A heterogeneous data integration and intelligent analysis method for patient whole-cycle health records

Abstract

The present application relates to the technical field of data analysis, in particular to a heterogeneous data integration and intelligent analysis method for patient whole-cycle health records, in the present application, the computer tomography image data, electronic medical record data and robot audit traceability data of the patient are respectively processed into image event segments, medical record semantic event segments and robot process event segments, diagnosis and treatment event anchors are generated based on patient diagnosis and treatment event association information, each event segment is mapped to the same event index structure, cross-dimension association relationship is established, traceability credibility markers are generated in combination with audit traceability fields, so that multi-source heterogeneous data can be associated, integrated and evaluated with credibility around the same diagnosis and treatment process; patient personalized diagnosis and treatment scheme basis is generated according to the cross-dimension association relationship and the traceability credibility markers, and iterative updating is realized by means of the backwriting of the new evaluation data corresponding to the new event segments, which helps to improve the association clarity, credible utilization degree and updating continuity of the patient personalized diagnosis and treatment scheme basis.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to a method for integrating and intelligently analyzing heterogeneous data from a patient's full-cycle health records. Background Technology

[0002] A patient's full-cycle health record typically consists of multi-source information, including imaging examination data, electronic medical record data, and equipment process data. Imaging examination data mainly reflects the location, morphology, volume of lesions, and imaging diagnostic conclusions. Electronic medical record data mainly records the diagnosis name, symptom description, examination and test results, treatment behavior, and medication information. Equipment process data records task number, operation time, executed actions, parameter settings, system feedback, and audit signatures. Because these data come from different business systems, their recording granularity, field naming, time expression methods, and semantic focus are inconsistent. Therefore, during the integration of health records, it is usually necessary to merge the data according to patient identification, visit time, or record number, and combine this with structured mapping, time sorting, or field matching methods to form a unified file.

[0003] In existing record organization methods, a common practice is to first aggregate data from different sources into the same patient file, and then arrange them according to the chronological order of examination, diagnosis, or treatment; or to standardize diagnostic names, examination items, and treatment records using standardized terminology to extract characteristics of disease changes, treatment trends, or risk indicators in subsequent analysis. For some equipment data with auditable attributes, basic verification can also be performed through task numbers, written records, or signature fields, thereby preserving source and process information.

[0004] However, in specific clinical scenarios, the correlation between imaging findings, medical record semantics, and equipment execution processes is often not accurately established solely by patient identification or chronological order. Especially within the same treatment phase, the recording times of data from different sources may differ, field expressions may not be entirely consistent, and some process data may be difficult to directly map to specific treatment actions due to different system recording methods. Consequently, while a significant amount of raw information can be obtained for subsequent analysis of personalized treatment plans, maintaining consistency in the correspondence, sources, and process reliability of this information is challenging, thus affecting the continuous organization, reliable use, and subsequent updates of the treatment basis. Summary of the Invention

[0005] The purpose of this invention is to provide a method for integrating and intelligently analyzing heterogeneous data from a patient's full-cycle health records, thereby addressing the problems mentioned in the background section. Specifically, the technical problem involves how to perform correlation integration and reliability evaluation of imaging data, medical record data, and robot audit traceability data from a patient's full-cycle health records, oriented towards the diagnosis and treatment process. This addresses the issues of unclear correlation criteria, difficulty in utilizing the reliability of sources, and insufficient continuity in the update process when generating personalized treatment plans based on heterogeneous diagnostic and treatment data.

[0006] To achieve the above objectives, the present invention aims to provide a method for heterogeneous data integration and intelligent analysis of a patient's full-cycle health records, specifically including the following method steps: S1. Acquire the patient's computed tomography (CT) image data, electronic medical record data, and robot audit traceability data, and process them into image event fragments, medical record semantic event fragments, and robot process event fragments, respectively; specifically including: Based on patient identification, computed tomography (CT) images, electronic medical records, and robotic auditing data are obtained from the image archiving system, electronic medical record system, and robotic auditing system, respectively. The CT images are preprocessed, and the location, volume, and boundary characteristics of the lesion are determined by the lesion region. The data are then combined with the image examination number, image examination time, examination site, image sequence number, and image diagnosis conclusion to encapsulate the data into an image event fragment. Read admission records, progress notes, examination and test records, medical orders, surgical records, discharge records and follow-up records from electronic medical records, perform word segmentation, entity recognition, time recognition and treatment behavior recognition on the text content, and merge the semantic content of the same medical record number, the same record time and belonging to the same treatment behavior into medical record semantic event fragments; Read the robot task number, operation time, operator identifier, device number, executed action, parameter settings, system feedback, data write record and audit signature from the robot audit traceability data, collect them according to robot task number, sort them according to operation time and encapsulate them into robot process event fragments.

[0007] Step S1 converts computed tomography (CT) image data, electronic medical record data, and robot audit traceability data into image event fragments, medical record semantic event fragments, and robot process event fragments, respectively, so that data from different sources have a unified event-based expression form, which facilitates subsequent mapping and association around the diagnosis and treatment process; this step provides a consistent basic data unit for the organization of the same event across data from different sources and subsequent credibility evaluation.

[0008] S2. Anchor point coding is performed based on patient treatment event association information to generate treatment event anchor points; specifically including: Obtain patient treatment event association information, which includes patient identifier, treatment stage number, treatment stage start time, treatment stage end time, treatment time, treatment department, treatment behavior type, imaging examination number, medical record number, and robot task number; The patient identifier, treatment stage number, treatment time, treatment department, and treatment behavior type are concatenated in a fixed order and according to field separators to form the original anchor string; Calculate a summary value for the original string of the anchor point, and use the calculated summary value as the anchor point encoding value; establish a binding relationship between the anchor point encoding value and the image examination number, medical record number and robot task number to generate a diagnosis and treatment event anchor point.

[0009] Step S2 generates treatment event anchors through patient treatment event association information and binds the anchor code value with the image examination number, medical record number and robot task number, so that the same treatment event has a locatable unified index; this step is used to establish the corresponding entry points between different types of event fragments to support subsequent aggregation according to the treatment process.

[0010] S3. Based on the diagnostic event anchor points, map image event fragments, medical record semantic event fragments, and robot process event fragments to the same event index structure; specifically including: Establish a unified event index structure that includes an anchor layer, an event fragment layer, a relationship layer, and a source field layer; Write the diagnosis and treatment event anchors into the anchor layer, and use the anchor code value as the main index in the anchor layer; Based on the image examination number, medical record number, and robot task number in the diagnosis and treatment event anchor point, the corresponding image event fragment, medical record semantic event fragment, and robot process event fragment are found and written into the event fragment layer. Under the anchor point of diagnosis and treatment event, a list of image event fragment identifiers, a list of medical record semantic event fragment identifiers, and a list of robot process event fragment identifiers are generated; Write the source system identifier, data generation time, data writing time, and field validation value corresponding to each event fragment into the source field layer.

[0011] Step S3 maps the diagnosis and treatment event anchor points and corresponding event fragments to the same event index structure, and synchronously writes the source system identifier, data generation time, data writing time and field verification value into the source field layer, so that the organization relationship and source information of event fragments can be managed at the same time. This step provides a unified index and source basis for the subsequent establishment of cross-dimensional relationship and source traceability credibility evaluation.

[0012] S4. Based on the same event index structure, establish cross-dimensional relationships between image event fragments, medical record semantic event fragments, and robot process event fragments; specifically including: Based on the same event index structure, image event fragments, medical record semantic event fragments, and robot process event fragments under the same diagnosis and treatment event anchor point are read; Extract lesion location, lesion volume, lesion boundary features and image diagnosis conclusions from image event fragments; extract diagnosis name, symptom description, examination items, treatment behavior and medication information from medical record semantic event fragments; and extract executed actions, parameter settings, system feedback and data writing records from robot process event fragments. Standardize the terminology of imaging diagnostic conclusions and diagnostic names, and calculate diagnostic correlation values; calculate feature correlation values ​​based on lesion location, lesion volume, lesion boundary characteristics, symptom description, and examination items. Calculate process-related values ​​based on treatment behaviors and performed actions; Based on anchor point encoding values, image event fragment identifiers, medical record semantic event fragment identifiers, robot process event fragment identifiers, diagnostic association values, feature association values, and process association values, cross-dimensional association relationships are generated and written into the association relationship layer of the same event index structure.

[0013] Step S4 calculates the diagnostic correlation value, feature correlation value, and process correlation value among image event fragments, medical record semantic event fragments, and robot process event fragments based on the same event index structure, thereby generating cross-dimensional correlation relationships. This step ensures that multi-source data under the same diagnosis and treatment event are no longer just stored side by side, but form a relational result with a clear correlation.

[0014] S5. Based on the diagnosis and treatment event anchor point, retrieve cross-dimensional relationships and corresponding robot process event fragments from the same event index structure, specifically including: Based on the anchor code value in the current diagnosis and treatment event anchor, locate the target anchor record in the anchor layer of the same event index structure; Read the fragment index set of the target anchor point recorded in the event fragment layer. The fragment index set includes image event fragment identifiers, medical record semantic event fragment identifiers, and robot process event fragment identifiers. Read the corresponding robot process event fragment based on the robot process event fragment identifier, and read the corresponding medical record semantic event fragment based on the medical record semantic event fragment identifier; Read the corresponding cross-dimensional relationship from the relationship layer based on the cross-dimensional relationship identifier; The consistency of the robot task number in the robot process event segment with the robot task number in the diagnosis and treatment event anchor point is verified, and the consistency of the anchor point code value in the cross-dimensional association relationship with the anchor point code value in the diagnosis and treatment event anchor point is verified. When the robot task number and anchor point code value are consistent, the read robot process event fragments, medical record semantic event fragments and cross-dimensional correlations will be input into the source traceability credibility evaluation process.

[0015] The above invocation process invokes the corresponding robot process event fragments and cross-dimensional relationships based on the anchor points of the diagnosis and treatment events. Input verification is completed through the consistency of robot task numbers and anchor point encoding values ​​to ensure that the data entering the evaluation process belongs to the same diagnosis and treatment event. This process provides a limited scope for subsequent source traceability credibility evaluation and reduces the impact of cross-event misuse on credible results.

[0016] Based on the audit traceability fields in robot process event fragments, the traceability credibility of cross-dimensional relationships is evaluated, and traceability credibility tags are generated; specifically including: The audit traceability fields are read from the robot process event fragments. The audit traceability fields include robot task number, operation time, operator identifier, device number, executed action, parameter settings, system feedback, data write record and audit signature. The task consistency evaluation value is obtained based on the consistency between the robot task number and the robot task number in the diagnosis and treatment event anchor point; The time consistency evaluation value is obtained based on the correspondence between the operation time and the start time and end time of the diagnosis and treatment stage in the diagnosis and treatment event anchor point; Based on the preset behavior action mapping table, the system verifies whether the executed action corresponds to the treatment behavior in the medical record semantic event fragment, and obtains the behavior consistency evaluation value. The write consistency evaluation value is obtained based on whether the data written record is consistent with the source field layer in the same event index structure; The audit signature is verified based on the pre-stored device certificate to obtain the signature integrity evaluation value; A traceability credibility score is generated based on the task consistency evaluation value, time consistency evaluation value, behavior consistency evaluation value, write consistency evaluation value, and signature integrity evaluation value. Based on the comparison results between the traceability credibility score and the first credibility threshold and the second credibility threshold, traceability credibility labels of high credibility, medium credibility, or low credibility are generated. The generated source traceability credibility markers are written into the association layer of the same event index structure, so that the source traceability credibility markers are bound to the corresponding cross-dimensional association identifiers.

[0017] The process of generating the source traceability credibility evaluation generates source traceability credibility scores through task consistency, time consistency, behavior consistency, write consistency, and signature integrity, and further forms source traceability credibility tags of high credibility, medium credibility, or low credibility; the results can be directly bound to cross-dimensional correlations, so that the source credibility can participate in the generation of subsequent diagnosis and treatment basis.

[0018] S6. Generate personalized treatment plans for patients based on cross-dimensional correlations and source traceability credibility markers; specifically including: Read the image event fragments, medical record semantic event fragments, robot process event fragments, cross-dimensional relationships, and source traceability credibility markers corresponding to each diagnosis and treatment event anchor point under the same patient identifier; The anchor points of each treatment event are sorted according to the treatment time to generate a patient treatment time sequence; Based on the patient's diagnosis and treatment time series, the trends of lesion volume change, symptom change, diagnosis change, treatment behavior change, and robot execution action feedback were calculated. The trends in lesion volume change, symptom change, treatment behavior change, and robot action feedback were converted into lesion volume change trend score, symptom change trend score, treatment behavior change score, and robot action feedback trend score, respectively. The treatment plan basis score is calculated based on the diagnostic correlation value, characteristic correlation value, process correlation value, lesion volume change trend score, symptom change trend score, and treatment behavior change trend score. The treatment plan is revised based on the source traceability credibility marker to obtain the recommended plan basis score; Based on the score of the recommended treatment plan, a personalized treatment plan is generated for the patient. The personalized treatment plan includes the patient's current treatment stage, main imaging data, main medical records, robotic process data, traceability credibility markers, risk warning information, and the score of the recommended treatment plan.

[0019] Step S6 reads the data corresponding to each treatment event anchor point under the same patient identifier and forms a time series according to the treatment time. Then, it extracts lesions, symptoms, diagnosis, treatment and robot execution trends to generate a basis for personalized treatment plan for the patient. This step incorporates cross-dimensional correlation and source traceability credibility mark into the calculation of the basis for the plan, so that the output results take into account both correlation and credibility.

[0020] S7. When the patient's personalized treatment plan does not meet the preset stability conditions, a simulation evaluation of the patient's personalized treatment plan is conducted to obtain new evaluation data; specifically including: When the patient's personalized treatment plan does not meet the preset stable conditions, the patient's current treatment stage is converted into a stage code, the main imaging data is converted into imaging status parameters, the main medical record data is converted into clinical status parameters, the robot process data is converted into execution process parameters, and the risk warning information is converted into risk status parameters. The simulation input parameter set is composed of stage coding, imaging status parameters, clinical status parameters, execution process parameters, risk status parameters, and scheme recommendation based on scoring; The simulation input parameter set is evaluated using preset simulation evaluation rules, which include treatment response prediction rules, lesion change prediction rules, risk change prediction rules, and protocol stability evaluation rules. Treatment response prediction values ​​are obtained through treatment response prediction rules, lesion change prediction values ​​are obtained through lesion change prediction rules, risk change prediction values ​​are obtained through risk change prediction rules, and protocol stability evaluation values ​​are obtained through protocol stability evaluation rules. The treatment response prediction values, lesion change prediction values, risk change prediction values, and protocol stability evaluation values ​​are used as new evaluation data.

[0021] The above simulation evaluation process transforms the patient's personalized treatment plan into a set of simulation input parameters and evaluates it when the pre-set stability conditions are not met, in order to obtain new evaluation data. This process is used to supplement subsequent change information and provide input for the dynamic correction of the treatment plan.

[0022] The newly added assessment data is processed into corresponding new event fragments and written into the same event index structure based on the anchor points of the diagnosis and treatment events. This process iteratively updates cross-dimensional relationships, source traceability credibility markers, and the basis for personalized patient treatment plans until preset stability conditions are met. Specifically, this includes: The newly added assessment data is processed into events to generate new event fragments; the anchor code value of the corresponding diagnosis and treatment event anchor point in the new event fragment is read, and the new event fragment is written into the event fragment layer of the same event index structure according to the anchor code value; Write the simulation evaluation rule version identifier, evaluation generation time, evaluation writing time and field verification value corresponding to the newly added event fragment into the source field layer; After writing the new event fragment, re-execute the cross-dimensional relationship update, the source traceability credibility mark update, and the patient personalized treatment plan basis update; When re-performing the cross-dimensional association update, based on the association calculation results between the newly added event fragments and image event fragments, medical record semantic event fragments, and robot process event fragments, the feature association value, process association value, and risk association value are updated, and the cross-dimensional association is also updated; When re-executing the traceability credibility tag update, an updated traceability credibility score is generated based on the robot process traceability evaluation value and the consistency evaluation value of the source of the newly added event fragment, and an updated traceability credibility tag is generated based on the updated traceability credibility score. When re-executing the update of the patient's personalized treatment plan basis, the treatment plan basis score and the plan recommendation basis score are recalculated based on the updated cross-dimensional correlation and the updated traceability credibility mark, generating a new patient personalized treatment plan basis, and then judging again whether the new patient personalized treatment plan basis meets the preset stability conditions, until the iterative update process ends.

[0023] The above iterative update process involves writing newly added assessment data into the same event index structure after it has been event-based, and updating cross-dimensional relationships, source traceability credibility markers, and patient-specific treatment plans. This process enables new information to be continuously integrated into the existing treatment chain, thereby enhancing the continuity of the update process.

[0024] Compared with the prior art, the beneficial effects of the present invention are: 1. Based on further analysis and research on the problems of unclear event correspondence, difficulty in using the credibility of sources to participate in the calculation of the basis for the treatment plan during the integration of existing heterogeneous diagnosis and treatment data, this invention recognizes that when image data, medical record data and robot audit traceability data are used to generate the basis for personalized diagnosis and treatment plans for patients, they should not be simply aggregated in parallel according to patient identification or time order, but should form an indexable, associative, evaluable and updatable data organization chain around the same diagnosis and treatment process; To this end, this invention processes computed tomography (CT) image data, electronic medical record data, and robot audit traceability data into image event fragments, medical record semantic event fragments, and robot process event fragments, respectively. It then generates treatment event anchors based on patient treatment event association information and maps each event fragment to the same event index structure. Within this same event index structure, cross-dimensional relationships are established, and traceability credibility markers are generated by combining the audit traceability fields in the robot process event fragments. Finally, based on the cross-dimensional relationships and traceability credibility markers, a basis for personalized patient treatment plans is generated. Thus, image findings, medical record semantics, and robot execution processes can form a relatively clear correspondence under the same treatment event, and this correspondence can further include the source credibility. This allows the generation process of personalized patient treatment plans to utilize not only the multi-source treatment content itself but also the correlation strength and credibility evaluation results between the various source data, thereby improving the interpretability, traceability, and continuous updating of the plan basis.

[0025] 2. The present invention further introduces a simulation evaluation and event-based write-back update mechanism when the patient's personalized treatment plan does not meet the preset stability conditions, so that the patient's personalized treatment plan can be iteratively corrected based on the newly added evaluation data. Specifically, this invention converts the patient's current treatment stage, key imaging data, key medical record data, robotic process data, risk warning information, and treatment plan recommendation scores into a set of simulation input parameters. New assessment data is obtained through preset simulation evaluation rules. This new assessment data is then processed into new event fragments and written into the same event index structure based on treatment event anchors. Subsequently, cross-dimensional correlations, source traceability credibility markers, and patient-specific treatment plan data are updated. Thus, the new assessment data can utilize existing treatment event anchors and the same event index structure for subsequent analysis, avoiding isolated records between the new data and existing imaging event fragments, medical record semantic event fragments, and robotic process event fragments. This helps enhance the dynamic adjustment capability and result stability of patient-specific treatment plan data in continuous treatment scenarios. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall method steps of the present invention; Figure 2 This is a schematic diagram of the core process of step S1 of the present invention; Figure 3 This is a schematic diagram of the core process of step S2 of the present invention; Figure 4 This is a schematic diagram of the core process of step S3 of the present invention; Figure 5 This is a schematic diagram of the core process of step S4 of the present invention; Figure 6 This is a schematic diagram of the core process of step S5 of the present invention; Figure 7 This is a schematic diagram of the core process of step S6 of the present invention; Figure 8 This is a schematic diagram of the core process of step S7 of the present invention. Detailed Implementation

[0027] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] In this embodiment, the data objects actually used include: patient identifiers, computed tomography (CT) image data, electronic medical record data, robot audit traceability data, medical event association information, image event fragments, medical record semantic event fragments, robot process event fragments, medical event anchor points, same event index structure, cross-dimensional association relationships, traceability credibility markers, patient personalized treatment plan basis, newly added assessment data, newly added event fragments, and preset stability conditions; wherein: Patient identifiers are used to uniquely correspond to a patient's full-cycle health record; computed tomography (CT) image data is used to form image event fragments; electronic medical record (EMR) data is used to form medical record semantic event fragments; robot audit traceability data is used to form robot process event fragments; diagnostic and treatment event association information is used to generate diagnostic and treatment event anchors; diagnostic and treatment event anchors are used to write event fragments from different sources into the same event index structure; the same event index structure is used to store diagnostic and treatment event anchors, event fragments, and their index relationships; cross-dimensional association relationships are used to express the correspondence between image event fragments, medical record semantic event fragments, and robot process event fragments; traceability credibility markers are used to characterize the credibility of cross-dimensional association relationships; patient-specific treatment plan basis is used to support the generation of treatment plans; newly added evaluation data is used to generate new event fragments when preset stability conditions are not met; preset stability conditions are used to determine whether the iteration process has ended.

[0029] Next, please refer to Figure 1 The purpose of this embodiment is to provide a method for heterogeneous data integration and intelligent analysis of a patient's full-cycle health records, including the following steps: S1. Acquire the patient's computed tomography image data, electronic medical record data, and robot audit traceability data, and process them into image event fragments, medical record semantic event fragments, and robot process event fragments, respectively. S2. Based on the patient's treatment event association information, perform anchor point coding to generate treatment event anchor points; S3. Based on the diagnostic event anchor point, map the image event fragments, medical record semantic event fragments, and robot process event fragments to the same event index structure; S4. Based on the same event index structure, establish cross-dimensional association relationships between image event fragments, medical record semantic event fragments, and robot process event fragments; S5. Based on the diagnosis and treatment event anchor point, call the cross-dimensional association relationship and the corresponding robot process event fragment from the same event index structure, and evaluate the traceability credibility of the cross-dimensional association relationship based on the audit traceability field in the robot process event fragment, and generate traceability credibility tag. S6. Generate personalized treatment plans for patients based on cross-dimensional relationships and source traceability credibility markers; S7. When the patient's personalized treatment plan does not meet the preset stability conditions, the patient's personalized treatment plan is simulated and evaluated to obtain new evaluation data. The new evaluation data is processed into corresponding new event fragments and written into the same event index structure according to the treatment event anchor point. This is done to iteratively update the cross-dimensional correlation, traceability credibility mark and patient's personalized treatment plan until the preset stability conditions are met.

[0030] The specific steps are as follows: Please see Figure 2 S1. Based on the patient identifier, obtain the patient's computed tomography (CT) image data from the image archiving system, the patient's electronic medical record data from the electronic medical record system, and the patient's robot audit traceability data from the robot audit system. Specifically, the CT image data is processed into image event fragments, including: Image preprocessing is performed on computed tomography (CT) images. The preprocessing includes slice thickness normalization, spatial registration, grayscale standardization, and lesion region determination. Slice thickness normalization converts CT images with different slice thicknesses into images with a uniform slice thickness. Spatial registration registers CT images of the same patient at different stages of diagnosis and treatment to a unified spatial coordinate system. Grayscale standardization converts image grayscale values ​​into a uniform grayscale range. After grayscale standardization, lesion regions are determined from computed tomography (CT) images. This process includes grayscale threshold screening, candidate region extraction, region growing, connected component analysis, and morphological correction. First, candidate voxels are selected from the CT images based on a preset grayscale range corresponding to the examined area. Then, region growing is performed using these candidate voxels as initial regions to obtain candidate lesion regions. Next, connected component analysis is performed on the candidate lesion regions to remove isolated regions with volumes smaller than a preset volume threshold. Finally, the remaining regions undergo dilation, erosion, and void filling processes to obtain the lesion regions. Based on these lesion regions, the lesion location, volume, and boundary characteristics are calculated. Extract the image examination number, image examination time, examination site, image sequence number, and image diagnosis conclusion from computed tomography image data; An image event fragment identifier is generated based on the patient identifier, image examination number, and image sequence number; the patient identifier, image examination number, image event fragment identifier, image examination time, examination site, image sequence number, lesion location, lesion volume, lesion boundary features, and image diagnosis conclusion are encapsulated into an image event fragment.

[0031] Electronic medical record data is processed into medical record semantic event fragments, specifically including: Read admission records, progress notes, examination and test records, medical orders, surgical records, discharge records, and follow-up records from electronic medical records; The text content in the electronic medical record data is sequentially segmented into words, recognized by entities, identified by time, and identified by treatment behavior. Entity recognition extracts symptom names, diagnosis names, medication names, and examination items; time recognition extracts the record time; and treatment behavior is extracted through treatment behavior recognition. Simultaneously, the recording department and medical record number are read. Entity recognition and treatment behavior recognition are performed using a pre-set medical terminology dictionary, medical thesaurus, symptom coding table, diagnosis coding table, medication coding table, examination item coding table, and treatment behavior coding table. When words in the electronic medical record text match standard terms in the aforementioned dictionaries or coding tables, the matched standard term is used as the corresponding entity or treatment behavior. When words in the electronic medical record text match synonyms in the medical thesaurus, the synonym is mapped to the corresponding standard term. The semantic content belonging to the same medical record number, the same record time, and the same treatment behavior is merged to generate a medical record semantic event fragment; Generate a medical record semantic event fragment identifier based on the patient identifier, medical record number, and record time; encapsulate the patient identifier, medical record number, medical record semantic event fragment identifier, record time, record department, diagnosis name, symptom description, examination items, treatment behavior, and medication information into a medical record semantic event fragment.

[0032] The data from robot auditing and tracing is processed into robot process event fragments, specifically including: Read the robot task number, operation time, operator ID, device number, executed actions, parameter settings, system feedback, data write records and audit signature from the robot audit traceability data; Robot audit traceability data is collected according to robot task number, and sorted according to operation time under the same robot task number; a group of actions executed consecutively under the same robot task number is recorded as a robot process event. Generate robot process event fragment identifiers based on patient identifier, robot task number, and operation time; encapsulate patient identifier, robot process event fragment identifier, robot task number, operation time, operator identifier, device number, executed action, parameter settings, system feedback, data write record, and audit signature into a robot process event fragment.

[0033] Please see Figure 3 S2. Based on the patient's treatment event association information, anchor point coding is performed to generate treatment event anchor points, specifically including: Obtain patient treatment event association information, which includes patient identifier, treatment stage number, treatment stage start time, treatment stage end time, treatment time, treatment department, treatment behavior type, imaging examination number, medical record number, and robot task number; The patient identifier, treatment stage number, treatment time, treatment department, and treatment behavior type are concatenated in a fixed order and according to field separators to form the original anchor string; the fixed order is patient identifier, treatment stage number, treatment time, treatment department, and treatment behavior type.

[0034] The SHA-256 algorithm is used to calculate the digest value of the original anchor string, and the calculated digest value is used as the anchor code value. The SHA-256 algorithm is a secure hash algorithm used to calculate a fixed-length digest value of the original anchor string, so that the same input will produce the same anchor code value and different inputs will hardly produce the same anchor code value. The anchor point encoding value is bound to the image examination number, medical record number and robot task number to generate the diagnosis and treatment event anchor point; Each treatment event anchor includes patient identifier, treatment stage number, treatment stage start time, treatment stage end time, treatment time, treatment department, treatment behavior type, anchor code value, imaging examination number, medical record number, and robot task number.

[0035] Please see Figure 4 S3. Based on the diagnostic event anchor points, map image event fragments, medical record semantic event fragments, and robot process event fragments to the same event index structure, specifically including: A unified event index structure is established, comprising an anchor layer, an event fragment layer, a relationship layer, and a source field layer. The anchor layer stores diagnostic event anchors; the event fragment layer stores image event fragments, medical record semantic event fragments, robot process event fragments, and newly added event fragments; the relationship layer stores cross-dimensional relationships and source credibility markers; and the source field layer stores the source system identifier, data generation time, data writing time, and field validation value for each event fragment.

[0036] The mapping process specifically includes: First, the diagnosis and treatment event anchors are written into the anchor layer, and the anchor code value is used as the main index in the anchor layer; Secondly, based on the image examination number in the diagnosis and treatment event anchor point, search for image event segments with the same image examination number from the image event segment set; when the same image examination number corresponds to multiple image event segments, write the multiple image event segments into the event segment layer in ascending order of image sequence number, and form an image event segment identifier list under the diagnosis and treatment event anchor point; Next, based on the medical record number in the diagnosis and treatment event anchor, search for medical record semantic event fragments with the same medical record number from the medical record semantic event fragment set; when the same medical record number corresponds to multiple medical record semantic event fragments, write the multiple medical record semantic event fragments into the event fragment layer in ascending order of record time, and form a medical record semantic event fragment identifier list under the diagnosis and treatment event anchor. Then, based on the robot task number in the diagnosis and treatment event anchor point, search for robot process event segments with the same robot task number from the robot process event segment set; when the same robot task number corresponds to multiple robot process event segments, write the multiple robot process event segments into the event segment layer in ascending order of operation time, and form a robot process event segment identifier list under the diagnosis and treatment event anchor point; Finally, the source system identifier, data generation time, data writing time, and field verification value corresponding to the image event fragment, medical record semantic event fragment, and robot process event fragment are written into the source field layer respectively; After mapping is completed, corresponding image event fragments, medical record semantic event fragments, and robot process event fragments are saved under the same diagnosis and treatment event anchor point.

[0037] Please see Figure 5 S4. Based on the same event index structure, establish cross-dimensional associations between image event fragments, medical record semantic event fragments, and robot process event fragments. The process of establishing cross-dimensional associations specifically includes: Based on the same event index structure, read image event fragments, medical record semantic event fragments, and robot process event fragments under the same diagnosis and treatment event anchor point; Extract lesion location, lesion volume, lesion boundary features, and imaging diagnosis conclusions from image event fragments; extract diagnosis name, symptom description, examination items, treatment behavior, and medication information from medical record semantic event fragments; extract executed actions, parameter settings, system feedback, and data writing records from robot process event fragments; The imaging diagnostic conclusions and diagnostic names are standardized using a pre-set medical terminology dictionary and a medical thesaurus; the standardized imaging diagnostic conclusions are converted into a set of imaging diagnostic terms, and the standardized diagnostic names are converted into a set of medical record diagnostic terms. Based on the preset medical terminology coding table, the set of image diagnosis terms is converted into an image diagnosis coding vector, and the set of medical record diagnosis terms is converted into a medical record diagnosis coding vector; each dimension in the coding vector corresponds to a preset medical term; if the corresponding medical term appears in the diagnosis text, the value of that dimension is 1; if the corresponding medical term does not appear in the diagnosis text, the value of that dimension is 0. Calculate the cosine similarity between the image diagnostic coding vector and the medical record diagnostic coding vector, and use the calculation result as the diagnostic association value; The location, volume, and boundary features of the lesion are normalized to form an image feature vector; After standardizing the terminology of symptom descriptions and examination items, they are converted into medical record feature vectors based on preset symptom coding tables and examination item coding tables. Each dimension of the medical record feature vector corresponds to a preset symptom term or examination item term. If the corresponding term appears in the symptom description or examination item, the value of that dimension is 1; if the corresponding term does not appear in the symptom description or examination item, the value of that dimension is 0. Based on the lesion location, lesion volume, and lesion boundary features in the image feature vector, and the symptom terms and examination item terms in the medical record feature vector, the similarity of lesion location, lesion volume, lesion boundary features, symptom matching, and examination item matching is calculated respectively. A weighted sum is then performed on these similarities, and the weighted sum is used as the feature association value. The feature association value is calculated according to the following formula: Feature association value = lesion location similarity × first feature weight + lesion volume similarity × second feature weight + lesion boundary feature similarity × third feature weight + symptom matching similarity × fourth feature weight + examination item matching similarity × fifth feature weight; where the sum of the first feature weight, second feature weight, third feature weight, fourth feature weight and fifth feature weight is 1; The treatment behaviors in the medical record semantic event fragments are converted into treatment behavior codes, and the execution actions in the robot process event fragments are converted into robot action codes. According to the preset behavior action mapping table, the treatment behavior codes and robot action codes are matched. When there is a correspondence between the treatment behavior codes and robot action codes in the preset behavior action mapping table, the process association value is 1; when there is no correspondence between the treatment behavior codes and robot action codes in the preset behavior action mapping table, the process association value is 0. Based on anchor point coding values, image event fragment identifiers, medical record semantic event fragment identifiers, robot process event fragment identifiers, diagnostic association values, feature association values, and process association values, cross-dimensional association relationships are generated. Write cross-dimensional relationships into the relationship layer of the same event index structure, and record the cross-dimensional relationship identifier under the corresponding diagnosis and treatment event anchor point.

[0038] Please see Figure 6 S5. Based on the diagnosis and treatment event anchor point, retrieve the cross-dimensional relationship and corresponding robot process event fragment from the same event index structure, and evaluate the traceability credibility of the cross-dimensional relationship based on the audit traceability field in the robot process event fragment, generating a traceability credibility tag. The specific retrieval process includes: Based on the anchor code value in the current diagnosis and treatment event anchor, locate the target anchor record in the anchor layer of the same event index structure; Read the fragment index set of the target anchor point recorded in the event fragment layer; the fragment index set includes image event fragment identifiers, medical record semantic event fragment identifiers, and robot process event fragment identifiers; Extract robot process event fragment identifiers from the fragment index set, and read the corresponding robot process event fragments based on the robot process event fragment identifiers; Extract medical record semantic event fragment identifiers from the fragment index set, and read the corresponding medical record semantic event fragments based on the medical record semantic event fragment identifiers; Read the cross-dimensional relationship identifier from the target anchor point record, and read the corresponding cross-dimensional relationship from the relationship layer based on the cross-dimensional relationship identifier; Perform consistency verification between the robot task number in the robot process event fragment and the robot task number in the diagnosis and treatment event anchor point; perform consistency verification between the anchor point code value in the cross-dimensional association relationship and the anchor point code value in the diagnosis and treatment event anchor point; When the robot task number and anchor code value are consistent, the robot process event fragments, medical record semantic event fragments and cross-dimensional correlations read will be input into the source traceability credibility evaluation process. When the robot task number in the robot process event segment is inconsistent with the robot task number in the diagnosis and treatment event anchor point, the source traceability credibility flag of the corresponding cross-dimensional association is set to low credibility, and the source traceability credibility flag is written into the association layer. When the anchor code value in the cross-dimensional relationship is inconsistent with the anchor code value in the diagnosis and treatment event anchor, the source traceability credibility mark of the corresponding cross-dimensional relationship is set to low credibility, and the source traceability credibility mark is written into the relationship layer. The process of generating traceability credibility markers specifically includes: Audit traceability fields are read from robot process event fragments. These fields include robot task number, operation time, operator identifier, device number, executed action, parameter settings, system feedback, data write records, and audit signature. The credibility of these audit traceability fields is evaluated, specifically including: First, verify whether the robot task number is consistent with the robot task number in the diagnosis and treatment event anchor point to obtain the task consistency evaluation value; when the robot task numbers are consistent, the task consistency evaluation value is 1; when the robot task numbers are inconsistent, the task consistency evaluation value is 0. Second, verify whether the operation time is between the start time and end time of the treatment phase in the treatment event anchor point to obtain the time consistency evaluation value; when the operation time is between the start time and end time of the treatment phase, the time consistency evaluation value is 1; when the operation time is not between the start time and end time of the treatment phase, the time consistency evaluation value is 0. Third, the behavior consistency evaluation value is obtained by verifying whether the executed action corresponds to the treatment behavior in the semantic event fragment of the medical record according to the preset behavior action mapping table; when there is a correspondence between the executed action and the treatment behavior, the behavior consistency evaluation value is 1; when there is no correspondence between the executed action and the treatment behavior, the behavior consistency evaluation value is 0. Fourth, verify whether the data write record is consistent with the source field layer in the same event index structure to obtain the write consistency evaluation value; when the write time, write object, and field verification value in the data write record are consistent with the source field layer, the write consistency evaluation value is 1; when the write time, write object, and field verification value in the data write record are inconsistent with the source field layer, the write consistency evaluation value is 0. Fifth, read the pre-stored device certificate based on the device number in the robot process event segment, and use the public key in the pre-stored device certificate to verify the audit signature to obtain the signature integrity evaluation value; when the audit signature passes the signature verification, the signature integrity evaluation value is 1; when the audit signature fails the signature verification, the signature integrity evaluation value is 0. The source tracing credibility score is calculated using the following formula: The traceability credibility score = task consistency evaluation value × first credibility weight + time consistency evaluation value × second credibility weight + behavior consistency evaluation value × third credibility weight + write consistency evaluation value × fourth credibility weight + signature integrity evaluation value × fifth credibility weight; where the sum of the first credibility weight, second credibility weight, third credibility weight, fourth credibility weight and fifth credibility weight is 1; When the traceability credibility score is greater than or equal to the first credibility threshold, a high-credibility traceability credibility tag is generated; when the traceability credibility score is less than the first credibility threshold but greater than or equal to the second credibility threshold, a medium-credibility traceability credibility tag is generated; when the traceability credibility score is less than the second credibility threshold, a low-credibility traceability credibility tag is generated. The generated traceability credibility tag is written into the association layer of the same event index structure, so that the traceability credibility tag is bound to the corresponding cross-dimensional association identifier.

[0039] Please see Figure 7 S6. Based on cross-dimensional correlations and source traceability credibility markers, generate personalized treatment plans for patients, specifically including: Read the image event fragments, medical record semantic event fragments, robot process event fragments, cross-dimensional relationships, and source traceability credibility markers corresponding to each diagnosis and treatment event anchor point under the same patient identifier; The anchor points of each treatment event are sorted according to the treatment time to generate a patient treatment time sequence; Based on the patient's diagnosis and treatment time series, the trends of lesion volume change, symptom change, diagnosis change, treatment behavior change, and robot action feedback were calculated. Specifically, the lesion volume change trend was calculated based on the difference in lesion volume corresponding to adjacent diagnosis and treatment event anchor points; the symptom change trend was calculated based on the change in symptom description corresponding to adjacent diagnosis and treatment event anchor points; the diagnosis change trend was calculated based on the change in diagnosis name corresponding to adjacent diagnosis and treatment event anchor points; the treatment behavior change trend was calculated based on the change in treatment behavior corresponding to adjacent diagnosis and treatment event anchor points; and the robot action feedback trend was calculated based on the change in system feedback corresponding to adjacent diagnosis and treatment event anchor points. The trend of lesion volume change is converted into a lesion volume change trend score; when the lesion volume decreases, the lesion volume change trend score is positive; when the lesion volume increases, the lesion volume change trend score is negative; when the lesion volume remains unchanged, the lesion volume change trend score is 0. The symptom change trend is converted into a symptom change trend score; when the number of symptoms decreases, the symptom change trend score is positive; when the number of symptoms increases, the symptom change trend score is negative; when the number of symptoms remains unchanged, the symptom change trend score is 0. The trend of changes in treatment behavior is converted into a treatment behavior change trend score; when the treatment behavior is consistent with the preset standard treatment path, the treatment behavior change trend score is positive; when the treatment behavior is inconsistent with the preset standard treatment path, the treatment behavior change trend score is negative. The robot's action feedback trend is converted into a robot action feedback trend score; when the system feedback indicates successful execution or improved feedback, the robot action feedback trend score is positive; when the system feedback indicates abnormal execution or worsened feedback, the robot action feedback trend score is negative; when the system feedback shows no significant change, the robot action feedback trend score is 0. The scores for lesion volume change trend, symptom change trend, and treatment behavior change trend were all normalized to the interval [-1, 1]; where a positive value indicates that the trend is favorable to the current treatment plan, a negative value indicates that the trend is unfavorable to the current treatment plan, and 0 indicates that there is no significant change in the trend. The robot's action feedback trend score was normalized to the range of [-1, 1]; where a positive value indicates that the robot's action feedback trend is favorable to the current treatment plan, a negative value indicates that the robot's action feedback trend is unfavorable to the current treatment plan, and 0 indicates that the robot's action feedback trend has no significant change. The first scheme weight is assigned to cross-dimensional relationships with high credibility traceability labels, the second scheme weight is assigned to cross-dimensional relationships with medium credibility traceability labels, and the third scheme weight is assigned to cross-dimensional relationships with low credibility traceability labels; the weight of the first scheme is greater than the weight of the second scheme, and the weight of the second scheme is greater than the weight of the third scheme. The treatment plan is based on the baseline score without source traceability and credibility correction, while the recommended plan is based on the final score after source traceability and credibility correction. The treatment plan basis score is calculated based on the diagnostic correlation value, characteristic correlation value, process correlation value, lesion volume change trend score, symptom change trend score, and treatment behavior change trend score; the treatment plan basis score is calculated according to the following formula: The treatment plan score is calculated as follows: Diagnostic correlation value × weight of the fourth plan + feature correlation value × weight of the fifth plan + process correlation value × weight of the sixth plan + lesion volume change trend score × weight of the seventh plan + symptom change trend score × weight of the eighth plan + treatment behavior change trend score × weight of the ninth plan; where the sum of the weights of the fourth, fifth, sixth, seventh, eighth and ninth plans is 1. When the source tracing credibility is marked as high credibility, the scheme recommendation basis score = treatment scheme basis score × weight of the first scheme; When the source tracing credibility is marked as medium credibility, the recommended treatment plan basis score = treatment plan basis score × weight of the second plan; When the source tracing credibility is marked as low credibility, the solution recommendation basis score = treatment plan basis score × weight of the third plan; Based on the recommended criteria score, a personalized treatment plan is generated for the patient. This personalized treatment plan includes the patient's current treatment stage, key imaging data, key medical records, robot process data, traceability credibility markers, risk warning information, and the recommended criteria score. Key imaging data is generated from lesion location, lesion volume, lesion boundary features, and imaging diagnostic conclusions. Key medical records are generated from diagnosis name, symptom description, examination items, treatment actions, and medication information. Robot process data is generated from executed actions, parameter settings, system feedback, and data writing records. Risk warning information is generated from lesion volume change trends, symptom change trends, robot action feedback trends, and traceability credibility markers.

[0040] Please see Figure 8S7. When the basis of the patient's personalized treatment plan does not meet the preset stability conditions, a simulation evaluation is performed on the basis of the patient's personalized treatment plan to obtain new evaluation data. The new evaluation data is processed into corresponding new event fragments and written into the same event index structure according to the treatment event anchor point. This is used to iteratively update the cross-dimensional correlation, traceability credibility marker, and patient's personalized treatment plan basis until the preset stability conditions are met. The preset stability conditions are: The recommended treatment plan based on the score change value obtained in three consecutive iterations is less than the preset score change threshold, and the traceability credibility marker obtained in three consecutive iterations remains consistent; when the number of iterations is less than three, it is determined that the patient's personalized treatment plan basis does not meet the preset stability condition.

[0041] When the patient's personalized treatment plan fails to meet the preset stability conditions, a simulation evaluation is performed on the patient's personalized treatment plan. The simulation evaluation process specifically includes: Convert the patient's current diagnosis and treatment stage into a stage code, convert the main imaging evidence into imaging status parameters, convert the main medical record evidence into clinical status parameters, convert the robot process evidence into execution process parameters, and convert risk warning information into risk status parameters. The simulation input parameter set is composed of stage codes, imaging status parameters, clinical status parameters, execution process parameters, risk status parameters, and scheme recommendation basis scores in a fixed order; The simulation input parameter set is evaluated using pre-defined simulation evaluation rules, which include rules for predicting treatment response, lesion changes, risk changes, and protocol stability. The treatment response prediction rule calculates the predicted treatment response value based on diagnostic correlation value, characteristic correlation value, process correlation value, treatment behavior change trend score, and protocol recommendation basis score; the predicted treatment response value is calculated according to the following formula: Treatment response prediction value = Diagnostic correlation value × First response weight + Feature correlation value × Second response weight + Process correlation value × Third response weight + Treatment behavior change trend score × Fourth response weight + Protocol recommendation basis score × Fifth response weight, where the sum of the first response weight, second response weight, third response weight, fourth response weight and fifth response weight is 1; The treatment response prediction value is used to represent the degree to which the patient will have an effective response to the current treatment plan.

[0042] The lesion change prediction rule calculates the predicted lesion change value based on the current lesion volume, the trend of lesion volume change between adjacent diagnostic event anchor points, lesion boundary features, and main imaging evidence. The normalized value of the current lesion volume change is obtained by normalizing the difference between the current lesion volume and the lesion volume corresponding to the previous diagnostic event anchor point. The lesion boundary feature score is assigned based on the regularity of the lesion boundary, the clarity of the boundary, and the state of boundary expansion. The main imaging evidence score is assigned based on the trend of lesion volume change and the consistency of the imaging diagnosis conclusion in the main imaging evidence. The predicted lesion change value is calculated according to the following formula: The predicted value of lesion change is calculated as follows: Normalized value of current lesion volume change × Weight of the first lesion + Score of lesion volume change trend × Weight of the second lesion + Score of lesion boundary feature × Weight of the third lesion + Score of main imaging evidence × Weight of the fourth lesion, where the sum of the weights of the first, second, third, and fourth lesions is 1; the threshold for the first lesion is greater than the threshold for the second lesion; a predicted value of lesion change greater than the threshold for the first lesion indicates that the volume of the subsequent lesion decreases, a predicted value of lesion change less than the threshold for the second lesion indicates that the volume of the subsequent lesion increases, and a predicted value of lesion change less than or equal to the threshold for the first lesion and greater than or equal to the threshold for the second lesion indicates that the volume of the subsequent lesion remains stable; The risk change prediction rule calculates the predicted risk change value based on the symptom change trend score, the robot's action feedback trend score, the source tracing credibility marker, and the risk warning information. The robot's action feedback trend score is assigned based on the system feedback changes corresponding to adjacent diagnostic and treatment event anchor points. The source tracing credibility marker score is assigned a preset credibility score according to high credibility, medium credibility, and low credibility. The risk warning information score is assigned based on the number of risk items in the risk warning information regarding lesion volume change trends, symptom change trends, and robot action feedback trends. The predicted risk change value is calculated using the following formula: The predicted risk change value is calculated as follows: Symptom Change Trend Score × First Risk Weight + Robot Execution Action Feedback Trend Score × Second Risk Weight + Source Tracing Credibility Marker Score × Third Risk Weight + Risk Warning Information Score × Fourth Risk Weight, where the sum of the first, second, third, and fourth risk weights is 1. The first risk threshold is greater than the second risk threshold. A predicted risk change value greater than the first risk threshold indicates a reduced treatment risk; a predicted risk change value less than the second risk threshold indicates an increased treatment risk; and a predicted risk change value less than or equal to the first risk threshold and greater than or equal to the second risk threshold indicates a stable treatment risk.

[0043] The scheme stability evaluation rule calculates the scheme stability evaluation value based on the changes in the scheme recommendation criteria score, the changes in the traceability credibility marker, and the predicted risk changes during continuous iterations; the scheme stability evaluation value is calculated according to the following formula: The stability evaluation value of the treatment plan is calculated as follows: Stable score change value × First stability weight + Stable source traceability credibility marker value × Second stability weight + Stable risk change value × Third stability weight, where the sum of the first, second, and third stability weights is 1. The stable score change value is determined based on the comparison between the change value of the recommended treatment plan score and the preset score change threshold. It is set to 1 when the change value of the recommended treatment plan score is less than the preset score change threshold, and 0 otherwise. The stable source traceability credibility marker value is determined based on whether the source traceability credibility markers are consistent in consecutive iterations. It is set to 1 when consistent and 0 when inconsistent. The stable risk change value is determined based on whether the predicted risk change value is between the first and second risk thresholds. It is set to 1 when it is in this range, and 0 otherwise. The stability evaluation value of the treatment plan is used to represent the stability of the current patient's personalized treatment plan during the iteration process.

[0044] Treatment response prediction, lesion change prediction, risk change prediction, and protocol stability evaluation value were added as new evaluation data.

[0045] The newly added assessment data is processed into events to generate new event fragments. The new event fragments include patient identifier, new event fragment identifier, simulation assessment time, treatment response prediction value, lesion change prediction value, risk change prediction value, protocol stability evaluation value, and corresponding diagnosis and treatment event anchor point. The iterative update process specifically includes: Based on the corresponding diagnosis and treatment event anchor point in the newly added event fragment, read the anchor point code value of the diagnosis and treatment event anchor point; Based on the anchor point encoding value, the new event fragment is written into the event fragment layer of the same event index structure, and the new event fragment identifier is recorded under the diagnosis and treatment event anchor point; Write the simulation evaluation rule version identifier, evaluation generation time, evaluation writing time and field verification value corresponding to the newly added event fragment into the source field layer; After writing the new event fragment, re-execute the cross-dimensional relationship update, the source traceability credibility mark update, and the patient personalized treatment plan basis update; When re-executing the cross-dimensional relationship update, read the image event fragments, medical record semantic event fragments, robot process event fragments, and newly added event fragments under the same diagnosis and treatment event anchor point; The predicted values ​​of lesion changes in newly added event segments are correlated with the lesion volume and lesion boundary features in the image event segments, and the feature correlation values ​​are updated. The predicted value of treatment response in the newly added event fragment is correlated with the treatment behavior in the medical record semantic event fragment and the execution action in the robot process event fragment, and the process correlation value is updated. The risk change prediction value in the newly added event segment is correlated with the symptom description in the medical record semantic event segment and the system feedback in the robot process event segment to generate a risk correlation value. The risk warning information and the score of the plan recommendation basis in the patient's personalized diagnosis and treatment plan are updated according to the risk correlation value. The above update process is performed using a weighted fusion method. The updated feature correlation value, updated process correlation value, and risk correlation value are obtained by weighting the original correlation value and the corresponding newly added assessment data according to the preset fusion weights, and the sum of the preset fusion weights for each group is 1. Update the cross-dimensional relationships based on the updated feature association values, process association values, and risk association values; When re-executing the traceability credibility mark update, the audit traceability field in the corresponding robot process event segment is called, and the simulation evaluation rule version identifier, evaluation generation time, evaluation writing time and field verification value in the source field layer of the newly added event segment are called; Perform robot process traceability evaluation on robot process event fragments to obtain robot process traceability evaluation values. The robot process traceability evaluation values ​​are obtained according to the calculation method of traceability credibility score in S5, that is, based on the task consistency evaluation value, time consistency evaluation value, behavior consistency evaluation value, write consistency evaluation value and signature integrity evaluation value and the corresponding credibility weights.

[0046] A source consistency evaluation is performed on the newly added event fragment to obtain the source consistency evaluation value of the newly added event fragment; the specific process of the source consistency evaluation of the newly added event fragment includes: Verify whether the simulation evaluation rule version identifier is consistent with the pre-registered rule version identifier, verify whether the evaluation generation time is earlier than the evaluation writing time, and verify whether the field verification value is consistent with the verification value recalculated by the newly added event fragment; when all three verifications are consistent, the consistency evaluation value of the source of the newly added event fragment is 1; when there is an inconsistency among the three verifications, the consistency evaluation value of the source of the newly added event fragment is 0. The updated source tracing credibility score is calculated using the following formula, specifically including: The updated traceability credibility score = robot process traceability evaluation value × first update weight + new event fragment source consistency evaluation value × second update weight; where the sum of the first update weight and the second update weight is 1; An updated source tracing credibility tag is generated based on the updated source tracing credibility score, and the updated source tracing credibility tag is written into the association layer of the same event index structure; When re-executing the update of the patient's personalized treatment plan, the treatment plan basis score and the plan recommendation basis score are recalculated based on the updated cross-dimensional correlation and the updated traceability credibility mark, and a new patient personalized treatment plan basis is generated. The process involves reassessing whether the new personalized treatment plan for patients meets the preset stability conditions; if the new personalized treatment plan for patients does not meet the preset stability conditions, the process continues to include simulation evaluation, generating new evaluation data, generating new event fragments, writing them into the same event index structure, updating cross-dimensional relationships, updating source traceability credibility markers, and updating the basis of the personalized treatment plan for patients. When the change in the recommended treatment plan score obtained in three consecutive iterations is less than the preset score change threshold, and the traceability credibility marker obtained in three consecutive iterations remains consistent, it is determined that the patient's personalized treatment plan meets the preset stability condition, and the iteration update process ends.

[0047] In this embodiment, the first confidence threshold, the second confidence threshold, the preset score change threshold, the first lesion threshold, the second lesion threshold, the first risk threshold, and the second risk threshold are all preset threshold parameters. Specifically, the first confidence threshold and the second confidence threshold both range from [0, 1], and the first confidence threshold is greater than the second confidence threshold. The preset score change threshold ranges from [0, 1]. The first lesion threshold and the second lesion threshold both range from [-1, 1], and the first lesion threshold is greater than the second lesion threshold. The first risk threshold and the second risk threshold both range from [-1, 1], and the first risk threshold is greater than... Second risk threshold; First credibility weight, second credibility weight, third credibility weight, fourth credibility weight, fifth credibility weight, first feature weight, second feature weight, third feature weight, fourth feature weight, fifth feature weight, first option weight, second option weight, third option weight, fourth option weight, fifth option weight, sixth option weight, seventh option weight, eighth option weight, ninth option weight, first response weight, second response weight, third response weight, fourth response weight, fifth response weight, first lesion weight, second lesion weight, third lesion weight, fourth lesion weight, first risk weight, second risk The weights, third risk weight, fourth risk weight, first stability weight, second stability weight, third stability weight, first update weight, second update weight, and preset fusion weight are all pre-set coefficient parameters, with values ​​ranging from [0, 1]. Specifically, the sum of all confidence weights in the same formula is 1; the sum of all feature weights in the same formula is 1; the sum of the weights of the fourth, fifth, sixth, seventh, eighth, and ninth schemes is 1; the sum of all response weights in the same formula is 1; the sum of all lesion weights in the same formula is 1; the sum of all risk weights in the same formula is 1; and the sum of all other weights in the same formula is 1. The sum of all stable weights is 1, the sum of the first update weight and the second update weight is 1, and the sum of the preset fusion weights for each group is 1. The weights of the first scheme, the second scheme, and the third scheme are used to correct the treatment plan based on the score according to the traceability credibility mark, and the weight of the first scheme is greater than the weight of the second scheme, and the weight of the second scheme is greater than the weight of the third scheme. The above threshold parameters and coefficient parameters are determined based on historical patient full-cycle health record samples, preset standard treatment paths, medical expert annotation results, robot audit traceability data verification results, and medical institution management rules, and are written into the parameter configuration table during system initialization, and are read directly from the parameter configuration table during calculation.

[0048] To further illustrate the working process of this embodiment, a specific operation process is given below: First, in step S1, the patient's computed tomography (CT) image data, electronic medical record data, and robot audit traceability data are obtained from the image archiving system, electronic medical record system, and robot audit traceability system, respectively, based on the patient's identifier. The CT image data is processed into image event fragments, the electronic medical record data is processed into medical record semantic event fragments, and the robot audit traceability data is processed into robot process event fragments. Then, in step S2, anchor point encoding is performed based on the patient's treatment event association information to generate treatment event anchor points; Next, in step S3, based on the diagnosis and treatment event anchor point, the image event fragment, medical record semantic event fragment, and robot process event fragment are mapped to the same event index structure, so that the corresponding image event fragment, medical record semantic event fragment, and robot process event fragment are stored under the same diagnosis and treatment event anchor point. Subsequently, in step S4, based on the same event index structure, image event fragments, medical record semantic event fragments, and robot process event fragments under the same diagnosis and treatment event anchor point are read, diagnostic correlation values, feature correlation values, and process correlation values ​​are calculated, and cross-dimensional correlation relationships are established between image event fragments, medical record semantic event fragments, and robot process event fragments. Subsequently, in step S5, based on the diagnosis and treatment event anchor point, the cross-dimensional association relationship and the corresponding robot process event fragment are called from the same event index structure, and the traceability credibility of the cross-dimensional association relationship is evaluated based on the audit traceability field in the robot process event fragment, and a traceability credibility tag is generated. Next, in step S6, based on the cross-dimensional correlation and source traceability credibility marker, combined with the lesion volume change trend score, symptom change trend score, treatment behavior change trend score, diagnostic correlation value, feature correlation value and process correlation value, the treatment plan basis score and the plan recommendation basis score are calculated, and the patient's personalized treatment plan basis is generated. Finally, in step S7, when the patient's personalized treatment plan does not meet the preset stability conditions, the patient's personalized treatment plan is simulated and evaluated to obtain new evaluation data. The new evaluation data is processed into corresponding new event fragments and written into the same event index structure according to the treatment event anchor point. This process iteratively updates the cross-dimensional correlation, traceability credibility marker, and patient's personalized treatment plan until the change value of the plan recommendation basis score obtained in three consecutive iterations is less than the preset score change threshold and the traceability credibility marker obtained in three consecutive iterations remains consistent. At this point, it is determined that the patient's personalized treatment plan meets the preset stability conditions, and the iterative update process ends.

[0049] As can be seen from the above description, the heterogeneous data integration and intelligent analysis method for a patient's full-cycle health record provided in this embodiment has the following technical effects: This embodiment provides a heterogeneous data integration and intelligent analysis method for a patient's full-cycle health record. It does not simply aggregate data from image archiving systems, electronic medical record systems, and robot auditing systems. Instead, it uses event-based processing of image event fragments, medical record semantic event fragments, and robot process event fragments, combined with diagnostic and treatment event anchors and a common event index structure, to enable computed tomography image data, electronic medical record data, and robot audit traceability data to form an indexable data organization chain around the same diagnostic and treatment process. Based on this, cross-dimensional relationships are established through diagnostic correlation values, feature correlation values, and process correlation values, and the audit traceability in the robot process event fragments is utilized. The method generates traceability credibility markers for fields, enabling personalized treatment plans for patients to simultaneously reflect the correlation strength and source credibility of multi-source data. Furthermore, when the personalized treatment plan for patients does not meet the preset stability conditions, new evaluation data is obtained through simulation evaluation. The new evaluation data is processed into corresponding new event fragments and written into the same event index structure to continuously iterate and update cross-dimensional correlations, traceability credibility markers, and personalized treatment plans for patients. This allows the method to achieve the correlation integration, credibility evaluation, and continuous updating of heterogeneous treatment data, improving the interpretability, traceability, dynamic adjustment capability, and result stability of personalized treatment plans for patients.

[0050] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for heterogeneous data integration and intelligent analysis of a patient's full-cycle health records, characterized in that, The methods and steps include the following: S1. Acquire the patient's computed tomography image data, electronic medical record data, and robot audit traceability data, and process them into image event fragments, medical record semantic event fragments, and robot process event fragments, respectively. S2. Based on the patient's treatment event association information, perform anchor point coding to generate treatment event anchor points; S3. Based on the diagnosis and treatment event anchor point, map the image event fragment, the medical record semantic event fragment, and the robot process event fragment to the same event index structure; S4. Based on the same event index structure, establish a cross-dimensional association between the image event fragment, the medical record semantic event fragment, and the robot process event fragment; S5. Based on the diagnosis and treatment event anchor point, retrieve the cross-dimensional association relationship and the corresponding robot process event fragment from the same event index structure, and evaluate the traceability credibility of the cross-dimensional association relationship based on the audit traceability field in the robot process event fragment, and generate a traceability credibility tag. S6. Generate a patient's personalized treatment plan based on the cross-dimensional correlation and the source traceability credibility marker; S7. When the patient's personalized treatment plan basis does not meet the preset stability conditions, the patient's personalized treatment plan basis is simulated and evaluated to obtain new evaluation data. The new evaluation data is processed into corresponding new event fragments and written into the same event index structure according to the treatment event anchor point, so as to iteratively update the cross-dimensional association, the source traceability credibility mark and the patient's personalized treatment plan basis until the preset stability conditions are met.

2. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 1, characterized in that, The processing procedure in step S1 specifically includes: Based on the patient identifier, the computed tomography (CT) image data, the electronic medical record data, and the robot auditing traceability data are obtained from the image archiving system, the electronic medical record system, and the robot auditing traceability system, respectively. The CT image data is preprocessed, and the lesion location, lesion volume, and lesion boundary characteristics are obtained by determining the lesion region. Then, the lesion is combined with the image examination number, image examination time, examination site, image sequence number, and image diagnosis conclusion to encapsulate the image event fragment. The system reads the admission records, progress notes, examination and test records, medical orders, surgical records, discharge records, and follow-up records from the electronic medical record data. It then performs word segmentation, entity recognition, time recognition, and treatment behavior recognition on the text content and merges the semantic content of the same medical record number, the same record time, and the same treatment behavior into the medical record semantic event fragment. Read the robot task number, operation time, operator identifier, device number, executed action, parameter settings, system feedback, data write record and audit signature from the robot audit traceability data, collect them according to the robot task number and sort them according to the operation time, and then encapsulate them into the robot process event fragment.

3. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 1, characterized in that, The process of generating the diagnostic event anchor points specifically includes: Obtain the patient's diagnosis and treatment event association information, which includes patient identifier, diagnosis and treatment stage number, diagnosis and treatment stage start time, diagnosis and treatment stage end time, diagnosis and treatment time, diagnosis and treatment department, diagnosis and treatment behavior type, imaging examination number, medical record number, and robot task number; The patient identifier, the treatment stage number, the treatment time, the treatment department, and the treatment behavior type are concatenated in a fixed order and according to field separators to form the original anchor string; A digest value is calculated for the original string of the anchor point, and the calculated digest value is used as the anchor point encoding value. The anchor point encoding value is then bound to the image examination number, the medical record number, and the robot task number to generate the diagnosis and treatment event anchor point.

4. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 1, characterized in that, The mapping process in step S3 specifically includes: Establish the same event index structure, including the anchor layer, event fragment layer, association layer, and source field layer; The diagnosis and treatment event anchor points are written into the anchor point layer, and the anchor point encoding value is used as the main index in the anchor point layer; Based on the image examination number, medical record number, and robot task number in the diagnosis and treatment event anchor point, the corresponding image event fragment, medical record semantic event fragment, and robot process event fragment are found respectively, and written into the event fragment layer; Under the aforementioned diagnostic and treatment event anchor point, a list of image event fragment identifiers, a list of medical record semantic event fragment identifiers, and a list of robot process event fragment identifiers are generated; Write the source system identifier, data generation time, data writing time, and field verification value corresponding to each event fragment into the source field layer.

5. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 4, characterized in that, The process of establishing the cross-dimensional relationship specifically includes: Based on the same event index structure, the image event fragment, the medical record semantic event fragment, and the robot process event fragment under the same diagnosis and treatment event anchor point are read; The lesion location, lesion volume, lesion boundary features, and image diagnosis conclusions are extracted from the image event segments. The diagnosis name, symptom description, examination items, treatment behavior, and medication information are extracted from the medical record semantic event segments. The executed actions, parameter settings, system feedback, and data writing records are extracted from the robot process event segments. The imaging diagnostic conclusions and diagnostic names are standardized in terms of terminology, and diagnostic correlation values ​​are calculated; feature correlation values ​​are calculated based on the lesion location, lesion volume, lesion boundary features, symptom description, and examination items. Calculate the process association value based on the treatment behavior and the execution action; Based on the anchor point encoding value, image event fragment identifier, medical record semantic event fragment identifier, robot process event fragment identifier, the diagnosis association value, the feature association value, and the process association value, the cross-dimensional association relationship is generated, and the cross-dimensional association relationship is written into the association relationship layer of the same event index structure.

6. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 5, characterized in that, The calling process in step S5 specifically includes: Based on the anchor code value in the current diagnosis and treatment event anchor, locate the target anchor record in the anchor layer of the same event index structure; Read the fragment index set of the target anchor point recorded in the event fragment layer. The fragment index set includes image event fragment identifiers, medical record semantic event fragment identifiers, and robot process event fragment identifiers. Read the corresponding robot process event segment according to the robot process event segment identifier, and read the corresponding medical record semantic event segment according to the medical record semantic event segment identifier; Read the corresponding cross-dimensional relationship from the relationship layer based on the cross-dimensional relationship identifier; The consistency of the robot task number in the robot process event segment with the robot task number in the diagnosis and treatment event anchor point is verified, and the consistency of the anchor point code value in the cross-dimensional association relationship with the anchor point code value in the diagnosis and treatment event anchor point is verified. When the robot task number and anchor point code value are consistent, the read robot process event fragment, the medical record semantic event fragment, and the cross-dimensional association relationship are input into the source traceability credibility evaluation process.

7. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 6, characterized in that, The process of generating the source tracing credibility marker specifically includes: The audit traceability field is read from the robot process event fragment. The audit traceability field includes robot task number, operation time, operator identifier, device number, executed action, parameter settings, system feedback, data write record and audit signature. The task consistency evaluation value is obtained based on the consistency between the robot task number and the robot task number in the diagnosis and treatment event anchor point; The time consistency evaluation value is obtained based on the correspondence between the operation time and the start time and end time of the treatment stage in the treatment event anchor point; Based on a preset behavior action mapping table, the corresponding behavior consistency evaluation value is obtained by verifying whether the executed action corresponds to the treatment behavior in the medical record semantic event fragment. The write consistency evaluation value is obtained based on whether the data write record is consistent with the source field layer in the same event index structure. The audit signature is verified based on the pre-stored device certificate to obtain a signature integrity evaluation value; A traceability credibility score is generated based on the task consistency evaluation value, the time consistency evaluation value, the behavior consistency evaluation value, the write consistency evaluation value, and the signature integrity evaluation value. Based on the comparison results between the source tracing credibility score and the first credibility threshold and the second credibility threshold, a source tracing credibility label of high credibility, medium credibility, or low credibility is generated. The generated source traceability credibility tag is written into the association layer of the same event index structure, so that the source traceability credibility tag is bound to the corresponding cross-dimensional association identifier.

8. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 7, characterized in that, The process of generating the personalized treatment plan for the patient specifically includes: Read the image event fragments, medical record semantic event fragments, robot process event fragments, cross-dimensional association relationships, and source traceability credibility markers corresponding to each diagnosis and treatment event anchor point under the same patient identifier; The anchor points of each treatment event are sorted according to the treatment time to generate a patient treatment time sequence; Based on the patient's diagnosis and treatment time series, the trends of lesion volume change, symptom change, diagnosis change, treatment behavior change, and robot execution action feedback were calculated. The trends in lesion volume change, symptom change, treatment behavior change, and robot action feedback are converted into lesion volume change trend scores, symptom change trend scores, treatment behavior change scores, and robot action feedback trend scores, respectively. The diagnostic and treatment plan basis score is calculated based on the diagnostic correlation value, the feature correlation value, the process correlation value, the lesion volume change trend score, the symptom change trend score, and the treatment behavior change trend score. The treatment plan is revised based on the source tracing credibility marker to obtain the plan recommendation basis score; The personalized treatment plan for the patient is generated based on the recommended criteria score. The personalized treatment plan for the patient includes the patient's current treatment stage, main imaging data, main medical records, robotic process data, traceability credibility markers, risk warning information, and the recommended criteria score.

9. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 8, characterized in that, The simulation evaluation process in step S7 specifically includes: When the patient's personalized treatment plan does not meet the preset stability conditions, the patient's current treatment stage is converted into a stage code, the main imaging data is converted into imaging status parameters, the main medical record data is converted into clinical status parameters, the robot process data is converted into execution process parameters, and the risk warning information is converted into risk status parameters. The simulation input parameter set is composed of the stage code, the image status parameters, the clinical status parameters, the execution process parameters, the risk status parameters, and the scheme recommendation basis score; The simulation input parameter set is evaluated using preset simulation evaluation rules, which include treatment response prediction rules, lesion change prediction rules, risk change prediction rules, and protocol stability evaluation rules. The treatment response prediction rule yields the predicted value of the treatment response, the lesion change prediction rule yields the predicted value of the lesion change, the risk change prediction rule yields the predicted value of the risk change, and the protocol stability evaluation rule yields the protocol stability evaluation value. The predicted value of the treatment response, the predicted value of the lesion change, the predicted value of the risk change, and the protocol stability evaluation value are then used as new evaluation data.

10. The method for heterogeneous data integration and intelligent analysis of patient's full-cycle health records according to claim 9, characterized in that, The iterative update process in step S7 specifically includes: The newly added evaluation data is processed into events to generate the new event fragments; the anchor code value of the corresponding diagnosis and treatment event anchor point in the new event fragments is read, and the new event fragments are written into the event fragment layer of the same event index structure according to the anchor code value; Write the simulation evaluation rule version identifier, evaluation generation time, evaluation writing time, and field verification value corresponding to the newly added event fragment into the source field layer; After writing the newly added event fragment, re-execute the cross-dimensional relationship update, the source traceability credibility marker update, and the patient personalized treatment plan basis update; When re-performing the cross-dimensional association update, the feature association value, the process association value, and the risk association value are updated based on the association calculation results between the newly added event fragment and the image event fragment, the medical record semantic event fragment, and the robot process event fragment, and the cross-dimensional association is also updated. When re-executing the traceability credibility tag update, an updated traceability credibility score is generated based on the robot process traceability evaluation value and the source consistency evaluation value of the newly added event fragment, and an updated traceability credibility tag is generated based on the updated traceability credibility score. When re-executing the update of the patient's personalized treatment plan basis, the treatment plan basis score and the plan recommendation basis score are recalculated based on the updated cross-dimensional correlation and the updated traceability credibility mark, generating a new patient personalized treatment plan basis, and the new patient personalized treatment plan basis is judged again to see if it meets the preset stability condition, until the iterative update process ends.

Images