A diagnosis and treatment data association verification method based on digital twinning

Abstract

The application discloses a kind of based on digital twin's diagnosis and treatment data association verification method, comprising the following steps: collecting the multi-source diagnosis and treatment data of target diagnosis and treatment process, generates diagnosis and treatment input structure;With four-dimensional variation assimilation as core, construct diagnosis and treatment data association verification framework;Based on diagnosis and treatment input structure, construct diagnosis and treatment process state layer and associated consistency state layer, generate double-layer twin state set;In combination with diagnosis and treatment event sequence, construct asymmetric time topology;Based on double-layer twin state set and asymmetric time topology, construct four-dimensional variation assimilation objective function;Execute four-dimensional variation assimilation solution, calculate associated consistency measure and construct diagnosis and treatment digital twin trajectory;Based on diagnosis and treatment digital twin trajectory and associated consistency measure generates association determination mark, and when association is not established, output association abnormal positioning information.The application improves the accuracy and traceability of multi-source diagnosis and treatment data association verification.

Description

Technical Field

[0001] This invention relates to the field of medical informatics, and in particular to a method for verifying the association of diagnostic and treatment data based on digital twins. Background Technology

[0002] With the continuous improvement of medical informatization, the data generated during diagnosis and treatment activities are characterized by diverse sources, complex structures, and inconsistent time distribution. This data is typically stored in various medical information systems, examination and testing systems, and different diagnostic and treatment devices. Differences exist between these systems in terms of data structure, field meaning, time recording methods, and identification rules, making it difficult to directly regard multi-source medical data as a unified representation of the same diagnostic and treatment process in practical applications.

[0003] In existing technologies, consistency verification of multi-source medical data mainly relies on rule matching, field comparison, or post-event manual verification. These methods typically rely on static fields or local time segments, lacking a systematic characterization of the overall temporal evolution of the medical process and failing to accurately reflect the inherent correlation of medical events in a continuous process. When there are temporal misalignments, semantic deviations, or inconsistent identity identifiers among data sources, existing methods often only provide simple matching results, unable to determine whether the inconsistency is within an acceptable range, or to pinpoint the specific stage at which the anomaly occurred.

[0004] In addition, although some technologies introduce models or optimization methods to process medical data, they often analyze different data as independent inputs and lack a mechanism for collaborative modeling of the status of the medical process and the status of data association under a unified time constraint. This makes the correlation verification results susceptible to local noise and difficult to meet the requirements for data reliability and traceability in complex medical scenarios.

[0005] Therefore, how to provide a method for verifying the correlation of medical data based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a method for verifying the association of medical data based on digital twins. This invention uses four-dimensional variational assimilation as its core to construct a framework for verifying the association of medical data. By modeling the structure of medical inputs, a two-layer twin state set is formed, consisting of a medical process state layer and an association consistency state layer. Assimilation calculations are performed under asymmetric temporal topological constraints to generate a medical digital twin trajectory and an association consistency metric, thereby achieving the association verification and anomaly localization of multi-source medical data. This invention can collaboratively constrain the evolution of the medical process and the data association state within a unified assimilation window, supporting the generation of association judgment identifiers and the determination of anomaly occurrence stage identifiers and anomaly type identifiers. It possesses the advantages of high accuracy in verifying the association of medical data, strong traceability of anomaly localization, and complete characterization of process consistency.

[0007] A method for verifying the association of medical data based on digital twins according to an embodiment of the present invention includes the following steps: Collect multi-source diagnosis and treatment data for the target diagnosis and treatment process, perform time alignment, field standardization and identity extraction processing, and generate a diagnosis and treatment input structure, including diagnosis and treatment event sequence, field mapping information and identity information; Based on four-dimensional variational assimilation, a framework for verifying the correlation of diagnostic and treatment data is constructed, including a diagnostic and treatment state modeling unit, a correlation state modeling unit, a time topology construction unit, and a potential function construction unit. The diagnostic and treatment input structure is input into the diagnostic and treatment state modeling unit and the associated state modeling unit to construct the diagnostic and treatment process state layer and the associated consistency state layer, forming a two-layer twin state set; The sequence of diagnostic and treatment events and the set of two-layer twin states are input into the temporal topology construction unit to construct an asymmetric temporal topology; By constructing units of two-layer twin state sets and asymmetric time topology input potential functions, a four-dimensional variational assimilation objective function is constructed, including process consistency potential function and correlation consistency potential function; Based on the diagnostic input structure and the four-dimensional variational assimilation objective function, the four-dimensional variational assimilation solution is performed, the correlation consistency metric is calculated, and a diagnostic digital twin trajectory is constructed. Based on the diagnosis and treatment digital twin trajectory and the correlation consistency measurement, a correlation determination identifier is generated. When the correlation determination identifier indicates that the correlation is not established, correlation anomaly location information is generated.

[0008] Optionally, the generated diagnostic input structure includes: Collect multi-source diagnostic and treatment data corresponding to the target diagnostic and treatment process from diagnostic and treatment information systems, examination and testing systems and diagnostic and treatment equipment records, perform time alignment processing, and map data records from different sources but pointing to the same diagnostic and treatment behavior to the same time axis position; Based on the multi-source medical data that has undergone time alignment processing, field normalization processing is performed on the fields in each data record to generate field mapping information; Based on the multi-source medical data after field standardization, identity information related to the medical subject in each data record is extracted and generated; Using a single medical treatment action as the smallest event granularity, the multi-source medical treatment data after completion time alignment, field standardization, and identity extraction are processed into event-based data, and each medical treatment action is constructed into a sequence of medical treatment events according to the time sequence. The diagnostic input structure is constructed based on the sequence of diagnostic events, field mapping information, and identity information.

[0009] Optionally, the construction of the diagnostic and treatment data association verification framework includes: After the diagnostic input structure is determined, a diagnostic data association verification framework is established based on the modeling and solution requirements of four-dimensional variational assimilation in the diagnostic data association verification task. Within the framework for verifying and linking diagnostic and treatment data, a diagnostic and treatment status modeling unit is configured to define the state space of the diagnostic and treatment process. Within the framework for verifying the association of diagnostic and treatment data, configure an association state modeling unit to define the association consistency state space; Within the framework for verifying the association of medical data, a time topology construction unit is configured to limit the time organization method in the process of verifying the association of medical data. Within the framework for verifying the correlation of diagnostic and treatment data, potential function construction units are configured to limit the source of constraints for the four-dimensional variational assimilation objective function.

[0010] Optionally, forming the two-layer twin state set includes: The diagnosis and treatment status modeling unit uses the sequence of diagnosis and treatment events as the basis for status modeling, identifies the stage attributes of diagnosis and treatment events in the diagnosis and treatment process, and constructs diagnosis and treatment stage variables; In the diagnosis and treatment status modeling unit, based on the occurrence order and completion status of each diagnosis and treatment event in the diagnosis and treatment event sequence, the degree of progress of the diagnosis and treatment events in the diagnosis and treatment process is characterized, and event progress variables are constructed. In the diagnosis and treatment status modeling unit, the temporal continuity of the diagnosis and treatment process is characterized based on the time interval relationship between adjacent diagnosis and treatment events in the sequence of diagnosis and treatment events, and a temporal continuity variable is constructed. Based on time continuity variables, diagnosis and treatment stage variables, and event progress variables, the diagnosis and treatment process state layer is composed of these variables. The associated state modeling unit uses field mapping information and identity identification information as the basis for state modeling, and describes the matching situation of different data records pointing to the same diagnosis and treatment object, and constructs identity consistency variables; In the associated state modeling unit, based on the field correspondence defined in the field mapping information, the degree of matching of multi-source diagnosis and treatment data at the field value level is characterized, and a field mapping consistency variable is constructed. In the associated state modeling unit, based on the semantic correspondence of fields constrained by field mapping information, the semantic matching of multi-source diagnosis and treatment data is characterized, and semantic consistency variables are constructed. The association consistency state layer is composed of semantic consistency variables, identity consistency variables, and field mapping consistency variables. A joint state organization is performed on the state layer of the diagnosis and treatment process and the state layer of correlation and consistency, and a unified arrangement is formed to form a two-layer twin state set.

[0011] Optionally, the construction of the asymmetric time topology includes: The sequence of diagnostic and treatment events and the set of two-layer twin states are input into the time topology construction unit. The order of events and the dependencies between events in the sequence of diagnostic and treatment events are used as the basis for time organization. The time range involved in the process of verifying the correlation of diagnostic and treatment data is limited, and an assimilation window is generated. Within the assimilation window, a forward treatment evolution time path is constructed according to the occurrence order and dependencies of treatment events in the treatment event sequence; Within the assimilation window, based on the dependencies of diagnosis and treatment events in the sequence of diagnosis and treatment events, a reverse correlation verification time path is constructed by tracing back from the later stages of the diagnosis and treatment process to the earlier stages. In the time topology construction unit, the time organization relationship between the forward diagnosis and treatment evolution time path and the reverse correlation verification time path is distinguished, and a time structure with asymmetrical time order is generated. The coverage of the forward diagnosis and treatment evolution time path and the reverse correlation verification time path in terms of time span is distinguished, and a time structure with asymmetric path length is generated. Distinguish the state objects affected by the forward diagnosis and treatment evolution time path and the reverse correlation verification time path, and generate an asymmetric time structure of the constraint objects. After completing the time organization processing for asymmetric time order, asymmetric path length, and asymmetric constraint objects, the assimilation window, the forward diagnosis and treatment evolution time path, and the reverse correlation verification time path are summarized to generate a complete asymmetric time topology.

[0012] Optionally, the construction of the four-dimensional variational assimilation objective function includes: Within the framework for verifying the association of diagnostic and treatment data, a two-layer twin state set and an asymmetric temporal topology are received. Based on the assimilation window, the participation range of the two-layer twin state set in the time dimension is constrained to determine the set of state variables. Based on the diagnosis and treatment stage variables, event progress variables, and time continuity variables in the state layer of the diagnosis and treatment process, a process consistency potential function is constructed. The process consistency potential function quantifies the state deviation between the state variables of the diagnosis and treatment process at adjacent time positions, forming a penalty term for the continuity of the time evolution of the state layer of the diagnosis and treatment process. Based on the identity consistency variables, field mapping consistency variables, and semantic consistency variables in the association consistency state layer, an association consistency potential function is constructed. The association consistency potential function quantifies the deviation of the association consistency state variables on the time position of the forward diagnosis and treatment evolution time path within the assimilation window and the backtracking position of the reverse association verification time path, forming a penalty term for association inconsistency. By combining the process consistency potential function and the correlation consistency potential function according to the time path interaction relationship in the asymmetric time topology, a four-dimensional variational assimilation objective function is constructed.

[0013] Optionally, the calculation of the correlation consistency metric and the construction of the diagnosis and treatment digital twin trajectory include: Within the framework for verifying and linking diagnostic and treatment data, the system receives the diagnostic and treatment input structure and the four-dimensional variational assimilation objective function. Based on the assimilation window, it maps the diagnostic and treatment event sequence, field mapping information, and identity information in the diagnostic and treatment input structure to the time position corresponding to the two-layer twin state set, and generates the initial state configuration. Within the assimilation window, according to the time organization relationship of the forward diagnosis and treatment evolution time path and the reverse correlation verification time path in the asymmetric time topology, the two-layer twin state set is solved by four-dimensional variational assimilation. By iteratively optimizing the four-dimensional variational assimilation objective function, the state values ​​of the diagnosis and treatment process state layer and the correlation consistency state layer within the assimilation window are jointly updated. In the process of solving the four-dimensional variational assimilation problem, based on the correlation consistency potential function, the deviation of the correlation consistency state variable between the time position on the forward diagnosis and treatment evolution time path and the backtracking position on the reverse correlation verification time path within the assimilation window is quantified, and a correlation consistency metric that changes with time position is generated. After completing the four-dimensional variational assimilation solution, a digital twin trajectory for diagnosis and treatment is constructed based on the state evolution results of the updated diagnosis and treatment process state layer within the assimilation window.

[0014] Optionally, generating associated anomaly location information includes: Based on the diagnosis and treatment digital twin trajectory and the correlation consistency measure, the maximum deviation value within the assimilation window is extracted from the correlation consistency measure to generate the correlation verification deviation value; Obtain a preset association judgment threshold, and generate an association judgment identifier based on the comparison result between the association verification deviation value and the preset association judgment threshold. When the association verification deviation value does not exceed the preset association judgment threshold, the association judgment identifier indicates that the association consistency measurement of multi-source diagnosis and treatment data under the same diagnosis and treatment digital twin trajectory constraint is within the allowable deviation range. When the association verification deviation value exceeds the preset association judgment threshold, the association judgment identifier indicates that the association consistency measurement of multi-source diagnosis and treatment data under the same diagnosis and treatment digital twin trajectory constraint exceeds the allowable deviation range. When the association determination identifier indicates that the association is not established, the peak position is determined based on the value distribution of the association consistency measure within the assimilation window, and mapped to the backtracking position of the reverse association verification time path in the asymmetric time topology. Based on the time index of the peak position in the digital twin trajectory of diagnosis and treatment, the peak position is mapped to the diagnosis and treatment stage variable in the state layer of the diagnosis and treatment process to generate an anomaly occurrence stage identifier; At the time position corresponding to the peak position, anomaly type discrimination is performed based on the degree of deviation of identity consistency variables, field mapping consistency variables and semantic consistency variables, and anomaly type identifiers are generated; When the association determination identifier indicates that the association is not established, the association anomaly location information is generated based on the anomaly occurrence stage identifier and the anomaly type identifier.

[0015] The beneficial effects of this invention are: First, this invention constructs a two-layer twin state set consisting of a treatment process state layer and an association consistency state layer, and introduces a four-dimensional variational assimilation mechanism under asymmetric temporal topological constraints to achieve joint verification of multi-source treatment data in both the time and state dimensions. Compared to existing methods based on single-rule verification or static consistency comparison, this invention can characterize the continuous evolution of treatment states and the changes in association consistency between multi-source data within the entire treatment process, thereby improving the systematicness and reliability of treatment data association verification as a whole and avoiding misjudgments or omissions caused by local verification.

[0016] Secondly, this invention constructs an asymmetric time constraint model for diagnosis and treatment verification tasks by setting process consistency potential functions and association consistency potential functions, and combining the forward diagnosis and treatment evolution time path with the reverse association verification time path. This allows four-dimensional variational assimilation to move beyond simply applying a general optimization framework, and instead deeply couple it with the characteristics of the diagnosis and treatment process and the association verification objective. This not only effectively suppresses discontinuities in the state of the diagnosis and treatment process during time evolution, but also quantitatively constrains unreasonable associations in multi-source diagnosis and treatment data at the levels of identity mapping, field consistency, and semantic consistency, thereby significantly improving the stability and interpretability of association verification results in complex diagnosis and treatment scenarios.

[0017] Furthermore, based on the assimilation solution, this invention introduces a correlation consistency measurement, peak position determination, and anomaly type discrimination mechanism, expanding the correlation verification result from a simple pass / fail judgment to a locationable and traceable anomaly analysis result. When the correlation is not established, it can accurately locate the diagnosis and treatment stage where the anomaly occurs and distinguish different types such as identity mapping anomalies, time alignment anomalies, semantic consistency anomalies, and process sequence anomalies. This provides a clear basis for subsequent diagnosis and treatment data governance, quality control, and problem tracing, significantly enhancing the practical value and engineering implementation capability of the diagnosis and treatment data correlation verification results. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of a digital twin-based diagnostic and treatment data association verification method proposed in this invention; Figure 2 This is a schematic diagram of the structure of the double-layer twin state set and the asymmetric time topology in this invention; Figure 3 This is a schematic diagram illustrating the principle of association verification and anomaly localization based on four-dimensional variational assimilation in this invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0020] refer to Figures 1-3 A method for verifying the association of medical data based on digital twins includes the following steps: Collect multi-source diagnosis and treatment data of the target diagnosis and treatment process, perform time alignment, field standardization and identity extraction processing on the multi-source diagnosis and treatment data, and generate a diagnosis and treatment input structure, which includes a diagnosis and treatment event sequence, field mapping information and identity information; Based on four-dimensional variational assimilation, a diagnostic and treatment data association verification framework is constructed. The diagnostic and treatment data association verification framework includes a diagnostic and treatment state modeling unit, an association state modeling unit, a time topology construction unit, and a potential function construction unit. The diagnostic and treatment input structure is input into the diagnostic and treatment state modeling unit and the associated state modeling unit. In the diagnostic and treatment state modeling unit, a diagnostic and treatment process state layer is constructed based on the diagnostic and treatment event sequence. In the associated state modeling unit, an associated consistency state layer is constructed based on field mapping information and identity identification information, forming a two-layer twin state set. The sequence of diagnostic and treatment events and the set of two-layer twin states are input into the temporal topology construction unit. Based on the occurrence order and event dependencies of the diagnostic and treatment event sequence, an asymmetric temporal topology is constructed. The asymmetric temporal topology includes an assimilation window, a forward diagnostic and treatment evolution time path, and a reverse correlation verification time path. The forward diagnostic and treatment evolution time path is used to constrain the temporal evolution of the state layer of the diagnostic and treatment process, and the reverse correlation verification time path is used to constrain the backtracking consistency of the correlation consistency state layer. By constructing units of the two-layer twin state set and the asymmetric time topology input potential function, a four-dimensional variational assimilation objective function is constructed. The four-dimensional variational assimilation objective function includes a process consistency potential function for constraining the consistency of the state layer evolution in the diagnosis and treatment process and a correlation consistency potential function for constraining the rationality of the correlation of the state layer. Based on the diagnostic input structure and the four-dimensional variational assimilation objective function, the four-dimensional variational assimilation solution is performed within the assimilation window. During the four-dimensional variational assimilation solution process, the correlation consistency metric is calculated based on the correlation consistency potential function, and the diagnostic digital twin trajectory is constructed based on the solution results. After the assimilation calculation is completed in the diagnosis and treatment data association verification framework, an association determination identifier is generated based on the diagnosis and treatment digital twin trajectory and the association consistency measure. The association determination identifier is used to characterize the association status of multi-source diagnosis and treatment data under the same diagnosis and treatment digital twin trajectory. When the association determination identifier indicates that the association is established, it means that the multi-source diagnosis and treatment data has passed the association verification. When the association determination identifier indicates that the association is not established, association anomaly location information is generated based on the reverse association verification time path and the peak position of the association consistency measure. The association anomaly location information includes anomaly occurrence stage identifier and anomaly type identifier. The anomaly type identifier includes identity mapping anomaly type identifier, time alignment anomaly type identifier, semantic consistency anomaly type identifier, and process sequence anomaly type identifier.

[0021] In this embodiment, generating the diagnostic input structure includes: Multi-source diagnostic and treatment data corresponding to the target diagnostic and treatment process are collected from diagnostic and treatment information systems, examination and testing systems, and diagnostic and treatment equipment records. The multi-source diagnostic and treatment data are time aligned according to a unified time benchmark. The time alignment process uses the time of occurrence of the diagnostic and treatment behavior as the alignment reference, mapping data records from different sources but pointing to the same diagnostic and treatment behavior to the same time axis position. The multi-source diagnostic and treatment data includes diagnostic and treatment record data, examination and testing data, and equipment operation data generated during the diagnostic and treatment process. Among them, the diagnostic and treatment record data includes diagnostic and treatment behavior identifier, diagnostic and treatment stage identifier, and diagnostic and treatment behavior occurrence time information; the examination and testing data includes examination and testing item identifier, examination and testing result data, and examination and testing time information; and the equipment operation data includes diagnostic and treatment equipment identifier, equipment acquisition result data, and equipment data acquisition time information. All types of data contain structured fields that can be used for field standardization processing and identity extraction. Based on the multi-source medical data that has undergone time alignment processing, field standardization processing is performed on the fields in each data record. Field standardization processing includes field name unification, field value range constraint and field semantic consistency processing, generating field mapping information. The field mapping information is used to describe the correspondence and semantic mapping relationship between fields in different data sources. Based on the multi-source medical data after field standardization, identity identification extraction processing is performed on the identity information related to the medical object in each data record. The identity identification extraction processing generates identity information by extracting and binding patient identifier, medical object identifier, and medical behavior identifier. The identity identification information is used to maintain the consistency of the medical object identifier in the multi-source medical data. Using a single medical treatment behavior as the smallest event granularity, the multi-source medical treatment data after completion time alignment processing, field standardization processing, and identity identification extraction processing are processed into events. Each medical treatment behavior is constructed into a medical treatment event sequence according to the time sequence, and each medical treatment event in the medical treatment event sequence is bound to the corresponding field mapping information and identity identification information. The diagnosis and treatment input structure is constructed based on the diagnosis and treatment event sequence, field mapping information and identity information. The diagnosis and treatment input structure serves as the common input for the state layer and the associated consistency state layer in the subsequent diagnosis and treatment process, and is used to support the construction of the two-layer twin state set and the subsequent four-dimensional variational assimilation solution process.

[0022] In this embodiment, constructing a diagnostic and treatment data association verification framework includes: After the diagnostic input structure is determined, a diagnostic data association verification framework is established to meet the modeling and solution requirements of four-dimensional variational assimilation in the diagnostic data association verification task. The diagnostic data association verification framework is used to coordinate the collaborative relationship between the diagnostic state modeling unit, the associated state modeling unit, the time topology construction unit, and the potential function construction unit within the same computational closed loop, so as to ensure that the two-layer twin state set, the asymmetric time topology, and the four-dimensional variational assimilation objective function are jointly constructed and used within a unified assimilation window. In the framework for verifying the correlation of diagnosis and treatment data, the functional units do not operate independently, but form a restricted information closed-loop structure around the goal of verifying the correlation of diagnosis and treatment data. The diagnosis and treatment state modeling unit and the correlation state modeling unit only generate a set of state variables that serve the consistency verification. The time topology construction unit is only used to limit the time organization method related to verification. The potential function construction unit only introduces the constraint terms that are directly related to the consistency of diagnosis and treatment and the rationality of correlation. Within the framework for verifying the association of diagnostic and treatment data, a diagnostic and treatment status modeling unit is configured. This unit is used to define the state space of the diagnostic and treatment process. The state space of the diagnostic and treatment process takes the sequence of diagnostic and treatment events as input and constrains the range of state representation and the dimension of state update of the state layer of the diagnostic and treatment process according to the stage attributes and time continuity attributes of the diagnostic and treatment events in the diagnostic and treatment process. Within the framework for verifying the association of medical data, an association state modeling unit is configured. This unit is used to define the association consistency state space. The association consistency state space takes field mapping information and identity information as input and constrains the range of state representation of multi-source medical data at the levels of identity consistency, field semantic consistency, and process association consistency, so as to support the construction of the association consistency state layer. Within the framework for verifying the association of diagnostic and treatment data, a time topology construction unit is configured. This unit is used to define the time organization method in the process of verifying the association of diagnostic and treatment data. The time organization method is based on the sequence of diagnostic and treatment events and constrains the construction rules of the assimilation window time span, the forward diagnostic and treatment evolution time path, and the reverse association verification time path to form an asymmetric time topology for four-dimensional variational assimilation modeling. Within the framework for verifying the correlation of diagnostic and treatment data, a potential function construction unit is configured. This unit is used to limit the source of constraints for the four-dimensional variational assimilation objective function. The source of constraints consists of process consistency constraints corresponding to the state layer of the diagnostic and treatment process and association consistency constraints corresponding to the state layer of the correlation consistency. This is used to support the modeling and subsequent solution of the four-dimensional variational assimilation objective function in the diagnostic and treatment data correlation verification task.

[0023] In this embodiment, forming a two-layer twin state set includes: Within the framework for verifying the correlation of diagnostic and treatment data, the diagnostic and treatment input structure is input into the diagnostic and treatment state modeling unit. The diagnostic and treatment state modeling unit uses the sequence of diagnostic and treatment events as the basis for state modeling, identifies the stage attributes of diagnostic and treatment events in the diagnostic and treatment process, constructs diagnostic and treatment stage variables to characterize the changes in the stages of the diagnostic and treatment process, and incorporates the diagnostic and treatment stage variables as a component of the diagnostic and treatment process state layer. In the diagnosis and treatment state modeling unit, based on the occurrence order and completion status of each diagnosis and treatment event in the diagnosis and treatment event sequence, the degree of progress of the diagnosis and treatment event in the diagnosis and treatment process is characterized, an event progress variable is constructed to characterize the execution progress of the diagnosis and treatment event, and the event progress variable is included in the state representation scope of the diagnosis and treatment process state layer; In the diagnosis and treatment status modeling unit, the temporal continuity of the diagnosis and treatment process is characterized based on the time interval relationship between adjacent diagnosis and treatment events in the sequence of diagnosis and treatment events, and a temporal continuity variable is constructed to characterize the temporal continuity of the diagnosis and treatment process. Based on time continuity variables, diagnosis and treatment stage variables, and event progress variables, the diagnosis and treatment process state layer is composed of these variables. After the state layer of the diagnosis and treatment process is constructed, the changes of each state variable at adjacent time positions are described by difference for the variables of diagnosis and treatment stage, event progress, and time continuity. The difference results are used as the input basis for constraining the time evolution continuity of the state layer of the diagnosis and treatment process in the process consistency potential function. Within the framework of diagnosis and treatment data association verification, the diagnosis and treatment input structure is input into the association state modeling unit. The association state modeling unit uses field mapping information and identity identification information as the basis for state modeling, and describes the matching situation of different data records in multi-source diagnosis and treatment data pointing to the same diagnosis and treatment object. It constructs an identity consistency variable to represent the consistency of the diagnosis and treatment object, and uses the identity consistency variable as a component of the association consistency state layer. In the associated state modeling unit, based on the field correspondence defined in the field mapping information, the degree of matching of multi-source diagnosis and treatment data at the field value level is characterized, a field mapping consistency variable is constructed to represent the consistency of field matching, and the field mapping consistency variable is included in the state representation scope of the associated consistency state layer. In the associated state modeling unit, based on the semantic correspondence of fields constrained by field mapping information, the matching of multi-source diagnosis and treatment data at the semantic level is characterized, and a semantic consistency variable is constructed to represent semantic consistency. The association consistency state layer is composed of semantic consistency variables, identity consistency variables, and field mapping consistency variables. In the association consistency state layer, the value specifications of identity consistency variables, field mapping consistency variables and semantic consistency variables are defined respectively, and the deviation of each variable is described by a unified scale at the same time position. The deviation results are used as the basic input for forming the penalty term in the association consistency potential function. A joint state organization is performed on the diagnosis and treatment process state layer and the association consistency state layer. The diagnosis and treatment stage variables, event progress variables and time continuity variables in the diagnosis and treatment process state layer are uniformly arranged with the identity consistency variables, field mapping consistency variables and semantic consistency variables in the association consistency state layer to form a two-layer twin state set for four-dimensional variational assimilation modeling and solution.

[0024] In this embodiment, constructing the asymmetric time topology includes: Within the framework for verifying the association of diagnostic and treatment data, the sequence of diagnostic and treatment events and the set of two-layer twin states are input into the temporal topology construction unit. The temporal topology construction unit uses the order of events in the sequence of diagnostic and treatment events and the dependencies between events as the basis for time organization, limits the time range involved in the process of verifying the association of diagnostic and treatment data, and generates an assimilation window for subsequent assimilation calculation. Within the assimilation window, a forward treatment evolution time path is constructed according to the occurrence order and dependency relationship of treatment events in the treatment event sequence. The forward treatment evolution time path is organized according to the natural time progression direction of treatment events, which is used to limit the time evolution order and state update direction of the treatment process state layer in the treatment process. Within the assimilation window, based on the dependencies of diagnosis and treatment events in the sequence of diagnosis and treatment events, the process traces back from the later stages of the diagnosis and treatment process to the earlier stages, constructing a reverse correlation verification time path. The reverse correlation verification time path is used to trace back the changes in the correlation consistency state layer in the time dimension to support the backtracking consistency constraints of the correlation consistency state layer. In the time topology construction unit, the time organization relationship between the forward diagnosis and treatment evolution time path and the reverse correlation verification time path is distinguished, and a time structure with asymmetrical time order is generated. The forward diagnosis and treatment evolution time path is expanded unidirectionally according to the occurrence order of diagnosis and treatment events, while the reverse correlation verification time path is traced back in reverse according to the dependency relationship of diagnosis and treatment events. In the time topology construction unit, the coverage of the forward diagnosis and treatment evolution time path and the reverse correlation verification time path in terms of time span is distinguished, and a time structure with asymmetric path length is generated. The forward diagnosis and treatment evolution time path covers the time span of the complete diagnosis and treatment process, while the reverse correlation verification time path covers the time sub-intervals related to the correlation consistency verification. In the time topology construction unit, the state objects affected by the forward diagnosis and treatment evolution time path and the reverse association verification time path are distinguished, and an asymmetric time structure of the constraint objects is generated. The forward diagnosis and treatment evolution time path is used to constrain the time evolution consistency of the state layer of the diagnosis and treatment process, and the reverse association verification time path is used to constrain the backtracking consistency of the association consistency state layer. After completing the time organization processing for asymmetric time order, asymmetric path length, and asymmetric constraint objects, the assimilation window, the forward diagnosis and treatment evolution time path, and the reverse association verification time path are summarized to generate a complete asymmetric time topology. The asymmetric time topology serves as the unified organization result of the time dimension in the diagnosis and treatment data association verification framework. It is used to simultaneously constrain the forward time evolution of the diagnosis and treatment process state layer and the reverse backtracking consistency of the association consistency state layer in the subsequent four-dimensional variational assimilation solution process.

[0025] In this embodiment, constructing the four-dimensional variational assimilation objective function includes: Within the framework of diagnosis and treatment data association verification, a two-layer twin state set and an asymmetric temporal topology are received. Based on the assimilation window limited by the asymmetric temporal topology, the participation range of the two-layer twin state set in the time dimension is constrained to determine the set of state variables participating in the four-dimensional variational assimilation modeling. Based on the diagnosis and treatment stage variables, event progress variables, and time continuity variables in the state layer of the diagnosis and treatment process, a process consistency potential function is constructed. The process consistency potential function quantifies the state deviation between the state variables of the diagnosis and treatment process at adjacent time positions, forming a penalty term for the time evolution continuity of the state layer of the diagnosis and treatment process, which is used to constrain the consistency of state changes of the state layer of the diagnosis and treatment process on the positive diagnosis and treatment evolution time path. Based on the identity consistency variables, field mapping consistency variables, and semantic consistency variables in the association consistency state layer, an association consistency potential function is constructed. The association consistency potential function quantifies the deviation of the association consistency state variables on the time position of the forward diagnosis and treatment evolution time path within the assimilation window and the backtracking position of the reverse association verification time path, forming a penalty term for association inconsistency, which is used to constrain the backtracking consistency of the association consistency state layer on the reverse association verification time path. The process consistency potential function and the association consistency potential function are combined according to the time path interaction relationship in the asymmetric time topology to construct a four-dimensional variational assimilation objective function. The four-dimensional variational assimilation objective function is used to perform joint constraint modeling of the diagnosis and treatment process state layer and the association consistency state layer within the assimilation window, providing a unified optimization objective for subsequent four-dimensional variational assimilation solution.

[0026] In this embodiment, calculating the correlation consistency metric and constructing the diagnosis and treatment digital twin trajectory includes: Within the framework for verifying the association of diagnostic and treatment data, the system receives the diagnostic and treatment input structure and the four-dimensional variational assimilation objective function. Based on the assimilation window constrained by asymmetric time topology, it maps the diagnostic and treatment event sequence, field mapping information and identity information in the diagnostic and treatment input structure to the time position corresponding to the two-layer twin state set, and generates the initial state configuration for solving the four-dimensional variational assimilation problem. Within the assimilation window, according to the time organization relationship of the forward diagnosis and treatment evolution time path and the reverse correlation verification time path in the asymmetric time topology, the two-layer twin state set is solved by four-dimensional variational assimilation. By iteratively optimizing the four-dimensional variational assimilation objective function, the state values ​​of the diagnosis and treatment process state layer and the correlation consistency state layer within the assimilation window are jointly updated. In the process of solving the four-dimensional variational assimilation problem, when the change of the four-dimensional variational assimilation objective function in continuous iterations is lower than the preset convergence threshold, or when the state update amplitude within the assimilation window tends to be stable overall, the assimilation iteration process is terminated, and stable diagnosis and treatment process state layer and association consistency state layer state results are output to ensure the executability and result certainty of the assimilation calculation in the scenario of diagnosis and treatment data association verification. In the four-dimensional variational assimilation process, based on the correlation consistency potential function, the deviation of the correlation consistency state variable between the time position on the forward diagnosis and treatment evolution time path and the backtracking position on the reverse correlation verification time path within the assimilation window is quantified, and a correlation consistency metric that changes with the time position is generated. The correlation consistency metric is represented in the form of a time series. After completing the four-dimensional variational assimilation solution, a digital twin trajectory for diagnosis and treatment is constructed based on the state evolution results of the updated diagnosis and treatment process state layer within the assimilation window. The digital twin trajectory for diagnosis and treatment is used to characterize the continuous state evolution of the target diagnosis and treatment process in the time dimension and maintains a one-to-one correspondence with the correlation consistency metric in terms of time position.

[0027] In this embodiment, generating associated anomaly location information includes: After completing the four-dimensional variational assimilation solution in the diagnostic and treatment data association verification framework, based on the diagnostic and treatment digital twin trajectory and the association consistency measure, the maximum deviation value within the assimilation window is extracted from the association consistency measure to generate the association verification deviation value. A preset association judgment threshold is obtained. An association judgment identifier is generated based on the comparison result between the association verification deviation value and the preset association judgment threshold. When the association verification deviation value does not exceed the preset association judgment threshold, the association judgment identifier indicates that the association consistency measurement of multi-source diagnosis and treatment data under the same diagnosis and treatment digital twin trajectory constraint is within the allowable deviation range. When the association verification deviation value exceeds the preset association judgment threshold, the association judgment identifier indicates that the association consistency measurement of multi-source diagnosis and treatment data under the same diagnosis and treatment digital twin trajectory constraint exceeds the allowable deviation range, and there is an unacceptable association inconsistency. The preset association judgment threshold is a judgment boundary value pre-set for the association consistency measurement, used to determine whether the association verification deviation value exceeds the allowable deviation range under the diagnosis and treatment digital twin trajectory constraint. When the association determination identifier indicates that the association is not established, the peak position is determined based on the value distribution of the association consistency measure within the assimilation window. The peak position is defined as the time position when the association consistency measure reaches the association verification deviation value, and the peak position is mapped to the backtracking position of the reverse association verification time path in the asymmetric time topology. Based on the time index of the peak position in the digital twin trajectory of diagnosis and treatment, the peak position is mapped to the diagnosis and treatment stage variable in the state layer of the diagnosis and treatment process to generate an anomaly occurrence stage identifier; At the time position corresponding to the peak position, anomaly type discrimination is performed based on the degree of deviation of identity consistency variables, field mapping consistency variables, and semantic consistency variables, and anomaly type identifiers are generated. Anomaly type identifiers include identity mapping anomaly type identifiers, time alignment anomaly type identifiers, semantic consistency anomaly type identifiers, and process sequence anomaly type identifiers. Among them, the identity mapping anomaly type identifier is generated when the deviation of the identity consistency variable reaches its maximum, the time alignment anomaly type identifier is generated when the deviation of the field mapping consistency variable reaches its maximum, the semantic consistency anomaly type identifier is generated when the deviation of the semantic consistency variable reaches its maximum, and the process sequence anomaly type identifier is generated when the time sequence of the diagnosis and treatment process state layer on the forward diagnosis and treatment evolution time path is inconsistent with the backtracking result of the reverse association verification time path. When the association determination identifier indicates that the association is not established, the association anomaly location information is generated based on the anomaly occurrence stage identifier and the anomaly type identifier.

[0028] Example 1: To verify the feasibility of this invention in practice, it was applied to a medical data management and verification application scenario. This scenario stems from a multi-system collaborative medical environment developed by a medical institution over a long period of operation. Medical data is dispersed across medical information systems, examination and testing systems, and various medical equipment recording systems. Due to different system construction phases, inconsistent data standards, and frequent cross-system transfers during the medical process, inconsistencies commonly arise in the time sequence, identity mapping, field semantics, and process relationships of multi-source data corresponding to the same medical process. These inconsistencies are difficult to accurately locate manually during post-audit, quality control, and medical dispute verification. Existing technologies often rely on rule comparison or single timestamp verification, which struggles to simultaneously depict the overall relationship between the evolution of the medical process and cross-source consistency, easily leading to missed or incorrect judgments.

[0029] In this application scenario, multi-source diagnostic and treatment data involved in a complete treatment process are used as input objects. This data includes medical records, examination and testing data, and operational data of diagnostic and treatment equipment. Using the method of this invention, time alignment processing under a unified time benchmark is performed on the aforementioned multi-source diagnostic and treatment data, and field normalization mapping processing is carried out. Simultaneously, identity identification information related to the treatment object is extracted and bound. Based on this, a diagnostic and treatment input structure is constructed, consisting of a sequence of diagnostic and treatment events, field mapping information, and identity identification information. The diagnostic and treatment input structure fully covers the core information of the diagnostic and treatment behavior in the three dimensions of time, semantics, and identity, providing a unified data foundation for subsequent modeling.

[0030] In practical applications, based on the diagnostic input structure, a diagnostic process state layer and a correlation consistency state layer are simultaneously constructed within the diagnostic data association verification framework. The diagnostic process state layer characterizes the evolution of the diagnostic process through diagnostic stage variables, event progress variables, and time continuity variables. The correlation consistency state layer characterizes the rationality of the association between multi-source data through identity consistency variables, field mapping consistency variables, and semantic consistency variables. The two-layer twin state set is jointly organized within the same modeling space, avoiding the information loss problem caused by separating diagnostic process modeling from data association verification.

[0031] Building upon this foundation, an asymmetric temporal topology is constructed by combining the sequence of diagnostic and treatment events. A forward-evolutionary time path characterizes the natural progression of the diagnostic and treatment process, while a reverse-correlation verification time path imposes backtracking constraints on correlation consistency. This ensures that the time organization method simultaneously serves both process consistency and correlation verification objectives. Subsequently, a four-dimensional variational assimilation objective function is constructed based on a two-layer twin state set and the asymmetric temporal topology. Joint optimization is performed within the assimilation window to obtain continuous digital twin trajectories of diagnostic and treatment events, and a correlation consistency metric that varies over time is generated simultaneously. By assessing the overall deviation of the correlation consistency metric, a correlation determination identifier is output. Furthermore, in cases where correlation is invalid, the stage of anomaly occurrence is located and the anomaly type is identified, thereby achieving traceable and locatable data correlation verification.

[0032] To verify the effectiveness of the method of this invention in practical applications, two common technical approaches in real-world systems were selected as control methods. One is a rule-based multi-source medical data consistency verification method, and the other is a medical data verification method based on single time alignment. Under the same dataset conditions, the three methods were compared and evaluated in terms of correlation verification accuracy, anomaly localization capability, and overall verification stability. The statistical results are shown in the table below.

[0033] Table 1. Comparison of Experimental Results of Different Methods for Verifying the Association of Diagnostic and Treatment Data

[0034] As shown in Table 1, rule-based verification methods exhibit low accuracy in both correlation verification and anomaly localization when faced with complex diagnostic and treatment processes and numerous cross-system relationships. In particular, they are prone to misjudgments regarding process sequence and semantic consistency. While time-aligned verification methods improve time-related issues to some extent, their limited improvement in anomaly localization accuracy and high misjudgment rate stem from a lack of joint modeling of the overall evolution of the diagnostic and treatment process and the state of correlation consistency.

[0035] In comparison, the method of this invention achieves improvements in both correlation verification accuracy and anomaly localization accuracy, with the correlation verification accuracy increasing to 92.3% and the anomaly localization accuracy increasing to 83.8%, while significantly reducing the false positive rate. This improvement does not rely on a single rule or simple threshold judgment, but rather stems from the joint modeling mechanism of a two-layer twin state set and asymmetric temporal topology, ensuring that the evolution of the diagnostic process state and the consistency of multi-source data correlation are constrained holistically within the same assimilation window. Although the method of this invention introduces four-dimensional variational assimilation solution in the computation process, slightly increasing the average verification latency, it does not significantly affect the system's real-time performance.

[0036] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for verifying the association of medical data based on digital twins, characterized in that, Includes the following steps: Collect multi-source diagnosis and treatment data for the target diagnosis and treatment process, perform time alignment, field standardization and identity extraction processing, and generate a diagnosis and treatment input structure, including diagnosis and treatment event sequence, field mapping information and identity information; Based on four-dimensional variational assimilation, a framework for verifying the correlation of diagnostic and treatment data is constructed, including a diagnostic and treatment state modeling unit, a correlation state modeling unit, a time topology construction unit, and a potential function construction unit. The diagnostic and treatment input structure is input into the diagnostic and treatment state modeling unit and the associated state modeling unit to construct the diagnostic and treatment process state layer and the associated consistency state layer, forming a two-layer twin state set; The sequence of diagnostic and treatment events and the set of two-layer twin states are input into the temporal topology construction unit to construct an asymmetric temporal topology; By constructing units of two-layer twin state sets and asymmetric time topology input potential functions, a four-dimensional variational assimilation objective function is constructed, including process consistency potential function and correlation consistency potential function; Based on the diagnostic input structure and the four-dimensional variational assimilation objective function, the four-dimensional variational assimilation solution is performed, the correlation consistency metric is calculated, and a diagnostic digital twin trajectory is constructed. Based on the diagnosis and treatment digital twin trajectory and the correlation consistency measurement, a correlation determination identifier is generated. When the correlation determination identifier indicates that the correlation is not established, correlation anomaly location information is generated.

2. The method for verifying the association of medical data based on digital twins according to claim 1, characterized in that, The generated diagnostic input structure includes: Collect multi-source diagnostic and treatment data corresponding to the target diagnostic and treatment process from diagnostic and treatment information systems, examination and testing systems and diagnostic and treatment equipment records, perform time alignment processing, and map data records from different sources but pointing to the same diagnostic and treatment behavior to the same time axis position; Based on the multi-source medical data that has undergone time alignment processing, field normalization processing is performed on the fields in each data record to generate field mapping information; Based on the multi-source medical data after field standardization, identity information related to the medical subject in each data record is extracted and generated; Using a single medical treatment action as the smallest event granularity, the multi-source medical treatment data after completion time alignment, field standardization, and identity extraction are processed into event-based data, and each medical treatment action is constructed into a sequence of medical treatment events according to the time sequence. The diagnostic input structure is constructed based on the sequence of diagnostic events, field mapping information, and identity information.

3. The method for verifying the association of diagnostic and treatment data based on digital twins according to claim 1, characterized in that, The framework for constructing a correlation verification framework for diagnostic and treatment data includes: After the diagnostic input structure is determined, a diagnostic data association verification framework is established based on the modeling and solution requirements of four-dimensional variational assimilation in the diagnostic data association verification task. Within the framework for verifying and linking diagnostic and treatment data, a diagnostic and treatment status modeling unit is configured to define the state space of the diagnostic and treatment process. Within the framework for verifying the association of diagnostic and treatment data, configure an association state modeling unit to define the association consistency state space; Within the framework for verifying the association of medical data, a time topology construction unit is configured to limit the time organization method in the process of verifying the association of medical data. Within the framework for verifying the correlation of diagnostic and treatment data, potential function construction units are configured to limit the source of constraints for the four-dimensional variational assimilation objective function.

4. The method for verifying the association of medical data based on digital twins according to claim 1, characterized in that, The formation of the two-layer twin state set includes: The diagnosis and treatment status modeling unit uses the sequence of diagnosis and treatment events as the basis for status modeling, identifies the stage attributes of diagnosis and treatment events in the diagnosis and treatment process, and constructs diagnosis and treatment stage variables; In the diagnosis and treatment status modeling unit, based on the occurrence order and completion status of each diagnosis and treatment event in the diagnosis and treatment event sequence, the degree of progress of the diagnosis and treatment events in the diagnosis and treatment process is characterized, and event progress variables are constructed. In the diagnosis and treatment status modeling unit, the temporal continuity of the diagnosis and treatment process is characterized based on the time interval relationship between adjacent diagnosis and treatment events in the sequence of diagnosis and treatment events, and a temporal continuity variable is constructed. Based on time continuity variables, diagnosis and treatment stage variables, and event progress variables, the diagnosis and treatment process state layer is composed of these variables. The associated state modeling unit uses field mapping information and identity identification information as the basis for state modeling, and describes the matching situation of different data records pointing to the same diagnosis and treatment object, and constructs identity consistency variables; In the associated state modeling unit, based on the field correspondence defined in the field mapping information, the degree of matching of multi-source diagnosis and treatment data at the field value level is characterized, and a field mapping consistency variable is constructed. In the associated state modeling unit, based on the semantic correspondence of fields constrained by field mapping information, the semantic matching of multi-source diagnosis and treatment data is characterized, and semantic consistency variables are constructed. The association consistency state layer is composed of semantic consistency variables, identity consistency variables, and field mapping consistency variables. A joint state organization is performed on the state layer of the diagnosis and treatment process and the state layer of correlation and consistency, and a unified arrangement is formed to form a two-layer twin state set.

5. The method for verifying the association of diagnostic and treatment data based on digital twins according to claim 1, characterized in that, The construction of the asymmetric time topology includes: The sequence of diagnostic and treatment events and the set of two-layer twin states are input into the time topology construction unit. The order of events and the dependencies between events in the sequence of diagnostic and treatment events are used as the basis for time organization. The time range involved in the process of verifying the correlation of diagnostic and treatment data is limited, and an assimilation window is generated. Within the assimilation window, a forward treatment evolution time path is constructed according to the occurrence order and dependencies of treatment events in the treatment event sequence; Within the assimilation window, based on the dependencies of diagnosis and treatment events in the sequence of diagnosis and treatment events, the reverse correlation verification time path is constructed by tracing back from the later stages of the diagnosis and treatment process to the earlier stages. In the time topology construction unit, the time organization relationship between the forward diagnosis and treatment evolution time path and the reverse correlation verification time path is distinguished, and a time structure with asymmetrical time order is generated. The coverage of the forward diagnosis and treatment evolution time path and the reverse correlation verification time path in terms of time span is distinguished, and a time structure with asymmetric path length is generated. Distinguish the state objects affected by the forward diagnosis and treatment evolution time path and the reverse correlation verification time path, and generate an asymmetric time structure of the constraint objects. After completing the time organization processing for asymmetric time order, asymmetric path length, and asymmetric constraint objects, the assimilation window, the forward diagnosis and treatment evolution time path, and the reverse correlation verification time path are summarized to generate a complete asymmetric time topology.

6. The method for verifying the association of medical data based on digital twins according to claim 1, characterized in that, The construction of the four-dimensional variational assimilation objective function includes: Within the framework for verifying the association of diagnostic and treatment data, a two-layer twin state set and an asymmetric temporal topology are received. Based on the assimilation window, the participation range of the two-layer twin state set in the time dimension is constrained to determine the set of state variables. Based on the diagnosis and treatment stage variables, event progress variables, and time continuity variables in the state layer of the diagnosis and treatment process, a process consistency potential function is constructed. The process consistency potential function quantifies the state deviation between the state variables of the diagnosis and treatment process at adjacent time positions, forming a penalty term for the continuity of the time evolution of the state layer of the diagnosis and treatment process. Based on the identity consistency variables, field mapping consistency variables, and semantic consistency variables in the association consistency state layer, an association consistency potential function is constructed. The association consistency potential function quantifies the deviation of the association consistency state variables on the time position of the forward diagnosis and treatment evolution time path within the assimilation window and the backtracking position of the reverse association verification time path, forming a penalty term for association inconsistency. By combining the process consistency potential function and the correlation consistency potential function according to the time path interaction relationship in the asymmetric time topology, a four-dimensional variational assimilation objective function is constructed.

7. The method for verifying the association of medical data based on digital twins according to claim 1, characterized in that, The calculation of the correlation consistency metric and the construction of the diagnosis and treatment digital twin trajectory include: Within the framework for verifying and linking diagnostic and treatment data, the system receives the diagnostic and treatment input structure and the four-dimensional variational assimilation objective function. Based on the assimilation window, it maps the diagnostic and treatment event sequence, field mapping information, and identity information in the diagnostic and treatment input structure to the time position corresponding to the two-layer twin state set, and generates the initial state configuration. Within the assimilation window, according to the time organization relationship of the forward diagnosis and treatment evolution time path and the reverse correlation verification time path in the asymmetric time topology, the two-layer twin state set is solved by four-dimensional variational assimilation. By iteratively optimizing the four-dimensional variational assimilation objective function, the state values ​​of the diagnosis and treatment process state layer and the correlation consistency state layer within the assimilation window are jointly updated. In the process of solving the four-dimensional variational assimilation problem, based on the correlation consistency potential function, the deviation of the correlation consistency state variable between the time position on the forward diagnosis and treatment evolution time path and the backtracking position on the reverse correlation verification time path within the assimilation window is quantified, and a correlation consistency metric that changes with time position is generated. After completing the four-dimensional variational assimilation solution, a digital twin trajectory for diagnosis and treatment is constructed based on the state evolution results of the updated diagnosis and treatment process state layer within the assimilation window.

8. The method for verifying the association of medical data based on digital twins according to claim 1, characterized in that, The generated associated anomaly location information includes: Based on the diagnosis and treatment digital twin trajectory and the correlation consistency measure, the maximum deviation value within the assimilation window is extracted from the correlation consistency measure to generate the correlation verification deviation value; Obtain a preset association judgment threshold, and generate an association judgment identifier based on the comparison result between the association verification deviation value and the preset association judgment threshold. When the association verification deviation value does not exceed the preset association judgment threshold, the association judgment identifier indicates that the association consistency measurement of multi-source diagnosis and treatment data under the same diagnosis and treatment digital twin trajectory constraint is within the allowable deviation range. When the association verification deviation value exceeds the preset association judgment threshold, the association judgment identifier indicates that the association consistency measurement of multi-source diagnosis and treatment data under the same diagnosis and treatment digital twin trajectory constraint exceeds the allowable deviation range. When the association determination identifier indicates that the association is not established, the peak position is determined based on the value distribution of the association consistency measure within the assimilation window, and mapped to the backtracking position of the reverse association verification time path in the asymmetric time topology. Based on the time index of the peak position in the digital twin trajectory of diagnosis and treatment, the peak position is mapped to the diagnosis and treatment stage variable in the state layer of the diagnosis and treatment process to generate an anomaly occurrence stage identifier; At the time position corresponding to the peak position, anomaly type discrimination is performed based on the degree of deviation of identity consistency variables, field mapping consistency variables and semantic consistency variables, and anomaly type identifiers are generated; When the association determination identifier indicates that the association is not established, the association anomaly location information is generated based on the anomaly occurrence stage identifier and the anomaly type identifier.

Images