Mobile inpatient registration information automatic processing method fusing online and offline data

Abstract

The application discloses a mobile inpatient registration information automatic processing method fusing online and offline data, comprising the following steps: collecting online and offline registration data through a cross-end registration information intelligent verification platform, extracting and encoding time sequence characteristics through a medical scene time sequence characteristic coding model, and screening abnormal registration information through a registration information abnormality detection and discrimination model; based on a registration field intelligent completion derivation algorithm, in combination with associated texture features extracted from a texture optimization model of multi-dimensional image fusion, missing fields are derived and completed; through a tourist visualized and interactive algorithm, an interactive verification channel is constructed to complete data cross comparison, effective data is again subjected to time sequence characteristic strengthening coding to generate a standardized data set, and after texture optimization processing, information meeting medical registration specifications is output. The method realizes deep fusion of online and offline data, improves abnormality detection accuracy and completion result consistency, optimizes processing flow efficiency, and adapts to the digital transformation needs of medical institutions.

Description

Technical Field

[0001] This invention relates to the field of automatic processing technology for hospital registration information, and in particular to an automatic processing method for mobile admission and discharge registration information that integrates online and offline data. Background Technology

[0002] In the process of digital transformation of healthcare services, hospital admission and discharge registration serves as a crucial starting and ending point in the patient treatment process, and its data processing efficiency directly impacts the service quality of medical institutions and the patient's medical experience. Currently, the parallel model of online mobile appointment registration and offline on-site registration at medical institutions is widely adopted. However, these two registration channels are independent of each other, with differences in data formats and collection standards, resulting in fragmented registration information that is difficult to integrate efficiently. Furthermore, medical registration data includes numerous time-related fields, treatment-related fields, and basic personal information, exhibiting complex data dimensions and significant temporal characteristics. Traditional processing methods rely on manual verification and completion, which not only consumes significant manpower but is also prone to data anomalies due to human error. This fails to meet the needs of medical institutions for precise and efficient processing of registration information, necessitating a technological solution that can integrate online and offline data and intelligently process registration information.

[0003] Existing technologies suffer from two major shortcomings when processing mobile hospital admission and discharge registration information: First, they lack a collaborative processing mechanism for multi-source registration data from both online and offline sources. They fail to effectively combine the temporal characteristics and data texture features of medical scenarios for comprehensive analysis, resulting in insufficient accuracy in data anomaly detection. This makes it difficult to fully identify issues such as missing fields and logical conflicts. Furthermore, the completion derivation process lacks reliable feature support, leading to poor consistency and accuracy in the completion results. Second, they have not built an efficient cross-platform data interaction verification and standardized processing flow. The data comparison process lacks visual interactive support, making it difficult to quickly verify data validity. At the same time, the standardized processing does not fully consider the requirements of medical registration specifications, resulting in the processed registration information being difficult to directly adapt to the medical institution's business system. This requires additional secondary processing, which seriously affects data processing efficiency and application effectiveness. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of existing technologies, this invention provides an automatic processing method for mobile hospital admission and discharge registration information that integrates online and offline data.

[0005] The technical solution adopted in this invention is an automatic processing method for mobile hospital admission and discharge registration information that integrates online and offline data, including the following steps: S1, collecting online mobile terminal input registration fields and offline medical institution terminal uploaded registration data through a cross-terminal registration information intelligent verification platform, and calling a medical scenario temporal feature encoding model to extract and encode temporal features of time-related fields in the two types of data; S2, using a registration information anomaly detection and discrimination model to identify multi-dimensional anomaly factors in the encoded data, and filtering out abnormal registration information with missing fields, logical conflicts, and data inconsistencies; S3, based on the intelligent completion derivation algorithm for registration fields, combined with texture optimization of multi-dimensional image fusion. The model extracts texture features of associated data to infer and complete missing fields in abnormal registration information; S4, a registration information interaction verification channel is built through visitor visualization and interaction algorithms, and the completed data is cross-compared with the associated verification data fed back by the cross-end verification platform; S5, the medical scene temporal feature encoding model is called to perform temporal feature enhancement encoding on the effective data after cross-comparison to generate a standardized registration information dataset; S6, the texture optimization model based on multi-dimensional image fusion is used to perform data texture optimization processing on the standardized dataset, and output admission and discharge registration information that conforms to the registration specifications of medical institutions, completing the automatic processing of online and offline data fusion.

[0006] Furthermore, the anomaly detection and discrimination model for registration information in S2 uses the following formula to construct the anomaly factor calculation model:

[0007] ,

[0008] in, To register abnormal factor values, These are the model weight parameters. The encoding results of the time field for the time-series feature encoding model in the medical scenario. Data texture feature values ​​extracted for texture optimization models of multidimensional image fusion. For the i-th registration field data on the online mobile terminal, For the i-th registered field data of the offline terminal, This is a function for calculating field similarity. For data variance, This represents the interactive verification feedback value output by the visitor visualization and interaction algorithm, where n is the total number of registration fields. Total data for online mobile registration fields. Register the total data for the offline terminal fields.

[0009] Furthermore, the intelligent completion derivation algorithm for the registration field in S3 uses the following formula to construct the completion derivation model:

[0010]

[0011] in, To complete the registration field data, The texture optimization results of the multidimensional image fusion texture optimization model on the associated registration data are shown. For the kth missing registration field, For the valid associated fields corresponding to the k-th missing field, This is a function to calculate field relevance, where m is the total number of missing fields. The algorithm for visualizing and interacting with tourists processes user feedback. The encoding result of the registration timestamp for the time-series feature encoding model in the medical scenario.

[0012] Furthermore, the cross-comparison process in S4 uses the following formula to construct the comparison confidence model:

[0013]

[0014] in, This represents the cross-comparison confidence level. The texture feature extraction results of the multidimensional image fusion texture optimization model on the completed data. The verification data is provided by the cross-platform verification platform. For data comparison similarity function, The field interaction validation values ​​output by the visitor visualization and interaction algorithm. is the encoding result of the time-related field j in the medical scenario temporal feature encoding model, where p is the total number of time-related fields.

[0015] Furthermore, the temporal feature enhancement coding in S5 uses the following formula to construct the coding enhancement model:

[0016]

[0017] in, To enhance the encoded feature values, The initial encoding result of the time field for the time-series feature encoding model in the medical scenario. To optimize the texture feature values ​​of the validation data for a multidimensional image fusion texture optimization model, For visitor visualization and interaction algorithms, the interaction confirmation value is provided. To complete the data for the q-th field, To verify the data in the q-th field, This is a function for calculating field correlation. The encoding result of the time difference of the q-th field in the time series feature encoding model for medical scenarios is given by r, where r is the total number of fields participating in the comparison.

[0018] Furthermore, the data texture optimization process in S6 uses the following formula to construct the optimization model:

[0019] ,

[0020] in, The registration information data is output in a standardized manner. The result of the texture optimization model for multidimensional image fusion optimizing the data completion process. To optimize the weight parameters, The encoding result of the time-series feature encoding model for medical scenarios for the final registration time. The output interaction values ​​of the visitor visualization and interaction algorithm. This refers to the data in the s-th standard field of the medical institution registration specifications. Here, t is the data similarity function, and t is the total number of standard fields. This is the total data for the completed field.

[0021] Further, S3 includes the following steps: S31, using a multi-dimensional image fusion texture optimization model to extract texture features from the online and offline associated registration data collected by the cross-end registration information intelligent verification platform, and filtering out texture feature data that is strongly correlated with the missing fields; S32, using a registration field intelligent completion derivation algorithm to mine field association rules for the filtered texture feature data, and establishing a derivation mapping relationship between the missing fields and associated data; S33, constructing a derivation result preview channel based on the tourist visualization and interaction algorithm, and visually comparing the initially completed field data with the associated verification data; S34, adjusting the completion parameters according to the comparison results to generate complete registration field data that meets the data consistency requirements.

[0022] Furthermore, S4 includes the following steps: S41, obtaining the association mapping relationship between the completed registration data and the original registration data of offline medical institutions through the cross-end registration information intelligent verification platform; S42, calling the medical scenario time-series feature encoding model to perform secondary encoding on the time series fields in the two types of data to strengthen the comparison basis of the time dimension; S43, using the texture optimization model of multi-dimensional image fusion to perform texture alignment processing on the encoded data to eliminate the comparison error caused by data format differences; S44, building a cross-comparison interactive interface based on tourist visualization and interaction algorithms, presenting the comparison results in the form of texture feature maps, and completing the data consistency verification.

[0023] Further, S5 includes the following steps: S51, classifying and organizing the valid registration data after cross-comparison, dividing it into basic information fields, diagnosis and treatment information fields, and time-series related fields according to medical registration standards; S52, calling the medical scene time-series feature encoding model to extract multi-dimensional time-series features from the time-series related fields, generating encoded features including timestamps, time intervals, and sequence relationships; S53, optimizing the texture features of all classified field data through a multi-dimensional image fusion texture optimization model to improve the data standardization; S54, combining the feedback results of the visitor visualization and interaction algorithm, fusing the encoded features and the optimized data to form a standardized registration information dataset.

[0024] An automated processing method for mobile hospital admission and discharge registration information integrating online and offline data is proposed. This method is implemented through different units, including: a cross-terminal data acquisition and transmission unit, which establishes a two-way data connection with online mobile terminals and offline medical institution terminals to collect and transmit registration data in real time and synchronize the data to subsequent processing units; a medical scenario temporal feature encoding unit, which receives data transmitted from the cross-terminal data acquisition and transmission unit, extracts and encodes temporal features from time-related fields, and outputs the encoded feature data to an anomaly detection unit; a registration information anomaly detection and discrimination unit, connected to the medical scenario temporal feature encoding unit, identifies multi-dimensional anomaly factors in the encoded data, filters abnormal registration information, and sends it to a field completion unit; and a registration field intelligent completion derivation unit, which receives… The system receives anomaly information from the anomaly detection unit, combines it with the associated texture features extracted by the texture optimization model of multi-dimensional image fusion, and derives and completes missing fields. The completed data is then transmitted to the cross-comparison unit. The cross-end data cross-comparison and verification unit connects the intelligent completion derivation unit for registration fields and the cross-end data acquisition and transmission unit. It constructs an interactive verification channel through visitor visualization and interaction algorithms, completes the cross-comparison of completed data and verification data, and outputs valid data to the data standardization unit. The data texture optimization and standardization output unit receives valid data from the cross-comparison unit, processes it through the texture optimization model of multi-dimensional image fusion, and outputs admission and discharge registration information that conforms to the registration specifications of medical institutions. All units interact with each other through the data bus, forming a closed-loop processing flow.

[0025] Beneficial Effects: This invention proposes an automatic processing method for mobile hospital admission and discharge registration information that integrates online and offline data. It achieves collaborative collection and transmission of online and offline registration data through a cross-platform intelligent verification platform. Utilizing a medical scenario temporal feature encoding model to mine data temporal correlation features, and combining a multi-dimensional image fusion texture optimization model to extract data texture features, it addresses the shortcomings of existing technologies that lack multi-source data collaborative processing mechanisms through an anomaly detection and discrimination model and intelligent completion derivation algorithm. This comprehensively identifies anomalies such as missing fields and logical conflicts, and performs accurate completion derivation, significantly improving the comprehensiveness of data anomaly detection and the consistency of completion results. Simultaneously, it constructs an efficient cross-platform data interaction verification channel through visitor visualization and interaction algorithms, coupled with standardized processing procedures and texture optimization mechanisms. This compensates for the lack of closed-loop verification and standardized processing in existing technologies, eliminating comparison errors caused by data format differences. Information conforming to medical registration standards can be output without additional secondary processing, significantly reducing labor costs, improving the accuracy and efficiency of registration information processing, achieving deep integration and intelligent processing of online and offline data, fully adapting to the digital transformation needs of medical institutions, and optimizing patient medical processes and healthcare service experiences. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention.

[0027] Figure 2 This is a flowchart of method step S3 of the present invention;

[0028] Figure 3 This is a flowchart of method step S4 of the present invention;

[0029] Figure 4 This is a flowchart of step S5 of the method of the present invention;

[0030] Figure 5 This is a diagram showing the unit composition for implementing the method of the present invention. Detailed Implementation

[0031] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0032] like Figure 1As shown, the automatic processing method for mobile hospital admission and discharge registration information integrating online and offline data includes the following steps: S1, collecting online mobile terminal registration fields and offline medical institution terminal uploaded registration data through a cross-terminal registration information intelligent verification platform, and calling a medical scenario temporal feature encoding model to extract and encode temporal features of time-related fields in the two types of data; S2, using a registration information anomaly detection and discrimination model to identify multi-dimensional anomaly factors in the encoded data, and filtering out abnormal registration information with missing fields, logical conflicts, and data inconsistencies; S3, based on the intelligent completion algorithm for registration fields, combined with a multi-dimensional image fusion texture optimization model to extract... The associated data texture features are used to deduce and complete missing fields in abnormal registration information; S4, a registration information interaction verification channel is built through visitor visualization and interaction algorithms, and the completed data is cross-compared with the associated verification data fed back by the cross-end verification platform; S5, the medical scene temporal feature encoding model is called to perform temporal feature enhancement encoding on the effective data after cross-comparison to generate a standardized registration information dataset; S6, the texture optimization model based on multi-dimensional image fusion is used to perform data texture optimization processing on the standardized dataset, and output admission and discharge registration information that conforms to the registration specifications of medical institutions, completing the automatic processing of online and offline data fusion.

[0033] The implementation process of step S1 is as follows: The cross-terminal registration information intelligent verification platform establishes API interface connections with at least 30 mainstream online mobile applications, and at the same time builds a transmission protocol compatible with more than 20 offline medical institution terminal data formats. This enables real-time collection of registration fields such as patient name, gender, age, treatment items, and appointment time entered on the online mobile terminal and registration data such as outpatient number, inpatient number, examination report number, and department uploaded by the offline medical institution terminal. The collection frequency is set to once every 0.5 seconds to ensure data timeliness. After data collection, the system automatically invokes the medical scenario temporal feature encoding model. This model has eight preset time-related field recognition dimensions, including registration timestamp, consultation interval duration, and treatment process node time. It extracts temporal features from the time-related fields in the two types of data. During the extraction process, the feature sliding window size is set to 12 data points, and the encoding adopts a 32-bit binary encoding method to convert the time features into computable coded data, providing a foundation for subsequent anomaly detection. This step achieves preliminary correlation and standardized processing of online and offline data through multi-channel data integration and temporal feature quantification, laying the data foundation for the entire processing flow.

[0034] The implementation process of step S2 is as follows: After obtaining the encoded data, the registration information anomaly detection and discrimination model is started. The model has 15 built-in anomaly detection dimensions, including core detection directions such as field integrity, data logical correlation, and cross-channel data consistency. Each detection dimension is set with a corresponding weight coefficient, of which field integrity accounts for 30%, data logical correlation accounts for 40%, and cross-channel data consistency accounts for 30%. The model scans the coded data field by field, calculating the matching degree of each field with the preset standard data template, and filtering out suspected abnormal data with a matching degree of less than 75%. The suspected abnormal data then undergoes a second verification, calling a data logic verification rule library. This rule library includes over 200 medical registration data logic rules, such as age and treatment item matching rules, and department and examination item correspondence rules. It identifies abnormal factors such as missing fields, data type conflicts, out-of-range numerical values, and cross-channel data contradictions. Finally, it outputs a list of abnormal registration information, including the location of abnormal fields, abnormal type, and abnormal confidence level. The abnormal confidence level threshold is set to 80% to ensure the accuracy of abnormal identification. This step, through a multi-dimensional and multi-level abnormal detection mechanism, comprehensively investigates problems in the registration data, providing a clear objective for subsequent data completion.

[0035] The implementation process of step S3 is as follows: Based on the intelligent completion derivation algorithm for registration fields, the algorithm presets 25 field association rules, including the association relationship of multiple types of fields such as patient basic information, diagnosis and treatment information, and time series information. At the same time, it calls the texture optimization model of multi-dimensional image fusion. This model sets 10 texture feature extraction channels to extract texture features from the online and offline associated registration data collected by the cross-end registration information intelligent verification platform. The extracted features include data distribution texture, field association texture, and time series change texture. During the extraction process, the feature resolution is set to 64×64 pixels. Texture feature data with a correlation degree of more than 85% with the missing field is selected by feature similarity calculation. Based on the filtered texture feature data and combined with the types of missing fields in the abnormal registration information, the algorithm uses multi-field collaborative deduction, that is, it uses the effective data of at least 5 related fields to build a deduction model. During the deduction process, the number of iterations is set to 12, and the convergence threshold of each iteration is 0.001. It gradually approaches the real data of the missing fields and completes the deduction and completion of the missing fields. The completed data must meet the logical consistency with the related fields. This step achieves accurate completion of missing data and improves the completeness of registration information through texture feature support and multi-rule collaborative deduction.

[0036] The implementation process of step S4 is as follows: The system calls the visitor visualization and interaction algorithm to build a registration information interaction and verification channel, including a data comparison interface, a feature display interface, and a feedback input interface. The interface supports 10 people to interact online at the same time, with a response latency of less than 1 second. The registration data completed in step S3 is divided into three categories according to field type: basic information, diagnosis and treatment information, and time series information. Each type of data is assigned an independent comparison identifier. At the same time, related verification data is retrieved from the cross-end registration information intelligent verification platform, including patient historical registration data, medical institution filing data, cross-department shared data, etc., totaling no less than 10 types of related verification data sources. The algorithm performs field-by-field cross-comparison between the completed data and the associated verification data. The comparison process adopts a two-way comparison mechanism, which combines the comparison of the completed data to the verification data and the comparison of the verification data to the completed data. The comparison similarity threshold is set to 90%. The comparison results are displayed in the form of texture feature maps on the interactive interface. The map includes information such as field matching degree curves, data difference point markers, and associated feature heatmaps. Staff can confirm the comparison results through the interactive interface, and the feedback confirmation information is synchronized to the system in real time. This step ensures the accuracy and consistency of the completed data through visual interaction and cross-comparison of multi-source data.

[0037] The implementation process of step S5 is as follows: After completing the cross-comparison and obtaining valid data, the system calls the medical scenario temporal feature encoding model again to perform temporal feature enhancement encoding on the valid data. Before encoding, the valid data is first screened, retaining fields with a time correlation higher than 70%, totaling no less than 15 time-related fields, including consultation start time, examination completion time, payment time, discharge application time, etc. The model adopts a multi-scale temporal feature extraction method, setting three different time windows: 1 hour, 6 hours, and 24 hours, to extract multi-dimensional temporal features from the time-related fields. The extracted features include time series trends, time interval distribution, and temporal correlation strength. During the encoding process, a dynamic encoding strategy is adopted, adjusting the encoding parameters according to the temporal change pattern of the data. The encoding length is set to 64 bits, doubling the feature dimension compared to the initial encoding, thus enhancing the expressive power of temporal features. After coding is completed, all field data are classified and integrated, arranged in the field order of medical registration standards, and a standardized dataset including patient registration information throughout the entire process is generated. The dataset adopts a structured storage format and supports fast querying and retrieval. This step improves the standardization and usability of the data by strengthening time-series coding and standardized integration.

[0038] Step S6 is implemented as follows: A multi-dimensional image fusion texture optimization model is invoked to perform texture optimization processing on the standardized registration information dataset. The model incorporates eight texture optimization algorithms, including texture smoothing, texture alignment, and texture enhancement algorithms. The appropriate optimization algorithm is automatically selected based on the data type. During optimization, the texture resolution adjustment coefficient is set to 1.2, the texture contrast enhancement ratio to 20%, and the texture noise filtering threshold to 0.05. The algorithms eliminate redundant textures and inconsistent texture alignment in the data, enhance effective texture features, and improve data stability and consistency. After optimization, the system performs final verification of the data according to the 12 core requirements of the medical institution registration specifications. Verification includes field completeness, logical consistency, format standardization, and data validity. Datasets that pass verification are organized according to a preset output format. This output format supports integration with more than 20 medical institution business systems and can be directly imported into hospital information management systems, electronic medical record systems, etc. The final output of admission and discharge registration information includes four major modules: patient basic information, treatment process information, time-series correlation information, and cost-related information, totaling no less than 50 core fields. This completes the automatic processing of online and offline data integration. This step ensures that the output data meets the actual application requirements through texture optimization and standardization verification, thus achieving a closed loop in the processing flow.

[0039] Preferably, the anomaly detection and discrimination model for registration information in S2 uses the following formula to construct the anomaly factor calculation model:

[0040] ,

[0041] in, To register abnormal factor values, These are the model weight parameters. The encoding results of the time field for the time-series feature encoding model in the medical scenario. Data texture feature values ​​extracted for texture optimization models of multidimensional image fusion. For the i-th registration field data on the online mobile terminal, For the i-th registered field data of the offline terminal, This is a function for calculating field similarity. For data variance, This represents the interactive verification feedback value output by the visitor visualization and interaction algorithm, where n is the total number of registration fields. Total data for online mobile registration fields. Register the total data for the offline terminal fields.

[0042] Specifically, in step S2, the anomaly factor calculation model of the registration information anomaly detection and discrimination model is based on three core influencing factors: temporal features, texture features, and interactive verification feedback of medical registration data. The model structure is constructed through a combination of weighted summation and product. Since temporal encoding results directly reflect the rationality of the data's time dimension, texture features reflect the inherent correlation patterns of the data, and interactive verification feedback supplements manual verification information, the three work together to comprehensively cover the dimensions of anomaly detection. The model weight parameters were determined through training with 1000 sets of medical registration sample data: temporal feature encoding weight is 0.35, texture feature weight is 0.45, and interactive verification feedback weight is 0.2, ensuring a higher weight ratio for key features. Similarity calculation uses the cosine similarity algorithm, and variance calculation uses the sample variance formula. By traversing all registration fields, the similarity and data variance of corresponding fields online and offline are calculated to avoid false detections caused by single-dimensional judgment. Interactive verification feedback values ​​are collected through visitor visualization and interactive algorithms, with a range of 0-1, and are quantified based on the verification results provided by staff. This model calculates abnormal factor values ​​by fusing multiple factors. When the abnormal factor value is higher than 0.8, it is judged as abnormal data. In the process, the coded data is preprocessed first, and then grouped by field type and substituted into the model for calculation one group at a time. The calculation efficiency is controlled to process 500 data points per second, so as to realize the rapid and accurate identification of abnormal data and provide a reliable basis for subsequent processing.

[0043] Preferably, the intelligent completion derivation algorithm for the registration field in S3 uses the following formula to construct the completion derivation model:

[0044]

[0045] in, For the completed registration field data, The texture optimization results of the multidimensional image fusion texture optimization model on the associated registration data are shown. For the k-th missing registration field, For the valid associated fields corresponding to the k-th missing field, This is a function to calculate field relevance, where m is the total number of missing fields. The algorithm for visualizing and interacting with tourists processes user feedback. The encoding result of the registration timestamp for the time-series feature encoding model in the medical scenario.

[0046] Specifically, in step S3, the intelligent field completion derivation model is based on the fusion of texture optimization results of associated data, field relevance, interactive feedback, and temporal coding. Texture optimization reduces noise interference in the associated data, field relevance directly determines the reliability of the derivation, and interactive feedback and temporal coding supplement dynamically adjusted information. The combination of these three factors improves completion accuracy. When processing associated registration data, the texture optimization model uses a multi-channel feature extraction and fusion method. The relevance calculation uses the Pearson correlation coefficient algorithm, and by analyzing the historical data association patterns between missing and associated fields, a relevance threshold of 0.85 is determined. The interactive processing value, ranging from 0 to 1, is generated by the visitor visualization and interactive algorithm based on user feedback on the preliminary completion results. The temporal coding result uses a 32-bit binary encoded numerical conversion value. The model is constructed by combining product and summation. First, the product of the texture optimization result of the associated data and the field correlation is calculated. Then, the product of the interaction processing value and the temporal coding result is superimposed. When deducing the missing field completion, all associated valid fields are traversed and substituted into the model to calculate the completed data. The number of iterations is set to 12 to approximate the true value. The convergence threshold of 0.001 ensures the stability of the derivation. During the implementation, the corresponding associated field data is called according to the missing field type. The completed data is logically verified to confirm consistency, thus achieving accurate completion of missing data.

[0047] Preferably, the cross-comparison process in S4 uses the following formula to construct the comparison confidence model:

[0048]

[0049] in, This represents the cross-comparison confidence level. The texture feature extraction results of the multidimensional image fusion texture optimization model on the completed data. The verification data is provided by the cross-platform verification platform. For data comparison similarity function, The field interaction validation values ​​output by the visitor visualization and interaction algorithm. The encoding result of the medical scene temporal feature encoding model for the j-th time-related field is given by p, where p is the total number of time-related fields.

[0050] Specifically, in step S4, the cross-comparison confidence model is based on four factors: texture features of the completed data, data similarity, interactive verification values, and the sum of squares of time-series codes. Because the data consistency is higher after texture feature optimization, similarity directly reflects the matching degree between the completed data and the verification data, interactive verification values ​​reflect the interface interaction feedback, and the sum of squares of time-series codes reflects the stability of the time dimension. The combination of division and square root can balance the influence of each factor. Texture feature extraction uses the gray-level co-occurrence matrix algorithm, extracting 16 texture feature parameters from the completed data and then fusing them to generate texture feature values. Similarity calculation uses the Jaccard similarity algorithm, comparing the overlap between the completed data and the verification data of the cross-platform verification platform. Interactive verification values ​​are generated through visitor visualization and interaction algorithms, ranging from 0 to 1. The sum of squares of time-series codes is obtained by calculating the sum of squares of the encoding results of all time-related fields. The model construction uses a numerator-denominator structure: the numerator is the product of the texture feature value and the data similarity, and the denominator is the sum of the interactive verification value and the square root of the sum of squares of time-series codes. The confidence value ranges from 0 to 1; a confidence value higher than 0.9 is considered a successful comparison. During implementation, the fields of the completed data and the verification data are first aligned, and then the data is substituted into the model for calculation by category. A caching mechanism is set up during the calculation process to improve the calculation efficiency of duplicate fields and achieve fast and accurate verification of the completed data.

[0051] Preferably, the temporal feature enhancement coding in S5 uses the following formula to construct the coding enhancement model:

[0052]

[0053] in, To enhance the encoded feature values, The initial encoding result of the original time field for the time-series feature encoding model in the medical scenario. To optimize the texture feature values ​​of the validation data for a multidimensional image fusion texture optimization model, For visitor visualization and interaction algorithms, the interaction confirmation value is provided. To complete the data for the q-th field, To verify the data in the q-th field, This is a function for calculating field correlation. The encoding result of the time difference of the q-th field in the time series feature encoding model for medical scenarios is given by r, where r is the total number of fields involved in the comparison.

[0054] Specifically, in step S5, the temporal feature enhancement coding model is based on the original temporal coding result, verification data texture features, interactive confirmation values, field correlation, and time-series difference coding results. The original temporal coding provides basic features, the verification data texture features ensure data reliability, the interactive confirmation values ​​supplement human feedback, and the field correlation and time-series difference coding enhance the details of the temporal dimension. The fusion of multiple factors improves the coding effect. The enhancement coding model uses a combination of product and summation structures. The original temporal coding result uses the initial 32-bit binary numerical result. The verification data texture features are extracted through a texture optimization model based on multi-dimensional image fusion. The interactive confirmation value, ranging from 0 to 1, is generated by the visitor visualization and interaction algorithm based on the user's confirmation result. The field correlation calculation uses a mutual information algorithm to analyze the correlation between the corresponding fields of the completed data and the verification data. The time-series difference coding result is generated by encoding the field time difference using a medical scenario temporal feature coding model. The model first calculates the product of the original temporal encoding and the texture features of the validation data, then superimposes the sum of the interaction confirmation value and the field correlation degree divided by the sum of the temporal difference encoding results. The weight parameters were determined to be 0.5 and 0.5 after training with 500 sets of samples. During implementation, effective data are classified according to temporal correlation fields, and each field is substituted into the model for enhanced encoding. The encoding length is extended to 64 bits, and the encoding efficiency is controlled to process 300 data points per second. The generated enhanced encoding features can more comprehensively reflect the temporal patterns of the data.

[0055] Preferably, the data texture optimization process in S6 uses the following formula to construct the optimization model:

[0056] ,

[0057] in, The registration information data is output in a standardized manner. The result of the texture optimization model for multidimensional image fusion optimizing the data completion process. To optimize the weight parameters, The encoding result of the time-series feature encoding model for medical scenarios for the final registration time. The output interaction values ​​of the visitor visualization and interaction algorithm. This refers to the data in the s-th standard field of the medical institution registration specifications. Here, t is the data similarity function, and t is the total number of standard fields. This is the total data for the completed field.

[0058] Specifically, in step S6, the data texture optimization model is based on the completed data texture optimization results, the final temporal encoding, the interactive output value, and the data standard matching degree. Since the texture optimization result is fundamental, the final temporal encoding ensures the standardization of the time dimension, the interactive output value supplements dynamically adjusted information, and the data standard matching degree directly determines whether the output conforms to the standard. Weighted summation and product combination can achieve comprehensive optimization. The optimization weight parameters were determined through training with 800 sets of medical registration standard data. The final temporal encoding weight is 0.4, the interactive output value weight is 0.3, and the data standard matching degree weight is 0.3. The texture optimization model adopts a multi-algorithm adaptive selection mechanism, automatically matching the optimal texture optimization algorithm according to the data type. The final temporal encoding uses a 64-bit binary encoded numerical result. The interactive output value is generated by the visitor visualization and interaction algorithm, ranging from 0 to 1. The data standard matching degree uses a cosine similarity algorithm to calculate the degree of matching between the completed data and the standard field data. The model constructs the optimization model by first calculating the weighted sum of the products of the final temporal encoding, the interactive output value, and the data standard matching degree, and then multiplying it by the completed data texture optimization result. During implementation, the standardized dataset is grouped by module, and each group is substituted into the model for texture optimization. The optimized data is arranged in the order of fields according to the medical registration specifications, and the output format is XML to ensure compatibility with the business systems of medical institutions and to achieve standardized output of registration information.

[0059] Preferred, such as Figure 2 As shown, S3 includes the following steps: S31, using a multi-dimensional image fusion texture optimization model to extract texture features from the online and offline associated registration data collected by the cross-end registration information intelligent verification platform, and filtering out texture feature data that are strongly correlated with the missing fields; S32, using a registration field intelligent completion derivation algorithm to mine field association rules for the filtered texture feature data, and establishing a derivation mapping relationship between the missing fields and associated data; S33, constructing a derivation result preview channel based on the tourist visualization and interaction algorithm, and visually comparing the initially completed field data with the associated verification data; S34, adjusting the completion parameters according to the comparison results to generate complete registration field data that meets the data consistency requirements.

[0060] Specifically, step S3 involves the step-by-step implementation of precise completion of missing fields. S31 calls a multi-dimensional image fusion texture optimization model, which sets up 10 parallel texture feature extraction channels, each corresponding to different types of associated registration data. During extraction, the feature sampling frequency is once every 0.2 seconds, and the feature resolution is set to 64×64 pixels. Comprehensive texture feature extraction is performed on the online and offline associated registration data collected by the cross-end registration information intelligent verification platform. By calculating the feature correlation, texture feature data with a correlation higher than 85% with the missing fields are selected, providing reliable feature support for completion derivation. S32 initiates the intelligent completion derivation algorithm for registration fields. The algorithm has a built-in 25-field association rule base, categorized and stored according to basic information, diagnosis and treatment information, and time-series information. It iterates through the rule base to mine the missing fields and their correlation with the missing fields. The derivation mapping relationship between related data is set with a confidence threshold of 90% to ensure the reliability of the derivation logic. S33 uses a tourist visualization and interaction algorithm to build a preview channel for the derivation results. The channel supports 8 people to view online at the same time with a response latency of less than 0.8 seconds. The preliminary completed field data and the associated verification data are visualized and compared with each other in the form of texture feature maps. The maps include information such as field matching curves and difference point markers. S34 automatically adjusts the completion parameters based on the visualization comparison results with an adjustment step size of 0.01. After 12 iterations of optimization, complete registration field data with consistent logic and unified data format with the associated fields is generated. The whole process achieves accurate completion of missing data and improves the completeness of registration information through step-by-step feature extraction, rule mining, visualization comparison and parameter optimization.

[0061] Preferred, such as Figure 3 As shown, S4 includes the following steps: S41, obtaining the association mapping relationship between the completed registration data and the original registration data of offline medical institutions through the cross-end registration information intelligent verification platform; S42, calling the medical scenario time series feature encoding model to perform secondary encoding on the time series fields in the two types of data to strengthen the comparison basis of the time dimension; S43, using the texture optimization model of multi-dimensional image fusion to perform texture alignment processing on the encoded data to eliminate the comparison error caused by data format differences; S44, building a cross-comparison interactive interface based on tourist visualization and interaction algorithms, presenting the comparison results in the form of texture feature map, and completing the data consistency verification.

[0062] Specifically, step S4 involves cross-checking and consistency verification of data. In S41, the cross-platform registration information intelligent verification platform establishes a bidirectional data index to quickly link the completed registration data with the original registration data from offline medical institutions. The index is updated every 0.3 seconds to ensure the real-time nature of the mapping relationship. Simultaneously, it obtains a table showing the field correspondence between the two types of data, clarifying the matching rules for various fields such as basic information and treatment information. In S42, the medical scenario time-series feature encoding model is used to perform secondary encoding on the time-series fields in both types of data. The encoding window size is set to 8 data points, and the encoding bit depth is 32 bits, strengthening the comparison basis in the time dimension and making the comparison of time-related fields more targeted. S43 initiates a multi-dimensional image fusion texture optimization model. This model uses a texture alignment algorithm with an alignment precision of 0.005 to perform texture alignment processing on the encoded data, eliminating comparison errors caused by differences in data acquisition equipment and formats, and improving the accuracy of data comparison. S44 builds a cross-comparison interactive interface based on tourist visualization and interaction algorithms. The interface includes functional modules such as a comparison result display area and a feedback input area. The comparison results are presented in the form of texture feature maps with a color differentiation threshold of 0.7, clearly displaying matching and non-matching fields. Staff can confirm the comparison results through the interface, and feedback information is synchronized to the system in real time to complete data consistency verification and ensure the reliability of the supplementary data.

[0063] Preferred, such as Figure 4 As shown, S5 includes the following steps: S51, classifying and organizing the valid registration data after cross-comparison, dividing it into basic information fields, diagnosis and treatment information fields, and time-series related fields according to medical registration standards; S52, calling the medical scene time-series feature encoding model to extract multi-dimensional time-series features from the time-series related fields, generating encoded features including timestamps, time intervals, and sequence relationships; S53, optimizing the texture features of all classified field data through a multi-dimensional image fusion texture optimization model to improve the data standardization; S54, combining the feedback results of tourist visualization and interaction algorithms, fusing the encoded features and optimized data to form a standardized registration information dataset.

[0064] Specifically, step S5 aims to generate a standardized registration information dataset. S51 involves classifying and organizing the valid registration data after cross-checking into three categories according to medical registration standards: basic information fields, treatment information fields, and time-series correlation fields. The basic information fields include 20 core items, the treatment information fields include 18 core items, and the time-series correlation fields include 15 core items. The field recognition accuracy is set to 99% during the classification process to ensure accuracy. S52 calls a medical scenario time-series feature encoding model to extract multi-dimensional time-series features from the time-series correlation fields. The extracted features include timestamps, time intervals, and sequence relationships. The feature dimension is set to 12 dimensions during extraction, generating multi-dimensional... The encoding features of the degree information comprehensively reflect the temporal correlation patterns; S53 optimizes the texture features of all classified field data through a multi-dimensional image fusion texture optimization model. The model adopts a texture enhancement algorithm with an enhancement coefficient set to 1.2 to improve the standardization of the data and make the data format more in line with the requirements of subsequent processing; S54 combines the feedback results of the visitor visualization and interaction algorithm with a weight of 0.3 to fuse the encoded features and the optimized data. The weighted average method is used in the fusion process, the number of data fusion iterations is 10, and the convergence threshold is 0.001. Finally, a standardized registration information dataset with a regular structure, consistent data, and clear temporal characteristics is formed, laying the foundation for subsequent data texture optimization and output.

[0065] The multidimensional image fusion texture optimization model in this application does not narrowly process visual images, but rather performs numerical and structured feature mapping and image representation on text field data from medical admission and discharge registration. It is an innovative application that borrows the algorithmic framework and feature processing logic of image texture optimization to perform deep feature mining and optimization on text-based registration data. Specifically, this application transforms heterogeneous online and offline text registration fields (such as patient basic information, diagnosis and treatment fields, time-series correlation fields, cross-platform data fields, etc.) into image-texture-like numerical feature matrices / feature maps through quantization encoding, dimensional mapping, and matrix transformation. The core attributes such as the correlation between text fields, data distribution patterns, time-series variation characteristics, cross-platform data consistency characteristics, and field logical matching degree are analogized to image texture features (such as texture gradient, feature co-occurrence, texture continuity, and feature correlation density), thus completing the transformation of text data into computable texture features. Building upon this foundation, the model employs mature image texture optimization techniques (multi-channel feature fusion, texture denoising, texture alignment, texture enhancement, and resolution optimization) to process the texture feature matrix mapped from the text fields. This process accurately extracts texture features strongly correlated with missing fields. These features are essentially hidden association rules between text fields, data distribution patterns, and consistency features across different data sources, rather than visual image textures. Instead of performing image manipulation on the text fields, this model leverages the algorithmic advantages of image texture optimization to address the issues of heterogeneity, missing data, logical conflicts, and data inconsistencies in medical text registration fields. It provides highly reliable feature support for the intelligent completion derivation algorithm for registration fields, making the derivation of missing fields more accurate and the completion results more consistent. Simultaneously, it eliminates format and feature differences between online and offline text data, achieving deep fusion and standardized optimization of multi-source text registration data. This represents a cross-disciplinary innovation in algorithmic framework reuse and data feature representation, adapting to the complex characteristics of medical registration text data.

[0066] like Figure 5As shown, an automatic processing method for mobile hospital admission and discharge registration information integrating online and offline data is implemented through different units, including: a cross-terminal data acquisition and transmission unit, which establishes a two-way data connection with online mobile terminals and offline medical institution terminals to collect and transmit registration data in real time and synchronize the data to subsequent processing units; a medical scenario temporal feature encoding unit, which receives data transmitted by the cross-terminal data acquisition and transmission unit, extracts and encodes temporal features from time-related fields, and outputs the encoded feature data to an anomaly detection unit; a registration information anomaly detection and discrimination unit, which connects to the medical scenario temporal feature encoding unit, identifies multi-dimensional anomaly factors in the encoded data, filters abnormal registration information, and sends it to the field completion unit; and a registration field intelligent completion derivation unit. The system receives anomaly information from the anomaly detection unit, combines it with the associated texture features extracted by the texture optimization model of multi-dimensional image fusion, and performs inference and completion on missing fields, transmitting the completed data to the cross-comparison unit. The cross-end data cross-comparison and verification unit connects the intelligent completion inference unit for registration fields and the cross-end data acquisition and transmission unit, constructs an interactive verification channel through visitor visualization and interaction algorithms, completes the cross-comparison of completed data and verification data, and outputs valid data to the data standardization unit. The data texture optimization and standardization output unit receives valid data from the cross-comparison unit, processes it through the texture optimization model of multi-dimensional image fusion, and outputs admission and discharge registration information that conforms to the registration specifications of medical institutions. All units interact with each other through the data bus, forming a closed-loop processing flow.

[0067] The formula in this invention integrates different scalar and vector parameters for unified calculation. Through standardization, dimensional adaptation, and weight allocation, a collaborative calculation logic is established to ensure the parameters are compatible in both numerical and physical dimensions. First, for scalar parameters such as field correlation and weight coefficients, normalization is used to map their values ​​to the 0-1 range, adapting them to the feature dimensions of vector parameters. For vector parameters such as temporal coding features and texture feature maps, feature dimensionality reduction or flattening is used to transform multidimensional vectors into single-dimensional numerical vectors, enabling direct computation with scalar parameters. For example, vector data generated from temporal feature encoding in medical scenarios, after flattening, is transformed into single-dimensional numerical values ​​reflecting temporal patterns, adapting to computation with scalar forms of field similarity and weight coefficients. Simultaneously, the formula balances the influence of different types of parameters by appropriately setting weight parameters, such as temporal feature encoding weights and texture feature weights, ensuring that the scalar weight adjustment effect complements the vector feature representation effect. Furthermore, by utilizing similarity calculation and variance analysis, a mathematical relationship between scalars and vectors is constructed. For example, the texture feature data of vectors and the correlation between scalar fields are fused through multiplication, preserving the feature details of vectors while controlling the degree of influence through scalars. This design comprehensively represents data features using vector parameters while optimizing calculation accuracy through scalar parameters, achieving the organic integration of different types of parameters and ensuring that the formula calculation results accurately reflect the core needs of registration information processing.

[0068] This method for automatically processing mobile hospital admission and discharge registration information, integrating online and offline data, addresses the lack of a multi-source data collaborative processing mechanism. It breaks down data barriers between online and offline channels through a cross-platform intelligent verification platform, enabling real-time collection and synchronization of both types of data. A medical scenario temporal feature encoding model is used to mine time-related features, and a multi-dimensional image fusion texture optimization model is combined to extract data texture features. This provides comprehensive feature support for an anomaly detection and discrimination model, accurately identifying anomalies such as missing fields and logical conflicts. Simultaneously, an intelligent field completion derivation algorithm establishes a correlation mapping relationship, ensuring the consistency and accuracy of the completion results. To address the lack of closed-loop verification and standardized processing, an interactive verification channel is constructed using visitor visualization and interactive algorithms. This allows for cross-comparison of completed and verified data, eliminating errors caused by format differences. Finally, temporal feature-enhanced encoding and texture optimization processing form a complete standardized process, outputting registration information directly compatible with medical institution business systems without secondary processing.

[0069] This method excels in data processing collaboration and accuracy. Through multi-model collaborative operation, it achieves deep integration of online and offline data, providing more comprehensive anomaly detection coverage and more sufficient basis for supplementary inferences, effectively reducing the data error rate. The processing workflow is highly efficient and adaptable, with a closed-loop data processing and interactive verification mechanism that significantly shortens the processing cycle. Standardized output directly meets medical registration standards, significantly reducing manpower input. The technology application is practical and scalable, adapting to the registration needs of different medical institutions. It can be adapted to diverse scenarios through parameter adjustments, improving the efficiency of medical services while optimizing the patient's medical experience, providing reliable technical support for the digital transformation of medical services.

[0070] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0071] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for automatically processing mobile hospital admission and discharge registration information that integrates online and offline data, characterized in that, Including the following steps: S1 collects online mobile terminal registration fields and offline medical institution terminal uploaded registration data through the cross-terminal registration information intelligent verification platform, and calls the medical scenario time-series feature coding model to extract and encode the time-related fields in the two types of data. S2, using the registration information anomaly detection and discrimination model to identify multi-dimensional anomaly factors in the encoded data, and filtering out abnormal registration information with missing fields, logical conflicts and inconsistent data; S3, based on the intelligent completion derivation algorithm for registration fields, combined with the texture features of associated data extracted by the texture optimization model of multi-dimensional image fusion, derivation and completion of missing fields in abnormal registration information; S4, through the construction of a registration information interaction verification channel by tourist visualization and interaction algorithms, cross-compares the completed data with the associated verification data fed back by the cross-end verification platform; S5, call the medical scenario temporal feature coding model to perform temporal feature enhancement coding on the effective data after cross-comparison, and generate a standardized registration information dataset; S6, based on a multi-dimensional image fusion texture optimization model, performs data texture optimization processing on standardized datasets, outputs admission and discharge registration information that conforms to the registration standards of medical institutions, and completes the automatic processing of online and offline data fusion; The intelligent completion derivation algorithm for the registration field in S3 uses the following formula to construct the completion derivation model: in, For the completed registration field data, The texture optimization results of the multidimensional image fusion texture optimization model on the associated registration data are shown. For the k-th missing registration field, For the valid associated fields corresponding to the k-th missing field, This is a function to calculate field relevance, where m is the total number of missing fields. The algorithm for visualizing and interacting with tourists processes user feedback. The encoding result of the registration timestamp for the time-series feature encoding model in the medical scenario.

2. The automatic processing method for mobile admission and discharge registration information integrating online and offline data according to claim 1, characterized in that, The anomaly detection and discrimination model for registration information in S2 uses the following formula to construct the anomaly factor calculation model: ， in, To register abnormal factor values, These are the model weight parameters. The encoding results of the time field for the time-series feature encoding model in the medical scenario. Data texture feature values ​​extracted for texture optimization models of multidimensional image fusion. For the i-th registration field data on the online mobile terminal, For the i-th registered field data of the offline terminal, This is a function for calculating field similarity. For data variance, This represents the interactive verification feedback value output by the visitor visualization and interaction algorithm, where n is the total number of registration fields. For the total data of the online mobile registration fields, Register the total data for the offline terminal fields.

3. The automatic processing method for mobile admission and discharge registration information integrating online and offline data according to claim 1, characterized in that, The cross-alignment process in S4 uses the following formula to construct the alignment confidence model: in, This represents the cross-comparison confidence level. The texture feature extraction results of the multidimensional image fusion texture optimization model on the completed data. The verification data is provided by the cross-platform verification platform. For data comparison similarity function, The field interaction validation values ​​output by the visitor visualization and interaction algorithm. The encoding result of the medical scene temporal feature encoding model for the j-th time-related field is given by p, where p is the total number of time-related fields.

4. The automatic processing method for mobile admission and discharge registration information integrating online and offline data according to claim 1, characterized in that, The temporal feature enhancement coding in S5 uses the following formula to construct the coding enhancement model: in, To enhance the encoded feature values, The initial encoding result of the original time field for the time-series feature encoding model in the medical scenario. To optimize the texture feature values ​​of the validation data for a multidimensional image fusion texture optimization model, For visitor visualization and interaction algorithms, the interaction confirmation value is provided. To complete the data for the q-th field, To verify the data in the q-th field, This is a function for calculating field correlation. The encoding result of the time difference of the q-th field in the time series feature encoding model for medical scenarios is given by r, where r is the total number of fields involved in the comparison.

5. The automatic processing method for mobile admission and discharge registration information integrating online and offline data according to claim 1, characterized in that, The data texture optimization process in S6 uses the following formula to construct the optimization model: ， in, The registration information data is output in a standardized manner. The result of the texture optimization model for multidimensional image fusion optimizing the data completion process. To optimize the weight parameters, The encoding result of the time-series feature encoding model for medical scenarios for the final registration time. The output interaction values ​​of the visitor visualization and interaction algorithm. This refers to the data in the s-th standard field of the medical institution registration specifications. Here, t is the data similarity function, and t is the total number of standard fields. This is the total data for the completed field.

6. The automatic processing method for mobile admission and discharge registration information integrating online and offline data according to claim 1, characterized in that, S3 includes the following steps: S31, call the multi-dimensional image fusion texture optimization model to extract texture features from the online and offline associated registration data collected by the cross-end registration information intelligent verification platform, and filter out texture feature data that are strongly correlated with the missing fields; S32, through the intelligent completion derivation algorithm of registration fields, field association rule mining is performed on the filtered texture feature data to establish the derivation mapping relationship between missing fields and associated data; S33, Based on the tourist visualization and interaction algorithm, a preview channel for the derivation results is constructed, and the preliminary supplemented field data is visually compared with the associated verification data; S34. Adjust the completion parameters based on the comparison results to generate complete registration field data that meets the data consistency requirements.

7. The automatic processing method for mobile admission and discharge registration information integrating online and offline data according to claim 1, characterized in that, S4 includes the following steps: S41, obtain the correlation mapping relationship between the completed registration data and the original registration data of offline medical institutions through the cross-terminal registration information intelligent verification platform; S42, invoke the medical scenario time series feature encoding model to perform secondary encoding on the time series fields in the two types of data, and strengthen the basis for comparison in the time dimension; S43 utilizes a multi-dimensional image fusion texture optimization model to perform texture alignment processing on the encoded data, eliminating comparison errors caused by differences in data format; S44, based on tourist visualization and interaction algorithms, builds a cross-comparison interactive interface, presents the comparison results in the form of texture feature maps, and completes data consistency verification.

8. The automatic processing method for mobile admission and discharge registration information integrating online and offline data according to claim 1, characterized in that, S5 includes the following steps: S51, after cross-comparison, the valid registration data is classified and organized by field, and the basic information field, diagnosis and treatment information field and time sequence association field are divided according to the medical registration standard; S52, call the medical scenario temporal feature encoding model to extract multi-dimensional temporal features from the temporal correlation fields, and generate encoded features including timestamps, time intervals and sequence relationships; S53 optimizes the texture features of all classified field data through a multi-dimensional image fusion texture optimization model, thereby improving the degree of data standardization. S54 combines the feedback results from tourist visualization and interaction algorithms to fuse the encoded features and optimized data, forming a standardized registration information dataset.

9. The automatic processing method for mobile admission and discharge registration information integrating online and offline data according to any one of claims 1-8, characterized in that, This method is implemented through different units, including: The cross-terminal data acquisition and transmission unit establishes a two-way data connection with online mobile terminals and offline medical institution terminals to collect and transmit registration data in real time, and synchronize the data to the subsequent processing unit. The medical scenario temporal feature encoding unit receives data transmitted from the cross-end data acquisition and transmission unit, extracts and encodes temporal features from time-related fields, and outputs the encoded feature data to the anomaly detection unit. The registration information anomaly detection and discrimination unit is connected to the medical scenario temporal feature encoding unit to identify multi-dimensional abnormal factors in the encoded data, filter abnormal registration information, and send it to the field completion unit. The registration field intelligent completion derivation unit receives the abnormal information from the abnormality detection unit, combines the associated texture features extracted by the texture optimization model of multi-dimensional image fusion, derives and completes the missing fields, and transmits the completed data to the cross-comparison unit. The cross-end data cross-comparison and verification unit connects the registration field intelligent completion derivation unit and the cross-end data collection and transmission unit. It constructs an interactive verification channel through tourist visualization and interactive algorithms, completes the cross-comparison of the completed data and the verification data, and outputs the valid data to the data standardization unit. The data texture optimization and standardization output unit receives valid data from the cross-comparison unit, processes it through a multi-dimensional image fusion texture optimization model, and outputs admission and discharge registration information that conforms to the registration standards of medical institutions. Each unit interacts with data through a data bus to form a closed-loop processing flow.

Images