Electronic medical record authenticity identification method and system

Abstract

The invention provides an electronic medical record authenticity identification method and system, and relates to the technical field of data processing, and the method comprises the steps: obtaining a to-be-identified electronic medical record; an auto-encoder is constructed in combination with a gradient descent method based on manifold embedding, and the auto-encoder comprises an encoder and a decoder; performing feature retention type desensitization on the electronic medical record to be identified in combination with an encoder weight matrix of the auto-encoder; inputting the desensitized to-be-identified electronic medical record into an auto-encoder, and outputting a reconstructed to-be-identified electronic medical record; calculating a reconstruction error between the to-be-identified electronic medical record and the reconstructed to-be-identified electronic medical record; determining a reconstruction error discriminant value in combination with a desensitization process; under the condition that the reconstruction error exceeds the reconstruction error discriminant value, abnormal content in the to-be-identified electronic medical record is positioned in combination with the reconstruction error; and outputting the abnormal content and an abnormal prompt about the abnormal content, otherwise, outputting that the to-be-identified electronic medical record is normal. And the authenticity identification of the electronic medical record is accurately completed under the condition of protecting privacy.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method and system for authenticating the authenticity of electronic medical records. Background Art

[0002] An electronic medical record (EMR for short) refers to a system that records, stores, and manages patients' health information through electronic means. It includes patients' personal basic information, medical history, diagnosis results, treatment plans, test and examination data, etc., and can achieve the sharing, convenient query, and electronic management of medical information.

[0003] With the continuous development of medical informatization, electronic medical records play an increasingly important role in the medical process. The authenticity and integrity of electronic medical records are directly related to the treatment effect and safety of patients. If there are tampering, forgery, or omission of information in the electronic medical record, it may lead to misdiagnosis and mistreatment by doctors, and even medical accidents. For example, when an electronic medical record is transferred among different physician nodes, it is necessary to conduct pre-review to avoid abnormal modification (including internal garbled codes or malicious modification) of the electronic medical record, which may cause abnormalities in the electronic medical record transferred to the current physician. If the current physician determines the patient's condition based on the modified electronic medical record, it is extremely easy to misjudge the patient's condition and lead to medical accidents. Therefore, ensuring the authenticity of electronic medical records can not only improve the accuracy of medical decisions, but also protect patients' privacy and life safety, and prevent the occurrence of medical disputes and risks.

[0004] However, the existing authentication of the authenticity of electronic medical records relies too much on manual work, with a large workload and low efficiency. Moreover, manual authentication is prone to subjective decision-making errors, with insufficient accuracy, and it is difficult for traditional authentication methods to conduct authentication while protecting patients' privacy. Summary of the Invention

[0005] In order to solve the technical problems in the prior art that the authentication of the authenticity of electronic medical records relies too much on manual work, with a large workload and low efficiency, and manual authentication is prone to subjective decision-making errors, with insufficient accuracy, and it is difficult for traditional authentication methods to conduct authentication while protecting patients' privacy, the present invention provides a method and system for authenticating the authenticity of electronic medical records.

[0006] The technical solutions provided by the embodiments of the present invention are as follows:

[0007] First Aspect

[0008] A method for authenticating the authenticity of an electronic medical record provided by an embodiment of the present invention includes:

[0009] S1: Obtain the electronic medical record to be authenticated;

[0010] S2: Construct an autoencoder by combining the gradient descent method based on manifold embedding, where the autoencoder includes an encoder and a decoder;

[0011] S3: Desensitize the electronic medical record to be identified in a feature-preserving manner by combining the encoder weight matrix of the autoencoder;

[0012] S4: Input the desensitized electronic medical record to be identified into the autoencoder and output the reconstructed electronic medical record to be identified;

[0013] S5: Calculate the reconstruction error between the electronic medical record to be identified and the reconstructed electronic medical record to be identified;

[0014] S6: Determine the reconstruction error discrimination value by combining the desensitization process;

[0015] S7: In the case where the reconstruction error exceeds the reconstruction error discrimination value, locate the abnormal content in the electronic medical record to be identified by combining the reconstruction error, and enter step S8; otherwise, enter step S9;

[0016] S8: Output the abnormal content and an abnormal reminder regarding the abnormal content;

[0017] S9: Output that the electronic medical record to be identified is normal.

[0018] Second aspect

[0019] An electronic medical record authenticity identification system provided by an embodiment of the present invention includes:

[0020] A processor;

[0021] A memory, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor, the electronic medical record authenticity identification method as in the first aspect is implemented.

[0022] Third aspect

[0023] A computer-readable storage medium provided by an embodiment of the present invention, on which a computer program is stored. When the program is executed by the processor, the electronic medical record authenticity identification method as in the first aspect is implemented.

[0024] The beneficial effects brought by the technical solutions provided by the embodiments of the present invention at least include:

[0025] In the embodiments of the present invention, an autoencoder is constructed by combining the gradient descent method based on manifold embedding, effectively improving the accuracy of the autoencoder in reconstructing electronic medical records. The autoencoder can more accurately capture the latent structure of the electronic medical records, reduce the influence of local minima, ensure that the model can identify anomalies or tampering in the data, and provide more reliable authenticity identification of electronic medical records. By automatically verifying the electronic medical records through the autoencoder, the workload of manual review is reduced and the efficiency is improved. At the same time, combined with the feature-preserving desensitization technology, it can not only effectively protect patient privacy, but also avoid increasing the reconstruction error of subsequent autoencoder processing, thereby avoiding the invalidation of the autoencoder-based authenticity identification of electronic medical records due to desensitization, and further ensuring the accuracy of the authenticity identification of electronic medical records. In addition, this solution can accurately locate abnormal content in electronic medical records by combining reconstruction errors, reducing the probability of medical accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0027] Figure 1 It is a schematic flowchart of a method for authenticating the authenticity of electronic medical records provided by an embodiment of the present invention;

[0028] Figure 2 It is a schematic structural diagram of a system for authenticating the authenticity of electronic medical records provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] The following will describe the technical solutions in the present invention with reference to the drawings.

[0030] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "example" in the present invention should not be construed as more preferred or more advantageous than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0031] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the drawings and specific embodiments.

[0032] Refer to the attached specification Figure 1 , which shows a schematic flowchart of a method for authenticating the authenticity of electronic medical records provided by an embodiment of the present invention.

[0033] An embodiment of the present invention provides a method for authenticating the authenticity of electronic medical records. This method can be implemented by an electronic medical record authenticity authentication device, which can be a terminal or a server. The processing flow of the electronic medical record authenticity authentication method may include the following steps:

[0034] S1: Obtain the electronic medical record to be authenticated.

[0035] S2: Construct an autoencoder by combining the gradient descent method based on manifold embedding.

[0036] Among them, the autoencoder includes an encoder and a decoder.

[0037] Among them, manifold embedding refers to mapping high-dimensional data to a low-dimensional space through mathematical methods, so that the internal structure and similarity of the data are preserved. In machine learning, manifold embedding technology enables the model to better capture the complexity and hidden patterns of the data by learning the nonlinear relationships between data points. In this solution, the main role of manifold embedding is to help optimize the training process of the autoencoder by mapping hyperparameters to a low-dimensional space. Specifically, manifold embedding is used to optimize hyperparameters during the optimization process, so that the autoencoder can better capture the internal structure and patterns of the data. Through manifold embedding, the optimization path can transition smoothly, avoiding falling into local optimal solutions, thereby improving the reconstruction effect of the autoencoder and enabling it to reconstruct and detect anomalies in electronic medical record data more accurately.

[0038] Among them, the autoencoder is an unsupervised learning model, usually used for data dimensionality reduction, feature learning, and anomaly detection. It consists of two main parts: an encoder and a decoder. The encoder is responsible for compressing the input data (such as electronic medical records) into a representation in the latent space (usually low-dimensional). The decoder is responsible for restoring this representation in the latent space to the reconstruction of the input data. The encoder weight matrix is the parameter matrix inside the encoder, which plays a key role in the transformation of the input data and determines how the input data is mapped to the latent space. The constructed autoencoder is trained based on publicly available electronic medical records.

[0039] In a possible implementation manner, S2 specifically includes:

[0040] S201: Obtain publicly available historical electronic medical records.

[0041] S202: Reconstruct the historical electronic medical records through the autoencoder to obtain the reconstructed historical electronic medical records.

[0042] S203: Optimize the hyperparameters of the autoencoder through the gradient descent method based on manifold embedding until the reconstruction error between the historical electronic medical records and the reconstructed historical electronic medical records is less than the preset reconstruction error, where the hyperparameters include the learning rate, the number of encoder layers, the number of decoder layers, and the dropout rate.

[0043] In a possible implementation, S203 specifically includes:

[0044] S203A: Map each hyperparameter to a Riemannian manifold.

[0045] The specific formula for the mapping process is:

[0046]

[0047] where μ and τ represent continuous hyperparameters and discrete hyperparameters respectively, represents the result of mapping μ and τ to the Riemannian manifold and e represents the natural constant.

[0048] Among them, hyperparameters refer to the parameters that need to be manually set in machine learning algorithms, usually determined before model training, and they affect the learning process and performance of the model. For example, the learning rate, the number of layers, etc. A Riemannian manifold is a mathematical concept, referring to a smooth surface with an inner product structure, used to describe the complex data space structure. In the optimization process, the manifold space can provide a structure more suitable for complex data distributions than the traditional Euclidean space. Continuous hyperparameters are those parameters that can vary continuously within the real number range, such as the learning rate, the dropout rate, etc. Discrete hyperparameters are those parameters that can only take a finite number of discrete values, such as the number of encoder layers, the number of decoder layers, etc. The hyperparameters are mapped to the Riemannian manifold in order to optimize in a data space with a more geometric structure, thereby improving the optimization efficiency, avoiding local optimal solutions, and better capturing the complex patterns of the data.

[0049] S203B: Construct a homotopy potential field on the Riemannian manifold for smoothing the hyperparameter optimization path.

[0050] The formula form of the homotopy potential field is specifically:

[0051]

[0052] where t ∈ [0, 1] represents the homotopy parameter that controls the smooth transition of the homotopy potential field, L rec represents the reconstruction error, μ represents the Riemannian manifold measure, represents the reconstruction error gradient with respect to the hyperparameter θ, represents the square of the reconstruction error gradient norm, Let \(M\) denote a Riemannian manifold, and \(H(\theta, t)\) denote a homotopy potential field with respect to the hyperparameter \(\theta\) and the homotopy parameter \(t\).

[0053] Among them, the hyperparameter optimization path refers to the trajectory followed during the optimization process when adjusting the hyperparameters. The homotopy potential field is an introduced energy function that is used to smoothly transition the optimization path. By adjusting the homotopy parameter, the transition can be made smoother during the optimization process, avoiding over-reliance on the reconstruction error and incorporating the geometric features of the manifold. The homotopy parameter is a parameter that controls the smooth transition of the function. In the homotopy potential field, \(t\) represents the degree of transition from the initial state to the target state during the optimization process, helping the optimization process to gradually approach the target smoothly. The Riemannian manifold measure refers to a measurement method on the Riemannian manifold, which defines the local structure of each point on the manifold, thus affecting the path selection during the optimization process. The reconstruction error gradient refers to the rate of change of the loss function with respect to the hyperparameters. During the optimization process, the gradient reflects the sensitivity of the error to the change of the hyperparameters and helps the optimization algorithm find the direction of the steepest descent. By constructing a homotopy potential field on the Riemannian manifold, the optimization path can be smoothed, ensuring that the optimization process not only focuses on the reconstruction error but also effectively considers the geometric features of the data. This method can avoid local optimal solutions, improve the optimization efficiency, ensure that the autoencoder can more accurately process the electronic medical record data, and ultimately improve the accuracy and reliability of the authenticity identification of electronic medical records.

[0054] Specifically, during the optimization process, the construction of the homotopy potential field is to smoothly transition the optimization path, so that the optimization not only focuses on the reconstruction loss but also considers the changes in the geometric features of the manifold. By introducing the manifold measure, the connectivity of the optimization path is ensured. The manifold measure provides a global perspective here, enabling the optimization to be carried out in the entire manifold space rather than just in local regions. Thus, the optimization efficiency is improved and the problem of local minima is avoided.

[0055] S203C: Calculate the corrected gradient direction for updating the hyperparameters along the geodesic line of the Riemannian manifold.

[0056] The execution formula of S203C is specifically:

[0057]

[0058] Among them, denotes the gradient of the homotopy potential field with respect to \(\theta\), \(A\) denotes the gauge field used to generate the curl, \(\text{curl}(A)\) denotes the curl of the gauge field \(A\), \(\alpha\) denotes an adjustable constant used to control the influence degree of the vorticity term \(\text{curl}(A)\), denotes projecting onto the tangent space of the Riemannian manifold at the point \(\theta\), denotes the corrected gradient direction.

[0059] Optionally, the adjustable constant can be set to 0.5.

[0060] Among them, the gauge field refers to a field in mathematics that is used to generate curl. It provides a vector field through the rotational part of which perturbations can be introduced to prevent instability or extreme value problems during the optimization process. Curl is a property of a vector field in mathematics, representing the local rotation degree of a field. During the optimization process, curl is used to generate perturbations to avoid getting stuck in local extrema prematurely. The tangent space refers to the set of all possible local directions at a certain point on a Riemannian manifold. The tangent space provides a space for linear approximation, allowing operations on the manifold near that point and ensuring that the optimization process proceeds along appropriate directions. The modified gradient direction is adjusted on the original gradient direction, adding a vortex term to avoid local extrema and ensuring through projection operations that the update direction remains within the tangent space of the manifold, thus avoiding inappropriate parameter adjustments.

[0061] By calculating the modified gradient direction along the geodesic line of the Riemannian manifold and combining the vortex term, local extrema and inappropriate optimization paths can be effectively avoided, ensuring the stability and efficiency of the optimization process. This method makes the autoencoder more robust during the training process, enabling it to more accurately identify abnormal content in electronic medical records and improving the accuracy and reliability of authenticity identification.

[0062] Among them, the vortex term is used to avoid extreme values. In addition, to ensure that the optimization always occurs on the manifold, the update direction needs to be projected onto the tangent space of the manifold through projection operations. This means that during the update process, all parameter adjustments must be kept within the local linear structure of the manifold, thus avoiding inappropriate update directions.

[0063] S203D: Update each hyperparameter according to the modified gradient direction and in combination with the Brouwer fixed point theorem to optimize the hyperparameters of the autoencoder.

[0064] The execution formula of S203D is specifically:

[0065]

[0066] Among them, θ n and θ n+1 represent the hyperparameters in the n-th iteration and the (n + 1)-th iteration respectively, γ represents the learning rate that controls the update step size in adjacent iteration processes, Γ represents the volume-preserving automorphism mapping of the Riemannian manifold that ensures the volume of the Riemannian manifold remains unchanged during the transformation, J Γ represents the Jacobian matrix of Γ, d represents the dimension of the Riemannian manifold space, and det(J Γ ) represents the determinant of H Γ .

[0067] Among them, the volume-preserving automorphism mapping is a mapping method that can ensure that the volume of the manifold remains unchanged during the manifold transformation. That is to say, after this mapping, the local structure of the manifold remains consistent, avoiding geometric distortion caused by the transformation, so as to maintain the overall stability of the data during the optimization process. The dimension of the Riemannian manifold space refers to the dimension of the local linear structure of the manifold, which describes the complexity or degree of freedom of the manifold. The dimension determines the number of directions in which points on the manifold can move freely around each point. The Brouwer fixed-point theorem is a classical theorem in mathematics, which describes that under certain conditions, any continuous function always has a fixed point.

[0068] Among them, det(J Γ ) represents the scaling factor of the automorphism mapping, and adjusts the update amplitude of each step through the determinant. The determinant of the automorphism mapping reflects the spatial deformation after applying this mapping on the manifold. By introducing the volume-preserving automorphism mapping and the Brouwer fixed-point theorem, it can be ensured that the update of hyperparameters in the optimization process will not cause damage to the local structure of the manifold, thus avoiding local extrema and fracture problems. This method effectively improves the stability of the autoencoder optimization, ensures smooth convergence during the training process, and thus improves the accuracy and reliability of the authenticity identification of electronic medical records. Through the adjustment of the Jacobian matrix, the optimization process can ensure stable convergence at each step.

[0069] It should be noted that the reconstruction error in step S203 is the reconstruction error during the training of the autoencoder by historical electronic medical records. By optimizing the hyperparameters of the autoencoder through the gradient descent method based on manifold embedding and combining the geometric structure of the Riemannian manifold, it can more effectively avoid local optimal solutions, smooth the optimization path, and improve the optimization efficiency. Introducing the homotopy potential field and modifying the gradient direction helps to avoid inappropriate optimization directions and ensure the stability of hyperparameter updates. By combining the Brouwer fixed-point theorem, it further ensures the stability of the local structure of the manifold during the optimization process, thereby improving the performance of the autoencoder and making it more accurate and reliable in the authenticity identification of electronic medical records.

[0070] S204: Set the autoencoder based on the optimized hyperparameters to complete the construction of the autoencoder.

[0071] It should be noted that those skilled in the art can set the size of the preset reconstruction error according to actual needs, and the present invention does not make any limitations here.

[0072] Specifically, the process of training the autoencoder is to learn how to reconstruct the input data through a large amount of normal electronic medical record data. During training, the model learns the latent structure and patterns (such as clinical features) of the data by minimizing the reconstruction error. The goal of the training process is to enable the autoencoder to learn to recover the original data from the latent space representation. It focuses on the common structures in the data, such as clinical features like disease types and treatment plans, so as to be able to effectively reconstruct the data subsequently. The reconstruction process of normal electronic medical records is captured by the autoencoder.

[0073] More specifically, the autoencoder uses a large amount of normal data for learning during the training process. Normal data refers to those data points that conform to a certain expected pattern. By minimizing the reconstruction error, the model learns how to effectively represent normal data. The autoencoder continuously adjusts its weights so that it can accurately reconstruct the input data. For normal data, it can better recover the original input and maintain a small reconstruction error. During the detection process: The reason why the autoencoder can effectively perform anomaly detection is that it learns the internal pattern of the input data by "reconstructing" the data. Since the autoencoder is only trained with normal data, it can capture the common features in the normal data and try to restore the data through "reconstruction". Therefore, when abnormal data that does not conform to these patterns appears, the model cannot effectively reconstruct it, resulting in an increase in the reconstruction error.

[0074] S3: Combine the encoder weight matrix of the autoencoder to desensitize the electronic medical record to be identified while retaining its features.

[0075] Among them, feature-retaining desensitization refers to processing sensitive information (such as patient names, medical record numbers, etc.) while protecting privacy, so that these sensitive information is desensitized while retaining other clinical features in the data. This method ensures that the desensitization operation does not damage the ability of the autoencoder to process other clinical features. That is, through this desensitization method, both patient privacy can be effectively protected and the increase in the reconstruction error during subsequent autoencoder processing can be avoided, thereby avoiding the invalidation of the authenticity identification of electronic medical records based on the autoencoder due to desensitization. Find the noise direction irrelevant to the reconstruction of the autoencoder so that the desensitization operation does not damage the features other than the sensitive features.

[0076] It should be noted that feature-retaining desensitization through the encoder weight matrix of the autoencoder not only protects patient privacy but also avoids the impact of the desensitization operation on subsequent autoencoder reconstruction. By identifying and utilizing the noise direction irrelevant to the autoencoder reconstruction, it is ensured that the desensitization process does not damage the clinical features and guarantees the accuracy and effectiveness of the authenticity identification process.

[0077] In a possible implementation manner, S3 specifically includes:

[0078] S301: Identify sensitive fields in the electronic medical record to be identified through a pre-trained model, where the sensitive fields include patient name, disease type, and identity information.

[0079] Among them, the pre-trained model includes a pre-trained NER model and a pre-trained BERT model.

[0080] The execution formula of S301 is specifically:

[0081]

[0082] Among them, P represents the set of sensitive fields, (i, j) represents the position of the i-th row and j-th column in the electronic medical record to be identified, and M ij represents the mask matrix that describes whether the field corresponding to the position of the i-th row and j-th column in the electronic medical record to be identified belongs to P.

[0083] It should be noted that the main purpose of identifying sensitive fields is to protect patient privacy, ensure that sensitive information (such as name, disease type, identity information) is not leaked, and at the same time ensure that when data processing and analysis are carried out, sensitive data is properly desensitized and protected, so as to avoid privacy risks and meet the requirements of relevant laws and regulations.

[0084] S302: Perform singular value decomposition on the weight matrix of the encoding layer of the autoencoder through the singular value decomposition algorithm, and extract the noise basis matrix orthogonal to the reconstruction direction of the autoencoder.

[0085] The execution formula of S302 is specifically:

[0086] N = V[:, h - k:]

[0087] W = UΣV T

[0088] h - k = a

[0089] Among them, N represents the noise basis matrix, W represents the encoder weight matrix, U represents the left singular vector matrix, Σ represents the singular value matrix, V represents the right singular vector matrix, h represents the dimension of the encoder hidden layer in the autoencoder, k represents the number of retained encoder hidden layer dimensions, V[:, h - k:] represents selecting the elements of the last h - k columns in V, a represents the dimension of the sensitive field, and the superscript T represents transpose.

[0090] Optionally, the sensitive fields include patient name, disease type, and identity information. At this time, the dimension of the sensitive field is 3.

[0091] Among them, the noise basis matrix is obtained from the training process of the autoencoder, which represents the directions of some perturbations that are learned during the training of the autoencoder. The column vectors of this matrix represent the directions of the perturbations, which are orthogonal to the reconstruction gradient. Orthogonality means that the directions of the perturbations do not affect how the autoencoder optimizes the reconstruction error during training. Therefore, the perturbations do not affect the autoencoder's learning of clinical features (such as diseases, treatment plans, etc.). The noise basis matrix represents the basic directions of the perturbations. By defining a space for the perturbations, it ensures that the perturbations can effectively affect the sensitive fields of the data while not affecting other information in the data.

[0092] It should be noted that the noise basis matrix orthogonal to the reconstruction direction of the autoencoder is extracted through singular value decomposition, which effectively perturbs the sensitive fields without affecting the clinical features. This method ensures privacy protection while maintaining the autoencoder's learning ability for medical record data and avoiding affecting the accuracy of the authenticity identification of electronic medical records.

[0093] S303: Combine the noise basis matrix to determine the noise perturbation orthogonal to the reconstruction direction of the autoencoder.

[0094] The execution formula of S303 is specifically:

[0095] η = N(N T (x ⊙ M ij ))

[0096] Among them, η represents the noise perturbation, x represents the electronic medical record to be identified, and ⊙ represents element-wise multiplication.

[0097] Among them, N T is the transpose of the matrix, which is used to adjust the direction of the noise to meet the requirements of the perturbation. In this formula, the transpose operation is used to construct the correct perturbation direction to ensure that the direction of the perturbation is orthogonal to the reconstruction process of the autoencoder, that is, it will not interfere with the learning of clinical features. The transpose operation of the noise basis matrix is used to construct the correct perturbation direction to ensure that the direction of the perturbation is orthogonal to the reconstruction process of the autoencoder, that is, it will not interfere with the learning of features other than sensitive features.

[0098] Among them, the reconstruction direction refers to when the autoencoder is trained, the model will learn how to effectively reconstruct the data according to the characteristics of the input data. If N is orthogonal to the reconstruction direction, then N T is also orthogonal to the reconstruction direction, and N Tis the transposed noise basis matrix. After transposition, the perturbation can act on sensitive data more precisely, while maintaining the structure and flexibility of the perturbation, ensuring privacy protection and avoiding negative impacts on the performance of the autoencoder in clinical feature learning. The transposition operation ensures that the perturbation still maintains orthogonality and can match the input data more flexibly. The perturbation is finely controlled and can adapt to different parts of the data (such as sensitive fields). While ensuring that the direction of the perturbation is orthogonal to the reconstruction direction of the autoencoder, it avoids learning features other than sensitive features. Then, N (N T (x⊙M ij )) uses the noise basis matrix to generate the final perturbation. This perturbation will be added to the sensitive fields, thus changing the sensitive data (such as patient names, ID numbers, etc.) without disturbing features other than sensitive features. While protecting privacy, it does not affect the autoencoder's learning of other features and subsequent verification of the authenticity of electronic medical records.

[0099] S304: Apply noise perturbation to sensitive fields to perform feature-preserving desensitization on the electronic medical records to be authenticated.

[0100] The execution formula of S304 is specifically:

[0101]

[0102] where represents the desensitized electronic medical records to be authenticated.

[0103] It should be noted that by combining the pre-trained model and the noise basis matrix, feature-preserving desensitization of the electronic medical records to be authenticated is achieved. This method first identifies sensitive fields in the electronic medical records (such as patient names, disease types, and identity information), and then uses noise perturbations orthogonal to the reconstruction direction of the autoencoder to ensure that the desensitization operation does not affect other clinical features. Through orthogonal perturbations, the privacy of sensitive data is effectively protected, while the autoencoder can continue to learn and process other clinical features, avoiding conflicts between privacy protection and model performance. Such a method not only improves the security of privacy protection but also ensures the accuracy and effectiveness of the authenticity identification of electronic medical records.

[0104] S4: Input the desensitized electronic medical records to be authenticated into the autoencoder and output the reconstructed electronic medical records to be authenticated.

[0105] Among them, the reconstructed electronic medical records to be authenticated refer to inputting the desensitized electronic medical records into the autoencoder, converting them into a latent space representation through the encoder, and then restoring them into reconstructed data with the same structure as the original electronic medical records through the decoder. This reconstruction process helps to detect whether the electronic medical records contain abnormal or tampered content. By comparing the original data and the reconstructed data, anomalies in the data can be further identified.

[0106] In a possible implementation, S4 specifically includes:

[0107] S401: Encode the de-identified electronic medical record to be identified through an encoder.

[0108] The execution formula of S401 is specifically:

[0109]

[0110] Wherein, represents the de-identified electronic medical record to be identified, Encoder() represents the encoder, W e and b e respectively represent the encoder weight and the encoder bias, σ represents the activation function, Flatten() represents the flattening operator, and z represents the encoding result.

[0111] Optionally, the activation function can be the ReLU activation function or the Sigmoid activation function.

[0112] After obtaining the encoding result, it is also necessary to decouple the features of the latent vector.

[0113] S402: Decode the encoding result through a decoder.

[0114] The execution formula of S402 is specifically:

[0115]

[0116] Wherein, Decoder() represents the decoder, W d and b d respectively represent the decoder weight and the decoder bias, represents the decoding result that is consistent with the dimension of the de-identified electronic medical record to be identified, that is, the reconstructed electronic medical record to be identified.

[0117] It should be noted that the de-identified electronic medical record is encoded and decoded through an autoencoder to generate reconstructed data. The encoder converts the de-identified data into a latent space representation, and the decoder then restores it to a structure similar to the original medical record. By comparing the original data and the reconstructed data, it is possible to detect whether there are abnormalities or tampering, ensuring the authenticity of the electronic medical record.

[0118] S5: Calculate the reconstruction error between the electronic medical record to be identified and the reconstructed electronic medical record to be identified.

[0119] In a possible implementation, S5 is specifically:

[0120] Determine the reconstruction error by calculating the sum of the squared differences of each field value between the electronic medical record to be identified and the reconstructed electronic medical record to be identified:

[0121] The calculation method of the reconstruction error is specifically as follows:

[0122]

[0123] Among them, E re represents the reconstruction error, and respectively represent the p-th field value in the electronic medical record to be identified and the reconstructed electronic medical record to be identified, where p = 1, 2,..., P, and P represents the total number of fields in the electronic medical record to be identified.

[0124] It should be noted that the reconstruction error is determined by calculating the sum of the squared differences of each field value between the electronic medical record to be identified and the reconstructed medical record. Quantifying the difference between the electronic medical record to be identified and the reconstructed data can effectively identify abnormalities or tampering in the data. This method improves the accuracy of the authenticity identification of electronic medical records and ensures the integrity and credibility of medical record data.

[0125] S6: Determine the reconstruction error discrimination value in combination with the desensitization process.

[0126] It should be noted that by combining the noise perturbation in the desensitization process, a reconstruction error discrimination value is calculated. This discrimination value is used to distinguish normal electronic medical records from medical records that may be tampered with or have abnormalities. By setting a threshold, it is ensured that when the reconstruction error exceeds the discrimination value, abnormal content can be effectively identified, thereby improving the accuracy and reliability of the authenticity identification of electronic medical records.

[0127] In a possible implementation manner, S6 specifically includes:

[0128] S601: Input the noise perturbation introduced in the desensitization process into the autoencoder to obtain the noise perturbation reconstruction error.

[0129] The execution formula of S601 is specifically:

[0130]

[0131] Among them, η q represents the q-th field value of the noise perturbation, represents η q the reconstructed noise perturbation obtained after being reconstructed by the autoencoder, where q = 1, 2,..., m, m represents the total number of noise perturbation fields, and E noise represents the noise perturbation reconstruction error.

[0132] Among them, the q-th field value of the noise perturbation refers to the q-th perturbed field introduced during the de-identification process of the electronic medical record. The noise perturbation is a modification or perturbation carried out to protect sensitive information (such as the patient's name, medical record number, etc.). These modifications impose perturbations on specific fields to ensure data privacy. In the autoencoder, the perturbation affects the values of these fields, generating perturbed data.

[0133] S602: Superimpose the reconstruction error of the noise perturbation and the average value of the reconstruction errors of historical electronic medical records during the construction process of the autoencoder to obtain a reconstruction error discrimination value.

[0134] The execution formula of S602 is specifically:

[0135]

[0136] Among them, E0 represents the reconstruction error discrimination value, represents the average value of the reconstruction errors.

[0137]

[0138] Among them, y r and respectively represent the r-th historical electronic medical record and the reconstructed historical electronic medical record, r = 1, 2,..., R, and R represents the total number of historical electronic medical records.

[0139] It should be noted that by introducing noise perturbation and calculating its reconstruction error, by calculating the reconstruction error of the noise perturbation, it can be judged whether unnecessary changes are introduced during the de-identification process, thus affecting the reconstruction effect of the autoencoder. Evaluate the impact of de-identification on data integrity, and at the same time ensure that the authenticity identification of electronic medical records can accurately identify abnormalities or tampering in the data under the premise of protecting privacy. At the same time, combined with the average value of the reconstruction errors of historical electronic medical records, a discrimination value is generated. This method can more accurately identify abnormal or tampered content in electronic medical records, improve the accuracy and reliability of authenticity identification, and ensure privacy protection without affecting data analysis.

[0140] S7: In the case where the reconstruction error exceeds the reconstruction error discrimination value, combine the reconstruction error to locate the abnormal content in the electronic medical record to be identified, and enter step S8; otherwise, enter step S9.

[0141] It should be noted that by comparing the reconstruction error with the discrimination value, if the reconstruction error exceeds the set threshold, it indicates that there is abnormal content in the electronic medical record. Using the reconstruction error, further locate and identify the specific abnormal part in the medical record. In this way, tampered or unreasonable information can be accurately identified to ensure the authenticity of the electronic medical record.

[0142] In a possible implementation manner, S7 specifically includes:

[0143] S701: Obtain the feature representations of the convolutional feature maps of each layer of the autoencoder for the electronic medical record to be identified.

[0144] S702: Calculate the contribution values of each feature representation to the reconstruction error.

[0145] S703: Combine the ReLU activation function and locate the abnormal content according to the contribution values.

[0146] The execution formula of S703 is specifically:

[0147]

[0148] where, H ij represents the abnormal value at the i-th row and j-th column position in the electronic medical record to be identified. If the abnormal value is greater than 0, the corresponding position is abnormal content; otherwise, the corresponding position is normal content. ReLU represents the ReLU activation function. represents taking the partial derivative, E re represents the reconstruction error, A ijl represents the feature representation of the i-th row and j-th column position in the electronic medical record to be identified in the convolutional feature map of the l-th layer.

[0149] Among them, the ReLU activation function can filter out the negative contribution area and only retain the area that has a positive impact on the abnormal determination. Integrate multi-layer feature information. Quantify the contribution degree of each position of the feature map to the abnormality, that is, the reconstruction error. Find the area most relevant to the abnormality and amplify the response of the abnormal field.

[0150] Optionally, the abnormal content can directly screen the abnormal content according to the positive or negative of the abnormal value, or first arrange the abnormal values from large to small, and then select the content corresponding to the preset number of abnormal values with the top-ranked abnormal values as the abnormal content. The size of the preset number can be set according to actual needs.

[0151] It should be noted that by obtaining the feature representations of the convolutional feature maps of each layer of the autoencoder and calculating the contribution values to the reconstruction error, the abnormal content in the electronic medical record can be accurately located. Combining the ReLU activation function can effectively filter out the irrelevant parts and retain the areas that have a positive impact on the abnormal determination, thereby enhancing the accuracy of abnormal detection. This method can more carefully identify and amplify the response of the abnormal field through multi-layer feature information fusion, improving the accuracy and reliability of the authenticity identification of electronic medical records.

[0152] S8: Output the abnormal content and an abnormal reminder regarding the abnormal content.

[0153] S9: Output that the electronic medical record to be identified is normal.

[0154] In the actual application process, through the autoencoder and feature-preserving desensitization technology, the authenticity identification process of electronic medical records is ensured to be efficient and secure. First, the system obtains the electronic medical record to be identified, constructs an autoencoder in combination with manifold embedding, and optimizes the model using the publicly available historical electronic medical record data. Then, a feature-preserving desensitization method is adopted to protect patient privacy without destroying clinical features. The desensitized medical record is input into the autoencoder for reconstruction, the reconstruction error is calculated, and combined with the error discrimination value to determine whether there is abnormal content. If an abnormality is found, the abnormal part is further located and a reminder is output. The advantage of this process is that it greatly reduces manual intervention through automated processing, improves the efficiency and accuracy of identification, effectively protects patient privacy, and reduces the risk of medical data leakage.

[0155] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0156] In the embodiments of the present invention, a gradient descent method based on manifold embedding is combined to construct an autoencoder, effectively improving the accuracy of the autoencoder in reconstructing electronic medical records. This autoencoder can more accurately capture the potential structure of electronic medical records, reduce the influence of local minima, ensure that the model can identify abnormalities or tampering in the data, and provide a more reliable authenticity identification of electronic medical records. Through the autoencoder, the automated verification of electronic medical records reduces the workload of manual review and improves efficiency. At the same time, combined with feature-preserving desensitization technology, it can not only effectively protect patient privacy, but also avoid increasing the reconstruction error of subsequent autoencoder processing, thereby avoiding the failure of the authenticity identification of electronic medical records based on the autoencoder due to desensitization, and further ensuring the accuracy of the authenticity identification of electronic medical records. In addition, this solution can accurately locate abnormal content in electronic medical records in combination with the reconstruction error, reducing the probability of medical accidents.

[0157] Refer to the attached Figure 2 illustrates a schematic structural diagram of an electronic medical record authenticity identification system provided by the present invention.

[0158] The present invention also provides an electronic medical record authenticity identification system 20, which is applied to the above-mentioned electronic medical record authenticity identification method, and includes:

[0159] A processor 201.

[0160] A memory 202, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor 201, the electronic medical record authenticity identification method as in the method embodiment is implemented.

[0161] The electronic medical record authenticity identification system 20 provided by the present invention can execute the above-mentioned electronic medical record authenticity identification method and achieve the same or similar technical effects. To avoid repetition, the present invention will not be elaborated here.

[0162] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0163] In the embodiments of the present invention, an autoencoder is constructed by combining the gradient descent method based on manifold embedding, effectively improving the accuracy of the autoencoder in reconstructing electronic medical records. The autoencoder can more accurately capture the potential structure of the electronic medical records, reduce the influence of local minima, ensure that the model can identify anomalies or tampering in the data, and provide more reliable authenticity identification of electronic medical records. By automatically verifying the electronic medical records through the autoencoder, the workload of manual review is reduced and the efficiency is improved. At the same time, combined with the feature-preserving desensitization technology, it can not only effectively protect patient privacy, but also avoid increasing the reconstruction error of subsequent autoencoder processing, thereby avoiding the invalidation of the autoencoder-based authenticity identification of electronic medical records due to desensitization, and further ensuring the accuracy of the authenticity identification of electronic medical records. In addition, this solution can accurately locate abnormal content in electronic medical records by combining the reconstruction error, reducing the probability of medical accidents.

[0164] It should be understood that the processor in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0165] It should also be understood that the memory in the embodiments of the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0166] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions according to the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, or a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0167] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context before and after.

[0168] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0169] It should be understood that in various embodiments of the present invention, the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0170] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0171] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described devices, apparatuses, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0172] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling, direct coupling, or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of devices or units can be in an electrical, mechanical, or other form.

[0173] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0174] In addition, the functional units in each embodiment of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0175] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0176] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the electronic medical record authenticity identification method as in the method embodiment.

[0177] The computer-readable storage medium provided by the present invention can implement the steps and effects of the electronic medical record authenticity identification method in the above method embodiment. To avoid repetition, the present invention will not elaborate further.

[0178] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include:

[0179] In the embodiment of the present invention, a gradient descent method based on manifold embedding is combined to construct an autoencoder, effectively improving the accuracy of the autoencoder in reconstructing electronic medical records. This autoencoder can more accurately capture the potential structure of electronic medical records, reduce the influence of local minima, ensure that the model can identify anomalies or tampering in the data, and provide a more reliable electronic medical record authenticity identification. Through the automatic verification of electronic medical records by the autoencoder, the workload of manual review is reduced and the efficiency is improved. At the same time, combined with the feature-preserving desensitization technology, it can not only effectively protect patient privacy, but also avoid increasing the reconstruction error of subsequent autoencoder processing, thereby avoiding the invalidation of the autoencoder-based electronic medical record authenticity identification due to desensitization, and further ensuring the accuracy of the electronic medical record authenticity identification. In addition, this solution can accurately locate abnormal content in electronic medical records in combination with the reconstruction error, reducing the probability of medical accidents.

[0180] As mentioned above, the above are only the specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0181] The following points need to be explained:

[0182] (1) The accompanying drawings of the embodiments of the present invention only relate to the structures involved in the embodiments of the present invention, and other structures can be referred to the general designs.

[0183] (2) For clarity, in the accompanying drawings used to describe the embodiments of the present invention, the thickness of layers or regions is enlarged or reduced, that is, these drawings are not drawn to actual scale. It can be understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element can be "directly" on or under the other element or there can be intervening elements.

[0184] (3) Without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0185] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for authenticating the authenticity of an electronic medical record, characterized in that: include: S1: Obtain the electronic medical records to be identified; S2: constructing an autoencoder by combining a gradient descent method based on manifold embedding, wherein the autoencoder includes an encoder and a decoder; S3: performing feature-preserving desensitization on the electronic medical record to be identified in combination with the encoder weight matrix of the autoencoder; S4: inputting the desensitized electronic medical record to be identified into the autoencoder, and outputting the reconstructed electronic medical record to be identified; S5: calculating a reconstruction error between the electronic medical record to be identified and the reconstructed electronic medical record to be identified; S6: Determine the reconstruction error discrimination value in combination with the desensitization process; S7: if the reconstruction error exceeds the reconstruction error discrimination value, locate the abnormal content in the electronic medical record to be identified in combination with the reconstruction error, and proceed to step S8; otherwise, proceed to step S9; S8: outputting the abnormal content and an abnormal reminder about the abnormal content; S9: Outputting the electronic medical record to be identified is normal.

2. The electronic medical record authenticity authentication method according to claim 1, characterized in that: The S2 specifically includes: S201: Access to publicly available historical electronic medical records; S202: reconstructing the historical electronic medical record by using the autoencoder to obtain a reconstructed historical electronic medical record; S203: optimizing the hyperparameters of the autoencoder by a gradient descent method based on manifold embedding until a reconstruction error between the historical electronic medical record and the reconstructed historical electronic medical record is less than a preset reconstruction error, wherein the hyperparameters include a learning rate, a number of encoder layers, a number of decoder layers, and a dropout rate; S204: Setting the autoencoder based on the hyperparameters obtained through optimization to complete the construction of the autoencoder.

3. The method for authenticating the authenticity of electronic medical records according to claim 2, characterized in that: The S203 specifically includes: S203A: Mapping each of the hyperparameters to a Riemann manifold; S203B: constructing a homotopic potential field on the Riemann manifold for smoothing the excessive hyperparameter optimization path; S203C: Calculate a corrected gradient direction for updating a hyperparameter along the bottom line of the Riemann manifold; S203D: According to the modified gradient direction, each hyperparameter is updated in combination with Brouwer's fixed point theorem to optimize the hyperparameters of the autoencoder.

4. The electronic medical record authenticity authentication method according to claim 1, characterized in that: The S3 specifically includes: S301: Identify sensitive fields in the electronic medical record to be identified through a pre-trained model, wherein the sensitive fields include patient name, disease type, and identity information; S302: performing singular value decomposition on the coding layer weight matrix of the autoencoder by a singular value decomposition algorithm to extract a noise basis matrix orthogonal to the reconstruction direction of the autoencoder; S303: Determine a noise disturbance orthogonal to the autoencoder reconstruction direction in combination with the noise basis matrix; S304: Apply the noise disturbance to the sensitive field to perform feature-preserving desensitization on the electronic medical record to be identified.

5. The method for authenticating the authenticity of electronic medical records according to claim 1, characterized in that: The S4 specifically includes: S401: Encoding the desensitized electronic medical record to be identified by the encoder; S402: Decode the encoding result by the decoder, and output the reconstructed electronic medical record to be identified.

6. The electronic medical record authenticity authentication method according to claim 1, characterized in that: The S5 is specifically: The reconstruction error is determined by calculating the sum of squares of differences between the electronic medical record to be identified and the reconstructed electronic medical record to be identified in each field value.

7. The method for verifying the authenticity of electronic medical records according to claim 4, characterized in that: The S6 specifically includes: S601: Input the noise disturbance applied in the desensitization process into the autoencoder to obtain a noise disturbance reconstruction error; S602: Superimpose the noise disturbance reconstruction error and the mean value of the reconstruction error of the historical electronic medical record in the process of constructing the autoencoder to obtain the reconstruction error discrimination value.

8. The method for verifying the authenticity of electronic medical records according to claim 7, characterized in that: The S7 specifically includes: S701: Obtain feature representations of the convolutional feature graphs of each layer of the autoencoder for the electronic medical record to be identified; S702: Calculate the contribution value of each of the feature representations to the reconstruction error; S703: In combination with the ReLU activation function, locate the abnormal content according to the contribution value.

9. An electronic medical record authenticity authentication system, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for authenticating the authenticity of an electronic medical record as claimed in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the electronic medical record authenticity authentication method as described in any one of claims 1 to 8 is implemented.

Images