, exercise plans, and regular monitoring.

[0282] In order to ensure the privacy and security of data, we use differential privacy protection technology to add noise to the data to ensure that user privacy is not leaked. In practical applications, the data set processed by differential privacy technology not only protects user privacy, but also provides high-quality data analysis results. For example, when analyzing the health data of residents in a community, the system can still accurately identify potential health risk groups in the community while protecting personal privacy, providing strong data support for community health service centers.

[0283] In terms of data sharing and interoperability, we use blockchain technology to achieve data sharing between different medical systems and institutions, ensuring the circulation and consistency of data among systems. Through smart contract technology, dynamic management of data access rights is achieved. For example, when a resident is referred to another hospital, his or her health records can be securely transferred to the new hospital through blockchain technology, ensuring the integrity and non-tamperability of the data. Specific data proves our beneficial effects as follows:

[0284] Table 1 Improvement report of health record management system in large medical institutions

[0285] Data Category Before application After application Monthly health data processing volume 500,000 1 million Data storage efficiency 10 6 Data encryption protection 5 50 Accuracy of hypertension risk prediction 60％ 90％ Customer satisfaction 70％ 95％

[0286] It can be seen from these data that the method of the present invention not only improves the efficiency and quality of residents' health record management, but also significantly improves the security and sharing capabilities of data, providing strong support for personalized health management and medical services.

[0287] In the first month of actual application, a hospital managed the health data of more than 5,000 residents through the method of the present invention. Through the integration and real-time processing of multi-source data, doctors in the hospital can have a more comprehensive understanding of the health status of patients and provide personalized medical advice based on the results of data analysis. Statistics show that among the patients with high blood pressure risks predicted by this system, 85% of them took effective preventive measures under the guidance of doctors and avoided the occurrence of high blood pressure.

[0288] In terms of data storage, the application of distributed databases and cloud storage systems has significantly improved data storage efficiency. Data that originally required 10TB of storage space only occupies 6TB of storage space after compression and partition storage, saving 40% of storage costs. At the same time, the application of data backup and disaster recovery systems ensures the security and recovery capabilities of data in emergency situations.

[0289] In terms of data security, through the application of AES encryption and RBAC model, the system prevented 30 unauthorized accesses in one quarter, protecting the health data security of residents. With the application of differential privacy protection technology, the system can perform efficient data analysis while protecting privacy. The community health service center identified 10 potential high-risk health groups by analyzing the privacy-protected data and carried out special health education and intervention activities for these groups.

[0290] In terms of data sharing, blockchain technology enables efficient data sharing between different medical institutions. After a resident is referred to another hospital, his health records are securely transmitted to the new hospital within a few minutes, and doctors can immediately obtain his complete health records and provide him with continuous medical services.

[0291] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A data processing method for residents' health records throughout their life cycle, characterized in that: The steps include: S1. Integrate data from multiple devices and systems through standardized interfaces, including wearable devices, medical institutions and personal health records, to collect multi-dimensional health data of residents in real time; S2, pre-processing the collected health data and storing it in a distributed database and cloud storage system; S3. Use data cleaning, deduplication and standardization technologies to process data from different sources in a consistent manner, and use ETL processes to achieve data integration; S4. Apply long short-term memory networks to conduct in-depth analysis of health data to generate health trend predictions, disease risk assessments, and personalized health recommendations; S5, data encryption through AES and RSA technology, and user privacy protection through role-based access control and differential privacy technology; S6. Integrate mobile health applications and smart wearable devices into the health record management system to support real-time health monitoring and telemedicine services, achieve data synchronization and interoperability between devices through the Internet of Things technology, and generate personalized health management plans based on the analysis results.

2. According to claim 1, a data processing method for life cycle resident health records is characterized in that: The S1 specifically includes: S11, receiving raw data from wearable devices, medical institutions and personal health records through a multi-protocol interface module; S12. Standardize the various data formats collected: T si =T i +ΔT; Among them, S(D i ) represents the standardized data item, D i represents the collected original data item, μ represents the mean of the data item, σ represents the standard deviation of the data item, T si Represents the standardized timestamp, T i represents the original timestamp, and ΔT represents the correction value of the time zone difference or time offset; S13, achieving real-time data collection, by setting the data collection module to synchronize data at a fixed frequency f, where the value of f is determined according to the data type and device characteristics; S14, data verification and quality control, verify the validity of the received data, eliminate errors and abnormal data, and use the verification algorithm V(d i ) to verify each data item, where d i Represents the i-th data item to be verified: Calculate the expected value E(d i ) and standard deviation σ(d i ): The validity range of the data item is L,U, where: L=E(d i )-k·σ(d i ); U=E(d i )+k·σ(d i ); k is a control parameter, which is determined according to actual needs and has a value range of 2 to 3; Verify data item d i Is it within the validity range? Among them, V(d i ) has an output of 1 for valid data and 0 for invalid data; S15. Use time series analysis techniques to preprocess continuous health data. t }, where d t Represents the data value at time t, and calculates the moving average MA of the time series t : Where n represents the size of the time window; Calculate the difference Δd of the time series t ; Δd t =d t -d t-1 ; Apply exponential smoothing to the time series after difference and calculate the smoothed data value S t : S t =αd t +(1-α)S t-1 ; Among them, α is the smoothing coefficient, and its value range is 0<α≤1; Apply the autoregressive integrated moving average model for time series modeling: ARIMA(p,d,q); Where p is the order of the autoregressive term, d is the number of differences, and q is the order of the moving average term: Among them, φ i represents the autoregressive coefficient, θ j represents the moving average coefficient, ∈ t represents the white noise term.

3. A data processing method for life-cycle resident health records according to claim 2, characterized in that: The S2 specifically includes: S21, convert the format of the collected health data and perform denoising processing, and use the Gaussian filtering algorithm to denoise the data. The original data sequence is {d i }, the filtered data sequence is {d′ i } Among them, σ is the standard deviation, k is the window size of the filter; S22: The preprocessed data is partitioned according to the data type and timestamp and stored in a distributed database, with the partition key P k A combination of data type T and timestamp t: P k =(T,t); S23. Use the cloud storage system to regularly back up the data in the distributed database and realize data disaster recovery in different locations. The number of copies of the backup data is n. b , the backup frequency is f b ; S24. Establish multidimensional indexes for stored data to speed up data query, and use B-tree and R-tree structures to index structured data and geospatial data respectively: I k =(T,t); Among them, T represents the data type, t represents the timestamp, and the R-tree index is used for fast query of spatial data. The bounding box of spatial data is B(x min ,y min ,x max ,y max ), where x and y represent geographic coordinates; S25. Use LZ77 compression algorithm to compress the stored data: B′ i =LZ77(B i ); Among them, LZ77 represents the application of data compression algorithm, and the original data block is {B i }, the compressed data block is {B′ i }; S26. Use a role-based access control model to set access control policies for stored data: Among them, user u can access the corresponding data only when he is granted access rights in the permission p corresponding to his role r. The role set is R, the permission set is P, and the user set is U; S27. Monitor the health status of the distributed database and cloud storage system, use the heartbeat mechanism to detect the status of the storage node, and the heartbeat interval is Δt. If the heartbeat signal is not received for n consecutive times, the node is determined to be failed and fault recovery is performed.

4. A data processing method for life-cycle resident health records according to claim 3, characterized in that: The S3 specifically includes: S31. Clean the collected health data and use the median filter algorithm to remove outliers, noise and outliers. The original data set is D = {d1, d2, ..., d n }, the cleaned data set is D ′ , where d i For the i-th data item: d′ i =median{d i-k ,…,d i ,…,d i+k }; Where k is the filter window size, d′ i is the data item after cleaning; S32. De-duplicate the data to avoid duplicate records. The data set after de-duplication is D ″ , use the hash algorithm to calculate the unique identifier h(d i ), if the hash value of a data item already exists in the data set, the data item is considered a duplicate and removed: h(d i )=hash(d i ); Among them, h(d i ) is the data item d i The hash value of S33, standardize the data from different sources, the original data item is D i , the standardized data item is S(D i ): Among them, μ is the mean of the data item, σ is the standard deviation of the data item; S34. Calibrate the timestamps of data from different sources. The original timestamp is T i , the calibrated timestamp is T si : T si =T i +ΔT; Wherein, ΔT is the correction value of the time offset; S35. Use the ETL process to integrate the cleaned, deduplicated and standardized data. The extracted data set is E, the converted data set is T, and the loaded data set is L. The ETL process is expressed as: E→T→L; Among them, E represents the data extracted from the source system, T represents the transformed and standardized data, and L represents the data loaded into the target database; S36. Perform consistency verification on the integrated data using a data consistency verification algorithm C(D), where D is the data set to be verified: Among them, the output of C(D) is 1, which means that the data consistency verification passes, and the output is 0, which means that the data consistency verification fails.

5. A data processing method for life-cycle resident health records according to claim 4, characterized in that: The S4 specifically includes: S41. Extract key features from the preprocessed health data. The data set is D = {d1, d2, ..., d n }, the feature set is F = {f1,f2,…,f m }, where each feature f i By function F i (d) Generate: f i =F i (d); Where d represents the data item, F i represents the feature extraction function; S42, the extracted features are standardized, and the standardized feature set is S(F): Among them, μ f The feature f i The mean value, σ f The feature f i The standard deviation of S43, the standardized feature set and the corresponding label Y = {y1, y2, ..., y n }, construct the training data set T = {(S(f1), y1), (S(f2), y2), …, (S(f n ),y n )}, where y i is the data item d i Labels; S44. Use the long short-term memory network to train the training data set. The input sequence is X t ={x t1 ,x t2 ,…,x tn }, the output is the prediction result i t =σ(W i ·[h t-1 ,x t ]+b i ); f t =σ(W f ·[h t-1 ,x t ]+b f ); the t =σ(W o ·[h t-1 ,x t ]+b o ); C t =f t ·C t-1 +i t ·tanh(W C ·[h t-1 ,x t ]+b C ); h t =o t ·tanh(C t ); Among them, i t is the input gate, f t For the forget gate, t is the output gate, C t is the unit state, h t is the hidden state, W and b represent the weight matrix and bias vector respectively, σ is the sigmoid activation function, tanh is the hyperbolic tangent activation function, x t is the current input data, h t-1 is the hidden state of the previous time step; S45. Use the validation data set to validate the trained model. The validation data set is V = {(S(f1), y1), (S(f2), y2), …, (S(f n ),y n )}, the error between the predicted result and the true label is calculated by the loss function L: in, is the model prediction result, y i is the true label, and the loss function L is minimized by adjusting the model parameters W and b; S46. Use the trained model to predict the health trend of the newly input health data. The new input data is X ′ ={x1 ′ ,x2 ′ ,…,x t ′ }, the model output prediction result is Among them, X′ t represents the new input sequence, Indicates predicted health trends; S47. Generate personalized health recommendations based on prediction results, and analyze predicted health trends Provide specific health management suggestions and intervention measures. The suggestion set is G = {g1, g2, …, g n }, where g i According to the prediction results Generated suggestions.

6. A data processing method for life-cycle resident health records according to claim 5, characterized in that: The S5 specifically includes: S51. Use the AES encryption standard to encrypt the health data. The original data set is D = {d1, d2, ..., d n }, the encrypted data set is E(D): E(d i )=AES(d i ,K); Among them, d i is the i-th data item, K is the encryption key, E(d i ) is the encrypted data item; S52. Decrypt the encrypted data using AES decryption technology: D ′ (d i )=AES -1 (E(d i ),K); Among them, AES -1 is the AES decryption algorithm, E(d i ) is the encrypted data item, K is the decryption key, D ′ (d i ) is the decrypted data item; S53. Set access control policies for the stored data, with the role set as R, the permission set as P, and the user set as U: Among them, user u can access the corresponding data only if he is granted access rights in the permission p corresponding to his role r; S54. Perform differential privacy processing on the data. The original data set is D, and the data set after differential privacy processing is D ′ , add noise ∈ to protect privacy, the processing process is: D ′ =D+∈; Among them, the noise ∈ follows the Laplace distribution: Among them, Δf is the sensitivity, ∈ is the privacy parameter; S55. Record the user's access operation to the data. The log record is L = {l1, l2, ..., l n }, each log record contains the access user u, access time t, access data item d and operation type o: l i =(u,t,d,o); Among them, l i is the i-th log record.

7. A data processing method for life-cycle resident health records according to claim 6, characterized in that: The S6 specifically includes: S61. Integrate mobile health applications and smart wearable devices into the health record management system, collect data through standardized interfaces, and the collected health data is D m ={d m1 ,d m2 ,…,d mn } and D w ={d w1 ,d w2 ,…,d wn }, where D m represents the data collected by mobile applications, D w Represents the data collected by smart wearable devices; S62. Use IoT technology to synchronize data between devices, including data transmission and timestamp alignment: D s =D m ∪D w ; Timestamp alignment: T si =T i +ΔT; Among them, T si is the timestamp after alignment, T i is the original timestamp, ΔT is the time offset; S63, the integrated data set D s To analyze and realize real-time monitoring of residents’ health status, the health monitoring index is H = {h1,h2,…,h k }, through the monitoring algorithm M(H,D s )Calculate health status: H i =M(H,D s ); Among them, H i is the calculated i-th health state; S64, transmit health data to telemedicine platform to support doctors to make remote diagnosis and suggestions; S65, generating a personalized health management plan P based on the health monitoring index H and the diagnosis result R of the telemedicine platform; S66. Push the personalized health management plan to the user through the mobile application. Each health suggestion corresponds to a management plan and the health management plan is adjusted according to user feedback.

Images