Anonymization method and system for user information of smart medical equipment based on artificial intelligence

Abstract

The present invention discloses an artificial intelligence-based user information anonymization method and system for smart medical devices, which belongs to the field of data processing technology. The present invention first ensures the pertinence and efficiency of data processing by accurately acquiring and screening the current user's neural regulation record text and its key data set. Secondly, by identifying the semantic value weight of each group of key text data, a refined assessment of the sensitivity of the data is achieved, providing an accurate basis for subsequent anonymous processing. Further, based on the semantic value recognition weight, the available disturbance index of the data is determined, and the key text data with the lowest index is screened out for desensitization. This strategy not only maximizes the protection of user privacy, but also ensures that the processed data still retains its value for medical analysis and research. While protecting personal privacy, it promotes the safe sharing and effective use of medical data, and provides a new solution for data management in the field of implantable neural regulation.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and specifically relates to an artificial intelligence-based smart medical device user information anonymization method and system. Background Art

[0002] With the increasing development of implantable neuromodulation technology, patients' neuromodulation reports have become an important basis for doctors to formulate treatment plans and monitor their condition. However, sensitive information contained in these reports, such as the patient's personal identity, medical details, etc., will seriously infringe on the patient's privacy rights if they are shared or analyzed without proper processing. Traditional anonymization methods often use simple data deletion or replacement. Although this method can protect privacy to a certain extent, it often causes the data to lose its original medical value, affecting subsequent analysis and research. Therefore, how to ensure the effective use of medical data while protecting patient privacy has become a technical problem that needs to be solved in the current field of implantable neuromodulation. Summary of the invention

[0003] The present invention provides an artificial intelligence-based method and system for anonymizing user information of smart medical devices, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0004] The embodiment of the present invention provides an artificial intelligence-based method for anonymous user information of a smart medical device, which is applied to an anonymous user information system. The method includes:

[0005] Acquire a current user's neural regulation record text of an implantable neural regulation report to be anonymized, and filter a key text data set of the current user's neural regulation record text from the implantable neural regulation report to be anonymized;

[0006] Filtering multiple groups of key text data from the key text data set, and identifying the semantic value weight of the current user's neural regulation record text based on each group of key text data to obtain the semantic value identification weight corresponding to each group of key text data;

[0007] Determining a data available disturbance index corresponding to each group of key text data based on a semantic value recognition weight corresponding to each group of key text data, and selecting a target group of key text data having a lowest data available disturbance index from the multiple groups of key text data;

[0008] The current user's neural regulation record text is privacy-desensitized based on the target group key text data.

[0009] In one implementation, the step of screening multiple groups of key text data from the key text data set includes:

[0010] Filtering a plurality of target key text data from the key text data set;

[0011] Taking each target key text data as a benchmark, a target number of initial key text data are screened from the key text data set by using a preset semantic transfer strategy and a preset cycle step;

[0012] Each target key text data and the initial key text data corresponding to each target key text data are classified into the same group to obtain the multiple groups of key text data.

[0013] In one implementation, the step of screening a plurality of target key text data from the key text data set includes:

[0014] Filtering subject text data from the current user's neural regulation record text;

[0015] A plurality of target key text data having different semantic differences from the subject text data are screened from the key text data set; wherein the plurality of target key text data include text data having the lowest semantic difference from the subject text data in the key text data set.

[0016] In one implementation, the semantic value recognition weights include clinical guidance semantic recognition variables;

[0017] The identifying the semantic value weight of the current user's neural regulation record text based on each set of key text data to obtain the semantic value identification weight corresponding to each set of key text data includes:

[0018] Determine a semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data;

[0019] Based on the semantic feature association map corresponding to each group of key text data and the implanted regulation operation semantic variables of the current user's neural regulation record text, the clinical guidance semantic variables of the current user's neural regulation record text are determined to obtain the clinical guidance semantic identification variables corresponding to each group of key text data.

[0020] In one implementation, determining the semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data includes:

[0021] Determine a first key text data cluster and a second key text data cluster from each group of key text data; wherein the key text data included in the first key text data cluster and the second key text data cluster are different;

[0022] Determine the mean clinical guidance semantic features and the mean implantation control operation semantic features of the first key text data cluster, and determine the mean clinical guidance semantic features and the mean implantation control operation semantic features of the second key text data cluster;

[0023] Based on the clinical guidance semantic mean features and implantation regulation operation semantic mean features corresponding to the first key text data cluster and the second key text data cluster respectively, a unique hot semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data is determined.

[0024] In one implementation, determining the first key text data cluster and the second key text data cluster from each group of key text data includes:

[0025] Filtering a first number of key text data from each group of key text data according to a descending rule of the corresponding implanted control operation semantic variable, and generating the first key text data cluster based on the first number of key text data;

[0026] A second number of key text data is selected from each group of key text data according to the ascending rule of the corresponding implanted control operation semantic variable, and the second key text data cluster is generated based on the second number of key text data.

[0027] In one implementation, determining the semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data includes:

[0028] Semantic mining is performed on the annotated implantation control operation semantic variables and the annotated clinical guidance semantic variables of the key user's neural regulation record text to which each group of key text data belongs, respectively, to obtain the implantation control operation description vector and the clinical guidance viewpoint description vector of the key user's neural regulation record text;

[0029] Based on the implantation control operation description vector and the clinical guidance viewpoint description vector of the key user's neural regulation record text, the implantation control operation semantic variables and the clinical guidance semantic variables of the key user's neural regulation record text are updated respectively to obtain the implantation control operation update semantics and the clinical guidance update semantics of the key user's neural regulation record text;

[0030] Acquire the implantation control operation update semantics and clinical guidance update semantics of each group of key text data from the implantation control operation update semantics and clinical guidance update semantics of the key user neural control record text;

[0031] Based on the implantation control operation update semantics and the clinical guidance update semantics of each group of key text data, a semantic feature association map corresponding to each group of key text data is determined.

[0032] In one implementation, the step of obtaining the implantation control operation update semantics and clinical guidance update semantics of each group of key text data from the implantation control operation update semantics and clinical guidance update semantics of the key user neural control record text includes:

[0033] Extracting the implantation control operation update semantics of the key user's neural control record text based on the semantic feature recognition corresponding to the clinical guidance update semantics of the key user's neural control record text;

[0034] The implantation regulation operation update semantics of each group of key text data is obtained from the extracted implantation regulation operation update semantics.

[0035] In one implementation, the determining of the clinical guidance semantic variables of the current user's neural regulation record text based on the semantic feature association map corresponding to each group of key text data and the implantation control operation semantic variables of the current user's neural regulation record text, and obtaining the clinical guidance semantic recognition variables corresponding to each group of key text data, includes:

[0036] Performing semantic mining on the annotated implantation control operation semantic variables of the current user's neural regulation record text to obtain an implantation control operation description vector of the current user's neural regulation record text;

[0037] Based on the implantation control operation description vector of the current user's neural regulation record text, the implantation control operation semantic variable of the current user's neural regulation record text is updated to obtain the implantation control operation update semantics of the current user's neural regulation record text;

[0038] Based on the semantic feature association map corresponding to each group of key text data and the implanted regulation operation update semantics of the current user's neural regulation record text, the clinical guidance semantic variables of the current user's neural regulation record text are determined to obtain the clinical guidance semantic identification variables corresponding to each group of key text data.

[0039] In one implementation, the privacy desensitizing processing of the current user's neural regulation record text based on the target group key text data includes:

[0040] Obtaining clinical guidance semantic recognition variables corresponding to the target group key text data;

[0041] Semantic fuzzy processing is performed based on the comparison result between the annotated clinical guidance semantic variables of the current user's neural regulation record text and the clinical guidance semantic recognition variables corresponding to the target group key text data.

[0042] In one implementation, the number of the key text data sets is multiple;

[0043] The performing privacy desensitization processing on the current user's neural regulation record text based on the target group key text data includes:

[0044] Determine the target group key text data with the lowest data available disturbance index from the target group key text data corresponding to the multiple key text data sets;

[0045] The current user's neural regulation record text is privacy-desensitized according to the determined target group key text data.

[0046] In one implementation, the privacy desensitizing processing of the current user's neural regulation record text based on the target group key text data includes:

[0047] Processing the semantic value weight of the current user's neural regulation record text based on the target group key text data to obtain a semantic value decision representation of the current user's neural regulation record text;

[0048] The index features of the target group key text data are configured into the semantic value decision representation of the current user's neural regulation record text, so that the anonymous state analysis system determines the target group key text data based on the index features in the semantic value decision representation, and obtains the privacy-desensitized text of the current user's neural regulation record text based on the target group key text data.

[0049] An embodiment of the present invention provides a user information anonymization system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above method.

[0050] An embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0051] The present invention first ensures the pertinence and efficiency of data processing by accurately acquiring and screening the current user's neural regulation record text and its key data set. Secondly, by identifying the semantic value weight of each group of key text data, a refined assessment of the data sensitivity is achieved, providing an accurate basis for subsequent anonymous processing. Furthermore, based on the semantic value recognition weight, the available data disturbance index is determined, and the key text data with the lowest index is screened out for desensitization. This strategy not only maximizes the protection of user privacy, but also ensures that the processed data still retains its value for medical analysis and research. In summary, while protecting personal privacy, it promotes the safe sharing and effective use of medical data, and provides a new solution for data management in the field of implantable neural regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 The present invention provides a flowchart of an artificial intelligence-based method for anonymous user information of smart medical devices.

[0053] Figure 2 A schematic diagram of the structure of a user information anonymity system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more. In addition, in the present invention, "and / or" means at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0056] Figure 1 An artificial intelligence-based method for anonymous user information of a smart medical device is shown, which is applied to an anonymous user information system. The method includes the following steps 110 to 140.

[0057] In modern medical technology, implantable neuromodulators are widely used to treat various neurological diseases, such as Parkinson's disease and epilepsy. Implantable neuromodulators (i.e., the smart medical devices of the present invention) can record and analyze the patient's neural activity and provide doctors with accurate treatment plans. However, as these devices are used more and more widely, the security and privacy protection of patient data are becoming increasingly prominent. In order to ensure that the patient's personal privacy is not leaked, the user information anonymization system came into being, which can protect the patient's privacy while ensuring the effective use of medical data.

[0058] The following is a specific application scenario example, which shows how the user information anonymization system can anonymize data in the implantable neuromodulation scenario. The Department of Neurology of a large hospital has introduced an advanced implantable neuromodulation device that can monitor and record the patient's neural activity data in real time. Doctors use this data to analyze the patient's condition and develop corresponding treatment plans. However, since this data contains sensitive information of the patient, such as name, age, condition, etc., it is necessary to ensure that the patient's privacy is fully protected when analyzing and sharing the data.

[0059] Step 110: Obtain the current user's neural regulation record text of the implantable neural regulation report to be anonymized, and filter the key text data set of the current user's neural regulation record text from the implantable neural regulation report to be anonymized.

[0060] The user information anonymization system first establishes a data connection with the hospital's implantable neuromodulation device to obtain the implantable neuromodulation report to be anonymized. This report contains the patient's neural activity records, the doctor's treatment recommendations, and the patient's personal information. After obtaining the report, the user information anonymization system begins to screen the key text data set in the current user's neuromodulation record text. The purpose of this step is to extract key information that is valuable for subsequent anonymous processing from a large amount of text data. For example, the system may screen out text fragments containing sensitive information such as the patient's name, age, and condition description.

[0061] Step 120: Filter multiple groups of key text data from the key text data set, and identify the semantic value weight of the current user's neural regulation record text based on each group of key text data to obtain the semantic value identification weight corresponding to each group of key text data.

[0062] Next, the user information anonymization system further filters multiple groups of key text data from the key text data set. These data may be directly related to the patient's privacy, such as name, age, etc., or they may be related to the description of the condition, such as the name of the disease, symptoms, etc. For each set of key text data, the user information anonymization system will identify the semantic value weight. The purpose of this step is to evaluate the importance of each set of data in the entire neuromodulation record text. For example, the patient's name may have the largest semantic value weight because once leaked, it will directly expose the patient's identity. Some text data related to the description of the condition, although also containing sensitive information, may have a relatively low semantic value weight.

[0063] Step 130: determine the data available disturbance index corresponding to each group of key text data based on the semantic value recognition weight corresponding to each group of key text data, and select a target group of key text data with the lowest data available disturbance index from the multiple groups of key text data.

[0064] After identifying the semantic value weights of each set of key text data, the user information anonymization system will determine the data availability perturbation index corresponding to each set of data based on these weights. This index reflects the extent to which the data can be perturbed or modified while protecting the privacy of patients. The system will filter out the target group of key text data with the lowest data availability perturbation index from multiple sets of key text data. These data are the focus of subsequent anonymization processing because they contain important sensitive information and can be perturbed or modified to a certain extent to protect the privacy of patients.

[0065] Step 140: Perform privacy desensitization processing on the current user's neural regulation record text based on the target group key text data.

[0066] Finally, the user information anonymization system performs privacy desensitization processing on the current user's neural regulation record text based on the target group's key text data. The purpose of this step is to ensure the effective use of medical data while protecting patient privacy. Specifically, the system may use methods such as replacement, deletion, or generalization to process the target group's key text data. For example, for the patient's name, the system may replace it with a unique identifier to ensure that the patient's true identity cannot be identified during the data analysis process. For sensitive information related to the description of the disease, the system may perform generalization processing, such as replacing specific disease names with more general disease categories.

[0067] Through the implementation of the above steps, the user information anonymization system can effectively anonymize sensitive information in the implantable neuromodulation report and protect the patient's personal privacy. At the same time, the processed data still retains its original medical value and can be used for doctors' analysis and research. In this way, not only the security of medical data is improved, but also the sharing and utilization of medical data is promoted. It can be seen that the data anonymization processing scheme of the user information anonymization system in the implantable neuromodulation scenario can ensure the effective use and sharing of medical data while protecting patient privacy, providing new ideas and methods for data security and privacy protection in the medical field.

[0068] In combination with the above technical solution, the desensitization process of user information is demonstrated through a specific example below.

[0069] 1. Example of Implantable Neuromodulation Report to be Anonymized

[0070] Patient Information:

[0071] Name: Zhang San

[0072] Age: 45

[0073] Gender: Male

[0074] Contact details: (...)

[0075] address:(...)

[0076] Disease Overview:

[0077] Zhang San has suffered from Parkinson's disease for a long time. In recent years, his symptoms have gradually worsened, with obvious symptoms such as tremor, stiffness and bradykinesia, which seriously affect his daily life. After a comprehensive evaluation by a team of neurology experts and full communication with the patient, it was decided to implant an advanced neuromodulation device for treatment in order to improve his condition.

[0078] Implantable neuromodulation device records data:

[0079] Device implantation date: March 15, 2023

[0080] Overview of the surgical procedure: The operation went smoothly and the device was successfully implanted in the patient. The initial setting parameters are as follows.

[0081] Initial device setting parameters: stimulation frequency 130 Hz, stimulation intensity 2.5 V, pulse width 60 μs, other settings (omitted).

[0082] Neural activity monitoring data (partial):

[0083] March 16: After the device was turned on, the patient's nerve activity fluctuated briefly, then gradually stabilized. The stimulation effect was initially assessed to be good, and the patient's tremor symptoms were alleviated.

[0084] March 17: The patient's neural activity showed regular changes under stimulation, the symptoms of Parkinson's disease were further alleviated, and motor ability was improved.

[0085] ...(Follow-up data are being continuously recorded, including daily changes in neural activity, evaluation of stimulation effects, etc.).

[0086] Doctor's treatment recommendations:

[0087] According to the monitoring data of the neuromodulation device, patient Zhang San responded well to the nerve stimulation and his Parkinson's disease symptoms were significantly relieved. It is recommended to continue to closely observe the treatment effect of the device and adjust the stimulation parameters in a timely manner according to the changes in the patient's symptoms. It is expected that in the next few months, the patient's symptoms will be further improved and his quality of life will be improved.

[0088] 2. Privacy Desensitization Report Example

[0089] Patient Information:

[0090] Patient number: 001

[0091] Age range: 45-50 years old

[0092] Gender: Male

[0093] Contact information: (Anonymous)

[0094] Address: (Anonymous)

[0095] Disease Overview:

[0096] The patient had been suffering from a certain neurological disease for a long time, and his symptoms had gradually worsened in recent years, seriously affecting his daily life. After a comprehensive evaluation by a team of neurology experts and full communication with the patient, it was decided to implant an advanced neuromodulation device for treatment in order to improve his condition.

[0097] Implantable neuromodulation device records data:

[0098] Device implantation date: Someday in March 2023

[0099] Overview of the surgical process: The operation was successful, the device was successfully implanted into the patient, and the initial setting parameters have been desensitized.

[0100] Initial device setting parameters: stimulation frequency XXX Hz, stimulation intensity X.XV, pulse width XX μs, other settings (all desensitized).

[0101] Neural activity monitoring data (partial):

[0102] Day 1 after implantation: After the device was activated, the patient's neural activity fluctuated briefly, then gradually stabilized. The stimulation effect was initially assessed to be good, and the symptoms of the neurological disease were alleviated.

[0103] The second day after implantation: The patient's neural activity showed regular changes under stimulation, and the symptoms of neurological diseases were further alleviated.

[0104] ...(Subsequent data is being recorded continuously, and specific values, times and detailed descriptions have all been anonymized).

[0105] Doctor's treatment recommendations:

[0106] According to the monitoring data of the neuromodulation device (desensitized), the patient responded well to the nerve stimulation and the symptoms of the neurological disease were significantly relieved. It is recommended to continue to closely observe the treatment effect of the device and adjust the stimulation parameters in time according to the changes in the patient's symptoms. It is expected that in the next few months, the patient's symptoms will be further improved and the quality of life will be improved.

[0107] In the privacy desensitization report, sensitive information such as the patient's name, specific age, contact information, address, specific date of device implantation, and specific values ​​of device setting parameters have been replaced, deleted, or blurred to ensure that the patient's personal privacy is strictly protected. At the same time, the desensitization report still retains enough information for doctors to accurately evaluate the treatment effect and develop a reasonable follow-up treatment plan. In this way, the effective use and sharing of medical data can be ensured while fully protecting the privacy of patients.

[0108] The present invention first ensures the pertinence and efficiency of data processing by accurately acquiring and screening the current user's neural regulation record text and its key data set. Secondly, by identifying the semantic value weight of each group of key text data, a refined assessment of the data sensitivity is achieved, providing an accurate basis for subsequent anonymous processing. Furthermore, based on the semantic value recognition weight, the available data disturbance index is determined, and the key text data with the lowest index is screened out for desensitization. This strategy not only maximizes the protection of user privacy, but also ensures that the processed data still retains its value for medical analysis and research. In summary, while protecting personal privacy, it promotes the safe sharing and effective use of medical data, and provides a new solution for data management in the field of implantable neural regulation.

[0109] It is worth mentioning that the core invention of the present invention is how to determine the data usability perturbation index corresponding to each set of key text data based on the semantic value recognition weight corresponding to these data. First of all, it is necessary to clarify the concept of "data usability perturbation index", which refers to the degree of data change that can be tolerated when the data is anonymized. The lower the index, the better the original information value and usability of the group of data can be maintained after the desensitization process.

[0110] In the specific implementation, an initial data availability disturbance index is first assigned to each set of key text data. This index can be a preset fixed value or dynamically adjusted according to the type and sensitivity of the data. Then, the user information anonymization system uses the semantic value recognition weight corresponding to each set of key text data that has been calculated to correct the initial index.

[0111] The correction process is as follows: the user information anonymization system first identifies the group with the largest semantic value recognition weight among all key text data and regards it as the benchmark. Then, for other groups of key text data, the user information anonymization system adjusts their data availability perturbation indicators according to the ratio of their semantic value recognition weights to the benchmark weights. Specifically, if the semantic value recognition weight of a group of data is low, it means that this group of data contributes less to the overall semantics, then the user information anonymization system can assign it a relatively high data availability perturbation indicator, allowing it to be modified more during the desensitization process. Conversely, if the semantic value recognition weight of a group of data is high, then the user information anonymization system needs to assign it a relatively low data availability perturbation indicator to ensure that after desensitization, this group of data can still better maintain its original information.

[0112] For example, there are three groups of key text data, whose semantic value recognition weights are 0.8, 0.5, and 0.2 respectively, and the initial data available perturbation index assigned to them by the user information anonymization system is 1.0. Then, after correction according to the weight ratio, the data available perturbation index of the first group of data may remain unchanged or slightly decrease because of its higher semantic value; the data available perturbation index of the second group of data may increase slightly because of its medium semantic value; and the data available perturbation index of the third group of data will increase significantly because of its relatively low semantic value, allowing for greater changes in the desensitization process. In this way, it can be ensured that during the desensitization process, the privacy of patients is effectively protected while the medical value and analytical usability of the data are retained as much as possible.

[0113] It is worth mentioning that, in an independent embodiment, the step 130 described in determining the data available perturbation index corresponding to each group of key text data based on the semantic value recognition weight corresponding to each group of key text data includes: based on the semantic value recognition weight corresponding to each group of key text data, performing identity-type semantic relationship mining and regulatory event-type semantic relationship mining on multiple text paragraphs in each group of key text data, respectively, to obtain an identity-type semantic relationship set and a regulatory event-type semantic relationship set; performing a first semantic paraphrase processing on the identity-type semantic relationship set through a first semantic paraphrase model to obtain a first partial text including an identity-type semantic relationship; performing a second semantic paraphrase processing on the regulatory event-type semantic relationship set through a second semantic paraphrase model to obtain a second partial text including a regulatory event-type semantic relationship; performing a merging process based on the first partial text and the second partial text to obtain an available text data set matching the target semantic relationship in each group of key text data, the target semantic relationship including at least one of an identity-type semantic relationship and a regulatory event-type semantic relationship; and generating a data available perturbation index for the available text data set using a long short-term memory model to obtain a data available perturbation index corresponding to each group of key text data.

[0114] It can be understood that the process of determining the data available disturbance index corresponding to each group of key text data based on the semantic value recognition weight corresponding to each group of key text data is particularly critical and important in the application scenario of implantable neuromodulation therapy.

[0115] During the implantable neuromodulation treatment, the patient's neuromodulation report contains a lot of sensitive information, such as the patient's identity information, condition details, treatment process, etc. This information is an important basis for doctors to formulate treatment plans and monitor the condition, but it also involves the privacy rights of patients. Therefore, how to ensure the semantic value and analytical usability of these medical text data while protecting patient privacy has become an urgent problem to be solved.

[0116] Based on this background, we first conduct identity-based semantic relationship mining and regulatory event-based semantic relationship mining on multiple text paragraphs in each set of key text data according to the semantic value recognition weights corresponding to each set of key text data. Identity-based semantic relationships mainly focus on information related to the patient's identity in the text, such as name, age, gender, etc., which is crucial to protecting patient privacy; while regulatory event-based semantic relationships mainly focus on medical behaviors or events described in the text, such as diagnosis, treatment, drug use, etc., which are of great value for medical analysis and research.

[0117] Through these two types of semantic relationship mining, the user information anonymization system can obtain identity-based semantic relationship sets and regulatory event-based semantic relationship sets. These two sets contain all the information related to the patient's identity and medical behavior in the text, providing an important reference for the subsequent desensitization processing of the user information anonymization system.

[0118] Next, the user information anonymization system performs semantic paraphrase processing on the two semantic relationship sets through the first semantic paraphrase model and the second semantic paraphrase model respectively. Semantic paraphrase refers to restating or rewriting the text without changing the original semantics to generate a text with different expressions but retaining the original semantics. In the present invention, the user information anonymization system uses two different semantic paraphrase models to process the identity-type semantic relationship set and the regulatory event-type semantic relationship set respectively, so as to obtain a first partial text including an identity-type semantic relationship and a second partial text including a regulatory event-type semantic relationship.

[0119] These two partial texts retain the patient identity information and medical behavior information in the original text, but the expression methods may be different. Such processing not only increases the diversity and complexity of the text, but also provides more possibilities for subsequent desensitization processing.

[0120] Then, the user information anonymization system merges the first partial text and the second partial text to obtain a usable text dataset that matches the target semantic relationship in each set of key text data. This dataset contains both the patient's identity information and medical behavior information, and the expression may have been re-stated. The target semantic relationship here includes at least one of an identity-type semantic relationship and a regulatory event-type semantic relationship, depending on the type of text information that the user information anonymization system wants to retain.

[0121] After obtaining the available text data set, the user information anonymization system uses the long short-term memory model (LSTM) to generate available perturbation indicators for the data. The long short-term memory model is a special recurrent neural network (RNN) that can process sequence data and capture long-term dependencies therein. In the present invention, the user information anonymization system uses the LSTM model to evaluate the semantic value and analytical usability of the available text data set that can still be retained after desensitization.

[0122] Specifically, the user information anonymization system uses the available text data set as the input of the LSTM model, and the model outputs a data available perturbation index. This index reflects the degree of data change that the text can tolerate after desensitization. In the application scenario of implantable neuromodulation therapy, this index is particularly important because it is directly related to whether the desensitized text data can still retain sufficient medical value and analytical usability.

[0123] If the data usability perturbation index is low, it means that the text can still maintain its original information value and usability after desensitization. This is very important for medical analysis and research, because the user information anonymization system can ensure that the desensitized text data still has sufficient semantic value and can be used for subsequent analysis and research.

[0124] On the contrary, if the data availability perturbation index is high, it may mean that the text will have a large information loss or reduced usability after desensitization. In this case, the user information anonymization system needs to be more cautious in selecting desensitization strategies and methods to ensure that while protecting patient privacy, information loss and the impact on medical analysis and research are minimized.

[0125] This design can determine the data available disturbance index corresponding to each group of key text data based on the semantic value recognition weight corresponding to each group of key text data. This index has important significance and value in the application scenario of implantable neuromodulation therapy, and it provides an important reference and basis for the subsequent desensitization processing of the user information anonymization system. The user information anonymization system can adjust the desensitization strategy and method based on the data available disturbance index to ensure that the semantic value and analytical usability of medical text data are retained to the maximum extent while protecting patient privacy.

[0126] In an independent embodiment, the multiple text paragraphs in each group of key text data are respectively subjected to identity-type semantic relationship mining and regulation event-type semantic relationship mining to obtain an identity-type semantic relationship set and a regulation event-type semantic relationship set, including: performing identity-type semantic relationship mining on the multiple text paragraphs in each group of key text data to obtain identity-type semantic relationship masks in each text paragraph and basic semantic relationship labels corresponding to each identity-type semantic relationship mask; determining an identity-type semantic relationship set based on the identity-type semantic relationship masks in each text paragraph and the corresponding basic semantic relationship labels; performing regulation event-type semantic relationship mining on the multiple text paragraphs in each group of key text data to obtain a regulation event-type semantic relationship set.

[0127] It can be understood that based on this embodiment, through the careful analysis of the text paragraphs, the purpose is to mine the identity semantic relationship and the control event semantic relationship therein, and then form two important sets: the identity semantic relationship set and the control event semantic relationship set

[0128] First, the user information anonymization system performs identity semantic relationship mining on multiple text paragraphs in each set of key text data. This process involves the refined extraction and identification of identity information in text paragraphs. To achieve this goal, the user information anonymization system uses advanced semantic analysis techniques that can deeply understand the context and semantics in the text, thereby accurately identifying semantic relationships related to identity information. The result of the mining is a series of identity semantic relationship masks that represent the existence and location of identity information in the text paragraph. At the same time, the user information anonymization system will also assign a basic semantic relationship label to each identity semantic relationship mask, which describes the specific type and meaning of the identity information represented by the mask. Through such processing, the user information anonymization system can clearly understand the distribution and characteristics of identity information in the text paragraph.

[0129] Secondly, based on the identity semantic relationship masks and corresponding basic semantic relationship labels in each text paragraph, the user information anonymization system begins to build an identity semantic relationship set. This set is a structured data set that stores the identity information in all text paragraphs in a specific organizational way. In this set, each identity semantic relationship is clearly represented, including the text paragraph in which it exists, the specific semantic relationship mask, and the corresponding basic semantic relationship label. Such a set provides great convenience for the subsequent data processing and privacy protection of the user information anonymization system. The user information anonymization system can easily query and access the identity information in the text, and can also desensitize this information when necessary to protect the privacy rights and interests of patients.

[0130] Then, the user information anonymization system performs regulatory event semantic relationship mining on multiple text paragraphs in each set of key text data. Similar to identity semantic relationship mining, regulatory event semantic relationship mining also uses advanced semantic analysis technology. However, the difference is that this process focuses on the medical behaviors or events described in the text paragraphs, especially regulatory events related to implantable neuromodulatory therapy. Through mining, the user information anonymization system can identify various regulatory events described in the text, such as the start of treatment, adjustment of treatment, and the effect of treatment. These regulatory events are of great value for medical analysis and research because they can reveal the details and dynamic changes of the treatment process. The result of mining is a regulatory event semantic relationship set, which contains all regulatory events described in the text paragraphs and their related semantic relationships.

[0131] After constructing the identity-based semantic relationship set and the regulatory event-based semantic relationship set, the user information anonymization system has the basis for in-depth analysis of each set of key text data. These two sets represent the identity information and regulatory event information in the text, respectively, and they are important objects for subsequent data processing and privacy protection in the user information anonymization system. Through further processing and analysis of these two sets, the user information anonymization system can better understand and utilize the semantic value in text data, and can also conduct effective data analysis and research under the premise of protecting patient privacy.

[0132] In the application scenario of implantable neuromodulatory therapy, this embodiment has important significance and value. First of all, through identity-type semantic relationship mining and regulatory event-type semantic relationship mining, the user information anonymization system can fully and deeply understand the sensitive information and important medical events in the text data. This is crucial for the user information anonymization system to formulate effective desensitization strategies and methods. The user information anonymization system can identify the parts that need to be desensitized based on the identity information in the identity-type semantic relationship set, and can also retain information that is valuable for medical analysis and research based on the regulatory events in the regulatory event-type semantic relationship set.

[0133] Secondly, by constructing identity-based semantic relationship sets and regulatory event-based semantic relationship sets, the user information anonymization system can provide a structured data foundation for subsequent text data processing and analysis. The data in these two sets are organized and stored according to specific semantic relationships, which makes it easier for the user information anonymization system to query, access, and analyze data. At the same time, since these two sets represent the identity information and regulatory event information in the text, respectively, the user information anonymization system can also perform data desensitization and privacy protection more flexibly.

[0134] Finally, the application of this embodiment also helps to promote the development and application of implantable neuromodulation therapy. Through deep mining and analysis of text data, the user information anonymization system can better understand the details and dynamic changes of the treatment process, thereby providing doctors with more accurate and comprehensive medical information support. At the same time, through effective desensitization and privacy protection processing, the user information anonymization system can also ensure that the privacy rights and interests of patients are fully protected, thereby enhancing patients' trust and satisfaction with medical services.

[0135] It can be seen that this embodiment can not only comprehensively and deeply mine the semantic relationship in text data, but also provide a structured data foundation for subsequent text data processing and analysis. At the same time, it also helps to promote the development and application of implantable neuromodulation therapy, provide doctors with more accurate and comprehensive medical information support, and protect the privacy rights of patients.

[0136] In an independent embodiment, the first semantic paraphrase model is used to perform a first semantic paraphrase process on the identity-type semantic relationship set to obtain a first partial text including identity-type semantic relationships, including: performing semantic relationship label sampling on each text paragraph in the identity-type semantic relationship set to obtain a core semantic relationship label corresponding to each text paragraph; performing mask reconstruction based on the size of the identity-type semantic relationship mask corresponding to the corresponding core semantic relationship label in each text paragraph to obtain a mask-reconstructed identity-type semantic relationship set; performing dynamic text paraphrase process on the mask-reconstructed identity-type semantic relationship set to obtain multiple first partial paraphrase texts including identity-type semantic relationships; and optimizing the first partial paraphrase texts belonging to the same individual label according to the individual labels to which the first partial paraphrase texts belong to obtain a first partial text including identity-type semantic relationships.

[0137] In an independent embodiment, the semantic relationship label sampling is performed on each text paragraph in the identity-type semantic relationship set to obtain the core semantic relationship label corresponding to each text paragraph, including: for each text paragraph in the identity-type semantic relationship set, when the number of basic semantic relationship labels of the text paragraph is at least two, the semantic relationship influence of each basic semantic relationship label is obtained; when the number of basic semantic relationship labels with the largest semantic relationship influence is one, the basic semantic relationship label with the largest semantic relationship influence is used as the core semantic relationship label of the corresponding text paragraph; when the number of basic semantic relationship labels with the largest semantic relationship influence is at least two, for each basic semantic relationship label with the largest semantic relationship influence, the mask influence of the corresponding identity-type semantic relationship mask is obtained; and the core semantic relationship label corresponding to the corresponding text paragraph is determined according to the basic semantic relationship label corresponding to the largest mask influence.

[0138] It can be understood that the core of this embodiment is that through the detailed analysis and processing of the identity semantic relationship set, the user information anonymization system can effectively extract and reconstruct the identity semantic relationship, and then generate local text with rich identity information, providing strong support for subsequent medical text analysis and research.

[0139] First, the user information anonymization system samples the semantic relationship labels for each text paragraph in the identity-type semantic relationship set. The purpose of this step is to extract the most representative and influential semantic relationship labels from each text paragraph. To achieve this goal, the user information anonymization system uses advanced semantic analysis techniques that can deeply understand the context and semantics in the text, thereby accurately identifying semantic relationships related to identity information. When the number of basic semantic relationship labels in a text paragraph is at least two, the user information anonymization system will further obtain the semantic relationship influence of each basic semantic relationship label. This influence is a quantitative indicator that reflects the importance and influence of the label in the text paragraph. By comparing the semantic relationship influence of different labels, the user information anonymization system can determine the most representative core semantic relationship labels.

[0140] Specifically, if there is only one basic semantic relationship label with the greatest semantic relationship influence, the user information anonymization system will use this label as the core semantic relationship label of the corresponding text paragraph. This label represents the most important and influential identity information in the text paragraph. If there are multiple basic semantic relationship labels with the greatest semantic relationship influence, the user information anonymization system will further consider the mask influence of the identity-type semantic relationship masks corresponding to these labels. Mask influence is an indicator that reflects the importance and uniqueness of the mask in the text. By comparing the mask influences of different masks, the user information anonymization system can determine the mask corresponding to the most influential identity information, and then determine the core semantic relationship label of the corresponding text paragraph.

[0141] Secondly, based on the size of the identity semantic relationship mask corresponding to the corresponding core semantic relationship label in each text paragraph, the user information anonymization system performs mask reconstruction processing. The purpose of this step is to reconstruct and optimize the identity semantic relationship to generate more accurate and complete identity information. To achieve this goal, the user information anonymization system adopts advanced mask reconstruction technology. These technologies can perform fine reconstruction and processing of the mask according to the size and characteristics of the mask, thereby restoring the original identity information. Through mask reconstruction processing, the user information anonymization system can obtain a set of identity semantic relationships after mask reconstruction. The identity information in this set is more accurate and complete, providing strong support for subsequent text retelling processing.

[0142] Then, the user information anonymization system performs dynamic text paraphrase processing on the identity semantic relationship set after mask reconstruction. The purpose of this step is to generate multiple first partial paraphrase texts containing identity semantic relationships. In order to achieve this goal, the user information anonymization system adopts advanced text paraphrase technology. These technologies can generate paraphrase texts with different expressions but retain the original semantics based on the identity information in the identity semantic relationship set. Through dynamic text paraphrase processing, the user information anonymization system can obtain multiple first partial paraphrase texts containing identity semantic relationships. These texts have rich identity information and diverse expressions, which provides more possibilities for subsequent optimization and processing.

[0143] Finally, the user information anonymization system optimizes the first partial repetition texts belonging to the same individual label according to the individual labels to which each first partial repetition text belongs. The purpose of this step is to further improve the accuracy and readability of the repetition text. In order to achieve this goal, the user information anonymization system adopts advanced text optimization technology. These technologies can perform fine optimization and processing of the repetition text according to individual labels, thereby generating more accurate and readable partial texts. Through optimization processing, the user information anonymization system can obtain the first partial text including identity-type semantic relationships. This text not only contains rich identity information, but also has accurate expression and strong readability, providing strong support for subsequent medical text analysis and research.

[0144] In the application scenario of implantable neuromodulation therapy, this embodiment has important significance and value. First, through semantic relationship label sampling and mask reconstruction processing, the user information anonymization system can effectively extract and reconstruct identity semantic relationships and generate local texts with rich identity information. This provides doctors with more accurate and comprehensive patient identity information support, which helps doctors better understand the patient's condition and treatment process. Secondly, through dynamic text retelling processing and text optimization technology, the user information anonymization system can generate multiple retelling texts with diverse expressions but retaining the original semantics. This not only increases the diversity and complexity of the text, but also provides more possibilities for subsequent text analysis and research. Finally, the application of this embodiment also helps to promote the development and application of the field of implantable neuromodulation therapy. Through in-depth mining and analysis of medical texts, the user information anonymization system can better understand the details and dynamic changes of the treatment process, thereby providing doctors with more accurate and comprehensive medical information support. At the same time, through effective privacy protection processing, the user information anonymization system can also ensure that the privacy rights and interests of patients are fully protected, thereby enhancing patients' trust and satisfaction with medical services.

[0145] In some optional embodiments, the filtering of multiple groups of key text data from the key text data set includes: filtering multiple target key text data from the key text data set; taking each target key text data as a benchmark, filtering a target number of initial key text data from the key text data set through a preset semantic transfer strategy and a preset cycle step; classifying each target key text data and the initial key text data corresponding to each target key text data into the same group to obtain the multiple groups of key text data.

[0146] In detail, the filtering of multiple target key text data from the key text data set includes: filtering the subject text data from the current user neural regulation record text; filtering multiple target key text data with different semantic differences from the subject text data from the key text data set; wherein the multiple target key text data include text data with the lowest semantic difference from the subject text data in the key text data set.

[0147] In actual application, for the application scenario of implantable neuromodulation therapy, the above embodiment aims to screen out multiple groups of key text data with specific semantic associations from a huge key text data set. The core of this is that through refined screening and processing, the user information anonymization system can effectively identify multiple groups of key text data that are closely related to the current user's neuromodulation record text, providing strong support for subsequent medical text analysis and research.

[0148] First, the user information anonymization system screens multiple target key text data from the key text dataset. The purpose of this step is to select text data with different semantic differences from the current user's neural regulation record text (the user information anonymization system calls it "topic text data") from the entire dataset. To achieve this goal, the user information anonymization system uses advanced semantic analysis technologies. These technologies can deeply understand the context and semantics in the text, so as to accurately calculate the semantic differences between texts.

[0149] Specifically, the user information anonymization system selects the subject text data from the current user's neuromodulation record text. These text data represent the core content and key information of the user's neuromodulation treatment. Then, the user information anonymization system uses a preset semantic analysis algorithm to compare the subject text data with each text data in the key text data set and calculate the semantic difference between them. In this way, the user information anonymization system can find multiple target key text data with different semantic differences from the subject text data.

[0150] It is worth noting that these target key text data not only include texts with large semantic differences from the subject text data, but also specifically include text data with the lowest semantic differences from the subject text data. This is because, in the research of implantable neuromodulation therapy, the user information anonymization system needs to understand cases that are very different from the user's treatment records to broaden the research perspective, and also needs to pay attention to cases that are highly similar to the user's treatment records to deeply analyze the commonalities and characteristics of the treatment.

[0151] Secondly, taking each target key text data as the benchmark, the user information anonymization system selects the target number of initial key text data from the key text data set through the preset semantic transfer strategy and the preset cycle step. The purpose of this step is to further expand and enrich the content of each set of key text data so that it contains more texts that are semantically related to the target key text data.

[0152] To achieve this goal, the user information anonymization system adopts a semantic transitive strategy. This strategy is based on the semantic association and transitivity between texts. By calculating the semantic similarity and association between texts, it finds texts that have a close semantic connection with the target key text data. At the same time, the user information anonymization system also sets a preset cycle step to control the scope and accuracy of the screening process. In this way, the user information anonymization system can screen out a certain amount of initial key text data from the key text data set for each target key text data.

[0153] Then, the user information anonymization system classifies each target key text data and the initial key text data corresponding to each target key text data into the same group, and obtains multiple groups of key text data. The purpose of this step is to group the screened text data according to specific semantic associations for subsequent analysis and processing.

[0154] Through this method, the user information anonymization system can obtain multiple groups of key text data with specific semantic associations. Each group of data contains a target key text data and initial key text data with close semantic associations. This grouping method not only helps the user information anonymization system to deeply understand the semantic features and associated information of each target key text data, but also provides rich text data support for the user information anonymization system to conduct more in-depth medical text analysis and research.

[0155] In the application scenario of implantable neuromodulation therapy, this technical solution has important significance and value. First, through refined screening and processing, the user information anonymization system can effectively identify multiple groups of key text data closely related to the current user's neuromodulation record text. This provides doctors with more comprehensive and accurate medical text information support, which helps doctors better understand the patient's condition and treatment process. Secondly, through the setting of semantic transmission strategy and preset cycle step, the user information anonymization system can further expand and enrich the content of each group of key text data, so that it contains more texts that are semantically associated with the target key text data. This not only increases the diversity and complexity of the text, but also provides more possibilities for subsequent text analysis and research. Finally, the application of this technical solution will also help promote the development and application of implantable neuromodulation therapy. Through in-depth mining and analysis of medical texts, the user information anonymization system can better understand the details and dynamic changes of the treatment process, thereby providing doctors with more accurate and comprehensive medical information support. At the same time, through effective text screening and grouping processing, the user information anonymization system can also improve the efficiency and accuracy of medical text analysis, and provide more powerful support for the research and application of implantable neuromodulation therapy.

[0156] In some preferred embodiments, the semantic value recognition weight includes a clinical guidance semantic recognition variable; the semantic value weight of the current user's neural regulation record text is identified based on each group of key text data to obtain the semantic value recognition weight corresponding to each group of key text data, including: determining a semantic feature association map between the clinical guidance semantic variables of each group of key text data and the implantation regulation operation semantic variables; determining the clinical guidance semantic variables of the current user's neural regulation record text based on the semantic feature association map corresponding to each group of key text data and the implantation regulation operation semantic variables of the current user's neural regulation record text to obtain the clinical guidance semantic recognition variables corresponding to each group of key text data.

[0157] In the embodiment of the present invention, the semantic value recognition weight is a comprehensive indicator, which includes multiple variables that affect the semantic value of the text. Among these variables, the clinical guidance semantic recognition variable is a crucial component. It represents the richness and accuracy of the clinical guidance information in the text, and is of great significance for evaluating the practical application value of the text in implantable neuromodulation therapy.

[0158] Next, the user information anonymization system identifies the semantic value weight of the current user's neural regulation record text based on each set of key text data. The purpose of this step is to find out the key text data with the highest semantic value for the current user's neural regulation record text through in-depth analysis and comparison. To achieve this goal, the user information anonymization system uses advanced semantic analysis technologies that can deeply understand the context and semantics in the text, thereby accurately calculating the semantic associations and differences between texts.

[0159] Specifically, the user information anonymization system first determines the semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each set of key text data. This map is a complex network structure that shows the association and mutual influence between various semantic variables in the text. By constructing this map, the user information anonymization system can clearly see the distribution and association of clinical guidance semantic variables and implantation regulation operation semantic variables in the text, providing strong support for subsequent analysis and processing.

[0160] Secondly, the user information anonymization system determines the clinical guidance semantic variables of the current user's neuromodulation record text based on the semantic feature association map corresponding to each set of key text data and the implanted control operation semantic variables of the current user's neuromodulation record text. The purpose of this step is to find out those semantic variables with the highest clinical guidance value in the current user's neuromodulation record text. By comparing and analyzing these variables with the clinical guidance semantic variables in each set of key text data, the user information anonymization system can accurately evaluate the semantic value of each set of key text data to the current user's neuromodulation record text.

[0161] Finally, the user information anonymization system obtains the clinical guidance semantic recognition variables corresponding to each set of key text data. This variable is a comprehensive indicator that reflects the clinical guidance value of each set of key text data in the current user's neuromodulation record text. By comparing the clinical guidance semantic recognition variables of different sets of key text data, the user information anonymization system can find out the key text data with the highest semantic value for the current user's neuromodulation record text, providing strong support for subsequent medical text analysis and research.

[0162] In the application scenario of implantable neuromodulation therapy, firstly, by introducing the clinical guidance semantic recognition variable, the user information anonymization system can more accurately evaluate the actual application value of the text in implantable neuromodulation therapy. This is very important information for doctors because it can help them better understand the differences and advantages and disadvantages of different texts in clinical guidance, so as to make more informed decisions.

[0163] Secondly, by building a semantic feature association graph, the user information anonymization system can gain a deeper understanding of the associations and mutual influences between various semantic variables in the text. This is a very important tool for text analysis and processing, because it can help the user information anonymization system understand the semantics and context of the text more accurately, thereby extracting more useful and accurate information.

[0164] Finally, the application of this embodiment also helps to promote the development and application of the field of implantable neuromodulation therapy. Through in-depth mining and analysis of medical texts, the user information anonymization system can better understand the details and dynamic changes of the treatment process, thereby providing doctors with more accurate and comprehensive medical information support. At the same time, through effective text screening and semantic value recognition, the user information anonymization system can also improve the efficiency and accuracy of medical text analysis, and provide more powerful support for the research and application of implantable neuromodulation therapy.

[0165] In other preferred embodiments, the method of determining the semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data includes: determining a first key text data cluster and a second key text data cluster from each group of key text data; wherein, there are differences in the key text data contained in the first key text data cluster and the second key text data cluster; determining the mean clinical guidance semantic features and the mean implantation regulation operation semantic features of the first key text data cluster, and determining the mean clinical guidance semantic features and the mean implantation regulation operation semantic features of the second key text data cluster; based on the mean clinical guidance semantic features and the mean implantation regulation operation semantic features corresponding to the first key text data cluster and the second key text data cluster respectively, determining the unique hot semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data.

[0166] It can be understood that this embodiment is applied to the intersection of implantable neuromodulation therapy and data anonymization. The uniqueness of this application scenario is that it requires the user information anonymization system to not only reveal the complex association between clinical guidance semantic variables and implanted regulation operation semantic variables, but also ensure that the patient's sensitive information is strictly protected in the process to achieve effective data anonymization.

[0167] First, the user information anonymization system selects the first key text data cluster and the second key text data cluster from each set of key text data. The division of these two data clusters is based on the differences between the text content, which is achieved through advanced text mining and clustering techniques. The first key text data cluster may focus mainly on the specific operation process, technical details and immediate reactions of patients for implantable neuromodulation therapy, while the second key text data cluster may be more about the long-term effects of treatment, the patient's rehabilitation progress and the adjustment of clinical guidance strategies. Although the two differ in content, they together constitute the cornerstone of the user information anonymization system's comprehensive understanding of implantable neuromodulation therapy.

[0168] Next, the user information anonymization system proceeded to determine the mean features of clinical guidance semantics and implantation control operation semantics for these two key text data clusters. The core of this step is to extract the "average" semantics that can represent the entire data cluster by calculating the common semantic features of all texts in each data cluster. This mean feature not only captures the main semantic tendencies in the data cluster, but also provides a quantitative basis for subsequent comparison and analysis of the user information anonymization system.

[0169] After determining the mean features of the two key text data clusters, the user information anonymization system constructs a unique hot semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each set of key text data based on these features. The use of unique hot encoding enables the user information anonymization system to express the association between different semantic variables in an intuitive and easy-to-handle way. In this map, each row represents a key text data cluster, and each column represents a specific semantic variable. Each element in the map reflects the unique hot encoding value of the corresponding data cluster on the corresponding semantic variable, thereby clearly revealing the semantic differences and connections between different data clusters in clinical guidance and implantation regulation operations.

[0170] However, in the process of building this association graph, the user information anonymization system always keeps in mind the importance of data anonymity. In order to protect the privacy of patients, the user information anonymization system adopts multiple encryption and anonymization processing technologies to ensure that the patient's sensitive information is strictly protected at every stage of text data processing and storage. The user information anonymization system uses advanced hash functions and pseudonym technology to completely separate the patient's real identity information from the text data, so that even in the most extreme data leakage cases, the text data cannot be linked to specific individual patients.

[0171] In addition, the user information anonymization system also pays special attention to the privacy risks that may be introduced when constructing the semantic feature association graph. In order to avoid indirectly leaking the patient's personal information through graph analysis, the user information anonymization system performs additional privacy protection on each element in the graph. The user information anonymization system uses differential privacy technology to inject an appropriate amount of random noise into the graph while maintaining the overall structure and analytical value of the graph to ensure that even the slightest data changes cannot be used to infer the patient's specific information.

[0172] This embodiment shows great potential and value in the application scenarios of implantable neuromodulation therapy and data anonymity. First, by constructing a semantic feature association map, the user information anonymization system can more deeply understand the complex associations between different key text data clusters in clinical guidance and implant regulation operations. This not only provides doctors with more comprehensive and detailed treatment guidance, but also provides researchers with valuable data support to promote the continuous development and improvement of implantable neuromodulation therapy technology.

[0173] Secondly, through strict data anonymization, the user information anonymization system ensures that the patient's sensitive information is fully protected during the process. This not only complies with the requirements of medical ethics and laws, but also enhances patients' trust in medical services and helps improve the overall quality and efficiency of medical services.

[0174] Finally, the application of this embodiment also helps to promote the development of medical informationization and intelligence. Through in-depth mining and analysis of medical texts, the user information anonymization system can extract more valuable information and knowledge, and provide more accurate and comprehensive data support for medical decision-making. At the same time, through effective data anonymization processing, the user information anonymization system can also ensure the security of these sensitive information during sharing and transmission, providing a more solid foundation for the development of medical informationization and intelligence.

[0175] In summary, this embodiment can not only help the user information anonymization system to more deeply understand the complex associations between different key text data clusters, but also ensure that the patient's sensitive information is fully protected in the process. It is believed that with the continuous advancement of technology and the continuous expansion of application scenarios, this embodiment will play a more important role in future medical text analysis and research, and provide more powerful support for the development and application of implantable neuromodulation therapy.

[0176] In a further embodiment, determining the first key text data cluster and the second key text data cluster from each group of key text data includes: screening a first number of key text data from each group of key text data by a descending rule of the corresponding implanted regulatory operation semantic variable, and generating the first key text data cluster based on the first number of key text data; screening a second number of key text data from each group of key text data by an ascending rule of the corresponding implanted regulatory operation semantic variable, and generating the second key text data cluster based on the second number of key text data.

[0177] First, the user information anonymity system selects the first key text data cluster from each group of key text data. This process is achieved through the descending rule of the corresponding implantation regulation operation semantic variable. Specifically, the user information anonymity system sorts each group of key text data according to the degree of association between the text data and the implantation regulation operation semantic variable. Then, the user information anonymity system selects a first number of key text data from the sorted text data, and these text data have a higher weight or importance on the implantation regulation operation semantic variable. Based on these screened text data, the user information anonymity system generates the first key text data cluster. This data cluster mainly focuses on those text data that are closely related to the implantation regulation operation, and provides the user information anonymity system with valuable information about the specific operation and effect of implantable neuromodulation therapy.

[0178] Next, the user information anonymization system uses a similar but opposite method to determine the second key text data cluster. This time, the user information anonymization system uses the ascending rule of the implanted regulatory operation semantic variable to filter the text data. This means that the user information anonymization system selects from each set of key text data those text data with lower weight or importance on the implanted regulatory operation semantic variable. Similarly, the user information anonymization system generates a second key text data cluster based on these filtered text data. This data cluster may involve more background information about the treatment, the patient's overall condition, and the relationship with other medical interventions, providing the user information anonymization system with information about the broader context of implantable neuromodulation therapy.

[0179] After determining these two key text data clusters, the user information anonymization system proceeded to analyze their clinical guidance semantic mean features and implantation control operation semantic mean features. The core of this step is to extract the "average" semantics that can represent the entire data cluster by calculating the common semantic features of all texts in each data cluster. This mean feature not only captures the main semantic tendencies in the data cluster, but also provides a quantitative basis for the subsequent comparison and analysis of the user information anonymization system. Through in-depth analysis of the mean features of these two data clusters, the user information anonymization system can better understand which semantic features are more prominent or representative in terms of clinical guidance and implantation control operations.

[0180] However, throughout the process, the user information anonymization system always keeps in mind the importance of data anonymity. In order to protect the privacy of patients, the user information anonymization system adopts multiple encryption and anonymization processing technologies. In each step of screening and generating key text data clusters, the user information anonymization system ensures that the patient's sensitive information is completely separated from the text data. The user information anonymization system uses advanced hash functions and pseudonym technology to effectively isolate the patient's real identity information from the text data, making it impossible to link the text data to a specific individual patient even in the most extreme data leakage situation.

[0181] In addition, the user information anonymization system also pays special attention to the privacy risks that may be introduced when constructing and analyzing key text data clusters. In order to avoid indirect leakage of patients' personal information through data cluster analysis, the user information anonymization system has carried out strict privacy protection design for the entire analysis process. The user information anonymization system uses differential privacy technology to inject an appropriate amount of random noise into the analysis process while maintaining the overall structure and analysis value of the data cluster. This approach ensures that even the slightest data changes cannot be used to infer the patient's specific information, thereby further enhancing data anonymity.

[0182] With this design, first, by constructing and analyzing two key text data clusters, the user information anonymization system can more comprehensively understand which semantic features are more prominent or representative in terms of clinical guidance and implant regulation operations. This not only provides doctors with more detailed and comprehensive treatment guidance, but also provides researchers with valuable data support to promote the continuous development and improvement of implantable neuromodulation treatment technology.

[0183] Secondly, through strict data anonymization, the user information anonymization system ensures that the patient's sensitive information is fully protected in the process. This not only complies with the requirements of medical ethics and laws, but also enhances patients' trust in medical services. Patients can safely share their medical experiences and data and contribute to medical research and treatment improvements without worrying about personal information being abused or leaked.

[0184] Finally, the application of this embodiment also helps to promote the development of medical informationization and intelligence. Through in-depth mining and analysis of medical texts, the user information anonymization system can extract more valuable information and knowledge, and provide more accurate and comprehensive data support for medical decision-making. At the same time, through effective data anonymization processing, the user information anonymization system can also ensure the security of these sensitive information during sharing and transmission, providing a more solid foundation for the development of medical informationization and intelligence.

[0185] In some optional embodiments, determining the semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data includes: performing semantic mining on the annotated implantation regulation operation semantic variables and the annotated clinical guidance semantic variables of the key user neuroregulation record text to which each group of key text data belongs, respectively, to obtain the implantation regulation operation description vector and the clinical guidance viewpoint description vector of the key user neuroregulation record text; updating the implantation regulation operation semantic variables and the clinical guidance semantic variables of the key user neuroregulation record text based on the implantation regulation operation description vector and the clinical guidance viewpoint description vector of the key user neuroregulation record text, respectively, to obtain the implantation regulation operation update semantics and the clinical guidance update semantics of the key user neuroregulation record text; obtaining the implantation regulation operation update semantics and the clinical guidance update semantics of each group of key text data from the implantation regulation operation update semantics and the clinical guidance update semantics of the key user neuroregulation record text; and determining the semantic feature association map corresponding to each group of key text data based on the implantation regulation operation update semantics and the clinical guidance update semantics of each group of key text data.

[0186] Next, the implantation control operation update semantics and clinical guidance update semantics of each group of key text data are obtained from the implantation control operation update semantics and clinical guidance update semantics of the key user neural regulation record text, including: extracting the implantation control operation update semantics of the key user neural regulation record text based on the semantic feature recognition corresponding to the clinical guidance update semantics of the key user neural regulation record text; and obtaining the implantation control operation update semantics of each group of key text data from the extracted implantation control operation update semantics.

[0187] It is understandable that in the implantable neuromodulation therapy scenario, the semantic analysis process before data anonymization is intended to balance privacy protection and data availability, ensuring that the value of medical text data is maximized and utilized while protecting patient privacy.

[0188] First, the user information anonymization system conducted in-depth semantic mining on the key user neuromodulation record texts belonging to each set of key text data. This step includes a detailed analysis of the text data labeled with semantic variables for implantation regulation operations and semantic variables for clinical guidance. Through advanced semantic mining technology, the user information anonymization system can extract description vectors of implantation regulation operations and viewpoint description vectors of clinical guidance from these texts. These two vectors represent the specific description of implantation regulation operations and the main viewpoints of clinical guidance in the text, respectively, providing rich semantic information for subsequent analysis by the user information anonymization system.

[0189] Next, the user information anonymization system updates the implantation control operation semantic variables and clinical guidance semantic variables of the key user neuroregulation record text based on these two vectors. The purpose of this step is to incorporate the latest information and opinions in the text into the semantic variables, so that the user information anonymization system can more accurately grasp the semantic content of the text's clinical guidance and implantation control operations. By updating the semantic variables, the user information anonymization system obtains the updated semantics of implantation control operations and clinical guidance updates for the key user neuroregulation record text. These two updated semantic variables not only contain the original information in the text, but also incorporate the latest descriptions and opinions, providing a more comprehensive and accurate semantic basis for the subsequent analysis of the user information anonymization system.

[0190] Then, the user information anonymization system obtains the corresponding semantics of each set of key text data from the implantation control operation update semantics and clinical guidance update semantics of the key user neural regulation record text. This step is to achieve the transition from individual text to the overall data set, which requires the user information anonymization system to extract representative semantic information from a large amount of individual text. In order to achieve this goal, the user information anonymization system extracts the implantation control operation update semantics based on the semantic feature recognition corresponding to the clinical guidance update semantics. Through this extraction process, the user information anonymization system can extract those implantation control operation semantics that are closely related to clinical guidance and representative from a large amount of text data.

[0191] After extracting the implantation control operation update semantics, the user information anonymization system further extracts the implantation control operation update semantics of each set of key text data. This step is to associate the extracted semantic information with the specific text data set, so that the user information anonymization system can accurately grasp the semantic features of each set of text data in terms of clinical guidance and implantation control operation. Through this association process, the user information anonymization system obtains the implantation control operation update semantics and clinical guidance update semantics of each set of key text data, providing a solid foundation for the construction of the subsequent semantic feature association map.

[0192] Based on the above-mentioned semantics of implantation control operation update and clinical guidance update, the user information anonymization system further determines the semantic feature association map corresponding to each set of key text data. This step is to visualize and structure the previous analysis results so that the user information anonymization system can more intuitively understand the semantic associations of different text data in clinical guidance and implantation control operations. By constructing a semantic feature association map, the user information anonymization system can clearly see the semantic connections and differences between different text data, providing doctors and researchers with more comprehensive and detailed guidance information.

[0193] In the process of building the semantic feature association graph, the user information anonymization system pays special attention to the design of privacy protection. First, the user information anonymization system ensures that all text data is strictly anonymized before semantic analysis, so that the patient's personal information is completely separated from the text data. The user information anonymization system uses advanced hash functions and pseudonym technology to effectively isolate the patient's real identity information from the text data, ensuring that even in the most extreme data leakage cases, the text data cannot be linked to specific individual patients.

[0194] Secondly, when constructing and analyzing the semantic feature association graph, the user information anonymization system also pays special attention to the privacy risks that may be introduced. In order to avoid indirectly leaking the patient's personal information through graph analysis, the user information anonymization system has carried out strict privacy protection design for the entire analysis process. The user information anonymization system uses differential privacy technology to inject an appropriate amount of random noise into the analysis process while maintaining the overall structure and analysis value of the graph. This approach ensures that even the slightest data changes cannot be used to infer the patient's specific information, thereby further enhancing the anonymity of the data.

[0195] Finally, the user information anonymization system also effectively combines this embodiment with data anonymization technology. After determining the semantic feature association map of each group of key text data, the user information anonymization system further anonymizes these text data. The user information anonymization system uses data anonymization technology based on homomorphic encryption, so that while protecting patient privacy, it can still perform effective semantic analysis and mining on the encrypted data. The introduction of this technology not only improves the security of the data, but also further expands the application scenarios and value of the data.

[0196] This design not only helps the user information anonymization system to more intuitively understand the semantic associations of different text data in clinical guidance and implant regulation operations, but also ensures that the patient's sensitive information is fully protected in the process. Through effective semantic analysis and anonymization processing, the user information anonymization system can balance the needs between privacy protection and data availability, providing new ideas and methods for the processing and analysis of medical text data.

[0197] In some other possible embodiments, the determining the clinical guidance semantic variables of the current user's neuroregulation record text based on the semantic feature association map corresponding to each group of key text data and the implantation regulation operation semantic variables of the current user's neuroregulation record text to obtain the clinical guidance semantic identification variables corresponding to each group of key text data includes: performing semantic mining on the annotated implantation regulation operation semantic variables of the current user's neuroregulation record text to obtain the implantation regulation operation description vector of the current user's neuroregulation record text; updating the implantation regulation operation semantic variables of the current user's neuroregulation record text based on the implantation regulation operation description vector of the current user's neuroregulation record text to obtain the implantation regulation operation update semantics of the current user's neuroregulation record text; determining the clinical guidance semantic variables of the current user's neuroregulation record text based on the semantic feature association map corresponding to each group of key text data and the implantation regulation operation update semantics of the current user's neuroregulation record text to obtain the clinical guidance semantic identification variables corresponding to each group of key text data.

[0198] Based on this embodiment, first, the user information anonymity system conducts in-depth semantic mining on the annotated implantation regulation operation semantic variables of the current user's neural regulation record text. The purpose of this step is to extract specific descriptions and key information about the implantation regulation operation from the text for subsequent analysis and processing. Through advanced semantic mining technology, the user information anonymity system can obtain the implantation regulation operation description vector of the current user's neural regulation record text. This vector not only contains the specific words and phrases about the implantation regulation operation in the text, but also contains the semantic relationships and contextual information between these words and phrases, providing a rich semantic basis for subsequent analysis by the user information anonymity system.

[0199] Secondly, the user information anonymity system updates the semantic variables of the implanted regulation operation of the text based on the description vector of the implanted regulation operation of the current user's neural regulation record text. The purpose of this step is to integrate the latest information and descriptions in the text into the semantic variables, so that the user information anonymity system can more accurately grasp the semantic content of the text regarding the implanted regulation operation. By updating the semantic variables, the user information anonymity system obtains the updated semantics of the implanted regulation operation of the current user's neural regulation record text. This updated semantic variable not only contains the original information in the text, but also incorporates the latest descriptions and opinions, providing a more comprehensive and accurate semantic basis for the subsequent analysis of the user information anonymity system.

[0200] Then, the user information anonymization system uses the semantic feature association map corresponding to each set of key text data and the implantation control operation update semantics of the current user's neuroregulation record text to determine the clinical guidance semantic variables of the record text. This step is the key to achieving the semantic mapping from implantation control operation to clinical guidance. The user information anonymization system analyzes the semantic feature association map to find out the clinical guidance semantic features associated with the implantation control operation update semantics of the current user's neuroregulation record text. These features not only reflect the specific descriptions and opinions about clinical guidance in the text, but also contain the semantic connection and contextual relationship between these descriptions and opinions and implantation control operations.

[0201] After determining the clinical guidance semantic features of the current user's neuromodulation record text, the user information anonymization system further integrates and abstracts these features to obtain the clinical guidance semantic variables of the text. This semantic variable not only contains the specific information and descriptions about clinical guidance in the text, but also contains the semantic relationship and contextual connection between this information and description and the implanted regulation operation. Through this step, the user information anonymization system successfully associates and maps the implanted regulation operation semantics of the current user's neuromodulation record text with the clinical guidance semantics.

[0202] Finally, the user information anonymization system obtains the clinical guidance semantic identification variables corresponding to each set of key text data based on the clinical guidance semantic variables of the current user's neuromodulation record text. This step is the key to achieving the transition from individual text to the overall data set. The user information anonymization system associates and compares the clinical guidance semantic variables of the current user's neuromodulation record text with each set of key text data, and finds out the features and variables that have similar or related clinical guidance semantics to the current user text. These features and variables not only reflect the semantic content and viewpoints of each set of key text data in clinical guidance, but also contain the semantic connection and contextual relationship between these contents and viewpoints and the current user text. Through this step, the user information anonymization system successfully associates and identifies the clinical guidance semantics of the current user's neuromodulation record text with each set of key text data.

[0203] Throughout the entire process, the user information anonymization system pays special attention to the design of privacy protection. First, the user information anonymization system ensures that all text data is strictly anonymized before semantic analysis, so that the patient's personal information is completely separated from the text data. The user information anonymization system uses advanced hash functions and pseudonym technology to effectively isolate the patient's real identity information from the text data, ensuring that even in the most extreme data leakage cases, the text data cannot be linked to a specific individual patient.

[0204] Secondly, when constructing and analyzing the semantic feature association graph, the user information anonymization system also pays special attention to the privacy risks that may be introduced. In order to avoid indirectly leaking the patient's personal information through graph analysis, the user information anonymization system has carried out strict privacy protection design for the entire analysis process. The user information anonymization system uses differential privacy technology to inject an appropriate amount of random noise into the analysis process while maintaining the overall structure and analysis value of the graph. This approach ensures that even the slightest data changes cannot be used to infer the patient's specific information, thereby further enhancing the anonymity of the data.

[0205] Finally, the user information anonymization system also effectively combines this embodiment with data anonymization technology. After determining the clinical guidance semantic recognition variables corresponding to each set of key text data, the user information anonymization system further anonymizes these variables. The user information anonymization system uses data anonymization technology based on homomorphic encryption, so that while protecting patient privacy, it can still perform effective semantic analysis and mining on the encrypted data.

[0206] In other optional embodiments, the privacy desensitizing processing of the current user's neural regulation record text based on the target group key text data includes: obtaining clinical guidance semantic recognition variables corresponding to the target group key text data; and performing semantic fuzzy processing based on the comparison results between the annotated clinical guidance semantic variables of the current user's neural regulation record text and the clinical guidance semantic recognition variables corresponding to the target group key text data.

[0207] It can be understood that this embodiment obtains the clinical guidance semantic recognition variables corresponding to the key text data of the target group, and compares them with the annotated clinical guidance semantic variables of the current user's neuromodulation record text, and then implements semantic fuzzy processing to achieve the purpose of privacy desensitization.

[0208] First, the user information anonymization system needs to obtain the clinical guidance semantic identification variables corresponding to the target group key text data. The target group key text data has been strictly screened and processed in the previous steps, and they represent the key information under a specific clinical scenario or disease type. Through in-depth analysis and mining of these text data, the user information anonymization system can extract the clinical guidance semantic features contained therein, and then form clinical guidance semantic identification variables. These variables not only contain the specific descriptions and opinions about clinical guidance in the text, but also contain the semantic relationships and contextual connections between these descriptions and opinions, which provides important reference and basis for the subsequent privacy desensitization processing of the user information anonymization system.

[0209] Secondly, the user information anonymization system compares the annotated clinical guidance semantic variables of the current user's neuromodulation record text with the clinical guidance semantic identification variables corresponding to the target group's key text data. This step is the key to achieving privacy desensitization. The annotated clinical guidance semantic variables are clinical guidance information that has been professionally annotated and identified in the current user's neuromodulation record text. They represent the specific content and viewpoints about clinical guidance in the text. By comparing the annotated clinical guidance semantic variables of the current user's text with the clinical guidance semantic identification variables of the target group's key text data, the user information anonymization system can find the similarities and differences between the two, and then determine the specific content and scope that need to be privacy desensitized.

[0210] During the comparison process, the user information anonymization system uses advanced semantic comparison algorithms and models to ensure the accuracy and efficiency of the comparison. These algorithms and models can fully consider the semantic relationships and contextual information in the text, making the comparison results more accurate and reliable. Through the comparison, the user information anonymization system can obtain the similarities and differences between the current user's neuromodulation record text and the target group's key text data in clinical guidance semantics, providing an important basis for subsequent privacy desensitization processing.

[0211] Then, the user information anonymization system performs semantic fuzzification based on the comparison results. Semantic fuzzification is one of the core steps of privacy desensitization. Its purpose is to modify or replace sensitive information in the text so that the desensitized text cannot be directly associated with a specific individual or sensitive information while maintaining the original semantics and contextual information. In this step, the user information anonymization system determines the specific content and scope that needs to be blurred based on the comparison results, and uses a variety of semantic fuzzification technologies and methods to achieve this goal.

[0212] For example, for the part of the current user's neuromodulation record text that is highly similar to the target group's key text data in clinical guidance semantics, the user information anonymization system can use synonym replacement, semantic generalization or context reorganization and other methods to perform fuzzy processing. Synonym replacement is to replace sensitive words or phrases in the text with synonyms or phrases with similar or identical semantics to achieve the purpose of fuzzy processing. Semantic generalization is to generalize the specific description or detailed information in the text into more general or abstract concepts or expressions to reduce the specificity and sensitivity of the text. Context reorganization is to cover up sensitive information in the text by changing the structure or order of sentences in the text, or adding some non-sensitive contextual information.

[0213] In addition to the above methods, the user information anonymization system can also use semantic fuzzification technology based on deep learning to achieve more complex and sophisticated fuzzy processing. These technologies use deep learning models to deeply analyze and understand text, and can automatically identify and replace sensitive information in the text while maintaining the semantic and contextual coherence of the text. By using these technologies, the user information anonymization system can achieve more efficient and accurate privacy desensitization processing.

[0214] Finally, after semantic fuzzification processing, the user information anonymization system obtains the desensitized neural regulation record text of the current user. While maintaining the original semantics and contextual information, this text can no longer be directly associated with a specific individual or sensitive information. It can be safely used in subsequent data analysis, mining or sharing scenarios without worrying about leaking the patient's personal privacy information.

[0215] Throughout the privacy desensitization process, the user information anonymization system pays special attention to the balance between privacy protection and data availability. On the one hand, the user information anonymization system ensures that the patient's personal privacy information will not be leaked or abused through strict privacy protection measures; on the other hand, the user information anonymization system also fully considers the availability and value of the data in subsequent applications. Through the application of reasonable semantic fuzzy processing and technical means, the user information anonymization system protects the patient's privacy while retaining the important information and value of the data in subsequent applications.

[0216] In addition, the user information anonymization system also conducted a comprehensive evaluation and test of this embodiment. Through experiments and verification on actual data sets, the user information anonymization system proved the effectiveness and feasibility of this embodiment. The experimental results show that the text data after privacy desensitization can no longer be directly associated with specific individuals or sensitive information while maintaining the original semantics and contextual information. At the same time, these data also show good usability and value in subsequent application scenarios such as data analysis, mining or sharing.

[0217] Such a design can not only help the user information anonymization system effectively protect the patient's personal privacy information from being leaked or abused, but also ensure the availability and value of the data in subsequent applications. Through the application of reasonable semantic fuzzy processing and technical means, the user information anonymization system achieves a balance and win-win situation between privacy protection and data availability.

[0218] In some optional embodiments, there are multiple key text data sets; the privacy desensitizing processing of the current user's neural regulation record text based on the target group key text data includes: determining the target group key text data with the lowest data available disturbance index from the target group key text data corresponding to the multiple key text data sets; and privacy desensitizing processing of the current user's neural regulation record text based on the determined target group key text data.

[0219] In the application scenario involved in this embodiment, there are multiple different data sets, each of which contains a series of key text data. In this context, the process of performing privacy desensitization processing on the current user's neural regulation record text based on the target group key text data requires more sophisticated and complex operations.

[0220] First, the user information anonymization system needs to determine the target group key text data with the lowest data available disturbance index from the target group key text data corresponding to multiple key text data sets. This step is the basis and key to the entire privacy desensitization process. The data available disturbance index is an indicator that measures the extent to which data can retain its original information and value after desensitization. The goal of the user information anonymization system is to find the target group key text data that can retain its original information and value to the greatest extent after desensitization, that is, the target group key text data with the lowest data available disturbance index.

[0221] To achieve this goal, the user information anonymization system needs to conduct in-depth analysis and evaluation of each key text data set. The user information anonymization system will preprocess each data set, including data cleaning, format unification, deduplication and other steps to ensure the accuracy and consistency of the data. Then, the user information anonymization system will use a series of data analysis tools and algorithms to conduct in-depth mining and analysis of each data set, extract key text data, and evaluate the data availability perturbation indicators of these data.

[0222] When evaluating the data availability perturbation index, the user information anonymization system will consider multiple factors, including data integrity, accuracy, consistency, interpretability, etc. The user information anonymization system will also use some specific indicators and measurement methods, such as data loss rate, information entropy, mutual information, etc., to quantify the availability and degree of perturbation of the data. By comprehensively considering these factors and indicators, the user information anonymization system can accurately evaluate and compare the data availability perturbation index of each data set.

[0223] After determining the target group of key text data with the lowest available data disturbance index, the user information anonymization system can perform privacy desensitization on the current user's neural regulation record text based on this data. This step is the core and purpose of the entire technical solution. The goal of the user information anonymization system is to protect user privacy while maintaining the original information and value of the text as much as possible.

[0224] To achieve this goal, the user information anonymization system will use a series of privacy desensitization technologies and methods. First, the user information anonymization system will use text replacement technology to replace sensitive information in the text with some common, meaningless words or symbols to achieve the purpose of desensitization. At the same time, the user information anonymization system will also use semantic generalization technology to generalize the specific details and descriptions in the text into more general and abstract concepts and expressions to reduce the specificity and sensitivity of the text.

[0225] In addition to text replacement and semantic generalization technologies, the user information anonymization system will also use some advanced natural language processing technologies and deep learning models to further improve the effect and quality of privacy masking. For example, the user information anonymization system can use word vector representation technology to convert words in the text into vector representations in a high-dimensional space, and then achieve masking by adjusting the distance and relationship between vectors. The user information anonymization system can also use deep learning models such as generative adversarial networks (GANs) to generate some masked texts that are similar to the original text but do not contain sensitive information.

[0226] When performing privacy masking, the user information anonymization system will pay special attention to maintaining the semantic coherence and consistency of the text. The user information anonymization system will use some semantic analysis and reasoning techniques to ensure that the desensitized text is still semantically coherent and consistent. For example, the user information anonymization system will use contextual information to infer some missing or ambiguous words and expressions to ensure the integrity and accuracy of the text. The user information anonymization system will also use some semantic similarity calculation methods to compare the similarity between different texts to ensure that the desensitized text still maintains a certain degree of semantic similarity and relevance with the original text.

[0227] Finally, after privacy desensitization, the current user's neural regulation record text can be safely used in subsequent data analysis, mining or sharing scenarios. Since the user information anonymization system has strictly desensitized the text, these texts can no longer be directly associated with specific individuals or sensitive information. At the same time, since the user information anonymization system pays special attention to maintaining the original information and value of the text during the desensitization process, these texts still have high usability and value in subsequent application scenarios.

[0228] In this way, by determining the target group of key text data with the lowest data availability disturbance index from multiple key text data sets, and performing privacy desensitization processing on the current user's neural regulation record text based on these data, the goal of protecting user privacy while maintaining the original information and value of the text as much as possible is achieved. This technical solution is not only highly practical and feasible, but also provides new ideas and methods for the balance between data privacy protection and data availability in the future.

[0229] In some further optional embodiments, the privacy-desensitizing processing of the current user's neural regulation record text based on the target group key text data includes: processing the semantic value weight of the current user's neural regulation record text based on the target group key text data to obtain a semantic value decision representation of the current user's neural regulation record text; configuring the index features of the target group key text data to the semantic value decision representation of the current user's neural regulation record text, so that the anonymous state analysis system determines the target group key text data based on the index features in the semantic value decision representation, and obtains the privacy-desensitized text of the current user's neural regulation record text based on the target group key text data.

[0230] It can be understood that this embodiment involves how to perform privacy desensitization processing on the current user's neural regulation record text based on the target group key text data. This process not only involves the processing of text data, but also involves the delicate balance between privacy protection and data utilization.

[0231] First, the user information anonymization system pays attention to the semantic value of the current user's neural regulation record text. Text data, especially neural regulation records, contain rich semantic information, which is of great significance for subsequent analysis and research. However, in the process of privacy desensitization, the user information anonymization system needs to retain this valuable semantic information as much as possible while protecting user privacy. Therefore, the user information anonymization system processes the semantic value weight of the current user's neural regulation record text based on the target group's key text data.

[0232] Processing semantic value weights is a complex and delicate process. The user information anonymization system uses natural language processing technology and machine learning algorithms to conduct an in-depth analysis of the current user's neural regulation record text. By identifying keywords, phrases, and sentences in the text, the user information anonymization system can evaluate their importance and value in the overall semantics. This evaluation is based not only on the frequency and distribution of vocabulary, but also takes into account the semantic relationship and contextual connection between vocabulary. In this way, the user information anonymization system can obtain a representation that reflects the semantic value weight of the current user's neural regulation record text.

[0233] Next, the user information anonymization system further transforms this semantic value weight representation into a semantic value decision representation. This process involves quantification and decision-making of semantic value. The user information anonymization system uses some advanced quantification techniques, such as vector space model and word embedding, to transform the semantic information in the text into computable numerical values. Then, based on these numerical values, the user information anonymization system uses a decision algorithm to determine which semantic information is critical and needs to be retained, and which can be discarded, in order to find the best balance between privacy desensitization and semantic retention.

[0234] However, it is not enough to simply obtain the semantic value decision representation. In order to ensure that the anonymous state analysis system can accurately identify the target group key text data and apply it to the privacy desensitization process, the user information anonymization system also needs to configure the index features of the target group key text data into the semantic value decision representation of the current user's neural regulation record text. Index features are a set of features that can uniquely identify the target group key text data, which may include specific vocabulary, phrases, sentence structures, etc. By embedding these index features into the semantic value decision representation, the user information anonymization system can ensure that the anonymous state analysis system can accurately identify the parts related to the target group key text data when processing the current user's neural regulation record text.

[0235] Finally, the anonymous state analysis system determines the key text data of the target group based on the index features in the semantic value decision representation, and obtains the privacy-desensitized text of the current user's neural regulation record text. This process is the core of privacy desensitization processing. By accurately identifying the parts related to the key text data of the target group, the anonymous state analysis system can perform specific desensitization processing on these parts, such as replacement, deletion or generalization, to ensure that the user's privacy information will not be leaked. At the same time, since the user information anonymity system has previously performed fine processing and decision-making on the semantic value of the text, the desensitized text still retains most of the original semantic information, which is of great significance for subsequent data analysis and research.

[0236] In this way, by processing the semantic value weight of the current user's neural regulation record text based on the target group's key text data, and configuring the index features of the target group's key text data into the semantic value decision representation, the goal of protecting user privacy while retaining text semantic information as much as possible is achieved. This process not only involves complex text processing technology and machine learning algorithms, but also requires in-depth consideration and decision-making on the balance between privacy protection and data utilization. Through this technical solution, the user information anonymization system can obtain desensitized text that protects user privacy and retains most of the original semantic information, providing strong support for subsequent data analysis and research.

[0237] Further, Figure 2 FIG. 2 is a schematic diagram of a user information anonymization system 200 provided in an embodiment of the present invention. Figure 2 The user information anonymization system 200 shown includes a processor 210, and the processor 210 can call and run a computer program from a memory to implement the method in the embodiment of the present invention.

[0238] Alternatively, if Figure 2 As shown, the user information anonymization system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present invention.

[0239] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0240] Alternatively, if Figure 2 As shown, the user information anonymization system 200 may further include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices, specifically, may send information or data to other devices, or receive information or data sent by other devices.

[0241] Optionally, the user information anonymization system 200 may implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or a device deployed with the storage engine in each method of the embodiments of the present invention, which will not be described in detail here for the sake of brevity.

[0242] It should be understood that the processor of the embodiment of the present invention may be an integrated circuit chip with signal processing capability.

[0243] It is understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that the memory of the systems and methods described herein is intended to include but is not limited to suitable types of memory.

[0244] Based on the above, a readable storage medium is provided, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0245] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the embodiment of the present invention is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0246] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0247] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the purpose of the present invention and the scope of protection of the present invention, all of which are within the protection of the present invention.

Claims

1. A method for anonymizing user information of smart medical equipment based on artificial intelligence, characterized in that: The method is applied to a user information anonymization system, and the method comprises: Acquire a current user's neural regulation record text of an implantable neural regulation report to be anonymized, and filter a key text data set of the current user's neural regulation record text from the implantable neural regulation report to be anonymized; A plurality of groups of key text data are screened from the key text data set, and the semantic value weight of the current user's neural regulation record text is identified based on each group of key text data to obtain the semantic value identification weight corresponding to each group of key text data; the semantic value identification weight includes a clinical guidance semantic identification variable; the identification of the semantic value weight of the current user's neural regulation record text based on each group of key text data to obtain the semantic value identification weight corresponding to each group of key text data includes: determining a semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data; determining the clinical guidance semantic variables of the current user's neural regulation record text based on the semantic feature association map corresponding to each group of key text data and the implantation regulation operation semantic variables of the current user's neural regulation record text to obtain the clinical guidance semantic identification variables corresponding to each group of key text data; Based on the semantic value recognition weight corresponding to each group of key text data, the data available disturbance index corresponding to each group of key text data is determined, and a target group of key text data with the lowest data available disturbance index is screened from the multiple groups of key text data; wherein, based on the semantic value recognition weight corresponding to each group of key text data, the data available disturbance index corresponding to each group of key text data is determined, including: based on the semantic value recognition weight corresponding to each group of key text data, identity-type semantic relationship mining and regulation event-type semantic relationship mining are respectively performed on multiple text paragraphs in each group of key text data to obtain an identity-type semantic relationship set and a regulation event-type semantic relationship set; through a first semantic paraphrase model, a first semantic paraphrase processing is performed on the identity-type semantic relationship set, A first partial text including identity-type semantic relations is obtained; a second semantic paraphrase processing is performed on the set of regulatory event-type semantic relations through a second semantic paraphrase model to obtain a second partial text including regulatory event-type semantic relations; a merging process is performed based on the first partial text and the second partial text to obtain an available text data set that matches the target semantic relations in each group of key text data, wherein the target semantic relations include at least one of identity-type semantic relations and regulatory event-type semantic relations; a data available perturbation index is generated for the available text data set using a long short-term memory model to obtain a data available perturbation index corresponding to each group of key text data; and a privacy desensitization process is performed on the current user's neural regulation record text based on the target group of key text data.

2. The method according to claim 1, characterized in that The step of screening multiple groups of key text data from the key text data set includes: Filtering a plurality of target key text data from the key text data set; Taking each target key text data as a benchmark, a target number of initial key text data are screened from the key text data set by using a preset semantic transfer strategy and a preset cycle step; Classifying each target key text data and the initial key text data corresponding to each target key text data into the same group to obtain the multiple groups of key text data; The step of screening a plurality of target key text data from the key text data set includes: Filtering subject text data from the current user's neural regulation record text; A plurality of target key text data having different semantic differences from the subject text data are screened from the key text data set; wherein the plurality of target key text data include text data having the lowest semantic difference from the subject text data in the key text data set.

3. The method according to claim 1, characterized in that The determining of the semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data includes: Determine a first key text data cluster and a second key text data cluster from each group of key text data; wherein the key text data included in the first key text data cluster and the second key text data cluster are different; Determine the mean clinical guidance semantic features and the mean implantation control operation semantic features of the first key text data cluster, and determine the mean clinical guidance semantic features and the mean implantation control operation semantic features of the second key text data cluster; Determine a unique-hot semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data based on the clinical guidance semantic mean features and the implantation regulation operation semantic mean features corresponding to the first key text data cluster and the second key text data cluster respectively; Determining the first key text data cluster and the second key text data cluster from each group of key text data includes: Filtering a first number of key text data from each group of key text data according to a descending rule of the corresponding implanted control operation semantic variable, and generating the first key text data cluster based on the first number of key text data; A second number of key text data is selected from each group of key text data according to the ascending rule of the corresponding implanted control operation semantic variable, and the second key text data cluster is generated based on the second number of key text data.

4. The method according to claim 1, characterized in that The determining of the semantic feature association map between the clinical guidance semantic variables and the implantation regulation operation semantic variables of each group of key text data includes: Semantic mining is performed on the annotated implantation control operation semantic variables and the annotated clinical guidance semantic variables of the key user's neural regulation record text to which each group of key text data belongs, respectively, to obtain the implantation control operation description vector and the clinical guidance viewpoint description vector of the key user's neural regulation record text; Based on the implantation control operation description vector and the clinical guidance viewpoint description vector of the key user's neural regulation record text, the implantation control operation semantic variables and the clinical guidance semantic variables of the key user's neural regulation record text are updated respectively to obtain the implantation control operation update semantics and the clinical guidance update semantics of the key user's neural regulation record text; Acquire the implantation control operation update semantics and clinical guidance update semantics of each group of key text data from the implantation control operation update semantics and clinical guidance update semantics of the key user neural control record text; Determine a semantic feature association map corresponding to each set of key text data based on the implantation control operation update semantics and the clinical guidance update semantics of each set of key text data; The step of obtaining the implantation control operation update semantics and clinical guidance update semantics of each group of key text data from the implantation control operation update semantics and clinical guidance update semantics of the key user neural control record text includes: Extracting the implantation control operation update semantics of the key user's neural control record text based on the semantic feature recognition corresponding to the clinical guidance update semantics of the key user's neural control record text; The implantation regulation operation update semantics of each group of key text data is obtained from the extracted implantation regulation operation update semantics.

5. The method according to claim 1, characterized in that The method of determining the clinical guidance semantic variables of the current user's neural regulation record text based on the semantic feature association map corresponding to each group of key text data and the implantation regulation operation semantic variables of the current user's neural regulation record text, and obtaining the clinical guidance semantic recognition variables corresponding to each group of key text data, includes: Performing semantic mining on the annotated implantation control operation semantic variables of the current user's neural regulation record text to obtain an implantation control operation description vector of the current user's neural regulation record text; Based on the implantation control operation description vector of the current user's neural regulation record text, the implantation control operation semantic variable of the current user's neural regulation record text is updated to obtain the implantation control operation update semantics of the current user's neural regulation record text; Based on the semantic feature association map corresponding to each group of key text data and the implanted regulation operation update semantics of the current user's neural regulation record text, the clinical guidance semantic variables of the current user's neural regulation record text are determined to obtain the clinical guidance semantic identification variables corresponding to each group of key text data.

6. The method according to claim 1, characterized in that The performing privacy desensitization processing on the current user's neural regulation record text based on the target group key text data includes: Acquire clinical guidance semantic recognition variables corresponding to the target group key text data; Semantic fuzzy processing is performed based on the comparison result between the annotated clinical guidance semantic variables of the current user's neural regulation record text and the clinical guidance semantic recognition variables corresponding to the target group key text data.

7. The method according to any one of claims 1 to 6, characterized in that The number of the key text data sets is multiple; The performing privacy desensitization processing on the current user's neural regulation record text based on the target group key text data includes: Determine the target group key text data with the lowest data available disturbance index from the target group key text data corresponding to the multiple key text data sets; The current user's neural regulation record text is privacy-desensitized according to the determined target group key text data.

8. The method according to any one of claims 1 to 6, characterized in that The performing privacy desensitization processing on the current user's neural regulation record text based on the target group key text data includes: Processing the semantic value weight of the current user's neural regulation record text based on the target group key text data to obtain a semantic value decision representation of the current user's neural regulation record text; The index features of the target group key text data are configured into the semantic value decision representation of the current user's neural regulation record text, so that the anonymous state analysis system determines the target group key text data based on the index features in the semantic value decision representation, and obtains the privacy-desensitized text of the current user's neural regulation record text based on the target group key text data.

9. A user information anonymity system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 8.

Images