Clinical test file-oriented integrated information system and data processing method thereof

Abstract

The invention relates to the technical field of computers, and discloses an integrated information system for clinical test archives and a data processing method thereof, and the method comprises the steps: constructing a standardized data model covering a whole process; multi-source data dynamic structured intake and integrity verification are carried out; performing multi-dimensional semantic association analysis based on a Bayesian network and a correlation coefficient; block chain type version tracing and difference comparison are carried out; performing role-sensitivity-operation-context four-dimensional access control; and task-driven collaborative sharing and auditing log records. The system comprises a unified modeling module, a data intake module, a semantic analysis module, a version tracing module, an access control module and a collaborative auditing module. The problems of data islands, semantic segmentation, poor dynamic adaptability, weak safety protection and the like in the prior art are solved, intelligent management, semantic interconnection, safety controllability and efficient collaboration of the whole life cycle of clinical test files are achieved, and data quality, audit compliance and multi-role collaboration efficiency are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a clinical trial archive-oriented integrated information system and a data processing method thereof. BACKGROUND

[0002] With the rapid development of modern medical research and the increasingly stringent regulatory requirements, clinical trials have become the core link of new drug research and development and medical technology evaluation. The clinical trial process involves a large number of heterogeneous data sources, including electronic data capture systems, laboratory information systems, medical image archiving systems, electronic medical record systems, and drug management systems. The massive archive data generated by these systems have the characteristics of diversity, complexity, and timeliness. At the same time, the regulatory compliance requirements of clinical trials require data to have complete traceability, strict access control, and accurate audit records. Therefore, it is of great significance to build a clinical trial archive management system that can effectively integrate multi-source heterogeneous data, ensure data quality and security, and support collaborative operation, in order to improve the efficiency of clinical trials, protect data integrity, and meet regulatory requirements.

[0003] In the prior art, each data source system lacks a unified data model and standardized architecture, resulting in serious data island phenomenon, difficult data integration, and problems such as inconsistent formats and semantic ambiguity. Secondly, the traditional data import process lacks a perfect data quality control mechanism, and the key links such as abnormal value identification, integrity verification, and time sequence logic verification are not handled finely enough, affecting data reliability. Thirdly, the version traceability function of existing systems is relatively simple and cannot provide fine-grained modification history records and difference comparison capabilities, making it difficult to meet the strict requirements of regulatory audits. In addition, the permission control mechanism is usually based on simple role allocation and lacks comprehensive consideration of data sensitivity, operation type, and access context, making it impossible to achieve fine-grained security control. Finally, there is a lack of effective conflict detection and merging mechanism during multi-user collaborative editing, and there is a lack of deep semantic analysis capability for archive data, making it impossible to fully exploit the association relationships and potential value of data.

[0004] Therefore, the present application provides a clinical trial archive-oriented integrated information system and a data processing method thereof, which builds a standardized data model, realizes intelligent data import and quality control, establishes a perfect version traceability mechanism, designs a multi-dimensional permission control strategy, provides collaborative editing and auditing functions, and builds a clinical trial archive knowledge network combined with semantic analysis technology, thereby fully solving the deficiencies of the prior art and providing strong support for the standardized management of clinical trials and data value mining. SUMMARY

[0005] The application aims to provide an integrated information system for clinical trial archives and a data processing method thereof to solve the problems of data islands, semantic fragmentation, poor dynamic adaptability and weak security protection in the prior art in clinical trial archive management.

[0006] To solve the above technical problems, the application provides the following technical solutions. An integrated information system for clinical trial archives comprises: A clinical modeling module is configured to build a standardized data model for the whole process of clinical trials, define data entities, attribute fields, data types, value constraints, correlation relationships and lifecycle state identifiers, form a unified metadata architecture, define a set of core data entities covering the whole process of clinical trials, including a subject basic information entity, an informed consent form entity, a screening period data entity, an enrollment data entity, a visit record entity, a laboratory test result entity, an adverse event report entity, a combined medication record entity, a researcher evaluation entity, a monitoring report entity, a data cleaning log entity, a statistical analysis plan entity and a final research report entity. For each type of data entity, a set of attribute fields, data types, value constraints and foreign key correlation relationships are defined, and a lifecycle state identifier is added to each data entity, which includes six states of draft, to-be-audited, confirmed, locked, archived and destroyed, and the legal migration path and trigger condition between the states are defined.

[0007] A data import module is configured to receive raw archive data from an electronic data collection system, a laboratory information system, a medical image archiving system, an electronic medical record system, a drug management system and a manual input terminal, perform field mapping, format conversion, missing value marking, abnormal value identification, data integrity checking and structured packaging according to the standardized data model, generate standardized archive data units and write them into a central archive storage. A semantic analysis module is configured to perform deep analysis and correlation mining of the standardized archive data units at the semantic level, build a subject longitudinal time axis, a causal correlation graph of visit events and laboratory indicators, a correlation network of adverse events and combined medications, a semantic consistency matrix of researcher evaluation and monitoring conclusions, a mapping relationship table of statistical analysis plans and data results, and form a multi-dimensional clinical trial archive knowledge network. A process tracing module is configured to establish a version evolution chain for each data unit, record the version identifier, timestamp, operator identity, data snapshots before and after modification, modification reason explanation of each modification operation, and store them in an independent version log database, support multi-dimensional version history backtracking and difference comparison. An access control module is configured to implement a four-dimensional permission control mechanism based on roles, data sensitivity, operation types, and access context, dynamically generate access control decisions through a strategy engine, and ensure that high-sensitivity data is legally accessed within an authorized range. A shared audit module is configured to provide a task-driven archive sharing workflow engine, a real-time collaborative editing conflict detection and merging mechanism, a complete operation audit log recording and multi-dimensional query function, support multi-role collaborative operation, and full-process behavior traceability.

[0008] A data processing method for clinical trial archives includes: Step S1, a standardized data model for the entire process of clinical trials is constructed, data entities, attribute fields, data types, value constraints, association relationships, and lifecycle state identifiers are defined, and a unified metadata architecture is formed; The data entities are defined, including: subject basic information entity, informed consent form entity, screening period data entity, enrollment data entity, visit record entity, laboratory test result entity, adverse event report entity, combined medication record entity, researcher evaluation entity, monitoring report entity, data cleaning log entity, statistical analysis plan entity, and final research report entity; For each type of data entity, a set of attribute fields, data types, value constraints, and foreign key association relationships are defined. A lifecycle state identifier is added to each data entity, which includes six states: draft, pending review, confirmed, locked, archived, and destroyed, and the legal migration path and trigger conditions between states are defined.

[0009] Step S2, raw archive data from electronic data collection systems, laboratory information systems, medical image archiving systems, electronic medical record systems, drug management systems, and manual input terminals is received, field mapping, format conversion, missing value marking, outlier identification, data integrity verification, and structured packaging are performed according to the standardized data model, standardized archive data units are generated and written into the central archive repository; A multi-source data access interface is established, and data transmission protocols, authentication methods, and data format templates are configured; Field mapping is performed to map the original data field names to the corresponding attribute fields in the standardized data model; Format conversion is performed to uniformly convert date formats, numerical formats, and coding formats in the original data to system internal standard formats; The missing value marking operation is performed, and a special marker "MISSING" is inserted into the field with no value and the missing reason code is recorded; Performing outlier identification operation, performing preset outlier detection rules on numerical fields, if an outlier is detected, inserting a special marker "OUTLIER" and recording the outlier type code, the expected value range, the actual value; Performing data integrity check operation, including six types of rules: field non-empty check, numerical range check, enumeration value legality check, foreign key consistency check, time sequence logic check, and signature integrity check; If any check fails, a structured error report is generated and the data unit is prevented from being written, and only when all checks pass, it is packaged as a standardized archive data unit and written into the central archive repository using a transaction mechanism.

[0010] The time sequence logic check in the data integrity check includes checking that the visit date cannot be earlier than the screening date, and the laboratory test date cannot be more than seven days later than the corresponding visit date; the signature integrity check includes checking whether the data unit that needs digital signature contains a valid and tamper-proof digital signature.

[0011] Step S3, performing deep analysis and correlation mining of the standardized archive data unit at the semantic level, constructing a longitudinal time axis of the subject, a causal correlation graph of visit events and laboratory indicators, a correlation network of adverse events and concomitant medications, a semantic consistency matrix of researcher assessments and monitoring conclusions, and a mapping relationship table of statistical analysis plans and data results, forming a multi-dimensional clinical trial archive knowledge network; Constructing a longitudinal time axis with the subject number as the unique identifier, arranging all visit records, laboratory test records, adverse event records, concomitant medication records, and researcher assessment records of the subject in chronological order; Using Bayesian network inference algorithm to construct a causal correlation graph of visit events and laboratory indicators, taking visit events as nodes, laboratory indicator changes as observation variables, and potential pathological mechanisms as hidden variables, and learning network parameters through maximum likelihood estimation; Using a dual-index fusion method of Pearson correlation coefficient and Spearman rank correlation coefficient to construct a correlation network of adverse events and concomitant medications, calculating the comprehensive correlation score between each pair of adverse events and concomitant medications, and retaining significant correlation pairs with a score higher than a preset threshold; Using a text matching algorithm based on word vector cosine similarity to construct a semantic consistency matrix of researcher assessments and monitoring conclusions, performing word segmentation, word vector average pooling, and cosine similarity calculation on the text, and marking inconsistent items with a similarity lower than a preset threshold; Constructing a mapping relationship table of statistical analysis plans and data results, recording the actual value, missing rate, distribution characteristics, and deviation degree from the planned expectation of each analysis variable in the data results; Integrate the actual value, missing rate, distribution characteristics, and deviation degree from the planned expectation of each analysis variable in the data results to form a multi-dimensional clinical trial archive knowledge network to support intelligent retrieval, trend prediction, and risk warning.

[0012] Step S4: Establish a version evolution chain for each data unit, record the version identifier, timestamp, operator identity, data snapshot before and after modification, and modification reason explanation of each modification operation, and store them in an independent version log database to support multi-dimensional version history backtracking and difference comparison. Assign a version identifier using a hierarchical coding structure to the data unit, with the first layer being the project number, the second layer being the subject number, the third layer being the data entity type code, the fourth layer being the data entity instance number, the fifth layer being the version sequence number, and the sixth layer being the timestamp hash value. Generate a version record containing the new version identifier, operation timestamp, operator identity, operation type code, modification reason explanation text, data snapshot before modification, and data snapshot after modification. Store the version record in an independent version log database using a blockchain-like chain storage structure, with each version record containing the previous version hash value, current version data hash value, and operation metadata hash value. Support version history backtracking by subject number, visit number, data type, and operation time interval, support field-level difference comparison and highlighting of any two version records, and support version rollback operation and generation of new version records.

[0013] Step S5: Implement a four-dimensional permission control mechanism based on role, data sensitivity, operation type, and access context, dynamically generate access control decisions through a strategy engine, and ensure that high-sensitive data is legally accessed within the authorized scope. Define a set of clinical trial participant roles, including sponsor project manager, research center researcher, independent data monitoring committee member, statistical analyst, auditor, and ethics committee member. Define data sensitivity levels, including public, restricted, confidential, and top secret. Public-level data includes project number, research center number, and visit schedule. Restricted-level data includes subject number, visit date, and laboratory test item name. Confidential-level data includes laboratory test specific values, adverse event description text, and combined medication name and dosage. Top secret-level data includes subject real identity information, informed consent form scan, and genetic sequencing raw data. Define operation type permissions, including read-only, edit, export, delete, and version rollback. Define access context constraints, including limited to specified IP address segment, limited to specified device fingerprint, limited to specified time period access, device fingerprint is generated by six parameters combination of device model, operating system version, browser type, screen resolution, font list, time zone setting; By policy engine dynamic combination role, data sensitivity, operation type, access context four-dimensional parameter, query preset policy rule library generates access control decision, only when all dimension constraints are met, access operation is allowed to execute.

[0014] Step S6, provide a task-driven archive sharing workflow engine, real-time collaborative editing conflict detection and merging mechanism, complete operation audit log recording and multi-dimensional query function, support multi-role collaborative operation and full-process behavior traceability.

[0015] Define sharing trigger conditions and sharing target roles, when the conditions are met, automatically generate sharing tasks and grant preset access permissions to target roles; Adopt operation conversion algorithm to solve multi-user concurrent editing conflict, decompose editing operation into atomic operation sequence and exchange sorting to ensure the consistency of the final data state; Record the performer, execution time, operation object, operation content, operation result state of each data access, modification, export, sharing, deletion operation, the operation object is recorded as the version identifier of the data unit, and the operation content is recorded as the specific modified field and value; Store the audit log in a write-once read-many storage medium and set read-only access permission, support audit query and report generation according to time range, operation type, user identity, data entity dimension.

[0016] Compared with the prior art, the beneficial effects of the present application are: The present application solves the integration difficulty problem caused by data structure heterogeneity in the prior art system by constructing a standardized data model covering the whole process of clinical trials, realizes seamless fusion and unified management of multi-source data; through dynamic structured data import mechanism and integrity checking rules, the high-quality conversion of original data to standardized archives is ensured, and data pollution and semantic distortion are avoided.

[0017] The present application realizes intelligent risk warning and decision assistance by a multi-dimensional semantic correlation analysis engine, which deeply mines the potential correlation between archive data, and surpasses the functional limitations of the prior art system which only provides static storage and simple retrieval; the present application realizes complete recording and non-tamperable of data modification history through full-process version traceability mechanism and block chain type log storage, which meets the stringent requirements of clinical trial data audit and supervision.

[0018] The application realizes accurate permission management of high-sensitive data, effectively prevents data leakage and unauthorized access risk through the four-dimensional fine-grained access control model; and the application improves multi-role collaboration efficiency and avoids data version confusion and operation conflict through the collaborative sharing workflow and real-time conflict resolution mechanism.

[0019] The overall system architecture of the application has high scalability and adaptability, can flexibly cope with different scales and different types of clinical trial project requirements, and provides a full life cycle, full element coverage and full chain controllable technical solution for clinical trial file management. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is an integrated information system architecture schematic diagram for clinical trial files proposed by the application; Figure 2 is a data processing method flow schematic diagram for clinical trial files proposed by the application. DETAILED DESCRIPTION

[0021] The features and exemplary embodiments of various aspects of the application will be described in detail below, in order to make the purposes, technical solutions and advantages of the application more clear and apparent, the application will be further described in detail below in combination with the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the application, but not to limit the application. The application can be implemented without some of these specific details by those skilled in the art. The following description of the embodiments is only to provide a better understanding of the application by showing examples of the application.

[0022] It should be noted that in this paper, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitations, the elements defined by the statement "include" do not exclude the presence of other identical elements in the process, method, article or equipment including the elements.

[0023] In the embodiments of the present application, the same reference signs represent the same components, and for the sake of brevity, detailed descriptions of the same components are omitted in different embodiments. It should be understood that the thickness, length, width and other dimensions of various components in the embodiments of the present application shown in the drawings, and the overall thickness, length, width and other dimensions of the integrated device are only exemplary and should not constitute any limitation on the present application.

[0024] An integrated information system for clinical trial archives, comprising: A clinical modeling module for constructing a standardized data model for the whole process of clinical trials, defining data entities, attribute fields, data types, value constraints, association relationships and lifecycle state identifiers, forming a unified metadata architecture; Defining a set of core data entities covering the whole process of clinical trials, including subject basic information entity, informed consent entity, screening period data entity, enrollment data entity, visit record entity, laboratory test result entity, adverse event report entity, combined medication record entity, researcher assessment entity, monitoring report entity, data cleaning log entity, statistical analysis plan entity, final research report entity; Defining a set of attribute fields, data types, value constraints, and foreign key association relationships for each type of data entity; adding a lifecycle state identifier to each data entity, which includes six states: draft, pending review, confirmed, locked, archived, and destroyed, and defining the legal migration path and trigger conditions between states.

[0025] A data ingestion module for receiving raw archive data from electronic data collection systems, laboratory information systems, medical image archiving systems, electronic medical record systems, drug management systems, and manual input terminals, performing field mapping, format conversion, missing value marking, outlier identification, data integrity checking, and structured packaging according to the standardized data model, generating standardized archive data units and writing them into a central archive repository; A semantic analysis module for performing in-depth analysis and correlation mining of standardized archive data units at the semantic level, constructing a subject longitudinal time axis, a causal association graph of visit events and laboratory indicators, a correlation network of adverse events and combined medications, a semantic consistency matrix of researcher assessments and monitoring conclusions, a mapping relationship table of statistical analysis plans and data results, forming a multi-dimensional clinical trial archive knowledge network; A process traceability module for establishing a version evolution chain for each data unit, recording the version identifier, timestamp, operator identity, data snapshots before and after modification, and modification reason explanation of each modification operation, and storing them in an independent version log database, supporting multi-dimensional version history backtracking and difference comparison; An access control module is configured to implement a four-dimensional permission control mechanism based on roles, data sensitivity, operation types, and access contexts, dynamically generate access control decisions through a strategy engine, and ensure that high-sensitivity data is legally accessed within an authorized range. A shared audit module is configured to provide a task-driven archive sharing workflow engine, a real-time collaborative editing conflict detection and merging mechanism, a complete operation audit log recording and multi-dimensional query function, support multi-role collaborative operation, and full-process behavior traceability.

[0026] A data processing method for clinical trial archives, comprising: Step S1, constructing a standardized data model for the whole process of clinical trials, defining data entities, attribute fields, data types, value constraints, relationship, and lifecycle state identifiers, and forming a unified metadata architecture.

[0027] First, define a set of core data entities covering the whole process of clinical trials, including subject basic information entity, informed consent entity, screening data entity, enrollment data entity, visit record entity, laboratory test result entity, adverse event report entity, combined medication record entity, researcher evaluation entity, inspection report entity, data cleaning log entity, statistical analysis plan entity, and final research report entity.

[0028] Each type of data entity needs to be clearly defined as a set of attribute fields, such as the subject basic information entity containing subject number, gender, birth date, height, weight, race, medical history, allergy history, etc. The informed consent entity contains the signing date, the identity of the signer, the version number, the ethics approval number, and the scanned copy storage path. The visit record entity contains visit number, visit date, research center number, researcher number, visit type, visit status, etc.

[0029] All attribute fields need to specify their data types, such as string, integer, float, boolean, date, and binary large object, and set their value constraints, for example, the gender field is only allowed to take the value of "male" or "female", the birth date field must be earlier than the current system date, the height field must be between 100 cm and 250 cm, and the weight field must be between 30 kg and 300 kg.

[0030] In addition, the association between each data entity needs to be defined, such as the visit record entity establishing a foreign key association with the subject basic information entity through the subject number field, the adverse event report entity establishing a foreign key association with the visit record entity through the visit number field, and the laboratory test result entity establishing a foreign key association with the visit record entity through the sample number field. All data entities need to be attached with a lifecycle state identifier, which includes six states: draft, pending review, confirmed, locked, archived, and destroyed.

[0031] The state migration path needs to be strictly defined. The data unit needs to be migrated from the draft state to the pending review state by the inputter submitting a review request, and the system automatically records the submission timestamp and operator identity. The data unit needs to be migrated from the pending review state to the confirmed state by the designated review role performing a confirmation operation, and the system automatically records the review timestamp, reviewer identity, and review opinion text. The data unit needs to be migrated from the confirmed state to the locked state by meeting the preset data freezing conditions, such as the current date being later than the project preset locking date or the current visit cycle being completed more than 90%. The data unit needs to be migrated from the locked state to the archived state by completing all visit cycles and being approved by the project leader, and the system automatically records the archiving timestamp and approver identity. The data unit needs to be migrated from the archived state to the destroyed state by meeting the statutory retention period and being authorized by the ethics committee in writing, and the system automatically records the destruction timestamp, authorization document number, and executor identity. All state migration operations need to trigger system event log recording and synchronously update the version evolution chain of the data entity.

[0032] In step S2, raw archival data is received from the electronic data collection system, the laboratory information system, the medical image archiving system, the electronic medical record system, the drug management system, and the manual input terminal. Field mapping, format conversion, missing value marking, outlier identification, data integrity verification, and structured packaging are performed according to the standardized data model to generate standardized archival data units and write them into the central archival repository.

[0033] A multi-source data access interface is established, each interface corresponding to a data source system, and the data transmission protocol, authentication method, and data format template are configured. When raw data arrives at the system entrance, the system first performs field mapping operation to map the field name in the raw data to the corresponding attribute field in the standardized data model, such as mapping the "SUBJID" field in the electronic data collection system to the "subject number" field in the subject basic information entity, and mapping the "TEST_DATE" field in the laboratory information system to the "test date" field in the laboratory test result entity.

[0034] Subsequently, a format conversion operation is performed to uniformly convert the date format, value format, and encoding format in the original data into a system internal standard format, convert a date in the "MM / DD / YYYY" format into a "YYYY-MM-DD" format, convert a scientific notation value in the "1.23E+03" format into a floating point number in the "1230.0" format, and convert text in the "UTF-8" encoding into "GBK" encoding.

[0035] After the format conversion is completed, the system performs a missing value marking operation to explicitly mark all fields for which no value is provided, the marking manner being to insert a special marker "MISSING" at the value position of the field and record a missing reason code such as "original data is not collected", "lost in the transmission process", and "field mapping fails". Subsequently, an outlier identification operation is performed to perform a preset outlier detection rule on all numerical value type fields, detect whether the height field exceeds the range of 100 cm to 250 cm, detect whether the weight field exceeds the range of 30 kg to 300 kg, and detect whether the laboratory test result field exceeds the medical reference range of the test item.

[0036] If an outlier is detected, the system inserts a special marker "OUTLIER" at the value position of the field and records an outlier type code, an expected value range, and an actual value. After the missing value marking and outlier identification are completed, the system performs a data integrity check operation, which includes six types of rules: a field non-empty check, which checks whether all mandatory fields all have non-empty values; a numerical value range check, which checks whether all numerical value type fields are within their preset ranges; an enumeration value legality check, which checks whether all enumeration type fields are within their preset enumeration value set; a foreign key association consistency check, which checks whether the primary key referenced by all foreign key fields exists in the target data entity; a time sequence logic check, which checks whether all time type fields satisfy a preset time sequence logic, such as that the visit date cannot be earlier than the screening date and the laboratory test date cannot be more than seven days later than the visit date; and a signature integrity check, which checks whether all data units that need to be digitally signed contain a valid digital signature and the signature has not been tampered with.

[0037] If any of the check rules fails, the system generates a structured error report containing error type, error field, error value, expected value, error location identifier, and prevents the data unit from entering the central archive repository, while sending an error notification to the data submitter, requiring it to correct the data and resubmit. Only when all checks pass, the system encapsulates the original data as a standardized archive data unit containing data entity identifier, attribute field value set, data source identifier, data collection timestamp, data submitter identity identifier, data check pass timestamp, and writes it into the central archive repository. The write operation adopts a transaction mechanism to ensure the atomicity and consistency of data writing. If any exception occurs during the writing process, the system automatically rolls back the entire transaction and records the rollback reason.

[0038] Step S3, the standardized archive data unit is subjected to deep semantic analysis and correlation mining, to build a subject longitudinal time axis, a causal correlation graph of visit events and laboratory indicators, a correlation network of adverse events and concomitant medications, a semantic consistency matrix of researcher evaluation and monitoring conclusions, a mapping relationship table of statistical analysis plan and data results, to form a multi-dimensional clinical trial archive knowledge network; A subject longitudinal time axis is built, which takes subject number as the unique identifier, arranges all visit records, laboratory test records, adverse event records, concomitant medication records, and researcher evaluation records of the subject in chronological order; each event on the time axis is labeled with its occurrence time, event type, event description, and associated data entity identifier. Subsequently, the system builds a causal correlation graph of visit events and laboratory indicators, which adopts Bayesian network inference algorithm, takes visit events as nodes, laboratory indicator changes as observation variables, and potential pathological mechanisms as hidden variables.

[0039] First, joint distribution samples of visit events and laboratory indicators are extracted from historical data, and then the structure parameters and conditional probability parameters of the Bayesian network are learned through maximum likelihood estimation; after learning is completed, the system can calculate the conditional probability distribution of laboratory indicator changes under the condition of any visit event occurrence, thereby quantifying the causal dependence strength between events and indicators. The system can calculate the probability of serum creatinine level rising after giving a certain drug or the probability of white blood cell count decreasing after a certain adverse event occurs, and then the system builds a correlation network of adverse events and concomitant medications, which adopts a dual-index fusion method of Pearson correlation coefficient and Spearman rank correlation coefficient.

[0040] The system first extracts the frequency sequence of adverse events and the combined drug dose sequence of all subjects, then calculates the Pearson correlation coefficient and the Spearman rank correlation coefficient between each pair of adverse events and combined drugs, and finally obtains the comprehensive correlation score by weighted average of the two coefficients. The system sets the correlation score threshold, only keeps the significant correlation pairs with a score higher than the threshold, and constructs a correlation network graph, where the nodes are adverse events or combined drugs, and the edges are correlation scores. Subsequently, the system constructs a semantic consistency matrix of investigator assessments and monitoring conclusions. This matrix uses a text matching algorithm based on word vector cosine similarity. The system first performs word segmentation on the investigator assessment text and the monitoring conclusion text, then converts each word into a high-dimensional vector using a pre-trained clinical medicine field word vector model, and then performs average pooling on all word vectors of each text to obtain the semantic vector representation of the text. Finally, the system calculates the cosine similarity between the semantic vectors of each pair of investigator assessment text and monitoring conclusion text as their semantic consistency score.

[0041] The system sets a similarity threshold, marks all inconsistent items with a score below the threshold, and generates an inconsistency report containing inconsistent text pairs, similarity scores, and keyword difference analysis. Subsequently, the system constructs a mapping table between statistical analysis plans and data results. This table takes the analysis variables in the statistical analysis plan as rows and the corresponding variables in the actual data results as columns, recording the actual value, missing rate, distribution characteristics, and deviation from the planned expectation of each analysis variable in the data results. All the above analysis results are integrated into a multi-dimensional clinical trial archive knowledge network. This network supports semantic-based intelligent retrieval, such as querying "all adverse events significantly related to drug A", the system will return all adverse event nodes connected to drug A in the correlation network. It also supports trend prediction, such as querying "the probability of subject B having liver function abnormalities in future visits", the system will return the conditional probability value based on Bayesian network inference; and risk warning, such as when the change pattern of a subject's laboratory indicators and the conditional probability of high-risk pathological mechanisms exceed the preset threshold, the system automatically triggers a risk warning notification.

[0042] Step S4: Establish a version evolution chain for each data unit, record the version identifier, timestamp, operator identity, data snapshots before and after modification, and modification reason explanation of each modification operation, and store them in an independent version log database to support multi-dimensional version history backtracking and difference comparison; The system assigns a unique version identifier to each data unit in the central archive repository, which adopts a hierarchical coding structure. The first layer is the project number, encoded with four digits. The second layer is the subject number, encoded with six digits. The third layer is the data entity type code, encoded with two letters, such as "SB" for subject basic information, "VS" for visit record, and "LB" for laboratory test result. The fourth layer is the data entity instance number, encoded with eight digits. The fifth layer is the version sequence number, encoded with three digits, starting from 001 and incrementing. The sixth layer is the timestamp hash value, calculated by hashing the operation timestamp using the SHA256 algorithm and taking the first eight hexadecimal characters.

[0043] Whenever a data unit undergoes any modification operation, the system automatically generates a new version record, which contains the new version identifier, operation timestamp, operator identity, operation type code, modification reason text, pre-modification data snapshot, and post-modification data snapshot. Both the pre-modification and post-modification data snapshots are stored in JSON format, serializing all field values of the data unit before and after modification. All version records are stored in a separate version log database, which adopts a blockchain-like chain storage structure. Each version record contains the previous version hash value, current version data hash value, and operation metadata hash value.

[0044] The previous version hash value is the hash value of the globally unique identifier of the previous version record. The current version data hash value is the hash value of the post-modification data snapshot in the current version record. The operation metadata hash value is the hash value of all fields in the current version record except the data snapshot. Through this chain structure, any tampering with historical version records will cause the hash values of all subsequent version records to mismatch, thereby achieving the tamper-proof nature of the version evidence chain.

[0045] The system supports version history backtracking based on multiple dimensions such as subject number, visit number, data type, and operation time interval. Users can specify any query conditions, and the system returns a list of all version records that meet the conditions. Users can also select any two version records for difference comparison, and the system automatically calculates the field-level differences between the two version data snapshots and displays the difference fields, original values, and new values in highlighted mode.

[0046] In addition, the system supports version rollback operations, and users can choose to roll back a data unit to any historical version. The system will automatically perform reverse operations to restore the data to the specified version state and generate a new version record, recording the timestamp, operator identity, and rollback target version identifier of the rollback operation.

[0047] Step S5, implement a four-dimensional permission control mechanism based on role, data sensitivity, operation type, and access context, dynamically generate access control decisions through a policy engine, and ensure that high-sensitivity data is legally accessed within the authorized scope; First, define a set of roles for clinical trial participants, including sponsor project manager, research center researcher, independent data monitoring committee member, statistical analyst, inspector, and ethics committee member. Each role is assigned a set of basic permissions, such as the sponsor project manager having read and write permissions for project-level metadata, the research center researcher having read and write permissions for data from the research center they are responsible for, the independent data monitoring committee member having read-only permissions for safety data, the statistical analyst having read and export permissions for analysis data sets, the inspector having read-only permissions for operation logs, and the ethics committee member having read-only permissions for informed consent forms and ethics approval documents.

[0048] Subsequently, the system defines data sensitivity levels, divided into four levels: public, restricted, confidential, and top secret. Public-level data includes project numbers, research center numbers, and visit schedules; restricted-level data includes subject numbers, visit dates, and laboratory test item names; confidential-level data includes laboratory test specific values, adverse event description text, and combined medication names and doses; and top-secret-level data includes subject real identity information, informed consent form scans, and genetic sequencing raw data. Each type of data entity is labeled with its default sensitivity level in the standardized data model, and the system supports dynamically adjusting the sensitivity level of specific data instances at runtime.

[0049] Subsequently, the system defines operation type permissions, including read-only, edit, export, delete, and version rollback. Read-only permission allows users to view data content but not modify it; edit permission allows users to modify data content and submit new versions; export permission allows users to export data to external file formats; delete permission allows users to mark data as logically deleted; and version rollback permission allows users to roll back data to a historical version. Subsequently, the system defines access context constraints, including access only from specified IP address segments, only from specified device fingerprints, and only within specified time periods. Device fingerprints are generated from six parameters: device model, operating system version, browser type, screen resolution, font list, and time zone setting, and are stored in a trusted device registration library. The system maintains a policy engine that dynamically combines the role of the requester, the data sensitivity level of the data being accessed, the operation type being performed, and the access context in which the request occurs, queries a pre-set policy rule library, and generates the final access control decision. The policy rule library is stored in a triple form, and each triple contains a combination of role, data sensitivity, and operation type conditions and allows or denies the decision result.

[0050] The rule "sponsor project manager + confidential level data + edit -> allow" indicates that the sponsor project manager can edit the confidential level data; and the rule "statistical analyst + top secret level data + export -> reject" indicates that the statistical analyst cannot export the top secret level data. If the access context contains an IP address segment constraint, the system additionally checks whether the request source IP is in the allowed list; if a device fingerprint constraint is contained, the system additionally checks whether the request device fingerprint is in the trusted device registration library; and if a time period constraint is contained, the system additionally checks whether the current system time is in the allowed time period. Only when the constraints of all dimensions are satisfied, the system allows the access operation to be executed. All access control decisions are recorded in the audit log, including the decision timestamp, the requester identity, the request data identification, the request operation type, the decision result, and the decision basis rule identification.

[0051] In step S6, a task-driven archive sharing workflow engine, a real-time collaborative editing conflict detection and merging mechanism, a complete operation audit log recording and multi-dimensional query function are provided to support multi-role collaborative operation and full-process behavior traceability.

[0052] The system provides a task-driven archive sharing workflow engine, which supports administrators to define sharing trigger conditions and sharing target roles. It can be defined that "when a certain subject completes the last visit, automatically share all data of the subject to the statistical analyst role"; or "when a certain research center completes the enrollment of all subjects, automatically share the enrollment data summary table of the center to the sponsor project manager role". The sharing trigger condition can be defined based on data state change, time condition, event trigger and various ways. When the trigger condition is met, the system automatically generates a sharing task, sends a notification to the target role, and grants it the preset access permission. The permission validity period can be set to a fixed length of time or until the task is completed. The system provides a real-time collaborative editing conflict detection and merging mechanism. When multiple users edit the same data unit at the same time, the system uses an operation conversion algorithm to solve the conflict.

[0053] Specifically, the system decomposes each editing operation into a sequence of atomic operations, such as "insert character X in field A" and "modify the value of field B from Y to Z". When concurrent operations are detected, the system performs exchange sorting on these atomic operation sequences to ensure that the final execution order satisfies the exchange law of operations, thereby guaranteeing the consistency of the final data state. For example, if user A performs "modify field B from Y to Z" and user B performs "modify field B from Y to W" at the same time, the system will detect the conflict and prompt the user to choose which modification to keep or automatically merge it into "modify field B from Y to Z+W". All conflict detection and merging operations are recorded in the operation log.

[0054] The system provides complete operation audit log recording function, records the performer, execution time, operation object, operation content, operation result state of each data access, modification, export, sharing, deletion operation. The operation object is recorded as the version identifier of the data unit, the operation content is recorded as the specific modified field and value, and the operation result state is recorded as success or failure and failure reason code. The audit log is stored in a write-once read-many storage medium, and read-only access permission is set to prevent the log from being tampered with or deleted. The system supports audit query according to time range, operation type, user identity, data entity and other dimensions, users can specify any query condition, the system returns a list of all audit records meeting the condition, and supports generating audit report, the report includes operation statistical summary, abnormal operation marking, permission change history, etc.

[0055] In the embodiment, the clinical modeling module is responsible for performing step S1, which internally includes a data entity definition submodule, an attribute field constraint submodule, an association relationship configuration submodule, and a lifecycle state management submodule; the data import module is responsible for performing step S2, which internally includes a multi-source data access submodule, a field mapping engine submodule, a format converter submodule, a missing value and abnormal value marking submodule, an integrity checker submodule, and a structured encapsulator submodule; the semantic analysis module is responsible for performing step S3, which internally includes a longitudinal time axis construction submodule, a causal association graph reasoning submodule, a correlation network calculation submodule, a semantic consistency matrix generation submodule, a mapping relationship table construction submodule, and a knowledge network integration submodule; the process tracing module is responsible for performing step S4, which internally includes a version identifier generation submodule, a version record construction submodule, a blockchain storage submodule, a historical backtracking interface submodule, a difference comparison engine submodule, and a version rollback execution submodule; the access control module is responsible for performing step S5, which internally includes a role permission definition submodule, a data sensitivity annotation submodule, an operation type authorization submodule, an access context verification submodule, a policy engine decision submodule, and a permission dynamic adjustment submodule; the sharing audit module is responsible for performing step S6, which internally includes a task-driven workflow engine submodule, a real-time collaboration conflict detection submodule, an operation conversion and merging submodule, an audit log recording submodule, a multi-dimensional query interface submodule, and a report generator submodule. All modules are deployed in a distributed server cluster, and data exchange and event notification are performed through an internal message bus to ensure the overall high availability and scalability of the system.

[0056] The above is only a preferred specific embodiment of the present application, not to limit the protection scope of the present application, therefore: all equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. An integrated information system for clinical trial records, characterized in that, include: The clinical modeling module is used to build a standardized data model for the entire clinical trial process, defining data entities, attribute fields, data types, value constraints, relationships, and lifecycle status identifiers to form a unified metadata architecture. The data intake module is used to receive raw archival data from electronic data acquisition systems, laboratory information systems, medical image archiving systems, electronic medical record systems, drug management systems, and manual data entry terminals. Based on the standardized data model, it performs field mapping, format conversion, missing value marking, outlier identification, data integrity verification, and structured encapsulation to generate standardized archival data units and write them into the central archival repository. The semantic analysis module is used to perform in-depth semantic analysis and correlation mining on the standardized archive data units, constructing a longitudinal timeline of subjects, a causal relationship map of visit events and laboratory indicators, a correlation network of adverse events and concomitant medications, a semantic consistency matrix of investigator assessment and monitoring conclusions, and a mapping relationship table between statistical analysis plans and data results, forming a multi-dimensional clinical trial archive knowledge network. The process traceability module is used to establish a version evolution chain for each data unit, record the version identifier, timestamp, operator identity, data snapshots before and after the modification, and explanation of the reason for the modification for each modification operation, and store them in an independent version log database, supporting multi-dimensional version history backtracking and difference comparison; The access control module is used to implement a four-dimensional permission control mechanism based on roles, data sensitivity, operation type, and access context. It dynamically generates access control decisions through a policy engine to ensure that highly sensitive data is legally accessed within the authorized scope. The shared audit module provides a task-driven file sharing workflow engine, a real-time collaborative editing conflict detection and merging mechanism, complete operation audit log recording and multi-dimensional query functions, and supports multi-role collaborative operation and full-process behavior traceability.

2. The integrated information system for clinical trial records according to claim 1, characterized in that: The clinical modeling module is used to define a set of core data entities covering the entire clinical trial process, including subject basic information entity, informed consent form entity, screening period data entity, enrollment data entity, visit record entity, laboratory test result entity, adverse event report entity, concomitant medication record entity, investigator evaluation entity, monitoring report entity, data cleaning log entity, statistical analysis plan entity, and final study report entity. Define the attribute field set, data type, value constraints, and foreign key relationships for each type of data entity; attach a lifecycle status identifier to each data entity, which includes six states: draft, pending review, confirmed, locked, archived, and destroyed, and define the legal migration paths and triggering conditions between states.

3. A data processing method for clinical trial archives, applied to the integrated information system for clinical trial archives as described in any one of claims 1-2, characterized in that, include: Step S1: Construct a standardized data model for the entire clinical trial process, defining data entities, attribute fields, data types, value constraints, relationships, and lifecycle status identifiers to form a unified metadata architecture; Step S2: Receive raw archive data from electronic data acquisition system, laboratory information system, medical image archiving system, electronic medical record system, drug management system and manual input terminal; perform field mapping, format conversion, missing value marking, outlier identification, data integrity verification and structured encapsulation according to the standardized data model; generate standardized archive data units and write them into the central archive repository. Step S3: Perform deep semantic analysis and correlation mining on the standardized archive data units to construct a longitudinal timeline of subjects, a causal relationship map of visit events and laboratory indicators, a correlation network of adverse events and concomitant medications, a semantic consistency matrix of investigator assessment and monitoring conclusions, and a mapping relationship table of statistical analysis plans and data results, forming a multi-dimensional clinical trial archive knowledge network. Step S4: Establish a version evolution chain for each data unit, record the version identifier, timestamp, operator identity, data snapshots before and after the modification, and explanation of the reason for the modification for each modification operation, and store them in an independent version log database to support multi-dimensional version history backtracking and difference comparison. Step S5: Implement a four-dimensional access control mechanism based on role, data sensitivity, operation type, and access context. The policy engine dynamically generates access control decisions to ensure that highly sensitive data is legally accessed within the authorized scope. Step S6 provides a task-driven file sharing workflow engine, a real-time collaborative editing conflict detection and merging mechanism, complete operation audit log recording and multi-dimensional query function, and supports multi-role collaborative operation and full-process behavior traceability.

4. The data processing method for clinical trial archives according to claim 3, characterized in that, In step S1, the defined data entities include: subject basic information entity, informed consent form entity, screening period data entity, enrollment data entity, visit record entity, laboratory test result entity, adverse event report entity, combined medication record entity, investigator evaluation entity, monitoring report entity, data cleaning log entity, statistical analysis plan entity, and final study report entity. Define the attribute field set, data type, value constraints, and foreign key relationships for each type of data entity; attach a lifecycle status identifier to each data entity, which includes six states: draft, pending review, confirmed, locked, archived, and destroyed, and define the legal migration paths and triggering conditions between states.

5. The data processing method for clinical trial archives according to claim 3, characterized in that, Step S2 includes: Establish multi-source data access interfaces and configure data transmission protocols, authentication methods, and data format templates; Perform field mapping operations to map the original data field names to the corresponding attribute fields in the standardized data model; Perform a format conversion operation to convert the date format, numeric format, and encoding format in the original data into the system's internal standard format; Perform a missing value marking operation, inserting the special marker "MISSING" for fields that do not provide values ​​and recording the missing reason code; Perform outlier identification. Apply preset outlier detection rules to numeric fields. If an outlier is detected, insert a special marker "OUTLIER" and record the outlier type code, expected value range, and actual value. Perform data integrity verification operations, including six categories of rules: field non-empty verification, numerical range verification, enumeration value validity verification, foreign key association consistency verification, time series logic verification, and signature integrity verification. If any check fails, a structured error report is generated and the data unit is blocked from being written. Only when all checks pass is the data unit packaged into a standardized archival data unit and written to the central archival repository using a transaction mechanism.

6. The data processing method for clinical trial archives according to claim 3, characterized in that, Step S3 includes: Construct a longitudinal timeline with subject ID as the unique identifier, and arrange all visit records, laboratory test records, adverse event records, concomitant medication records, and investigator evaluation records for that subject in chronological order; A causal relationship graph between visit events and laboratory indicators was constructed using a Bayesian network inference algorithm. Visit events were used as nodes, changes in laboratory indicators were used as observed variables, and potential pathological mechanisms were used as latent variables. The network parameters were learned through maximum likelihood estimation. A correlation network between adverse events and concomitant medications was constructed using a dual-index fusion method of Pearson correlation coefficient and Spearman rank correlation coefficient. The comprehensive correlation score between each pair of adverse events and concomitant medications was calculated, and significant correlation pairs with scores higher than a preset threshold were retained. A semantic consistency matrix for researchers' evaluation and monitoring conclusions is constructed using a text matching algorithm based on word vector cosine similarity. The text is segmented, word vector average pooling is performed, and cosine similarity is calculated. Inconsistent entries with similarity below a preset threshold are marked. Construct a mapping table between the statistical analysis plan and the data results, and record the actual value, missing rate, distribution characteristics, and degree of deviation from the planned expectations for each analysis variable in the data results; By integrating the actual values, missing rates, distribution characteristics, and deviations from planned expectations of each analytical variable in the data results, a multi-dimensional clinical trial archive knowledge network is formed to support intelligent retrieval, trend prediction, and risk warning.

7. The data processing method for clinical trial archives according to claim 3, characterized in that, Step S4 includes: The data units are assigned version identifiers with a hierarchical coding structure. The first layer is the project number, the second layer is the subject number, the third layer is the data entity type code, the fourth layer is the data entity instance number, the fifth layer is the version number, and the sixth layer is the timestamp hash value. Generate a version record, including the new version identifier, operation timestamp, operator identity identifier, operation type code, text explaining the reason for modification, data snapshot before modification, and data snapshot after modification; Version records are stored in an independent version log database that uses a blockchain-style chain storage structure. Each version record contains the hash value of the previous version, the hash value of the current version data, and the hash value of the operation metadata. It supports version history tracing by subject number, visit number, data type, and operation time interval; it supports field-level difference comparison and highlighting of any two version records; and it supports version rollback operations and the generation of new version records.

8. The data processing method for clinical trial archives according to claim 3, characterized in that, Step S5 includes: Define the set of roles for clinical trial participants, including sponsor project managers, research center researchers, independent data monitoring committee members, statistical analysts, auditors, and ethics committee members; Define data sensitivity levels, including public, restricted, confidential, and top secret. Public data includes project number, research center number, and visit schedule. Restricted data includes subject number, visit date, and laboratory test item name. Confidential data includes specific laboratory test values, adverse event description text, and names and dosages of concomitant drugs. Top secret data includes the subject's real identity information, scanned copy of informed consent form, and raw gene sequencing data. Define operation type permissions, including read-only, edit, export, delete, and version rollback; Define access context constraints, including access is limited to a specified IP address range, access is limited to a specified device fingerprint, and access is limited to a specified time period. The device fingerprint is generated by a combination of six parameters: device model, operating system version, browser type, screen resolution, font list, and time zone setting. The strategy engine dynamically combines four dimensions of parameters: role, data sensitivity, operation type, and access context. It queries the preset policy rule base to generate access control decisions and allows access operations to be executed only when all dimensional constraints are met.

9. The data processing method for clinical trial archives according to claim 3, characterized in that, Step S6 includes: Define the sharing trigger conditions and the sharing target role. When the conditions are met, automatically generate a sharing task and grant preset access permissions to the target role. An operation transformation algorithm is used to resolve multi-user concurrent editing conflicts. Editing operations are decomposed into atomic operation sequences and swapped and sorted to ensure the consistency of the final data state. Record the executor, execution time, operation object, operation content, and operation result status for each data access, modification, export, sharing, and deletion operation. The operation object is recorded as the version identifier of the data unit, and the operation content is recorded as the specific fields and values ​​that are modified. The audit logs are stored on a write-once, read-many storage medium with read-only access permissions, supporting audit queries and report generation by time range, operation type, user identity, and data entity dimension.

10. The data processing method for clinical trial archives according to claim 5, characterized in that: The time-series logic verification in the data integrity verification includes checking that the visit date is not earlier than the screening date and the laboratory testing date is not more than seven days later than the corresponding visit date; the signature integrity verification includes checking whether the data unit that needs to be digitally signed contains a valid and tamper-proof digital signature.

Images