Data synchronization method and system based on medical consumables

Abstract

The invention discloses a data synchronization method and system based on medical consumables, and relates to the technical field of data synchronization, and the method comprises the steps: collecting monitoring target list data of medical equipment, carrying out the arrangement and structuring of the sampled data, forming an independent data group, extracting the collected data of different sensor data items, and forming an organization matrix; and carrying out linear coding on the organization matrix by using a random linear network coding (RLNC) method, generating a redundant coding packet data matrix by using matrix multiplication, and carrying out homomorphic encryption on coding packet data. According to the method, data protection and calculation operation are seamlessly combined through homomorphic encryption, encryption monitoring data can be directly processed, and encryption potential features are generated through combination of time random features and mapping hidden features, so that the hidden space features can express time dependence features and global features of random coding at the same time; and the multi-dimensional characteristic fusion capability in a medical consumable monitoring scene is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data synchronization, and in particular to a data synchronization method and system based on medical consumables. BACKGROUND

[0002] With the continuous development of medical industry technology, more and more medical devices are equipped with multi-dimensional sensors for collecting data of their operating parameters and consumable states. In order to effectively monitor the use of medical device consumables, medical institutions are increasingly relying on data synchronization and analysis technology to ensure that real-time and reliable information is obtained, thereby assisting medical decision-making, improving resource management efficiency, and reducing the risk of device consumable shortages.

[0003] However, traditional medical data collection and synchronization technology often faces problems such as insufficient data collection accuracy, real-time monitoring delay, insufficient data transmission reliability, and privacy leakage. In multi-dimensional high-frequency data collected by sensors, the transmission of different data packets may result in a lack of data consistency due to network packet loss, data delay, and other reasons, affecting the integrity and reliability of the data. The encoding and transmission methods commonly used often lack sufficient redundancy coding design, and cannot efficiently recover data when monitoring data packets are lost or partially damaged, thereby causing the consistency stream of monitoring data to be damaged or even interrupted, affecting the real-time nature of medical consumable monitoring. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a data synchronization method and system based on medical consumables to solve the problem of lack of data consistency in multi-dimensional high-frequency data collected by sensors, affecting the integrity and reliability of the data. The encoding and transmission methods commonly used often lack sufficient redundancy coding design, and cannot efficiently recover data when monitoring data packets are lost or partially damaged, thereby causing the consistency stream of monitoring data to be damaged or even interrupted, affecting the real-time nature of medical consumable monitoring.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a data synchronization method based on medical consumables, comprising:

[0008] The monitoring target list data of the medical equipment is collected, the collected data is structured, an independent data group is formed, the collection data of different sensor data items is extracted, an organization matrix is composed, a random linear network coding (RLNC) method is used to linearly encode the organization matrix, a matrix multiplication is used to generate a redundant coding package data matrix, homomorphic encryption is performed on the coding package data, a sliding window is constructed to determine the context features of the homomorphic encryption data, a linear combination vector is extracted from the random coding matrix, and data packet encapsulation is performed;

[0009] The contribution weight is calculated by the norm-driven normalization according to the linear combination vector, the time-dependent weight is calculated by the smoothness contribution degree according to the context features, the linear mapping matrix is constructed by combining the output time random features, the homomorphic encryption data is extracted to an implicit feature space, the initial weight is calibrated based on the global statistical information of the redundant coding package data matrix, and the initial weight is updated based on the statistical distribution of the homomorphic encryption data as a correction coefficient value, and the encrypted potential features are comprehensively generated;

[0010] The distribution parameters of the encrypted potential features are extracted, the potential distribution is constructed, and the sampling is performed through the reparameterization technique, the reconstructed coding package is obtained through the decoder, the decoder parameters are optimized through the gradient descent algorithm, and the optimized reconstructed coding package is obtained.

[0011] The consistency data stream is output according to the optimized reconstructed coding package, and data synchronization and storage are performed.

[0012] As a preferred scheme of the medical consumable-based data synchronization method, the random linear network coding (RLNC) method is used to linearly encode the organization matrix, a matrix multiplication is used to generate redundant coding package data, and homomorphic encryption is performed on the coding package data, including,

[0013] Based on the collection data of the monitoring target list, the time stamp is unified, and the consumable monitoring target data items of the medical equipment are uniformly collected, the collected data is structured, the matching relationship between different sensor data items and different medical equipment consumables is determined, and an independent data group is formed;

[0014] According to the independent data group, the collection data of different sensor data items is extracted, and an organization matrix is composed.

[0015] The random linear network coding (RLNC) method is used to generate a random coding matrix based on the organization matrix, the independence of the random coding matrix is verified, the organization matrix is linearly encoded, and a matrix multiplication is used to generate a redundant coding package data matrix.

[0016] The Paillier encryption algorithm is used to perform homomorphic encryption on each coding package data of the redundant coding package data matrix.

[0017] The homomorphic encrypted encoded packet data is constructed into a sliding window according to time steps, and a time context feature is defined for each time step;

[0018] The linear combination vector is extracted for the random encoding matrix and normalized, and the context feature is combined to package the encrypted redundant encoded packet data.

[0019] As a preferred scheme of the medical consumable-based data synchronization method, the norm-driven normalization calculation contribution weight is calculated according to the linear combination vector, the time-dependent weight is extracted according to the smoothness contribution degree calculation of the context feature, the linear mapping matrix is constructed by combining the output time random feature, and the homomorphic encrypted data is extracted to the implicit feature space, including,

[0020] Based on the data packet packaging, the norm-driven normalization calculation contribution weight is calculated for the normalized linear combination vector;

[0021] The time-dependent weight is extracted according to the smoothness contribution degree calculation of the context feature;

[0022] Based on the time-dependent weight and the contribution weight, a time random feature extraction function is constructed based on the data packet packaging;

[0023] The homomorphic encrypted data is extracted to the implicit feature space using the linear mapping matrix, wherein the initial weight calibration is performed based on the global statistical information of the redundant encoded packet data matrix, the initial weight update is performed based on the statistical distribution of the homomorphic encrypted data as a correction coefficient value, the initial bias is obtained based on the global mean characteristic normalization of the homomorphic encrypted data, and the update of the initial weight and the initial bias is optimized and updated by the gradient descent method;

[0024] The encrypted latent feature is generated by integrating the time random feature and the mapped latent feature.

[0025] As a preferred scheme of the medical consumable-based data synchronization method, the distribution parameter of the encrypted latent feature is extracted, the latent distribution is constructed and sampled by the reparameterization trick, and the reconstructed encoded packet is obtained by the decoder, including,

[0026] The distribution parameter is extracted based on the encrypted latent feature, including a mean vector and a standard deviation vector, and the latent distribution is constructed;

[0027] The latent distribution is sampled by the reparameterization trick based on the variational autoencoder, and a noise term is introduced;

[0028] The decoder latent feature is used to sample back to the original data space to obtain the reconstructed encoded packet.

[0029] As a preferred scheme of the medical consumable-based data synchronization method, wherein: the decoder parameters are optimized by gradient descent using the mean square error algorithm to obtain the optimized reconstruction encoding package, comprising,

[0030] The reconstruction error is calculated using the mean square error (MSE), the weights and biases of the mean vector and the standard deviation vector, and the decoder parameters are optimized by gradient descent, and the iteration is stopped when the error calculated in the continuous iteration process no longer decreases significantly, and the optimized weights and biases of the mean vector and the standard deviation vector and the decoder parameters are output.

[0031] The latent distribution is re-determined according to the encrypted latent feature, and the latent feature sample is calculated to obtain the optimized reconstruction encoding package.

[0032] As a preferred scheme of the medical consumable-based data synchronization method, wherein: the consistent data stream is output according to the optimized reconstruction encoding package, and data synchronization and storage are performed, comprising,

[0033] The optimized reconstruction encoding package is integrated according to the time step according to the encoding package order of the redundant encoding package data matrix to form a complete consistent data stream.

[0034] Push to the server and the local storage module for data synchronization and storage.

[0035] As a preferred scheme of the medical consumable-based data synchronization method, wherein: the monitoring target list data of the medical device is collected, comprising,

[0036] Based on the hardware binding relationship between the medical device and the consumable, the device and consumable identification enumeration are performed, and the binding identification of different devices and consumables is recorded.

[0037] For different consumables of the medical device, the corresponding monitoring target data item of the consumable is extracted to form a monitoring target list.

[0038] In a second aspect, the present application provides a medical consumable-based data synchronization system, comprising,

[0039] Data acquisition module: responsible for collecting monitoring target list data of medical devices;

[0040] Organizational matrix construction module: convert the structured data into an organizational matrix, linearly encode the organizational matrix using the RLNC method, and generate a redundant encoding package data matrix through matrix multiplication;

[0041] Homomorphic encryption module: encrypt the redundant encoding package data, extract the context features of the homomorphic encryption data through a sliding window, extract the linear combination vector of the random encoding matrix, and calculate the contribution weight based on the norm normalization;

[0042] a time random feature generation module: combined with the time-dependent weight and the smoothness feature, a comprehensive time random feature is generated to comprehensively represent the context time characteristics;

[0043] an implicit feature space mapping module: the homomorphic encryption data is projected to an implicit feature space through a linear mapping matrix, the time random characteristics and the global statistical information are integrated to extract potential features;

[0044] a potential distribution construction module: the distribution parameters of the encrypted potential features are extracted, and the reparameterization technique is used to sample the feature description distribution;

[0045] a decoder module: the data is decoded according to the sampling results of the potential distribution to generate a reconstructed encoding package, and the mean square error algorithm gradient is used to optimize the decoder parameters;

[0046] a data stream generation module: based on the optimized reconstructed encoding package, a consistent data stream is output, the stream data is formed in the time step order, and the consistent data stream is pushed to the server and the local storage module.

[0047] In a third aspect, the present application provides a computer device comprising a memory and a processor, the memory storing a computer program, wherein the computer program is executed by the processor to implement any step of the medical consumable-based data synchronization method according to the first aspect of the present application.

[0048] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, wherein the computer program is executed by the processor to implement any step of the medical consumable-based data synchronization method according to the first aspect of the present application.

[0049] The present application has the following beneficial effects: the data protection is seamlessly combined with the computing operation through homomorphic encryption, the encrypted monitoring data can be directly processed, the encrypted potential features are generated through the combination of the time random features and the mapped implicit features, the implicit space features can express the time-dependent characteristics and the global characteristics of the random coding at the same time, the multi-dimensional feature fusion ability in the medical consumable monitoring scene is strengthened, the overall feature extraction can be realized in view of the common action of the heterogeneous structure, the consistent time sequence and the random disturbance of the data, the potential distribution is re-determined and the potential feature sampling is re-calculated based on the optimized parameters, the implicit feature sampling distribution is re-constructed by using the weight and bias of the optimized mean vector and standard deviation vector, and the reconstructed encoding package after decoding and the original redundant encoding package have more significant consistency. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0051] Fig. 1 A flowchart of a medical consumable-based data synchronization method in Embodiment 1.

[0052] Fig. 2 A structural diagram of a medical consumable-based data synchronization system in Embodiment 1. DETAILED DESCRIPTION

[0053] In order to make the above objectives, features and advantages of the present application more apparent and comprehensible, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0054] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, so the present application is not limited to the specific embodiments disclosed below.

[0055] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0056] Embodiment 1, with reference to Figs. 1-2 , as the first embodiment of the present application, the embodiment provides a medical consumable-based data synchronization method, comprising the following steps:

[0057] S1, collecting the monitoring target list data of the medical equipment, structuring the collected data, forming an independent data group, extracting the collected data of different sensor data items, composing an organization matrix, using a random linear network coding (RLNC) method to linearly encode the organization matrix, using matrix multiplication to generate a redundant encoding package data matrix, homomorphically encrypting the encoding package data, constructing a sliding window to determine the context features of the homomorphic encryption data and extracting a linear combination vector from the random encoding matrix, and performing data packet encapsulation;

[0058] Preferably, the monitoring target list data of the medical equipment comprises,

[0059] Based on the hardware binding relationship between medical devices and consumables, the device and consumable identification enumeration is performed, and the binding identification of different devices and consumables is recorded.

[0060] For different consumables of medical devices, the corresponding monitoring target data items of the consumables are extracted to form a monitoring target list.

[0061] Preferably, the random linear network coding (RLNC) method is used to linearly encode the organization matrix, generate redundant encoding package data using matrix multiplication, and homomorphically encrypt the encoding package data, including,

[0062] Based on the monitoring target list, the collected data is timestamped and collected for the consumable monitoring target data items of the medical devices. The sampled data is structured and structured to determine the matching relationship between different sensor data items and different medical device consumables, forming an independent data group, represented as:

[0063]

[0064] Among them, represents the kth medical device consumable, represents the mapping of the kth medical device consumable in the pre-defined monitoring target data item-medical device consumable mapping table, represents the different sensor data item set corresponding to the kth medical device consumable as an independent data group, represents the xth sensor data value in the kth medical device consumable, represents the yth sensor data value in the kth medical device consumable;

[0065] According to the independent data group, the collection data of different sensor data items is extracted to form an organization matrix, represented as:

[0066]

[0067] Among them, represents the organization matrix of the kth medical device consumable extracted from , each row corresponds to the monitoring sensor parameter data corresponding to the consumable, each column corresponds to the time step of the collection, , and represent the xth, yth and zth sensor data values of the kth medical device consumable in the total number of time steps T, respectively;

[0068] Using the random linear network coding (RLNC) method, a random encoding matrix is generated based on the organization matrix (based on Python), the independence of the random encoding matrix is verified, and the organization matrix is linearly encoded. Redundant encoding package data matrix is generated using matrix multiplication, represented as:

[0069]

[0070]

[0071] wherein, represents a redundant encoding package data matrix, represents a random encoding matrix of the kth type of medical device consumable, wherein , represents the number of redundant packages, must satisfy to ensure sufficient redundancy capability, represents the number of rows, i.e. the number of sensor data items, represents the i th encoding package in the redundant encoding package data matrix , represents the element in the i th row and j th column of the random encoding matrix , represents the time series data of the j th sensor parameter of the organization matrix ;

[0072] For each encoding package data of the redundant encoding package data matrix, homomorphic encryption is performed using the Paillier encryption algorithm, represented as:

[0073]

[0074] wherein, represents homomorphic encryption data, the calculation result retains the structural characteristics of the linear encoding package, and the property of homomorphic encryption allows the data to participate in linear operation in an encrypted state, thereby realizing seamless continuation of the function, represents that the elements in the i th row and j th column of the random encoding matrix are encrypted first;

[0075] The homomorphic encrypted encoding package data is constructed into a sliding window according to the time step, and a time context feature is defined for each time step, represented as:

[0076]

[0077] wherein, represents a set of context features of time step t, and the width of the time window can be defined as 2q+1, and respectively represent the encoding package data of which are subjected to homomorphic encryption operation at time steps t-q and t+q, wherein q represents the half-width of the sliding window, which determines the range of the time context;

[0078] The linear combination vector is extracted from the random coding matrix and normalized, and the context features are combined to package the encrypted redundant coding packet data, which is represented as:

[0079]

[0080]

[0081] wherein, represents the data packet packaging identifier of the i-th coding packet, represents the i-th row of the random coding matrix , which is used to describe the linear combination weight of the i-th coding packet, represents the normalized linear combination vector of the i-th coding packet.

[0082] By matching the uniform timestamp and the monitoring target data item, the time sequence consistency of the medical consumable data and the accurate association of the corresponding data items of the medical equipment are ensured, effectively avoiding the data invalidation problem caused by the asynchronous device acquisition time or the confusion of sensor data item classification. By generating an organization matrix, the structured relationship between different sensors and corresponding time steps is clearly defined, and the complex monitoring consumable data representation is converted into a clear matrix form. This structure enhances the operability of the medical consumable monitoring data, enabling stable processing of multi-dimensional data in the random linear coding process, and facilitating subsequent coding optimization based on matrix operations;

[0083] By random linear network coding (RLNC), the medical equipment sensor data items in the organization matrix are converted into a random coding matrix. After generating redundant coding packet data using matrix multiplication, the recovery capability of lost packet data is significantly improved. By using the Paillier encryption algorithm to homomorphically encrypt the redundant coding packet data, the data obtains complete privacy protection during network transmission while retaining the structural characteristics and homomorphism of linear coding packets. This allows the encrypted data to still participate in linear operations such as matrix multiplication, ensuring the seamless continuation of data functions. At the same time, homomorphic encryption can still maintain high computational efficiency in high-dimensional consumable data, effectively combining security and computational efficiency;

[0084] By constructing a sliding window and defining time context features, the local correlation of redundant coding packets at different time steps is obtained. The key information support for the dynamic changes of medical consumable monitoring is provided by combining the feature data within the time range. The time context operation of the sliding window enhances the capture of data time-dependent characteristics while avoiding the limitations of relying solely on static monitoring data for feature analysis;

[0085] The linear combination vector is extracted by the random coding matrix for normalization processing, combined with the context feature to construct the redundant data packet encapsulation in the encrypted state, unify the data packet identification and the final characteristics after processing, so that the monitoring data packets of different medical device consumables have clear characteristic association, and it is convenient for efficient analysis during network transmission, data storage and subsequent decoding. In addition to the combination of randomness and redundancy characteristics, the introduction of context characteristics realizes the multi-dimensional adaptability of data packet definition

[0086] The data protection is seamlessly combined with the computing operation by Paillier homomorphic encryption, which can directly process encrypted monitoring data, and the RLNC coding redundancy enhances the anti-packet loss property, and a high security and high robustness synchronous process suitable for medical scene is constructed.

[0087] The randomness and security of the monitoring data are unified through the seamless connection between the construction of random coding and the homomorphic encryption process, and the data packet structure with unique identification is encapsulated, realizing efficient transmission and rapid recovery ability of data under limited bandwidth.

[0088] S2, according to the norm-driven normalization calculation contribution weight of linear combination vector, according to the calculation of time-dependent weight of smoothness contribution degree of context feature, combined with output time random feature, construct linear mapping matrix, extract homomorphic encryption data to implicit feature space, based on the global statistical information of redundant coding packet data matrix, carry out initial weight calibration, and based on the statistical distribution of homomorphic encryption data as correction coefficient value, carry out initial weight update, comprehensive generate encrypted potential feature;

[0089] Preferably, according to the norm-driven normalization calculation contribution weight of linear combination vector, according to the calculation of time-dependent weight of smoothness contribution degree of context feature, combined with output time random feature, construct linear mapping matrix, extract homomorphic encryption data to implicit feature space, including,

[0090] Based on the data packet encapsulation, the norm-driven normalization calculation contribution weight of normalized linear combination vector is calculated, the contribution intensity of random coding characteristics to the overall feature is described, and is expressed as:

[0091]

[0092] Among them, Contribution weight, describes the contribution intensity of random coding characteristics to the overall feature, the higher the value, the stronger the randomness dependence, , Indicates the ℓ1 norm calculation, which is used to measure the average influence of absolute weight in random coding;

[0093] The calculation of the smoothness contribution degree of the context feature is performed, and the time-dependent weight of the time characteristic extraction is extracted, which is represented as:

[0094]

[0095] wherein, represents the time-dependent weight, represents the context feature value at the s-th time point, represents the variance function;

[0096] Based on the time-dependent weight and the contribution weight, a time random feature extraction function is constructed based on the data packet encapsulation, which is represented as:

[0097]

[0098] wherein, represents the generation of specific time random features combined with random coding and time characteristics, represents the context feature homomorphic encryption data mean at different times;

[0099] The homomorphic encryption data is extracted to the implicit feature space using the linear mapping matrix, wherein the initial weight calibration is performed based on the global statistical information of the redundant coding packet data matrix, the initial weight update is performed based on the statistical distribution of the homomorphic encryption data as the correction coefficient value, the initial bias is obtained based on the global mean characteristic normalization of the homomorphic encryption data, and the gradient descent method is used to update and optimize the updated initial weight and initial bias, which is represented as:

[0100]

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107] wherein, represents the mapping function of the implicit feature, represents the weight matrix of the updated mapping layer, represents the transpose of represents the bias term, represents the initial mapping layer weight, ​represents the initial mapping layer weight of the i-th encoding package and time step j, represents the optimized mapping layer weight, with a dimension of , d represents the dimension of the compressed hidden feature representation, which can be determined according to , represents the compression ratio, which is determined based on experience, represents the optimization calculation loss, i.e., the calculation difference between the target hidden feature and the function output, represents the learning rate, Q represents the total time step, represents the data of the i-th encoding package and time step t in the redundant encoding package data matrix, represents the correction coefficient matrix, represents the correction coefficient value of the i-th encoding package and time step j in the correction coefficient matrix, represents the homomorphic encryption data value of the pair at the j-th time point, represents the initial bias of the homomorphic encryption data value of the pair , represents the optimized bias, Y represents the total number;

[0108] The integrated time random feature and the mapped latent feature generate an encrypted latent feature, represented as:

[0109]

[0110] wherein, represents the encrypted latent feature of the pair homomorphic encryption data value.

[0111] The contribution weight is calculated based on the normalization of the ℓ1 norm through the random encoding characteristic, so that the randomness contribution in each encoding package is explicitly quantified, and the specific influence of the random combination in different encoding packages on the overall feature can be exhibited. Through the time-dependent weight calculation of the context feature, the characteristic smoothness or mutation intensity of each time period in the time series is embodied, thereby quantifying the sensitivity of the time characteristics to the device behavior changes in the consumable monitoring data. The time-dependent weight combines the difference and change degree of the time series, providing explicit support for comprehensively mastering the dynamic changes of the time series trend in the consumable monitoring process, especially for capturing the periodicity or mutability of the device state.

[0112] By combining the time-dependent weight and the contribution weight, a specific time random feature is generated, which contains both the spatial dimension (contribution of random encoding) and emphasizes the time dimension (dynamic nature of the time-dependent weight). The combination of the two dimensions reflects the synergistic relationship between different sensor data in the medical consumable monitoring process and its dynamic changes over time.

[0113] The linear mapping weight initialization and correction update based on the global statistical information of the redundant encoding package and the statistical distribution of the encrypted data makes the initial mapping weight parameter more consistent with the actual distribution in the process of mapping the homomorphic encryption data to the implicit feature space. The sensitivity of the model is dynamically adjusted through the correction coefficient value, which solves the feature extraction deviation problem in the implicit space caused by distribution deviation, thereby improving the accuracy and interpretability of the mapping process.

[0114] The global mean normalization of the initial bias value, combined with the optimization update of the bias, enhances the output balance of the mapping layer, ensures that the mapping result can be kept within a reasonable range and can be dynamically optimized according to the implicit feature distribution. The combination of time random features and mapping implicit features generates encrypted latent features, so that the implicit space features can express both time-dependent characteristics and global characteristics of random coding, and the multi-dimensional feature fusion capability in the medical consumable monitoring scene is strengthened. It can realize comprehensive feature extraction for the combined action of heterogeneous structure, consistent time series and random disturbance of data.

[0115] S3, extract the distribution parameters of the encrypted latent features, construct the latent distribution and sample through the reparameterization trick, get the reconstructed encoding package through the decoder, and optimize the decoder parameters through gradient descent using the mean square error algorithm to get the optimized reconstructed encoding package.

[0116] Preferably, the distribution parameters of the encrypted latent features are extracted, the latent distribution is constructed and sampled through the reparameterization trick, and the reconstructed encoding package is obtained through the decoder, including,

[0117] The distribution parameters of the encrypted latent features are extracted, the latent distribution is constructed and sampled through the reparameterization trick, and the reconstructed encoding package is obtained through the decoder, including,

[0118]

[0119]

[0120]

[0121] wherein, represents the latent distribution of the encrypted latent features, represents the mean vector of the latent distribution and the standard deviation vector , and represent the weight and bias of the mean vector, and represent the weight and bias of the standard deviation vector, wherein the dimension of the encrypted latent features is , and the dimension of the weight of the mean vector and the weight of the standard deviation vector is v represents the description ability of the latent feature implicit distribution, which can be determined according to The dimensions of the bias of the mean vector and the bias of the standard deviation vector are ;

[0122] Based on the variational autoencoder, the latent feature sampling is obtained from the latent distribution by the reparameterization technique, and a noise term is introduced, which is represented as:

[0123]

[0124] wherein, represents the latent feature sampling, represents the noise term, represents the scaling and offset of the sampling result of the implicit distribution to obtain the actual feature;

[0125] The decoder is used to map the latent feature sampling back to the original data space to obtain the reconstructed encoding package.

[0126] By constructing the latent distribution of the encrypted latent feature, dynamically learning the mean and standard deviation weights combined with the bias, the distribution modeling has higher description ability, especially capable of capturing the implicit feature distribution rule of the medical consumable monitoring data. Through the reparameterization technique of the variational autoencoder, random feature sampling of the latent distribution is realized, and a noise term of Gaussian distribution is introduced in the model, so that the sampled feature not only retains the global characteristic distribution, but also has a certain random flexibility.

[0127] By using the decoder to map the latent feature sampling back to the original data space, a high-quality reconstructed encoding package is generated. The reconstruction process combines the encoding strategy and the characteristics of homomorphic encryption, so that the reconstructed data package retains the structural characteristics of the original data and can be further restored to the real monitoring data through decryption. The decoding process is particularly significant for packet recovery and noise correction, which can improve the integrity and usability of the medical consumable monitoring data, thereby ultimately achieving the efficient, safe and accurate goal of "data synchronization based on medical consumables" as expected by the scheme.

[0128] Further, the mean square error algorithm is used to optimize the decoder parameters and obtain the optimized reconstructed encoding package, including,

[0129] The mean square error (MSE) is used to calculate the reconstruction error, and the weights and biases of the mean vector and the standard deviation vector and the decoder parameters are optimized by gradient descent. In the continuous iteration process, if the error calculated is no longer significantly decreased, the iteration is stopped and the optimized weights and biases of the mean vector and the standard deviation vector and the decoder parameters are outputted.

[0130] According to the encrypted latent feature, the latent distribution is re-determined and the latent feature sampling is calculated to obtain the optimized reconstructed encoding package.

[0131] The reconstruction error is calculated by mean square error (MSE), which can accurately measure the deviation between the decoding result after latent feature sampling and the actual encrypted encoding package, so that the optimization target has a specific mathematical basis, and the reconstruction error can be stably reduced in the form of gradient descent;

[0132] By re-determining the latent distribution and re-calculating the latent feature sampling based on the optimization parameters, the hidden feature sampling distribution is reconstructed using the weights and biases of the optimized mean vector and standard deviation vector, so that the new features sampled can be closer to the real distribution of the encrypted latent features, and the reconstructed encoding package after decoding and the original redundant encoding package have more significant consistency.

[0133] S4, outputting a consistency data stream according to the optimized reconstructed encoding package, and performing data synchronization and storage;

[0134] Preferably, the consistency data stream is output according to the optimized reconstructed encoding package, and data synchronization and storage are performed, including,

[0135] The optimized reconstructed encoding package is integrated according to the encoding package order of the redundant encoding package data matrix according to the time step, and a complete consistency data stream is formed;

[0136] Push to the server and the local storage module for data synchronization and storage.

[0137] The optimized reconstructed encoding package is integrated according to the time step, and a complete consistency data stream is formed, which solves the problem of inconsistent time steps caused by the diversity of device monitoring data sources. The data stream construction method based on time step integration can be compatible with the time sequence characteristics of redundant encoding package data, avoiding the characteristic fragmentation problem that may be caused by too many redundant relationships of original data, and providing global consistency guarantee for medical consumable monitoring stream data processing;

[0138] Through pushing to the server and the local storage module for data synchronization and storage, the uniformity and practicality of medical consumable data in the cloud and the local are ensured, and the security and reliability of data transmission and storage are further enhanced.

[0139] The embodiment also provides a data synchronization system based on medical consumables, including,

[0140] Data acquisition module: responsible for acquiring monitoring target list data of medical devices;

[0141] Organizational matrix construction module: converting the structured data into an organizational matrix, linearly encoding the organizational matrix using the RLNC method, and generating a redundant encoding package data matrix through matrix multiplication;

[0142] Homomorphic encryption module: encrypt the redundant coded packet data, extract the context features of the homomorphic encryption data through the sliding window, extract the linear combination vector of the random coding matrix, and calculate the contribution weight based on the norm normalization;

[0143] Time random feature generation module: combine the time-dependent weight and the smoothness feature to generate comprehensive time random features, and comprehensively represent the context time characteristics;

[0144] Implicit feature space mapping module: project the homomorphic encryption data into the implicit feature space through a linear mapping matrix, integrate the time random characteristics and the global statistical information to extract potential features;

[0145] Potential distribution construction module: extract the distribution parameters of the encrypted potential features, and sample the feature description distribution by using the reparameterization technique;

[0146] Decoder module: decode the data according to the sampling results of the potential distribution to generate a reconstructed coded packet, and use the mean square error algorithm to gradient optimize the decoder parameters;

[0147] Data stream generation module: output a consistent data stream based on the optimized reconstructed coded packet, form a streaming data in time step order, and push the consistent data stream to the server and the local storage module.

[0148] The embodiment also provides a computer device suitable for the case of the data synchronization method based on medical consumables, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the data synchronization method based on medical consumables proposed in the above embodiment.

[0149] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0150] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement a medical consumable-based data synchronization method as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0151] To sum up, the present application seamlessly combines data protection and computing operation through homomorphic encryption, can directly process encrypted monitoring data, generates encrypted potential features through the combination of time random features and mapping hidden features, so that the hidden space features can express time-dependent characteristics and globally random coded characteristics at the same time, strengthens the multi-dimensional feature fusion capability in the medical consumable monitoring scene, can realize comprehensive feature extraction in view of the combined action of heterogeneous structure, consistent time sequence and random disturbance of data, re-determines the potential distribution and re-calculates the potential feature sampling through the optimization parameter, re-constructs the hidden feature sampling distribution by using the weight and bias of the optimized mean vector and standard deviation vector, and guarantees that the reconstructed coded package after decoding and the original redundant coded package have more significant consistency.

[0152] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A data synchronization method based on medical consumables, characterized in that, include: Data on the monitoring target list of medical equipment is collected, the sampled data is organized and structured into independent data groups, the collected data of different sensor data items are extracted to form an organization matrix, the organization matrix is ​​linearly encoded using the Random Linear Network Coding (RLNC) method, redundant encoded packet data matrix is ​​generated using matrix multiplication, and the encoded packet data is homomorphically encrypted. A sliding window is constructed to determine the contextual features of the homomorphically encrypted data and a linear combination vector is extracted from the random encoding matrix for data packet encapsulation. The contribution weights are calculated by norm-driven normalization based on linear combination vectors, and time-dependent weights are extracted by calculating the smoothness contribution based on context features. Combined with the output time random features, a linear mapping matrix is ​​constructed to extract homomorphic encrypted data into the implicit feature space. Initial weight calibration is performed based on the global statistical information of the redundant coded packet data matrix, and initial weight updates are performed based on the statistical distribution of homomorphic encrypted data as the correction coefficient value. The encrypted latent features are generated in a comprehensive manner. Extract the distribution parameters of the latent encryption features, construct the latent distribution and sample it using the reparameterization technique, obtain the reconstructed encoded packet through the decoder, and use the mean squared error algorithm to optimize the decoder parameters through gradient descent to obtain the optimized reconstructed encoded packet; Based on the optimized refactored encoding package, a consistent data stream is output for data synchronization and storage.

2. The data synchronization method based on medical consumables as described in claim 1, characterized in that: The method of using Random Linear Network Coding (RLNC) to linearly encode the organization matrix, using matrix multiplication to generate redundant encoded packet data, and then homomorphically encrypting the encoded packet data includes: Based on the data collected from the list of monitoring targets, a unified timestamp is used to collect data on the target data items of medical equipment consumables. The sampled data is then organized and structured to determine the matching relationship between different sensor data items and different medical equipment consumables, forming independent data groups. Based on the independent data groups, the collected data from different sensor data items are extracted and organized into an organization matrix; The Random Linear Network Coding (RLNC) method is used to generate a random coding matrix based on the organization matrix, verify the independence of the random coding matrix, perform linear coding on the organization matrix, and use matrix multiplication to generate a redundant coding packet data matrix. For each coded packet data in the redundant coded packet data matrix, homomorphic encryption is performed using the Paillier encryption algorithm; A sliding window is constructed for the homomorphically encrypted encoded packet data according to time steps, and a time context feature is defined for each time step; Linear combination vectors are extracted from the random coding matrix and normalized. Combined with contextual features, the encrypted redundant coding packet data is then encapsulated into data packets.

3. The data synchronization method based on medical consumables as described in claim 2, characterized in that: The process includes: calculating contribution weights using norm-driven normalization based on linear combination vectors; extracting time-dependent weights by calculating smoothness contribution based on contextual features; constructing a linear mapping matrix by combining output temporal random features; and extracting homomorphically encrypted data into an implicit feature space. Based on data packet encapsulation, norm-driven normalization is used to calculate the contribution weights of normalized linear combination vectors. Calculate the smoothness contribution of context features and extract the time dependency weights of time features; Based on time-dependent weights and contribution weights, a time-random feature extraction function is constructed based on data packet encapsulation. The homomorphic encrypted data is extracted into the implicit feature space using a linear mapping matrix. Initial weight calibration is performed based on the global statistical information of the redundant coded packet data matrix, and the initial weight is updated based on the statistical distribution of the homomorphic encrypted data as the correction coefficient value. The initial bias is obtained by normalization based on the global mean characteristics of the homomorphic encrypted data, and the initial weight and initial bias are optimized and updated using the gradient descent method. By combining the random features of time with the latent features of the mapping, a latent cryptographic feature is generated.

4. The data synchronization method based on medical consumables as described in claim 3, characterized in that: The process involves extracting distribution parameters of latent encryption features, constructing a latent distribution, sampling using reparameterization techniques, and obtaining a reconstructed encoded packet via a decoder. Based on encrypted latent features, distribution parameters including mean vector and standard deviation vector are extracted, and a latent distribution is constructed; Variational autoencoders sample from the latent distribution using reparameterization techniques and introduce noise terms. The latent feature samples from the decoder are used to map back to the original data space to obtain the reconstructed encoded packet.

5. The data synchronization method based on medical consumables as described in claim 4, characterized in that: The mean squared error algorithm is used to optimize the decoder parameters through gradient descent and obtain the optimized reconstructed encoded packet. include, The reconstruction error is calculated using mean squared error (MSE). The weights and biases of the mean vector and standard deviation vector, as well as the decoder parameters, are optimized by gradient descent. If the calculated error no longer decreases significantly during continuous iteration, the iteration stops and the optimized weights and biases of the mean vector and standard deviation vector, as well as the decoder parameters, are output. The potential distribution is redefined based on the latent cryptographic features, and the latent feature sampling is calculated to obtain the optimized reconstructed encoding packet.

6. The data synchronization method based on medical consumables as described in claim 5, characterized in that: The process of outputting a consistent data stream based on the optimized reconstructed encoding packet, and performing data synchronization and storage, includes: The optimized reconstructed encoding packets are integrated step by step according to the encoding packet order of the redundant encoding packet data matrix to form a complete and consistent data stream; Data is pushed to the server and local storage module for synchronization and storage.

7. The data synchronization method based on medical consumables as described in claim 6, characterized in that: The data collected includes the list of monitoring targets for medical devices. Based on the hardware binding relationship between medical devices and consumables, the device and consumable identifiers are enumerated, and the binding identifiers of different devices and consumables are recorded. For different consumables of medical equipment, the corresponding monitoring target data items of the consumables are extracted and compiled into a monitoring target list.

8. A data synchronization system based on medical consumables, based on the data synchronization method based on medical consumables according to any one of claims 1 to 7, characterized in that: include, Data acquisition module: responsible for collecting data on the list of monitoring targets for medical equipment; Organization Matrix Construction Module: Transforms structured data into an organization matrix, uses the RLNC method to linearly encode the organization matrix, and generates a redundant encoded packet data matrix through matrix multiplication; Homomorphic encryption module: Encrypts redundant encoded packet data, extracts contextual features of homomorphically encrypted data through a sliding window, extracts linear combination vectors from random encoding matrices, and calculates contribution weights based on norm normalization; Temporal random feature generation module: Combines time-dependent weights and smoothness features to generate comprehensive temporal random features, fully representing the temporal characteristics of the context; Implicit Feature Space Mapping Module: Projects homomorphic encrypted data onto the implicit feature space through a linear mapping matrix, integrating temporal randomness and global statistical information to extract latent features; Latent distribution building module: Extracts the distribution parameters of encrypted latent features and uses reparameterization techniques to sample features to describe the distribution; Decoder module: Decodes the data based on the latent distribution sampling results to generate reconstructed encoded packets, and optimizes the decoder parameters using the mean square error algorithm gradient. Data stream generation module: Based on the optimized reconstructed encoding package, output a consistent data stream, construct streaming data in time step order, and push the consistent data stream to the server and local storage module.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the data synchronization method based on medical consumables as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the data synchronization method based on medical consumables as described in any one of claims 1 to 7.

Images