Sensor monitoring data optimization method and system for smart medical care

Abstract

The present application discloses a sensor monitoring data optimization method and system for smart medical treatment, which belongs to the field of data processing technology. By constructing and debugging the target data optimization strategy matching network, the present application can automatically extract and match the optimal data optimization strategy from the complex sensor monitoring information flow, significantly improving the accuracy and reliability of the data. The network is debugged using past learning information, so that the network can accurately identify defective states in the data, such as noise, abnormal fluctuations, etc., and assign corresponding voting weights to them, thereby ensuring the pertinence and effectiveness of the data optimization strategy. Finally, by determining the data optimization strategy matching viewpoint corresponding to each implantable device sensor monitoring information, that is, the preprocessing strategy keyword, real-time and customized preprocessing of the target vital sign monitoring data is achieved, providing solid data support for in-depth research and application in the fields of smart medical treatment, monitoring and / or regulation and control of implantable devices.

Description

Technical Field

[0001] The present application belongs to the field of data processing technology, and specifically relates to a sensor monitoring data optimization method and system for smart medical care. Background Art

[0002] In frontier fields such as smart healthcare and neural monitoring and control, implantable device technology is increasingly becoming a hot topic in research and practice. However, due to the complexity of the physiological environment and the physical limitations of implantable devices (implantable devices include implantable devices for performing monitoring functions, implantable devices for performing control functions, and implantable devices for performing both monitoring and control functions), the original sensor monitoring information flow often contains a large amount of noise and abnormal data, which seriously restricts the accuracy of subsequent data analysis and application. Although traditional data preprocessing methods can improve data quality to a certain extent, they often lack pertinence and real-time performance and cannot meet the needs of high-precision medical monitoring and control. Therefore, how to achieve efficient and accurate data denoising, abnormality correction and optimization for the original implantable device sensor monitoring information flow, especially the target vital sign monitoring data that needs to be preprocessed, has become a key issue that needs to be urgently solved in the current technological development. Summary of the invention

[0003] The present application provides a method and system for optimizing sensor monitoring data for smart medical care, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0004] The present application embodiment provides a sensor monitoring data optimization method for smart medical care, which is applied to a sensor monitoring data optimization system. The method includes:

[0005] Obtaining the original implantable device sensor monitoring information stream to be optimized;

[0006] Inputting the original implantable device sensor monitoring information flow into a pre-debugged target data optimization strategy matching network to obtain a data optimization strategy matching viewpoint set;

[0007] Determining the data optimization strategy matching viewpoint corresponding to each implantable device sensor monitoring information in the original implantable device sensor monitoring information flow based on the data optimization strategy matching viewpoint set;

[0008] in:

[0009] The original implantable device sensor monitoring information flow includes target vital sign monitoring data that needs to be preprocessed;

[0010] The target data optimization strategy matching network is a data optimization strategy matching network obtained by inputting past learning information into the original data optimization strategy matching network to be debugged for debugging, the past learning information includes a batch of past data quality characterization vectors respectively mined from a batch of past implantable device sensor monitoring information and data defect states and voting weights corresponding to each past data quality characterization vector, the data defect state reflects an abnormal state keyword of a past data quality characterization vector, the voting weight reflects a reinforcement coefficient corresponding to the abnormal state keyword, the batch of past implantable device sensor monitoring information is a batch of uninterrupted implantable device sensor monitoring information, and each past data quality characterization vector includes a data quality characterization vector determined by a past implantable device sensor monitoring information and a previous past implantable device sensor monitoring information of the past implantable device sensor monitoring information;

[0011] The data optimization strategy matching viewpoint is used to represent the preprocessing strategy keywords of the target vital sign monitoring data.

[0012] In one implementation, a first characterization vector is mined on the batch of past implantable device sensor monitoring information to obtain a first batch of data quality characterization vectors, wherein a data quality characterization vector in the first batch of data quality characterization vectors corresponds to a piece of past implantable device sensor monitoring information in the batch of past implantable device sensor monitoring information;

[0013] Performing second characterization vector mining on the batch of past implantable device sensor monitoring information to obtain a second batch of data quality characterization vectors, wherein a first data quality characterization vector in the second batch of data quality characterization vectors corresponds to a first past implantable device sensor monitoring information in the past implantable device sensor monitoring information, a previous frame of implantable device sensor monitoring information of the first past implantable device sensor monitoring information is the second past implantable device sensor monitoring information, and the first data quality characterization vector represents contextual embedded semantics between the first past implantable device sensor monitoring information and the second past implantable device sensor monitoring information;

[0014] The first batch of data quality characterization vectors and the second batch of data quality characterization vectors are respectively integrated to obtain the batch of past data quality characterization vectors.

[0015] In one implementation, the performing second characterization vector mining on the batch of past implantable device sensor monitoring information to obtain a second batch of data quality characterization vectors includes:

[0016] The second characterization vector mining is performed on the batch of past implantable device sensor monitoring information through the following steps to obtain a second batch of data quality characterization vectors, wherein the past implantable device sensor monitoring information for which the second characterization vector mining is performed each time is used as the current past implantable device sensor monitoring information, and the obtained data quality characterization vector is used as the current data quality characterization vector, and the second batch of data quality characterization vectors includes the current data quality characterization vector;

[0017] Performing a sensor monitoring information analysis operation on the current past implantable device sensor monitoring information to determine a first batch of distribution feature variables, wherein the first batch of distribution feature variables is used to describe the area of ​​the target vital sign monitoring data in the current past implantable device sensor monitoring information;

[0018] Performing a sensor monitoring information analysis operation on the previous past implantable device sensor monitoring information of the current past implantable device sensor monitoring information to determine a second batch of distribution feature variables, wherein the second batch of distribution feature variables is used to describe the area of ​​the target vital sign monitoring data in the previous past implantable device sensor monitoring information;

[0019] The current data quality characterization vector is determined based on the first batch of distribution feature variables and the second batch of distribution feature variables, wherein the current data quality characterization vector is used to represent the difference between corresponding distribution feature variables in the first batch of distribution feature variables and the second batch of distribution feature variables.

[0020] In one implementation, the method further includes:

[0021] The data defect state and voting weight corresponding to each past data quality characterization vector in the batch of past data quality characterization vectors are determined respectively by the following steps, wherein the past data quality characterization vector for determining the data defect state and voting weight each time is used as the current past data quality characterization vector, and the data defect state corresponding to the current past data quality characterization vector is used as the current data defect state and the current voting weight:

[0022] Inputting the current past data quality characterization vector into the original data optimization strategy matching network to obtain the current data defect state, wherein the original data optimization strategy matching network is a data optimization strategy matching network obtained by performing a reset process in advance;

[0023] The current voting weight is determined based on the current data defect state and the prior learning annotation corresponding to the current past data quality characterization vector, wherein the prior learning annotation carries the authentication tag in the past implantable device sensor monitoring information corresponding to the current past data quality characterization vector, and the authentication tag includes the target preprocessing strategy keywords of the target vital sign monitoring data, and the voting weight is used to indicate whether the current data defect state matches the target preprocessing strategy keywords.

[0024] In one implementation, the step of inputting the current past data quality characterization vector into the original data optimization strategy matching network to obtain the current data defect state includes:

[0025] Input each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set and the current past data quality characterization vector into the first discrimination algorithm branch in order to obtain a first batch of quantitative discrimination indicators, wherein the first batch discrimination indicators include the quantitative discrimination indicators corresponding to each preprocessing strategy keyword;

[0026] The candidate preprocessing strategy keyword corresponding to the first quantitative discrimination indicator with the largest indicator value in the first batch discrimination indicators is determined as the current preprocessing strategy keyword corresponding to the current data defect state.

[0027] In one implementation, before inputting each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set in sequence together with the current past data quality characterization vector into the first discrimination algorithm branch to obtain the first batch of quantitative discrimination indicators, the method further includes:

[0028] Determine the thermodynamic characteristic variables through the target mode;

[0029] Determining whether the thermal characteristic variables conform to a set thermal distribution rule;

[0030] On the basis that the thermal characteristic variables conform to the set thermal distribution rules, dynamically extracting candidate preprocessing strategy keywords from the pre-stored preprocessing strategy keyword set as the current data defect state;

[0031] On the basis that the thermal characteristic variable does not conform to the set thermal distribution rule, the current past data quality characterization vector is input into the original data optimization strategy matching network to obtain the current data defect state.

[0032] In one implementation, determining the candidate preprocessing strategy keyword corresponding to the first quantitative discrimination indicator with the largest indicator value in the first batch discrimination indicators as the current preprocessing strategy keyword corresponding to the current data defect state includes:

[0033] On the basis that the alternative preprocessing strategy keywords corresponding to the first quantization discrimination index are the same as the target preprocessing strategy keywords, the current voting weight is determined as the first voting weight, where the first voting weight indicates that the original data optimization strategy matching network successfully outputs a matching result;

[0034] On the basis that the alternative preprocessing strategy keywords corresponding to the first quantization discrimination index are different from the target preprocessing strategy keywords, the current voting weight is determined as the second voting weight, where the second voting weight indicates that the original data optimization strategy matching network does not successfully output a matching result.

[0035] In one implementation, the method further includes:

[0036] Screen a number of past linkage monitoring records from the past learning information, where the i-th past linkage monitoring record among the number of past linkage monitoring records includes the i-th past data quality characterization vector, the i-th data defect state corresponding to the i-th past data quality characterization vector, the i-th voting weight, and the (i + 1)-th past data quality characterization vector, and i is a positive integer;

[0037] Debug the original data optimization strategy matching network to be debugged based on the number of past linkage monitoring records to obtain the target data optimization strategy matching network. Among them, on the basis that the number of iterations for debugging the original data optimization strategy matching network reaches the set iteration threshold, the original data optimization strategy matching network is determined as the target data optimization strategy matching network. On the basis that the number of iterations for debugging the original data optimization strategy matching network does not reach the set iteration threshold, improve the neural network weights of the original data optimization strategy matching network according to the pre-configured training error expression. The input of each round of debugging process is one of the number of past linkage monitoring records.

[0038] In one implementation, the method further includes:

[0039] Debug the original data optimization strategy matching network to be debugged based on multiple groups of past linkage monitoring records through the following steps to obtain the target data optimization strategy matching network, where the past linkage monitoring record input to the original data optimization strategy matching network is the i-th linkage monitoring record:

[0040] Determine whether the past implantable device sensing monitoring information corresponding to the (i + 1)-th past data quality characterization vector is the last past implantable device sensing monitoring information in the past implantable device sensing monitoring information flow;

[0041] On the basis that the past implantable device sensor monitoring information corresponding to the i+1th past data quality characterization vector is the last past implantable device sensor monitoring information, determining the training parameters generated by the original data optimization strategy matching network based on the i-th voting weight;

[0042] On the basis that the past implantable device sensor monitoring information corresponding to the i+1th past data quality characterization vector is not the last past implantable device sensor monitoring information, the training parameter is determined based on the i-th voting weight and the output of the associated data optimization strategy matching network, wherein the associated data optimization strategy matching network is a data optimization strategy matching network obtained by performing a reset process in advance, and the neural network weights of the associated data optimization strategy matching network and the original data optimization strategy matching network are different;

[0043] Determining an error variable of the training error expression based on the training parameter and the output of the original data optimization strategy matching network, and adjusting a neural network weight of the original data optimization strategy matching network based on the error variable of the training error expression;

[0044] On the basis that the number of iterations of executing the above steps reaches the set iteration threshold, the original data optimization strategy matching network is determined as the target data optimization strategy matching network.

[0045] In one implementation, the determining the training parameter based on the output of the i-th voting weight and the associated data optimization strategy matching network includes:

[0046] Input each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set and the i+1th past data quality characterization vector into the second discrimination algorithm branch in order to obtain a second batch discrimination index, wherein the second batch discrimination index includes a quantitative discrimination index corresponding to each preprocessing strategy keyword;

[0047] A weighted result of a second quantitative discrimination index having the largest index value among the second batch discrimination indexes and the i-th voting weight is determined as the training parameter.

[0048] An embodiment of the present application provides a sensor monitoring data optimization 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.

[0049] An embodiment of the present application 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.

[0050] In the embodiments of the present application, for the original implanted device sensing and monitoring information flow, especially the target vital sign monitoring data that needs to be preprocessed, efficient and accurate data optimization processing is achieved. By constructing and debugging the target data optimization strategy matching network, the optimal data optimization strategy can be automatically extracted and matched from the complex sensing and monitoring information flow, significantly improving the accuracy and reliability of the data. Using past learning information to debug the network enables the network to accurately identify the defective states in the data, such as noise, abnormal fluctuations, etc., and assign corresponding voting weights to them, thus ensuring the pertinence and effectiveness of the data optimization strategy. Finally, by determining the data optimization strategy matching viewpoints corresponding to each implanted device sensing and monitoring information, that is, the preprocessing strategy keywords, real-time and customized preprocessing of the target vital sign monitoring data is achieved, providing solid data support for in-depth research and application in fields such as intelligent healthcare, monitoring and / or regulation processing of implanted devices. Brief Description of the Drawings

[0051] Figure 1 It is a flowchart of a sensing and monitoring data optimization method for intelligent healthcare provided by an embodiment of the present application.

[0052] Figure 2 It is a schematic structural diagram of a sensing and monitoring data optimization system provided by an embodiment of the present application. Detailed Embodiments

[0053] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0054] The terms "first", "second", etc. in the present application are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are usually of the same category, and the number of objects is not limited. For example, the first object can be one or multiple. In addition, "and / or" in the present application means at least one of the connected objects, and the character " / " generally indicates an "or" relationship between the associated objects before and after.

[0055] Figure 1 A sensing and monitoring data optimization method for intelligent healthcare is shown, which is applied to a sensing and monitoring data optimization system. The method includes the following steps 110-step 130.

[0056] Step 110: Obtain the original implantable device sensor monitoring information flow to be optimized.

[0057] The original implantable device sensor monitoring information flow includes target vital sign monitoring data that needs to be preprocessed.

[0058] Step 120: Input the original implantable device sensor monitoring information flow into the pre-debugged target data optimization strategy matching network to obtain a data optimization strategy matching viewpoint set.

[0059] Among them, the target data optimization strategy matching network is a data optimization strategy matching network obtained by inputting past learning information into the original data optimization strategy matching network to be debugged for debugging, the past learning information includes a batch of past data quality characterization vectors respectively mined from a batch of past implantable device sensor monitoring information and the data defect state and voting weight corresponding to each past data quality characterization vector, the data defect state reflects an abnormal state keyword of a past data quality characterization vector, the voting weight reflects the reinforcement coefficient corresponding to the abnormal state keyword, the batch of past implantable device sensor monitoring information is a batch of uninterrupted implantable device sensor monitoring information, and each past data quality characterization vector includes a data quality characterization vector determined by a past implantable device sensor monitoring information and a previous past implantable device sensor monitoring information of the past implantable device sensor monitoring information.

[0060] Step 130: Determine the data optimization strategy matching viewpoint corresponding to each implantable device sensor monitoring information in the original implantable device sensor monitoring information flow based on the data optimization strategy matching viewpoint set.

[0061] The data optimization strategy matching viewpoint is used to represent the preprocessing strategy keywords of the target vital sign monitoring data.

[0062] In recent years, sensor data collection for neural monitoring and regulation through implantable brain-computer interface chips has become one of the development directions of smart / digital medicine. However, due to the complexity of the physiological environment and the limitations of implantable sensor monitoring technology, the original sensor monitoring information often contains noise and interference, and needs to be optimized to serve subsequent tasks.

[0063] In step 110, the sensor monitoring data optimization system first obtains the original sensor monitoring information flow from the implanted device in the brain of the patient with chronic neurological diseases. The original sensor monitoring information flow includes electroencephalogram (EEG) signals, local field potential (LFP) signals, etc., which reflect the state of neural activity in the patient's brain. In particular, the sensor monitoring data optimization system focuses on those monitoring data related to the target signs (such as tremor frequency, muscle stiffness, etc.), which are the focus of subsequent optimization processing.

[0064] Next, in step 120, the sensor monitoring data optimization system inputs the original implantable device sensor monitoring information flow into a pre-debugged target data optimization strategy matching network. The core function of the target data optimization strategy matching network is to output a series of data optimization strategy matching viewpoints based on the input information flow.

[0065] In order to build the network, the sensor monitoring data optimization system has done a lot of learning and debugging work in advance. Specifically, the sensor monitoring data optimization system has collected a batch of past implantable device sensor monitoring information, which is continuous and covers monitoring data under various physiological states and stimulation parameters. For this batch of past information, the sensor monitoring data optimization system extracts a batch of past data quality characterization vectors through data mining technology. Each vector not only contains the data features of the current monitoring information, but also incorporates the data features of the previous monitoring information, so that it can better capture the trends and patterns of data changes.

[0066] Subsequently, the sensor monitoring data optimization system analyzes these past data quality characterization vectors and identifies the data defect states. The data defect states are represented by abnormal state keywords, such as "excessive noise" and "signal drift". The abnormal state keywords reflect the specific types of data quality problems. At the same time, the sensor monitoring data optimization system also assigns a voting weight to each abnormal state keyword. The voting weight reflects the importance and frequency of the defect state in the overall data set, which can also be understood as the reinforcement coefficient for optimizing the defect state.

[0067] Based on this batch of past learning information, the sensor monitoring data optimization system debugged and trained the original data optimization strategy matching network, and finally obtained a network model that can accurately match the data optimization strategy. The model can output a series of targeted data optimization strategy matching viewpoints based on the input original implantable device sensor monitoring information flow.

[0068] In step 130, the system determines a specific data optimization strategy matching viewpoint for each monitoring information in the original implantable device sensor monitoring information flow based on the viewpoint set output by the data optimization strategy matching network. The data optimization strategy matching viewpoint includes optimization suggestions for different data defect states, such as filtering and denoising, baseline correction, signal enhancement, etc.

[0069] For example, for a certain monitoring information, the viewpoint set output by the target data optimization strategy matching network may include the defect state of "too much noise" and give the optimization strategy of "applying wavelet transform filtering". The sensor monitoring data optimization system will perform the corresponding wavelet transform filtering operation on the monitoring information according to this viewpoint to remove the noise component.

[0070] In this way, the sensor monitoring data optimization system can optimize the original sensor monitoring information from the implantable device in real time to improve the accuracy and reliability of the data.

[0071] The technical solution recorded in steps 110-130 can effectively optimize the sensor monitoring data and improve the accuracy and reliability of the data by obtaining the original implantable device sensor monitoring information flow, inputting the pre-debugged target data optimization strategy matching network, and determining the data optimization strategy matching viewpoint.

[0072] It should be noted that the above technical solution is aimed at optimizing the processing of sensor monitoring data, rather than directly diagnosing or treating diseases in living human or animal bodies. The processing object is the original sensor monitoring information stream collected by the implantable device. Although these data come from the human or animal body, they have become technical data that can be analyzed and processed after digital conversion. Furthermore, the purpose of the above technical solution is to improve the accuracy and reliability of sensor monitoring data so as to better serve subsequent tasks, such as neural monitoring and regulation. It does not directly involve the diagnosis or treatment process of the disease, nor does it produce any diagnosis or treatment results. In addition, the above technical solution optimizes the original sensor monitoring information through a series of data processing steps, such as filtering and denoising, baseline correction, signal enhancement, etc. These steps all belong to the category of technical processing and do not involve the selection and application of medical diagnosis or treatment methods.

[0073] In addition, compared to disease diagnosis and treatment methods that directly use living human or animal bodies as the implementation object, the above-mentioned sensor monitoring data optimization system processes technical data that has been digitally converted. Compared to disease diagnosis and treatment methods that aim to identify, determine or eliminate the cause or lesion and produce diagnostic or treatment results, the above-mentioned sensor monitoring data optimization system aims to improve the accuracy and reliability of data and does not produce any direct diagnostic or treatment results. Compared to disease diagnosis and treatment methods that belong to the medical field and require doctors to make judgments and operations based on professional knowledge and experience, the methods of the above-mentioned sensor monitoring data optimization system belong to the field of data processing technology and are automatically processed through preset algorithms and models.

[0074] It is worth noting that the technical solution recorded in step 120 is the core invention of this application in the three technical fields of smart healthcare, neural monitoring and regulation, and implantable devices. From the perspective of data processing, it provides an innovative solution for optimizing the original implantable device sensor monitoring information flow.

[0075] First, the technical solution constructs a target data optimization strategy matching network. This network is not generated out of thin air, but is debugged and trained based on a large amount of past learning information. This past learning information comes from a batch of uninterrupted implantable device sensor monitoring information, which covers monitoring data under various physiological states and stimulation parameters, and has extremely high representativeness and practical value.

[0076] Secondly, for this batch of past implantable device sensor monitoring information, a batch of past data quality characterization vectors were extracted through data mining technology. Each vector not only contains the data features of the current monitoring information, but also cleverly incorporates the data features of the previous monitoring information. This design enables each vector to better capture the trends and patterns of data changes, providing a richer information basis for subsequent data optimization strategy matching.

[0077] Then, we conducted an in-depth analysis of these past data quality characterization vectors and identified the data defect states. These defect states are represented by abnormal state keywords, such as "excessive noise" and "signal drift", which intuitively reflect the specific types of data quality problems. At the same time, in order to more accurately measure the importance and frequency of these defect states, a voting weight is assigned to each abnormal state keyword. This weight can be understood as a reinforcement coefficient for optimizing the defect state, which will play an important role in the subsequent data optimization strategy matching.

[0078] Finally, based on this batch of past learning information, the original data optimization strategy matching network was repeatedly debugged and trained. Through this process, the network gradually learned how to output a series of targeted data optimization strategy matching opinions based on the input original implantable device sensor monitoring information flow. These opinion sets are like tailor-made optimization solutions for each monitoring information, which will guide the subsequent data processing process to ensure that the accuracy and reliability of sensor monitoring data are significantly improved.

[0079] Therefore, the technical solution recorded in step 120 realizes accurate optimization of the original implantable device sensor monitoring information flow by constructing a target data optimization strategy matching network and debugging and training based on past learning information. This innovative solution not only improves the accuracy and reliability of data, but also provides strong technical support for the development of smart medical care, monitoring and / or control processing of implantable devices, and implantable devices.

[0080] Furthermore, step 130, as one of the core inventions of the present application in the three technical fields of smart healthcare, neural monitoring and regulation, and implantable devices, provides an innovative solution for optimizing the original implantable device sensor monitoring information flow from the perspective of NLP (natural language processing) data processing.

[0081] First, we need to understand the meaning of the data optimization strategy matching opinion set. The data optimization strategy matching opinion set is generated by the target data optimization strategy matching network based on the input raw implantable device sensor monitoring information flow, and it contains optimization strategy suggestions for each monitoring information. These suggestions are expressed in natural language, such as "apply wavelet transform filtering" or "perform baseline correction", which directly correspond to the possible defect states in the sensor monitoring data.

[0082] Secondly, in order to extract the optimization strategy specific to each implantable device sensor monitoring information from these data optimization strategy matching opinions, NLP technology was used. NLP technology plays a key role here, as it can analyze and understand the natural language text in the opinion set, thereby identifying the optimization strategy corresponding to each monitoring information.

[0083] Then, through further processing with NLP technology, the system is able to match the identified optimization strategy with each monitoring information in the original implantable device sensor monitoring information stream. This matching process is completed based on information such as the unique identifier or timestamp of the monitoring information, ensuring that each monitoring information can find the corresponding optimization strategy.

[0084] Finally, based on the matching results, the system applies the corresponding data optimization strategy to each monitoring information in the original implantable device sensor monitoring information flow. For example, if the optimization strategy for a certain monitoring information is "apply wavelet transform filtering", the system will perform a wavelet transform filtering operation on the monitoring information to remove the noise component. In this way, the system can optimize the original sensor monitoring information from the implantable device in real time to improve the accuracy and reliability of the data.

[0085] It can be seen that the technical solution recorded in step 130 realizes the precise optimization of each monitoring information in the original implantable device sensor monitoring information flow by combining NLP data processing technology. This innovative solution not only improves the accuracy and reliability of data, but also provides strong technical support for the development of smart medical care, monitoring and / or control processing of implantable devices, and implantable devices.

[0086] In summary, the embodiments of the present application achieve efficient and accurate data optimization processing for the original implantable device sensor monitoring information flow, especially the target vital sign monitoring data that needs to be preprocessed. By constructing and debugging the target data optimization strategy matching network, it is possible to automatically extract and match the optimal data optimization strategy from the complex sensor monitoring information flow, significantly improving the accuracy and reliability of the data. The network is debugged using past learning information so that the network can accurately identify defective states in the data, such as noise, abnormal fluctuations, etc., and assign corresponding voting weights to them, thereby ensuring the pertinence and effectiveness of the data optimization strategy. Finally, by determining the data optimization strategy matching viewpoint corresponding to each implantable device sensor monitoring information, that is, the preprocessing strategy keyword, real-time and customized preprocessing of the target vital sign monitoring data is achieved, providing solid data support for in-depth research and application in the fields of smart medical care, monitoring and / or regulation and control of implantable devices.

[0087] In some optional embodiments, the method also includes: performing a first characterization vector mining on the batch of past implantable device sensor monitoring information to obtain a first batch of data quality characterization vectors, wherein a data quality characterization vector in the first batch of data quality characterization vectors corresponds to a past implantable device sensor monitoring information in the batch of past implantable device sensor monitoring information; performing a second characterization vector mining on the batch of past implantable device sensor monitoring information to obtain a second batch of data quality characterization vectors, wherein a first data quality characterization vector in the second batch of data quality characterization vectors corresponds to a first past implantable device sensor monitoring information in the past implantable device sensor monitoring information, the previous frame of implantable device sensor monitoring information of the first past implantable device sensor monitoring information is the second past implantable device sensor monitoring information, and the first data quality characterization vector represents the contextual embedded semantics between the first past implantable device sensor monitoring information and the second past implantable device sensor monitoring information; integrating the first batch of data quality characterization vectors and the second batch of data quality characterization vectors respectively to obtain the batch of past data quality characterization vectors.

[0088] Based on this embodiment, the method is not limited to simple data input and output, but includes a series of complex and sophisticated data processing steps to ensure that the final data optimization strategy matching viewpoint set can accurately reflect the data characteristics and optimization requirements in the original implantable device sensor monitoring information flow.

[0089] First, in order to conduct in-depth analysis and learning of past implantable device sensor monitoring information, the sensor monitoring data optimization system needs to perform the first characterization vector mining on this batch of information. The purpose of this step is to extract its unique data quality characterization vector from each past implantable device sensor monitoring information. These vectors are abstract representations of the data, and they capture the most critical data quality features in each monitoring information. Through such a mining process, a batch of data quality characterization vectors that correspond one-to-one to past implantable device sensor monitoring information are obtained, that is, the first batch of data quality characterization vectors. This step provides a basic data feature set for subsequent data optimization strategy matching.

[0090] Secondly, in addition to considering the data quality characteristics of each past implantable device sensor monitoring information itself, the sensor monitoring data optimization system further considers the contextual relationship between these monitoring information. In order to achieve this point, the sensor monitoring data optimization system performs a second characterization vector mining. The purpose of this step is to mine the contextual embedding semantics between adjacent past implantable device sensor monitoring information, that is, the data change patterns and associations between them. Through such a mining process, the second batch of data quality characterization vectors are obtained. Each of these vectors represents the contextual relationship between a pair of adjacent past implantable device sensor monitoring information, providing richer information for subsequent data optimization strategy matching.

[0091] Then, the sensor monitoring data optimization system needs to integrate the first batch of data quality characterization vectors and the second batch of data quality characterization vectors. The purpose of this step is to fuse the data quality characteristics of a single monitoring information with the contextual relationship characteristics between adjacent monitoring information to obtain a more comprehensive set of data quality characterization vectors. Through such an integration process, a batch of past data quality characterization vectors are obtained. This batch of vectors not only contains the data quality characteristics of each past implantable device sensor monitoring information itself, but also contains the contextual relationship characteristics between them, providing a more accurate and comprehensive information basis for subsequent data optimization strategy matching.

[0092] Next, based on this batch of past data quality characterization vectors, the sensor monitoring data optimization system constructed a target data optimization strategy matching network. The purpose of this step is to train a network that can automatically match data optimization strategies by learning the relationship between past data quality characterization vectors and their corresponding data defect states and voting weights. To achieve this, the sensor monitoring data optimization system uses past data quality characterization vectors as input and their corresponding data defect states and voting weights as output, and repeatedly debugs and trains the network. Through this learning process, the network gradually learns how to output corresponding data optimization strategy matching views based on the input data quality characterization vectors.

[0093] Finally, when the new original implantable device sensor monitoring information flow needs to be optimized, the sensor monitoring data optimization system only needs to input it into the trained target data optimization strategy matching network to obtain a series of targeted data optimization strategy matching viewpoints. These viewpoints are jointly determined based on past learning information and the data characteristics of the current monitoring information flow, so they can accurately reflect the data problems and optimization needs in the current monitoring information flow. In this way, the sensor monitoring data optimization system achieves accurate optimization of the original implantable device sensor monitoring information flow and improves the accuracy and reliability of the data.

[0094] It can be seen that by combining multiple steps such as first characterization vector mining, second characterization vector mining, data quality characterization vector integration, and target data optimization strategy matching network construction, a comprehensive analysis and precise optimization of the original implantable device sensor monitoring information flow is achieved. It not only takes into account the data quality characteristics of a single monitoring information, but also takes into account the contextual relationship characteristics between adjacent monitoring information, so it can more accurately identify problems and optimization needs in the data. At the same time, by constructing a target data optimization strategy matching network, the sensor monitoring data optimization system realizes the automatic matching and real-time application of data optimization strategies, greatly improving the efficiency and accuracy of data processing.

[0095] In combination with the above, the data quality characterization vector is a core concept in the embodiments of the present application, which is the result of deep mining and abstract representation of the original implantable device sensor monitoring information.

[0096] The data quality characterization vector is a vector that quantifies the data quality characteristics of past implantable device sensor monitoring information. It captures the most critical data features of the monitoring information, such as signal stability, noise level, abnormal fluctuations, etc., and expresses these features in the form of numerical values. Through this representation, it is easier to analyze and process the monitoring information, and then extract useful information for matching data optimization strategies.

[0097] A data quality characterization vector usually contains multiple numerical features, each of which corresponds to a specific data quality aspect in the monitoring information. For example, a data quality characterization vector may contain the following numerical features:

[0098] Mean: represents the average signal level of monitoring information;

[0099] Standard Deviation: Indicates the signal fluctuation degree of monitoring information;

[0100] Maximum: indicates the highest signal level in the monitoring information;

[0101] Minimum: Indicates the lowest signal level in the monitoring information;

[0102] Noise Level: indicates the noise intensity in the monitoring information;

[0103] Anomaly Index: Indicates the possibility of abnormal fluctuations in monitoring information.

[0104] For example, there is a past implantable device sensor monitoring information. After the first characterization vector is mined, the following data quality characterization vector is obtained: [0.5, 0.1, 1.0, -0.5, 0.05, 0.2]. This vector represents the data quality characteristics of the monitoring information. Specifically:

[0105] The mean is 0.5, which means that the average signal level of the monitoring information is 0.5;

[0106] The standard deviation is 0.1, which means that the signal fluctuation of the monitoring information is small;

[0107] The maximum value is 1.0, which means the highest signal level in the monitoring information is 1.0;

[0108] The minimum value is -0.5, which means the lowest signal level in the monitoring information is -0.5;

[0109] The noise level is 0.05, indicating that the noise intensity in the monitoring information is low;

[0110] The anomaly index is 0.2, indicating that the possibility of abnormal fluctuations in the monitoring information is small.

[0111] Through such a data quality characterization vector, the data quality characteristics of the past implantable device sensor monitoring information can be clearly understood, and the corresponding data optimization strategy can be matched for it. For example, if the noise level of the monitoring information is considered to be high, a denoising strategy such as wavelet transform filtering can be selected to optimize the data.

[0112] Therefore, by comparing the data quality characterization vectors of different monitoring information, we can find out the similarities and differences between them, and then match them with the most suitable data optimization strategy. At the same time, the data quality characterization vector can also be used as the input feature of the machine learning algorithm to train and optimize the data optimization strategy matching network to achieve more accurate data optimization processing.

[0113] In some other preferred embodiments, the second characterization vector mining of the batch of past implantable device sensor monitoring information to obtain the second batch of data quality characterization vectors includes: performing the second characterization vector mining on the batch of past implantable device sensor monitoring information to obtain the second batch of data quality characterization vectors through the following steps, wherein the past implantable device sensor monitoring information subjected to the second characterization vector mining each time is used as the current past implantable device sensor monitoring information, and the obtained data quality characterization vector is used as the current data quality characterization vector, and the second batch of data quality characterization vectors includes the current data quality characterization vector; performing a sensor monitoring information analysis operation on the current past implantable device sensor monitoring information to determine the first batch of distribution feature variables, wherein, The first batch of distribution characteristic variables are used to describe the area of ​​the target vital sign monitoring data in the current past implantable device sensor monitoring information; a sensor monitoring information analysis operation is performed on the previous past implantable device sensor monitoring information of the current past implantable device sensor monitoring information to determine the second batch of distribution characteristic variables, wherein the second batch of distribution characteristic variables are used to describe the area of ​​the target vital sign monitoring data in the previous past implantable device sensor monitoring information; the current data quality characterization vector is determined based on the first batch of distribution characteristic variables and the second batch of distribution characteristic variables, wherein the current data quality characterization vector is used to represent the difference between the corresponding distribution characteristic variables in the first batch of distribution characteristic variables and the second batch of distribution characteristic variables.

[0114] In this embodiment, mining a second characterization vector of a batch of past implantable device sensor monitoring information to obtain a second batch of data quality characterization vectors is an in-depth and complex process. The process involves not only the analysis of the current past implantable device sensor monitoring information, but also the analysis of the previous past implantable device sensor monitoring information, and determining the data quality characterization vector by comparing the difference between the two.

[0115] First, each time the second characterization vector mining is performed, the past implanted device sensor monitoring information is regarded as the current past implanted device sensor monitoring information, and the mined data quality characterization vector is regarded as the current data quality characterization vector. This process is iterative until all the past implanted device sensor monitoring information is processed. In this way, a second batch of data quality characterization vector sets containing all current data quality characterization vectors can be obtained.

[0116] Next, for each current past implantable device sensor monitoring information, a sensor monitoring information analysis operation needs to be performed. The purpose of this step is to determine a batch of distribution feature variables, which are used to describe the area of ​​target vital sign monitoring data in the current past implantable device sensor monitoring information. These distribution feature variables may include the mean, variance, maximum value, minimum value, etc. of the signal, which can fully reflect the distribution of target vital sign monitoring data in the current past implantable device sensor monitoring information.

[0117] At the same time, it is also necessary to perform the same sensor monitoring information analysis operation on the previous past implantable device sensor monitoring information of the current past implantable device sensor monitoring information. The purpose of this step is to determine another batch of distribution feature variables, which are used to describe the area of ​​the target vital sign monitoring data in the previous past implantable device sensor monitoring information. Through such analysis, the distribution of the target vital sign monitoring data in two adjacent past implantable device sensor monitoring information can be obtained.

[0118] Then, based on these two batches of distribution feature variables, the current data quality characterization vector can be determined. This step is the core, which involves the comparison and analysis of the two batches of distribution feature variables. Specifically, it is necessary to calculate the differences between the corresponding distribution feature variables in the two batches of distribution feature variables, such as the difference in mean, the difference in variance, etc. These differences can reflect the changes in the target vital sign monitoring data between the adjacent past implantable device sensor monitoring information.

[0119] Through this process, a current data quality characterization vector can be obtained, which represents the change characteristics of the target vital sign monitoring data between adjacent past implantable device sensor monitoring information. This vector not only contains the data characteristics of the current past implantable device sensor monitoring information, but also contains the contextual relationship characteristics between it and the previous past implantable device sensor monitoring information.

[0120] Finally, after processing all the past implantable device sensor monitoring information, a complete second batch of data quality characterization vector sets can be obtained. This vector set not only contains the data features of each past implantable device sensor monitoring information, but also contains the contextual relationship features between them. Such a feature set provides a richer and more accurate information basis for subsequent data optimization strategy matching.

[0121] In practical applications, the mining process of the second batch of data quality characterization vectors may involve complex algorithms and calculations. For example, it may be necessary to use machine learning algorithms to analyze and operate sensor monitoring information to determine distribution feature variables. At the same time, it is also necessary to use appropriate mathematical methods to calculate the differences between the two batches of distribution feature variables and convert these differences into a current data quality characterization vector with practical significance.

[0122] In addition, in order to improve the accuracy and reliability of the second batch of data quality characterization vectors, the mining process can also be iterated and optimized multiple times. For example, methods such as cross-validation can be used to evaluate the stability and consistency of the mining results, and the mining process can be adjusted and improved according to the evaluation results.

[0123] Thus, mining the second characterization vector for a batch of past implantable device sensing and monitoring information to obtain the second batch of data quality characterization vectors is a process involving multiple steps and complex calculations. Through this process, a feature set containing rich data features and context relationship features can be obtained, providing strong support for subsequent data optimization strategy matching. The proposal and application of this technical solution will further promote the development of implantable device technology and the field of intelligent healthcare.

[0124] In an alternative embodiment, the method further includes: respectively determining the data defect status and voting weight corresponding to each past data quality characterization vector in the batch of past data quality characterization vectors through the following steps, where each past data quality characterization vector for which the data defect status and voting weight are determined each time is used as the current past data quality characterization vector, and the data defect status corresponding to the current past data quality characterization vector is used as the current data defect status and the current voting weight: inputting the current past data quality characterization vector into the original data optimization strategy matching network to obtain the current data defect status, where the original data optimization strategy matching network is a data optimization strategy matching network obtained by performing a reset process in advance; determining the current voting weight based on the current data defect status and the prior learning annotation corresponding to the current past data quality characterization vector, where the prior learning annotation carries the authentication label in the past implantable device sensing and monitoring information corresponding to the current past data quality characterization vector, and the authentication label includes the target preprocessing strategy keyword of the target vital sign monitoring data, and the voting weight is used to indicate whether the current data defect status matches the target preprocessing strategy keyword.

[0125] In an alternative embodiment, the method not only covers the previously mentioned steps 110 to 130, but also further expands its function by a series of refined operations to determine the data defect status and voting weight corresponding to each past data quality characterization vector in a batch of past data quality characterization vectors.

[0126] First of all, this embodiment is carried out on the basis of the original method framework, and it inherits the processing flow of the past implantable device sensing and monitoring information in the previous steps, including the mining and generation of data quality characterization vectors. On this basis, the sensing and monitoring data optimization system further introduces the concepts of data defect status and voting weight to more comprehensively evaluate and optimize the quality of past data.

[0127] Specifically, for each of the past data quality characterization vectors, the sensor monitoring data optimization system regards it as the current past data quality characterization vector and determines a corresponding data defect state and voting weight for it. This process is iterative until the sensor monitoring data optimization system has processed all the past data quality characterization vectors.

[0128] When determining the data defect state of the current past data quality characterization vector, the sensor monitoring data optimization system inputs it into a special data optimization strategy matching network. The network is obtained by prior reset processing, and it can output a corresponding data defect state according to the input data quality characterization vector. The state reflects the problems or defects in the data quality of the past implantable device sensor monitoring information represented by the current past data quality characterization vector.

[0129] After obtaining the current data defect state, the sensor monitoring data optimization system needs to further determine the voting weight corresponding to the current past data quality characterization vector. This process is based on the current data defect state and the prior learning annotations corresponding to the current past data quality characterization vector. Prior learning annotation is an important concept, which carries the authentication label in the past implantable device sensor monitoring information corresponding to the current past data quality characterization vector. The authentication label contains the target preprocessing strategy keywords for the target vital sign monitoring data, which reflects the target state or strategy that the sensor monitoring data optimization system expects to achieve when preprocessing the past implantable device sensor monitoring information.

[0130] By comparing the current data defect state and the target preprocessing strategy keywords in the authentication tag, the sensor monitoring data optimization system can evaluate the degree of match between the data quality of the past implantable device sensor monitoring information represented by the current past data quality characterization vector and the target state. The degree of match is the voting weight to be determined by the sensor monitoring data optimization system, which is used to indicate whether the current data defect state matches the target preprocessing strategy keywords and the degree of match.

[0131] In practical applications, the determination of voting weights may involve complex algorithms and calculations. For example, the sensor monitoring data optimization system can use a machine learning algorithm to train a model that can predict voting weights based on the current data defect status and the target preprocessing strategy keywords in the authentication tag. Alternatively, the sensor monitoring data optimization system can also design a rule-based algorithm to calculate voting weights by comparing the similarity or difference between the current data defect status and the target preprocessing strategy keywords.

[0132] Regardless of which method is used, the goal of the sensor monitoring data optimization system is to obtain a voting weight that can accurately reflect the degree of matching between the data quality of the past implantable device sensor monitoring information represented by the current past data quality characterization vector and the target state. This weight will play an important role in the subsequent data optimization strategy selection and matching process, helping the sensor monitoring data optimization system to more accurately select and optimize data optimization strategies for past implantable device sensor monitoring information.

[0133] In addition, in order to improve the accuracy and reliability of the determination of data defect status and voting weights, the sensor monitoring data optimization system can also continuously update and optimize the original data optimization strategy matching network and prior learning annotations. For example, the sensor monitoring data optimization system can regularly use new past implantable device sensor monitoring information to train and optimize the original data optimization strategy matching network to improve its ability to identify data defect status. At the same time, the sensor monitoring data optimization system can also continuously collect and organize new authentication tags and target preprocessing strategy keywords to enrich and improve the content of prior learning annotations.

[0134] In this way, by introducing the concepts of data defect status and voting weight, the function and application scope of the original method are further expanded. It enables the sensor monitoring data optimization system to more comprehensively evaluate and optimize the data quality of past implantable device sensor monitoring information, and provide a more accurate and reliable information basis for subsequent data optimization strategy selection and matching.

[0135] In the next step, the current past data quality characterization vector is input into the original data optimization strategy matching network to obtain the current data defect state, including: each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set is input into the first discrimination algorithm branch in sequence together with the current past data quality characterization vector to obtain a first batch of quantitative discrimination indicators, wherein the first batch discrimination indicator includes the quantitative discrimination indicator corresponding to each preprocessing strategy keyword; the alternative preprocessing strategy keyword corresponding to the first quantitative discrimination indicator with the largest indicator value in the first batch discrimination indicator is determined as the current preprocessing strategy keyword corresponding to the current data defect state.

[0136] Based on this embodiment, the process of inputting the current past data quality characterization vector into the original data optimization strategy matching network and obtaining the current data defect state is a complex task involving multiple links and precise calculations.

[0137] First, the original data optimization strategy matching network is a specially designed and trained network model that can convert the input past data quality characterization vector into the corresponding data defect state. The network model has been reset in advance to ensure that it can accurately match the current data optimization strategy. In this process, the sensor monitoring data optimization system uses a pre-stored preprocessing strategy keyword set, which contains all possible preprocessing strategy keywords, which represent the various strategies that the sensor monitoring data optimization system may adopt when preprocessing the past implantable device sensor monitoring information.

[0138] In order to determine the data defect status, the sensor monitoring data optimization system first inputs each preprocessing strategy keyword in the preprocessing strategy keyword set into the first discrimination algorithm branch in order together with the current past data quality characterization vector. The discrimination algorithm branch is a carefully designed algorithm model, which can calculate a corresponding quantitative discrimination index based on the input preprocessing strategy keyword and data quality characterization vector. The quantitative discrimination index is a numerical value that reflects the matching degree or similarity between the current past data quality characterization vector and the input preprocessing strategy keyword.

[0139] When the sensor monitoring data optimization system inputs all the preprocessing strategy keywords and the current past data quality characterization vector into the first discrimination algorithm branch, the sensor monitoring data optimization system will obtain a batch of quantitative discrimination indicators. The number of these batched discrimination indicators is the same as the number of keywords in the preprocessing strategy keyword set, and each quantitative discrimination indicator corresponds to a preprocessing strategy keyword. The numerical values ​​of these quantitative discrimination indicators reflect the degree of match between the current past data quality characterization vector and each preprocessing strategy keyword.

[0140] Next, the sensor monitoring data optimization system needs to determine a first quantitative discrimination indicator with the largest indicator value from these batch discrimination indicators. The quantitative discrimination indicator with the largest indicator value represents the highest matching degree between the current past data quality characterization vector and a preprocessing strategy keyword. The sensor monitoring data optimization system determines the preprocessing strategy keyword corresponding to the quantitative discrimination indicator with the largest indicator value as the current preprocessing strategy keyword corresponding to the current data defect state.

[0141] The current preprocessing strategy keyword is the current data defect state that the sensor monitoring data optimization system wants. It reflects the problems or defects in the data quality of the past implantable device sensor monitoring information represented by the current past data quality characterization vector, and which preprocessing strategy keyword best matches the problem or defect. Through the current preprocessing strategy keyword, the sensor monitoring data optimization system can know which preprocessing strategy should be adopted for the past implantable device sensor monitoring information to optimize its data quality.

[0142] In practical applications, the implementation of the first discrimination algorithm branch may involve complex algorithms and calculations. For example, the sensor monitoring data optimization system can use a machine learning algorithm to train a model that can predict quantitative discrimination indicators based on the input preprocessing strategy keywords and data quality characterization vectors. Alternatively, the sensor monitoring data optimization system can also design a set of rule-based algorithms to calculate quantitative discrimination indicators by comparing the features or attributes of the data quality characterization vector and the preprocessing strategy keywords. Regardless of which method is used, the goal of the sensor monitoring data optimization system is to obtain a quantitative discrimination indicator that can accurately reflect the degree of match between the data quality characterization vector and the preprocessing strategy keywords.

[0143] In addition, in order to improve the accuracy and reliability of determining the data defect state, the sensor monitoring data optimization system can also continuously update and optimize the original data optimization strategy matching network and the first discrimination algorithm branch. For example, the sensor monitoring data optimization system can regularly use new past implantable device sensor monitoring information to train and optimize the original data optimization strategy matching network to improve its ability to identify data defect states. At the same time, the sensor monitoring data optimization system can also continuously collect and organize new preprocessing strategy keywords and data quality characterization vectors to enrich and improve the training data set of the first discrimination algorithm branch and improve its discrimination accuracy.

[0144] Thus, the process of inputting the current past data quality characterization vector into the original data optimization strategy matching network and obtaining the current data defect state is a complex task involving multiple links and fine calculations. By introducing the preprocessing strategy keyword set and the first discrimination algorithm branch, the sensor monitoring data optimization system can accurately determine the problems or defects in the data quality of the past implantable device sensor monitoring information represented by the current past data quality characterization vector, and provide an accurate information basis for the subsequent data optimization strategy selection and matching.

[0145] Under some other design ideas, before inputting each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set into the first discrimination algorithm branch in sequence together with the current past data quality characterization vector to obtain the first batch of quantitative discrimination indicators, the method also includes: determining the thermal characteristic variable through the target mode; determining whether the thermal characteristic variable meets the set thermal distribution rules; on the basis that the thermal characteristic variable meets the set thermal distribution rules, dynamically extracting the alternative preprocessing strategy keywords in the pre-stored preprocessing strategy keyword set as the current data defect state; on the basis that the thermal characteristic variable does not meet the set thermal distribution rules, inputting the current past data quality characterization vector into the original data optimization strategy matching network to obtain the current data defect state.

[0146] Under some other design ideas, for the process of matching each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set with the current past data quality characterization vector and obtaining quantitative judgment indicators, the sensor monitoring data optimization system introduces a new pre-step, that is, first determining the thermal characteristic variables through the target mode, and determining the subsequent processing flow based on whether the thermal characteristic variables conform to the set thermal distribution rules.

[0147] First, the purpose of this pre-step is to provide a more efficient and accurate way to determine the current data defect status. By introducing thermal characteristic variables and thermal distribution rules, the sensor monitoring data optimization system can more directly reflect the data quality problems or defects of the past implantable device sensor monitoring information represented by the current past data quality characterization vector, and select more appropriate preprocessing strategy keywords accordingly.

[0148] Specifically, in this pre-step, the sensor monitoring data optimization system first determines a thermal characteristic variable through a target pattern. The target pattern is a predefined pattern that represents the target state or strategy that the sensor monitoring data optimization system expects to achieve when preprocessing the past implantable device sensor monitoring information. By comparing the current past data quality characterization vector with the target pattern, the sensor monitoring data optimization system can extract a thermal characteristic variable that can reflect the difference or similarity between the two.

[0149] After obtaining the thermal characteristic variable, the sensor monitoring data optimization system needs to further determine whether the variable conforms to the set thermal distribution rule. The thermal distribution rule is a predefined rule that describes the preprocessing strategy keywords or data defect states corresponding to the thermal characteristic variables under different numerical ranges. By comparing the numerical value of the thermal characteristic variable and the threshold of the thermal distribution rule, the sensor monitoring data optimization system can determine whether there are problems or defects in the data quality of the past implantable device sensor monitoring information represented by the current past data quality characterization vector, and what the preprocessing strategy keywords corresponding to the problem or defect are.

[0150] If the thermal characteristic variable meets the set thermal distribution rule, the sensor monitoring data optimization system can directly extract the alternative preprocessing strategy keyword from the pre-stored preprocessing strategy keyword set according to the rule as the current data defect state. The alternative preprocessing strategy keyword corresponds to the numerical range of the thermal characteristic variable, and it represents the strategy that the sensor monitoring data optimization system should adopt when preprocessing the past implantable device sensor monitoring information. In this way, the sensor monitoring data optimization system can determine the current data defect state more directly and accurately without the need for subsequent complex calculations or matching processes.

[0151] However, if the thermal characteristic variables do not conform to the set thermal distribution rules, the sensor monitoring data optimization system needs to take another approach to determine the current data defect state. Specifically, the sensor monitoring data optimization system will input the current past data quality characterization vector into the original data optimization strategy matching network, and obtain a corresponding current data defect state through network processing. The original data optimization strategy matching network is a specially designed and trained network model that can output a corresponding data defect state based on the input past data quality characterization vector. Through the processing of this network, the sensor monitoring data optimization system can obtain a more accurate and reliable current data defect state to reflect the problems or defects in the data quality of the past implantable device sensor monitoring information represented by the current past data quality characterization vector.

[0152] In this way, by introducing thermal characteristic variables and thermal distribution rules, a more efficient and accurate way to determine the current data defect state is provided. When the thermal characteristic variables meet the set thermal distribution rules, the sensor monitoring data optimization system can directly extract the preprocessing strategy keywords as the current data defect state according to the rules; when the thermal characteristic variables do not meet the set thermal distribution rules, the sensor monitoring data optimization system can obtain a more accurate and reliable current data defect state through the original data optimization strategy matching network. This method not only improves the accuracy and efficiency of determining the data defect state, but also provides a more accurate and reliable information basis for the subsequent data optimization strategy selection and matching.

[0153] Under some preferred design ideas, the alternative preprocessing strategy keyword corresponding to the first quantitative discrimination indicator with the largest indicator value in the first batch discrimination indicator is determined as the current preprocessing strategy keyword corresponding to the current data defect state, including: on the basis that the alternative preprocessing strategy keyword corresponding to the first quantitative discrimination indicator is the same as the target preprocessing strategy keyword, the current voting weight is determined as the first voting weight, wherein the first voting weight indicates that the original data optimization strategy matching network successfully outputs a matching result; on the basis that the alternative preprocessing strategy keyword corresponding to the first quantitative discrimination indicator is different from the target preprocessing strategy keyword, the current voting weight is determined as the second voting weight, wherein the second voting weight indicates that the original data optimization strategy matching network did not successfully output a matching result.

[0154] Under this design idea, the sensor monitoring data optimization system introduces a more sophisticated and dynamic weight allocation mechanism for the process of determining the candidate preprocessing strategy keyword corresponding to the first quantitative discrimination indicator with the largest indicator value in the first batch of quantitative discrimination indicators as the current preprocessing strategy keyword corresponding to the current data defect state. This mechanism not only considers the size of the quantitative discrimination indicator, but also combines the matching of the target preprocessing strategy keyword, so that the preprocessing strategy corresponding to the current data defect state can be determined more accurately and flexibly.

[0155] First, the design idea aims to improve the accuracy and reliability of data defect status determination. By introducing the concept of voting weight, the sensor monitoring data optimization system can more finely control the influence of different preprocessing strategy keywords in the matching process, thereby ensuring that the final determined current preprocessing strategy keywords are more in line with actual needs.

[0156] Specifically, under this design idea, the sensor monitoring data optimization system will first determine whether the alternative preprocessing strategy keyword corresponding to the first quantitative discrimination index is the same as the target preprocessing strategy keyword. This target preprocessing strategy keyword is a predefined keyword, which represents the target strategy that the sensor monitoring data optimization system expects to achieve when preprocessing past implantable device sensor monitoring information. If the alternative preprocessing strategy keyword corresponding to the first quantitative discrimination index is the same as the target preprocessing strategy keyword, then the sensor monitoring data optimization system will consider that the original data optimization strategy matching network has successfully output a matching result, and at this time the sensor monitoring data optimization system will determine the current voting weight as the first voting weight.

[0157] This first voting weight is a relatively high weight value, which indicates that the original data optimization strategy matching network highly recognizes and trusts the current candidate preprocessing strategy keyword. By assigning a higher voting weight, the sensor monitoring data optimization system can ensure that in the process of determining the current data defect state, the candidate preprocessing strategy keyword that is the same as the target preprocessing strategy keyword has greater influence and priority.

[0158] However, if the alternative preprocessing strategy keyword corresponding to the first quantitative discrimination indicator is different from the target preprocessing strategy keyword, then the sensor monitoring data optimization system will believe that the original data optimization strategy matching network has not successfully output a matching result. At this time, the sensor monitoring data optimization system will determine the current voting weight as the second voting weight. This second voting weight is a lower weight value, which indicates that the original data optimization strategy matching network has a lower recognition and trust in the current alternative preprocessing strategy keyword. By assigning a lower voting weight, the sensor monitoring data optimization system can reduce the influence and priority of the alternative preprocessing strategy keyword that is different from the target preprocessing strategy keyword in the current data defect status determination process.

[0159] Through this design idea, the sensor monitoring data optimization system can more flexibly and accurately determine the preprocessing strategy corresponding to the current data defect state. When the keywords of the candidate preprocessing strategy are the same as the target preprocessing strategy keywords, the sensor monitoring data optimization system will give a higher voting weight to ensure that this strategy has a greater influence in the process of determining the current data defect state; when the keywords of the candidate preprocessing strategy are different from the target preprocessing strategy keywords, the sensor monitoring data optimization system will give a lower voting weight to reduce the influence of this strategy on the process of determining the current data defect state.

[0160] In addition, this design idea also has good scalability and adaptability. In practical applications, the sensor monitoring data optimization system can flexibly adjust the voting weight allocation mechanism and numerical value according to specific needs and scenarios to achieve better results and application value. For example, the sensor monitoring data optimization system can set different voting weight allocation rules and thresholds according to the different types, sources or quality levels of past implantable device sensor monitoring information to more accurately reflect the importance and priority of preprocessing strategy keywords in different situations.

[0161] In this way, by introducing the concept of voting weight, a more sophisticated and dynamic way is provided to determine the preprocessing strategy corresponding to the current data defect state. It not only improves the accuracy and reliability of determining the data defect state, but also provides a more accurate and reliable information basis for the subsequent data optimization strategy selection and matching.

[0162] In some other optional embodiments, the method also includes: selecting a number of past linkage monitoring records from the past learning information, wherein the i-th past linkage monitoring record among the several past linkage monitoring records includes the i-th past data quality characterization vector, the i-th data defect state corresponding to the i-th past data quality characterization vector, the i-th voting weight and the i+1-th past data quality characterization vector, and i is a positive integer; debugging the original data optimization strategy matching network to be debugged based on the several past linkage monitoring records to obtain the target data optimization strategy matching network, wherein, on the basis that the number of iterations of debugging the original data optimization strategy matching network reaches a set iteration threshold, the original data optimization strategy matching network is determined as the target data optimization strategy matching network, and on the basis that the number of iterations of debugging the original data optimization strategy matching network does not reach the set iteration threshold, the neural network weights of the original data optimization strategy matching network are improved according to a preconfigured training error expression, and the input of each round of debugging process is one past linkage monitoring record among the several past linkage monitoring records.

[0163] In this embodiment, in order to further improve the accuracy and efficiency of the data optimization strategy matching network, the sensor monitoring data optimization system introduces past linkage monitoring records to debug the network. The technical solution of this embodiment aims to iteratively debug the original data optimization strategy matching network by utilizing past monitoring data and learning information to obtain a more accurate and efficient target data optimization strategy matching network.

[0164] First, the sensor monitoring data optimization system needs to filter out several past linkage monitoring records from past learning information. These past linkage monitoring records are valuable data generated in the past implantable sensor monitoring process. They contain rich information, such as data quality characterization vectors, data defect status, voting weights, etc. Specifically, the i-th past linkage monitoring record includes the i-th past data quality characterization vector, the i-th data defect status corresponding to the i-th past data quality characterization vector, the i-th voting weight, and the i+1-th past data quality characterization vector, where i is a positive integer. These data provide the sensor monitoring data optimization system with rich samples and features for subsequent network debugging and training.

[0165] Next, the sensor monitoring data optimization system will debug the original data optimization strategy matching network based on these past linkage monitoring records. The goal of debugging is to obtain a target data optimization strategy matching network that can match the data optimization strategy more accurately and efficiently. To achieve this goal, the sensor monitoring data optimization system will set an iteration threshold as the termination condition of the debugging process.

[0166] In the process of debugging the original data optimization strategy matching network, the sensor monitoring data optimization system will continuously use the data in the past linkage monitoring records as input to train and test the network. The input of each round of debugging process is one of several past linkage monitoring records, which can ensure that the network can fully learn the characteristics and rules in the past monitoring data.

[0167] When the number of iterations of debugging the original data optimization strategy matching network reaches the set iteration threshold, the sensor monitoring data optimization system believes that the network has been fully trained and debugged, and then determines the original data optimization strategy matching network as the target data optimization strategy matching network. This target network has high accuracy and efficiency and can be used for actual data optimization strategy matching tasks.

[0168] However, if the number of iterations for debugging the original data optimization strategy matching network does not reach the set iteration threshold, the sensor monitoring data optimization system needs to improve the neural network weights of the network according to the preconfigured training error expression. The training error expression is a function that measures the difference between the network output and the actual data defect state. By optimizing this function, the sensor monitoring data optimization system can continuously adjust the network weights so that the network output is closer to the actual data defect state.

[0169] Through such a debugging process, the sensor monitoring data optimization system can gradually improve the performance of the original data optimization strategy matching network, so that it can more accurately and efficiently match the corresponding data optimization strategy when processing new implanted sensor monitoring data. Finally, when the number of iterations reaches the set threshold, the target data optimization strategy matching network obtained by the sensor monitoring data optimization system will have excellent performance and stability, providing strong support for subsequent data processing and analysis.

[0170] In this way, by introducing past linkage monitoring records and iteratively debugging the original data optimization strategy matching network, the accuracy and efficiency of data optimization strategy matching have been successfully improved. In practical applications, the sensor monitoring data optimization system can flexibly adjust and optimize this technical solution according to specific needs and scenarios to achieve better results and application value. For example, the sensor monitoring data optimization system can set different screening rules and debugging parameters according to different types, sources or quality levels of past implantable sensor monitoring data to more accurately reflect the characteristics and laws of data optimization strategy matching under different circumstances. At the same time, the sensor monitoring data optimization system can also be combined with other advanced data processing and analysis technologies, such as machine learning, deep learning, etc., to further improve the accuracy and efficiency of data optimization strategy matching.

[0171] In some further optional embodiments, the method further includes: debugging the original data optimization strategy matching network to be debugged based on multiple groups of past linkage monitoring records through the following steps to obtain the target data optimization strategy matching network, wherein the past linkage monitoring record input into the original data optimization strategy matching network is the i-th linkage monitoring record: determining whether the past implantable device sensor monitoring information corresponding to the i+1-th past data quality characterization vector is the last past implantable device sensor monitoring information in the past implantable device sensor monitoring information flow; on the basis that the past implantable device sensor monitoring information corresponding to the i+1-th past data quality characterization vector is the last past implantable device sensor monitoring information, determining the training parameters generated by the original data optimization strategy matching network based on the i-th voting weight; determining the training parameters generated by the original data optimization strategy matching network based on the i+1-th past data quality characterization vector. On the basis that the corresponding past implantable device sensor monitoring information is not the last past implantable device sensor monitoring information, the training parameter is determined based on the i-th voting weight and the output of the associated data optimization strategy matching network, wherein the associated data optimization strategy matching network is a data optimization strategy matching network obtained by pre-reset processing, and there is a difference between the neural network weights of the associated data optimization strategy matching network and the original data optimization strategy matching network; based on the training parameter and the output of the original data optimization strategy matching network, the error variable expressed by the training error is determined, and the neural network weights of the original data optimization strategy matching network are adjusted based on the error variable expressed by the training error; based on the number of iterations of executing the above steps reaching the set iteration threshold, the original data optimization strategy matching network is determined as the target data optimization strategy matching network.

[0172] In this embodiment, in order to further improve the accuracy and efficiency of the data optimization strategy matching network, the sensor monitoring data optimization system adopts a debugging method based on multiple sets of past linkage monitoring records. This method makes full use of past monitoring data and learning information to iteratively debug the original data optimization strategy matching network to obtain a more accurate and efficient target data optimization strategy matching network.

[0173] First, the sensor monitoring data optimization system needs to clearly input the past linkage monitoring records of the original data optimization strategy matching network as the i-th linkage monitoring record. This means that the sensor monitoring data optimization system will use the data in the i-th linkage monitoring record to debug and optimize the original data optimization strategy matching network.

[0174] Next, the sensor monitoring data optimization system needs to determine whether the past implantable device sensor monitoring information corresponding to the i+1th past data quality characterization vector is the last past implantable device sensor monitoring information in the past implantable device sensor monitoring information flow. The purpose of this step is to determine whether the sensor monitoring data optimization system has processed all past linkage monitoring records. If it is the last past implantable device sensor monitoring information, the sensor monitoring data optimization system will determine the training parameters generated by the original data optimization strategy matching network based on the i-th voting weight. This training parameter will be used for subsequent network training and weight adjustment.

[0175] If the past implantable device sensor monitoring information corresponding to the i+1th past data quality characterization vector is not the last past implantable device sensor monitoring information, then the sensor monitoring data optimization system needs to determine the training parameters based on the i-th voting weight and the output of the associated data optimization strategy matching network. Here, the associated data optimization strategy matching network is a data optimization strategy matching network that has been reset in advance, and it is different from the neural network weights of the original data optimization strategy matching network. By using the output of the associated data optimization strategy matching network, the sensor monitoring data optimization system can introduce more diversity and differences, thereby further improving the accuracy and efficiency of debugging.

[0176] Once the sensor monitoring data optimization system determines the training parameters, the sensor monitoring data optimization system can determine the error variable expressed by the training error based on the training parameters and the output of the original data optimization strategy matching network. This error variable reflects the difference between the network output and the actual data defect state. The sensor monitoring data optimization system will use this error variable to adjust the neural network weights of the original data optimization strategy matching network to reduce the error and improve the accuracy of the network.

[0177] Then, the sensor monitoring data optimization system will repeat the above steps until the number of iterations reaches the set iteration threshold. In each iteration, the sensor monitoring data optimization system will use different past linkage monitoring records as input and continuously adjust the network weights to make the network output closer to the actual data defect state.

[0178] Finally, when the number of iterations reaches the set threshold, the sensor monitoring data optimization system determines the original data optimization strategy matching network as the target data optimization strategy matching network. This target network has been fully trained and debugged, has high accuracy and efficiency, and can be used for actual data optimization strategy matching tasks.

[0179] It can be seen that the accuracy and efficiency of data optimization strategy matching have been successfully improved by iteratively debugging the original data optimization strategy matching network based on multiple groups of past linkage monitoring records. In practical applications, the sensor monitoring data optimization system can flexibly adjust and optimize this technical solution according to specific needs and scenarios. For example, the sensor monitoring data optimization system can set different screening rules and debugging parameters according to different types, sources or quality levels of past implantable sensor monitoring data. At the same time, the sensor monitoring data optimization system can also be combined with other advanced data processing and analysis technologies, such as machine learning, deep learning, etc., to further improve the accuracy and efficiency of data optimization strategy matching.

[0180] In addition, in practical applications, the sensor monitoring data optimization system can flexibly increase or decrease the number of past linkage monitoring records according to specific needs and scenarios to adjust the accuracy and efficiency of debugging. At the same time, the sensor monitoring data optimization system can also adjust the size of the set iteration threshold according to actual needs to balance the accuracy and time cost of debugging.

[0181] In addition, it is worth noting that in this embodiment, the sensor monitoring data optimization system uses the associated data optimization strategy matching network to introduce more diversity and differences. This method can effectively improve the accuracy and efficiency of debugging and avoid the problems of overfitting and local optimal solutions. In practical applications, the sensor monitoring data optimization system can select different associated data optimization strategy matching networks according to specific needs and scenarios to achieve better effects and application value.

[0182] In this way, the accuracy and efficiency of data optimization strategy matching are successfully improved by iteratively debugging the original data optimization strategy matching network based on multiple sets of past linkage monitoring records. This technical solution has good scalability and adaptability, and can be flexibly adjusted and optimized according to specific needs and scenarios. In practical applications, the sensor monitoring data optimization system can be combined with other advanced data processing and analysis technologies to further improve the accuracy and efficiency of data optimization strategy matching.

[0183] In some exemplary embodiments, the training parameter is determined based on the output of the i-th voting weight and the associated data optimization strategy matching network, including: inputting each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set into the second discrimination algorithm branch in sequence together with the i+1-th past data quality characterization vector to obtain a second batch discrimination index, wherein the second batch discrimination index includes a quantitative discrimination index corresponding to each preprocessing strategy keyword; and determining the weighted result of the second quantitative discrimination index with the largest indicator value in the second batch discrimination index and the i-th voting weight as the training parameter.

[0184] In this embodiment, the sensor monitoring data optimization system focuses on how to determine the specific implementation of the training parameters based on the i-th voting weight and the output of the associated data optimization strategy matching network. The core of this process is to make full use of the information of the i-th voting weight and the output of the associated data optimization strategy matching network to generate a training parameter that can effectively guide the training of the original data optimization strategy matching network.

[0185] First, the sensor monitoring data optimization system needs to input each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set into the second discrimination algorithm branch in order together with the i+1th past data quality characterization vector. The purpose of this step is to obtain the second batch discrimination index, which can reflect the degree of association or matching between each preprocessing strategy keyword and the current data quality characterization vector. Specifically, the second batch discrimination index includes the quantitative discrimination index corresponding to each preprocessing strategy keyword, and these quantitative discrimination indexes can be in numerical form for subsequent comparison and selection.

[0186] Secondly, after obtaining the second batch discrimination index, the sensor monitoring data optimization system needs to select the second quantitative discrimination index with the largest index value. The second quantitative discrimination index with the largest index value represents the preprocessing strategy keyword that best matches or has the highest degree of correlation with the current data quality characterization vector. This indicator is selected to ensure that the sensor monitoring data optimization system can focus on the preprocessing strategies that are most relevant to the current data quality characterization vector in the subsequent training process, thereby improving the accuracy and efficiency of training.

[0187] Then, the sensor monitoring data optimization system determines the weighted result of the second quantitative discrimination index with the largest value of this index and the i-th voting weight as the training parameter. This step combines the results of the previous two steps to generate a comprehensive training parameter. Specifically, the sensor monitoring data optimization system performs a weighted operation on the second quantitative discrimination index and the i-th voting weight to obtain a weighted result, which is used as the training parameter for the subsequent training of the original data optimization strategy matching network of the sensor monitoring data optimization system. In this way, the sensor monitoring data optimization system can effectively integrate the information of the voting weights and the matching degree between the preprocessing strategy keywords and the data quality characterization vector into the training parameters, thereby guiding the network training process.

[0188] The advantage of this implementation is that it fully considers the information of voting weights and the degree of match between the preprocessing strategy keywords and the data quality characterization vector. By combining these two factors to generate training parameters, the sensor monitoring data optimization system can more accurately guide the training process of the original data optimization strategy matching network, thereby improving the accuracy and efficiency of the network. At the same time, this method also has certain flexibility and scalability. The sensor monitoring data optimization system can adjust the content of the preprocessing strategy keyword set and the specific implementation method of the discrimination algorithm branch according to actual conditions to adapt to different application scenarios and needs.

[0189] In practical applications, the sensor monitoring data optimization system can select appropriate preprocessing strategy keyword sets and discrimination algorithm branches according to specific data sets and task requirements. For example, for different implantable sensor monitoring tasks, the sensor monitoring data optimization system can customize the preprocessing strategy keyword set according to the characteristics and requirements of the task to ensure that the sensor monitoring data optimization system can capture key information related to the task. At the same time, the sensor monitoring data optimization system can also adjust the specific implementation method of the discrimination algorithm branch according to different data sets and task requirements to improve the accuracy and effectiveness of the discrimination indicators.

[0190] In addition, the sensor monitoring data optimization system can also combine other advanced data processing and analysis technologies to further optimize the implementation. For example, the sensor monitoring data optimization system can use machine learning or deep learning methods to train the discriminant algorithm branch to improve its accuracy and efficiency. At the same time, the sensor monitoring data optimization system can also use feature selection or dimensionality reduction methods to optimize the preprocessing strategy keyword set to improve its representativeness and effectiveness.

[0191] In this way, by fully considering the information of voting weights and the degree of match between the preprocessing strategy keywords and the data quality characterization vector, training parameters are generated to guide the training process of the original data optimization strategy matching network. This method has the advantages of high accuracy, high efficiency, strong flexibility and good scalability. In practical applications, it can be flexibly adjusted and optimized according to specific needs and scenarios. By combining other advanced data processing and analysis technologies, the sensor monitoring data optimization system can further improve the performance and effect of this implementation method.

[0192] In some independent embodiments, the method further includes: based on the preprocessing strategy keywords of the target vital sign monitoring data, performing data preprocessing optimization on the target vital sign monitoring data to obtain vital sign monitoring optimization data; and performing desensitizing and visualization processing on the vital sign monitoring optimization data.

[0193] It can be understood that the core of the embodiments of this application lies in comprehensively optimizing the preprocessing of target vital sign monitoring data based on the keyword of the preprocessing strategy for the target vital sign monitoring data, and then obtaining optimized vital sign monitoring data. Subsequently, desensitization and visualization processing are performed on these optimized data to ensure the security and interpretability of the data.

[0194] First of all, the keyword of the preprocessing strategy plays a crucial role. These keywords are carefully selected and optimized, and they can accurately reflect the key points and optimization directions that need to be concerned in the preprocessing of target vital sign monitoring data. Based on these keywords, the sensing monitoring data optimization system can build an efficient data preprocessing strategy to ensure the accuracy and consistency of the data.

[0195] In actual operation, the optimization of data preprocessing may include multiple specific processing steps. For example, the sensing monitoring data optimization system may need to clean the original data to remove noise and outliers. This step is crucial for improving the accuracy of subsequent analysis results. At the same time, the sensing monitoring data optimization system also needs to standardize or normalize the data to ensure that data from different sources or of different magnitudes can be compared and analyzed on the same scale. In addition, for certain specific vital sign monitoring data, such as time series data, the sensing monitoring data optimization system may also need to perform calibration and synchronization processing of timestamps to ensure the timeliness of the data.

[0196] After completing the optimization of data preprocessing, the sensing monitoring data optimization system will obtain a set of optimized vital sign monitoring data. This set of data has been significantly improved in terms of accuracy and consistency, providing a solid foundation for subsequent analysis and applications. However, before presenting the data to the end user or conducting further analysis, the sensing monitoring data optimization system also needs to consider the security and interpretability of the data.

[0197] This leads to the next important step: desensitization and visualization processing. In this step, the main goal of the sensing monitoring data optimization system is to ensure that the optimized vital sign monitoring data can be presented to the end user in an intuitive and easy-to-understand manner while protecting personal privacy and data security. To achieve this goal, the sensing monitoring data optimization system needs to adopt a series of data desensitization techniques, such as data masking, data generalization, etc., to hide or replace sensitive information that may disclose personal privacy. At the same time, the sensing monitoring data optimization system also needs to use data visualization techniques, such as charts, dashboards, etc., to present the data in an intuitive and easy-to-understand way so that users can quickly capture the key information and trends in the data.

[0198] In practical applications, desensitized visualization processing may face many challenges. For example, the sensor monitoring data optimization system needs to retain the original information and value of the data as much as possible while protecting personal privacy and data security. This requires the sensor monitoring data optimization system to perform fine control and adjustment during the data desensitization process to ensure that the desensitized data can still meet the needs of analysis and application. In addition, for different types of vital sign monitoring data, the sensor monitoring data optimization system may need to adopt different visualization technologies and methods to ensure the interpretability and ease of use of the data.

[0199] In this way, by optimizing data preprocessing based on the preprocessing strategy keywords of the target vital sign monitoring data, and then desensitizing and visualizing the optimized data, the comprehensive optimization and safe presentation of the vital sign monitoring data are achieved. This solution not only improves the accuracy and consistency of the data, but also ensures the security and interpretability of the data, providing a solid foundation for subsequent analysis and application. In practical applications, the sensor monitoring data optimization system can flexibly adjust and optimize the solution according to specific needs and scenarios to adapt to different data types and application scenarios. For example, the sensor monitoring data optimization system can adjust the selection and optimization process of the preprocessing strategy keywords according to the characteristics and analysis requirements of the data; the sensor monitoring data optimization system can also adjust the methods and techniques of data desensitization and visualization according to different user groups and needs. Through such flexible adjustment and optimization, the sensor monitoring data optimization system can better meet the diverse and personalized needs in practical applications and further improve the efficiency and effect of data processing.

[0200] Further, Figure 2 Schematic diagram of the structure of a sensor monitoring data optimization system 200 provided in an embodiment of the present application. Figure 2 The sensor monitoring data optimization system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0201] Alternatively, if Figure 2 As shown, the sensor monitoring data optimization 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 application.

[0202] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0203] Alternatively, if Figure 2As shown, the sensor monitoring data optimization 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.

[0204] Optionally, the sensor monitoring data optimization system 200 can 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 a storage engine in each method of the embodiments of the present application. For the sake of brevity, they are not repeated here.

[0205] It should be understood that the processor of the embodiment of the present application may be an integrated circuit chip with signal processing capabilities.

[0206] It is understood that the memory in the embodiments of the present application 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.

[0207] 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.

[0208] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do 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 methods and devices in the embodiments of the present application 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.

[0209] 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 application, 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 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 application.

[0210] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application 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 application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the present application, all of which are within the protection of the present application.

Claims

1. A sensor monitoring data optimization method for smart medical care, characterized in that: The method is applied to a sensor monitoring data optimization system, and the method comprises: Obtaining the original implantable device sensor monitoring information stream to be optimized; Inputting the original implantable device sensor monitoring information flow into a pre-debugged target data optimization strategy matching network to obtain a data optimization strategy matching viewpoint set; Determining the data optimization strategy matching viewpoint corresponding to each implantable device sensor monitoring information in the original implantable device sensor monitoring information flow based on the data optimization strategy matching viewpoint set; in: The original implantable device sensor monitoring information flow includes target vital sign monitoring data that needs to be preprocessed; The target data optimization strategy matching network is a data optimization strategy matching network obtained by inputting past learning information into the original data optimization strategy matching network to be debugged for debugging, the past learning information includes a batch of past data quality characterization vectors respectively mined from a batch of past implantable device sensor monitoring information and data defect states and voting weights corresponding to each past data quality characterization vector, the data defect state reflects an abnormal state keyword of a past data quality characterization vector, the voting weight reflects a reinforcement coefficient corresponding to the abnormal state keyword, the batch of past implantable device sensor monitoring information is a batch of uninterrupted implantable device sensor monitoring information, and each past data quality characterization vector includes a data quality characterization vector determined by a past implantable device sensor monitoring information and a previous past implantable device sensor monitoring information of the past implantable device sensor monitoring information; The data optimization strategy matching viewpoint is used to represent the preprocessing strategy keywords of the target vital sign monitoring data; The method further comprises: The data defect state and voting weight corresponding to each past data quality characterization vector in the batch of past data quality characterization vectors are determined respectively by the following steps, wherein the past data quality characterization vector for determining the data defect state and voting weight each time is used as the current past data quality characterization vector, and the data defect state corresponding to the current past data quality characterization vector is used as the current data defect state and the current voting weight: Inputting the current past data quality characterization vector into the original data optimization strategy matching network to obtain the current data defect state, wherein the original data optimization strategy matching network is a data optimization strategy matching network obtained by performing a reset process in advance; Determining the current voting weight based on the current data defect state and the prior learning annotation corresponding to the current past data quality characterization vector, wherein the prior learning annotation carries the authentication tag in the past implantable device sensor monitoring information corresponding to the current past data quality characterization vector, the authentication tag includes the target preprocessing strategy keyword of the target vital sign monitoring data, and the voting weight is used to indicate whether the current data defect state matches the target preprocessing strategy keyword; The step of inputting the current past data quality characterization vector into the original data optimization strategy matching network to obtain the current data defect state includes: Input each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set and the current past data quality characterization vector into the first discrimination algorithm branch in order to obtain a first batch of quantitative discrimination indicators, wherein the first batch discrimination indicators include the quantitative discrimination indicators corresponding to each preprocessing strategy keyword; Determine the candidate preprocessing strategy keyword corresponding to the first quantitative discrimination indicator with the largest indicator value among the first batch discrimination indicators as the current preprocessing strategy keyword corresponding to the current data defect state; Before inputting each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set in sequence together with the current past data quality characterization vector into the first discrimination algorithm branch to obtain the first batch of quantitative discrimination indicators, the method further includes: Determine the thermodynamic characteristic variables through the target mode; Determining whether the thermal characteristic variables conform to a set thermal distribution rule; On the basis that the thermal characteristic variables conform to the set thermal distribution rules, dynamically extracting candidate preprocessing strategy keywords from the pre-stored preprocessing strategy keyword set as the current data defect state; On the basis that the thermal characteristic variable does not conform to the set thermal distribution rule, the current past data quality characterization vector is input into the original data optimization strategy matching network to obtain the current data defect state; The step of determining the candidate preprocessing strategy keyword corresponding to the first quantitative discrimination indicator with the largest indicator value in the first batch discrimination indicators as the current preprocessing strategy keyword corresponding to the current data defect state includes: On the basis that the candidate preprocessing strategy keyword corresponding to the first quantitative discrimination indicator is the same as the target preprocessing strategy keyword, determining the current voting weight as the first voting weight, wherein the first voting weight indicates that the original data optimization strategy matching network successfully outputs a matching result; On the basis that the candidate preprocessing strategy keyword corresponding to the first quantitative discrimination indicator is different from the target preprocessing strategy keyword, determining the current voting weight as a second voting weight, wherein the second voting weight indicates that the original data optimization strategy matching network has not successfully output a matching result; The method further includes: based on the preprocessing strategy keywords of the target vital sign monitoring data, performing data preprocessing optimization on the target vital sign monitoring data to obtain vital sign monitoring optimization data; and performing desensitization and visualization processing on the vital sign monitoring optimization data.

2. The method according to claim 1, characterized in that The method further comprises: Performing first characterization vector mining on the batch of past implantable device sensor monitoring information to obtain a first batch of data quality characterization vectors, wherein a data quality characterization vector in the first batch of data quality characterization vectors corresponds to a piece of past implantable device sensor monitoring information in the batch of past implantable device sensor monitoring information; Performing second characterization vector mining on the batch of past implantable device sensor monitoring information to obtain a second batch of data quality characterization vectors, wherein a first data quality characterization vector in the second batch of data quality characterization vectors corresponds to a first past implantable device sensor monitoring information in the past implantable device sensor monitoring information, a previous frame of implantable device sensor monitoring information of the first past implantable device sensor monitoring information is the second past implantable device sensor monitoring information, and the first data quality characterization vector represents contextual embedded semantics between the first past implantable device sensor monitoring information and the second past implantable device sensor monitoring information; The first batch of data quality characterization vectors and the second batch of data quality characterization vectors are respectively integrated to obtain the batch of past data quality characterization vectors.

3. The method according to claim 2, characterized in that The second characterization vector mining is performed on the batch of past implantable device sensor monitoring information to obtain a second batch of data quality characterization vectors, including: The second characterization vector mining is performed on the batch of past implantable device sensor monitoring information through the following steps to obtain a second batch of data quality characterization vectors, wherein the past implantable device sensor monitoring information for which the second characterization vector mining is performed each time is used as the current past implantable device sensor monitoring information, and the obtained data quality characterization vector is used as the current data quality characterization vector, and the second batch of data quality characterization vectors includes the current data quality characterization vector; Performing a sensor monitoring information analysis operation on the current past implantable device sensor monitoring information to determine a first batch of distribution feature variables, wherein the first batch of distribution feature variables is used to describe the area of ​​the target vital sign monitoring data in the current past implantable device sensor monitoring information; Performing a sensor monitoring information analysis operation on the previous past implantable device sensor monitoring information of the current past implantable device sensor monitoring information to determine a second batch of distribution feature variables, wherein the second batch of distribution feature variables is used to describe the area of ​​the target vital sign monitoring data in the previous past implantable device sensor monitoring information; The current data quality characterization vector is determined based on the first batch of distribution feature variables and the second batch of distribution feature variables, wherein the current data quality characterization vector is used to represent the difference between corresponding distribution feature variables in the first batch of distribution feature variables and the second batch of distribution feature variables.

4. The method according to claim 1, characterized in that The method further comprises: Selecting a number of past linkage monitoring records from the past learning information, wherein the i-th past linkage monitoring record among the number of past linkage monitoring records includes the i-th past data quality characterization vector, the i-th data defect state corresponding to the i-th past data quality characterization vector, the i-th voting weight, and the i+1-th past data quality characterization vector, where i is a positive integer; The original data optimization strategy matching network to be debugged is debugged based on the several past linkage monitoring records to obtain the target data optimization strategy matching network, wherein, on the basis that the number of iterations for debugging the original data optimization strategy matching network reaches a set iteration threshold, the original data optimization strategy matching network is determined as the target data optimization strategy matching network, and on the basis that the number of iterations for debugging the original data optimization strategy matching network does not reach the set iteration threshold, the neural network weights of the original data optimization strategy matching network are improved according to a preconfigured training error expression, and the input of each round of debugging process is one past linkage monitoring record among the several past linkage monitoring records.

5. The method according to claim 4, characterized in that The method further comprises: The target data optimization strategy matching network is obtained by debugging the original data optimization strategy matching network to be debugged based on multiple groups of past linkage monitoring records through the following steps, wherein the past linkage monitoring record input into the original data optimization strategy matching network is the i-th linkage monitoring record: Determine whether the past implantable device sensor monitoring information corresponding to the (i+1)th past data quality characterization vector is the last past implantable device sensor monitoring information in the past implantable device sensor monitoring information stream; On the basis that the past implantable device sensor monitoring information corresponding to the i+1th past data quality characterization vector is the last past implantable device sensor monitoring information, determining the training parameters generated by the original data optimization strategy matching network based on the i-th voting weight; On the basis that the past implantable device sensor monitoring information corresponding to the i+1th past data quality characterization vector is not the last past implantable device sensor monitoring information, the training parameter is determined based on the i-th voting weight and the output of the associated data optimization strategy matching network, wherein the associated data optimization strategy matching network is a data optimization strategy matching network obtained by performing a reset process in advance, and the neural network weights of the associated data optimization strategy matching network and the original data optimization strategy matching network are different; Determining an error variable of the training error expression based on the training parameter and the output of the original data optimization strategy matching network, and adjusting a neural network weight of the original data optimization strategy matching network based on the error variable of the training error expression; On the basis that the number of iterations of executing the above steps reaches the set iteration threshold, the original data optimization strategy matching network is determined as the target data optimization strategy matching network; The determining of the training parameter based on the output of the i-th voting weight and the associated data optimization strategy matching network includes: Input each preprocessing strategy keyword in the pre-stored preprocessing strategy keyword set and the i+1th past data quality characterization vector into the second discrimination algorithm branch in order to obtain a second batch discrimination index, wherein the second batch discrimination index includes a quantitative discrimination index corresponding to each preprocessing strategy keyword; A weighted result of a second quantitative discrimination index having the largest index value among the second batch discrimination indexes and the i-th voting weight is determined as the training parameter.

6. A sensor monitoring data optimization 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 5.

Images