Neuropathic pain assessment system and method based on medical big data

Abstract

The present application relates to a kind of neuropathic pain assessment system and method based on medical big data.Method is executed by computer, the surgery process record formed, postoperative pain record and postoperative local abnormal event record are structured coding, construct surgery area disturbance graph, generate pain migration path, extract candidate neural correlation path, utilize abnormal event covering data to carry out three-dimensional covering segmentation to pain migration unit, output intermediate result data containing candidate neural correlation interval and data source identification.System includes data acquisition module, structured processing module, disturbance graph construction module, pain path generation module, correlation path extraction module, abnormality covering generation module, hierarchical processing module and result output module.The present application realizes the unified structured processing of formed medical record and the output of traceable intermediate result data.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, specifically to a neuropathic pain assessment system and method based on medical big data, and in particular to a computer information processing method for performing structured coding, surgical area perturbation map construction, pain migration path generation, candidate neural pathway extraction, abnormal event coverage segmentation, and intermediate result data output on existing postoperative medical records. Background Technology

[0002] The organization and analysis of information related to neuropathic pain is an important auxiliary step in postoperative pain management. Medical information systems typically contain data such as surgical procedure records, postoperative pain records, and postoperative local abnormal event records. These data come from different business processes, and the recording time, location descriptions, pain characteristics, and abnormal state descriptions are often inconsistent. They need to be processed by computer structure before they can be correlated and analyzed under the same data structure.

[0003] Existing medical information systems can typically store, query, and display postoperative pain records, surgical records, and abnormal event records. Some systems can also perform statistical analysis based on pain scores, pain locations, or medication use. However, the above processing is mostly limited to data aggregation, display of single indicators, or simple rule filtering, and has the following problems.

[0004] First, the data generated from different business processes is not expressed consistently. The time expressions, anatomical location descriptions, pain nature descriptions, and abnormal state descriptions in surgical procedure records, pain records, and local abnormal event records are often inconsistent, making it difficult to directly perform cross-source correlation processing.

[0005] Second, there is a lack of a computable relational structure between surgical area manipulation information and adjacent neural regions. Existing systems typically do not organize surgical area manipulation nodes, adjacent neural nodes, region mapping edges, and neural adjacency edges into a searchable surgical area perturbation graph, resulting in the inability to perform path-based retrieval of the relational information between surgical area manipulation information and adjacent neural regions at the computer level.

[0006] Third, there is a lack of a path-based processing mechanism for changes in pain location over time. Existing systems typically display pain records as independent records, without converting adjacent pain records into pain transfer units or further forming pain transfer paths. This makes it impossible to continuously search and match the temporal changes in pain location with surgical operation nodes.

[0007] Fourth, there is a lack of a mechanism for segmenting the coverage of pain migration units by abnormal events. Local abnormal events such as incision abnormalities, inflammatory responses, tissue swelling, postural traction, dressing change stimulation, or drainage stimulation may partially cover pain migration units in terms of location, time, and state. Existing systems have difficulty distinguishing pain migration units in candidate neural pathways into fully covered units, uncovered units, and partially covered units.

[0008] Fifth, the output data lacks a traceable intermediate result structure. The existing system's output pain statistics or scoring results lack the correspondence between candidate nerve association intervals, surgical site operation node identifiers, pain record identifiers, adjacent nerve node identifiers, and abnormal event source identifiers, which is not conducive to verifying the data source during subsequent medical judgment. Summary of the Invention

[0009] To address the shortcomings of the existing technologies, the technical problem to be solved by this invention is to provide a neuropathic pain assessment system and method based on medical big data, which solves the problems of inconsistencies in the expression of time, location and state in existing postoperative medical records, and the inability of existing systems to process surgical operation information, pain migration information and abnormal event coverage information in a unified data structure. The computer generates intermediate result data containing candidate nerve association intervals and data source identifiers for subsequent medical judgment.

[0010] To solve the above-mentioned technical problems, the present invention adopts the following technical solution.

[0011] The first aspect of this invention provides a method for assessing neuropathic pain based on medical big data, executed by a computer, comprising the following steps: The system retrieves postoperative patient surgical procedure records, postoperative pain records, and postoperative local abnormal event records from the medical information system, and performs time stamping, anatomical location coding, and status feature coding to obtain structured medical data. Based on the structured medical data, a surgical area perturbation map is constructed, which includes surgical area operation nodes, adjacent neural nodes, and node relationships. Pain migration units are generated based on adjacent postoperative pain records, and multiple pain migration units are connected to form a pain migration path; The pain migration path is matched with the surgical area perturbation map for node matching and continuity retrieval to extract candidate neural association paths; Abnormal event coverage data is generated based on the postoperative local abnormal event records, and the abnormal event coverage data is used to segment the pain migration units in the candidate neural association pathways to obtain fully covered units, uncovered units, and partially covered units. Based on the distribution of the uncovered and partially covered units in the surgical area perturbation map, candidate neural association intervals are generated, and intermediate result data containing the candidate neural association intervals and their data source identifiers are output.

[0012] Furthermore, the process involves acquiring pre-existing postoperative patient surgical records, postoperative pain records, and postoperative local abnormal event records from the medical information system, and then performing time stamping, anatomical location coding, and state feature coding to obtain structured medical data, including: The surgical time, surgical procedure name, surgical area description, and operation duration are extracted from the surgical procedure record. The surgical area description is mapped to a unified anatomical site code to obtain structured surgical data. The recording time, pain location, pain nature, and pain intensity are extracted from the postoperative pain records. The pain location is mapped to the unified anatomical site code, and the pain nature is converted into a pain nature category code to obtain structured pain data. The location, start time, end time and abnormal state of the abnormal event are extracted from the postoperative local abnormal event record, and the location of the abnormal event is mapped to the unified anatomical site code to obtain structured abnormal event data. Establish patient identification associations and time-based associations among the structured surgical data, structured pain data, and structured abnormal event data.

[0013] Furthermore, the step of constructing a surgical area perturbation map based on the structured medical data, including surgical area operation nodes, adjacent neural nodes, and node relationships, includes: The structured surgical data was divided into multiple operation segments according to the operation time, and each operation segment was converted into a surgical area operation node. Based on the time interval between adjacent surgical area operation nodes and the adjacency relationship of anatomical sites, a directed operation edge is generated between surgical area operation nodes. Query the mapping table between anatomical regions and nerve innervation regions, generate neighboring nerve nodes for each surgical region operation node, and generate region mapping edges between the surgical region operation node and neighboring nerve nodes. Based on the anatomical adjacency relationship between the nerve innervation areas corresponding to adjacent nerve nodes, nerve adjacency edges between adjacent nerve nodes are generated to form the surgical area perturbation map.

[0014] Furthermore, the step of generating pain migration units based on adjacent postoperative pain records and connecting multiple pain migration units into a pain migration path includes: Two postoperative pain records with adjacent times are combined into a pain record pair. The code of the previous pain location in the pain record pair is used as the migration starting point, and the code of the next pain location is used as the migration ending point. Query the anatomical site adjacency data, generate the site adjacency sequence between the migration start point and the migration end point, and write the site adjacency sequence into the pain migration unit; Pain state features are generated based on changes in pain nature category coding and changes in pain intensity parameters in the pain record pairs; Based on the time interval between adjacent pain migration units, the location adjacency relationship, and the pain state characteristics, a migration continuity identifier is generated, and multiple pain migration units are connected into a pain migration path according to the migration continuity identifier.

[0015] Further, the step of performing node matching and continuity retrieval between the pain migration path and the surgical area perturbation map to extract candidate neural association paths includes: The pain migration units in the pain migration path are matched with the surgical area operation nodes in the surgical area perturbation map by location encoding and time interval matching to obtain the matching nodes. Using the matching node as the starting point for retrieval, the corresponding neighboring neural nodes are read along the region mapping edge, and a sequence of neighboring neural nodes corresponding to the pain migration direction is generated along the neural adjacency edge. Generate neural continuity identifiers based on the sequence of neighboring neural nodes corresponding to continuous pain migration units; According to the neural continuity identifier, continuous pain migration units and their corresponding neighboring neural node sequences are written into the candidate neural association path, and path break identifiers are written at the locations where the neural continuity identifier is interrupted.

[0016] Furthermore, the step of generating abnormal event coverage data based on postoperative local abnormal event records includes: Based on the location of the abnormal event location code in the structured abnormal event data in the adjacent data of the anatomical region, generate the location coverage range; Generate a time coverage range based on the start and end times of the abnormal event; Based on the mapping relationship between abnormal event types and pain nature categories, generate state coverage; The location coverage, time coverage, and status coverage of the same abnormal event are associated and stored together to form abnormal event coverage data; For multiple abnormal events with overlapping coverage areas and overlapping time coverage areas, separate event source identifiers are set.

[0017] Furthermore, the step of using the abnormal event coverage data to segment the pain migration units in the candidate neural pathways to obtain fully covered units, uncovered units, and partially covered units includes: For pain migration units in candidate neural pathways, generate location coverage vectors, time coverage vectors, and state coverage vectors between them and anomalous event coverage data; Write the pain migration unit of the complete coverage unit set into which the location coverage vector, time coverage vector and state coverage vector all point to the same abnormal event coverage data. Write the pain migration units that do not have any abnormal event coverage data in any dimension of location, time or state into the set of uncovered units; Pain migration units that do not belong to the complete coverage unit but have one-dimensional or two-dimensional coverage are segmented according to the coverage boundary and written into the partial coverage unit set. Fully covered units, uncovered units, and partially covered units are associated with and stored as anomaly source identifier and neighboring neural node identifier, respectively.

[0018] Further, the step of generating candidate neural association intervals based on the distribution of the uncovered and partially covered units in the surgical area perturbation map, and outputting intermediate result data containing the candidate neural association intervals and their data source identifiers, includes: Read continuously arranged uncovered units along the candidate neural association path, and merge them into the first association interval based on the continuity relationship of adjacent neural nodes and the direction of pain migration; Read the segments in the partial coverage unit that are not covered by abnormal events, and merge them into a second associated interval according to the connection relationship of the corresponding neighboring neural nodes; The first association interval was identified as the primary candidate neural association interval, and the second association interval was identified as the candidate neural association interval to be reviewed. Add surgical site operation node identifiers, pain record identifiers, and adjacent nerve node identifiers to the main candidate nerve association intervals, and further add abnormal event source identifiers to the candidate nerve association intervals to be reviewed; The output includes intermediate result data containing the main candidate neural association intervals, candidate neural association intervals to be reviewed, fully covered units, and their data source identifiers.

[0019] The second aspect of the present invention provides a neuropathic pain assessment system based on medical big data, including a data acquisition module, a structured processing module, a perturbation graph construction module, a pain path generation module, an association path extraction module, an anomaly coverage generation module, a hierarchical processing module, and a result output module, wherein each module cooperates to execute the method described in the first aspect.

[0020] A third aspect of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the method described in the first aspect.

[0021] Compared with the prior art, the present invention has the following beneficial effects.

[0022] Firstly, it transforms existing medical records from different sources into structured medical data under a unified anatomical site coding system, enabling surgical procedure records, postoperative pain records, and postoperative local abnormal event records to be processed by computer under the same time reference and coding system, thus solving the problem of inconsistent expression of cross-source data.

[0023] Secondly, a perturbation graph of the surgical area is constructed, which includes surgical area operation nodes, neighboring neural nodes, directed operation edges, region mapping edges, and neural adjacency edges, so that the relationship between surgical area operation information and neighboring neural regions can be retrieved and traced by computer.

[0024] Third, adjacent postoperative pain records are converted into pain migration units and connected into pain migration paths, realizing path-based processing of temporal changes in pain locations, and providing structured input for subsequent node matching and continuous retrieval.

[0025] Fourth, by extracting candidate neural pathways through node matching and continuity retrieval, path-level association between pain migration pathways and surgical area perturbation maps was achieved.

[0026] Fifth, by using the three dimensions of location coverage vector, time coverage vector and state coverage vector, pain migration units in candidate neural pathways are segmented for coverage, which can distinguish between fully covered units, uncovered units and partially covered units, and retain the source identifier of abnormal events.

[0027] Sixth, the output intermediate results data includes candidate neural association intervals and data source identifiers, which facilitates source tracing and interval verification in subsequent medical judgments. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 The overall flowchart of the neuropathic pain assessment method based on medical big data provided in the embodiments of the present invention is shown.

[0030] Figure 2 This is a schematic diagram of candidate neural association path extraction provided in an embodiment of the present invention.

[0031] Figure 3 This is a schematic diagram of coverage segmentation processing provided in an embodiment of the present invention.

[0032] Figure 4 The structural block diagram of the neuropathic pain assessment system based on medical big data provided in the embodiments of the present invention. Detailed Implementation

[0033] This embodiment provides a computer-executed method for assessing neuropathic pain based on big data in medical science, such as... Figure 1 As shown, the postoperative patient medical records stored in a certain medical information system are used as the processing object. This medical information system already contains surgical procedure records from the surgical record module, postoperative pain records from the nursing record module, and postoperative local abnormal event records from the abnormal event record module. The specific implementation methods of each processing step are explained in detail below.

[0034] Step S1: Obtain the existing postoperative patient surgical process records, postoperative pain records, and postoperative local abnormal event records from the medical information system, and perform time stamping, anatomical location coding, and status feature coding to obtain structured medical data.

[0035] The purpose of step S1 is to convert three types of original medical records from different sources and with inconsistent formats into structured medical data that can be processed under the same data structure, specifically including structured surgical data, structured pain data, and structured abnormal event data.

[0036] First, the surgical procedure record is structured. The computer reads the existing postoperative patient surgical procedure record from the surgical record interface of the medical information system. Each surgical procedure record contains four core fields: surgical time, surgical operation name, surgical area description, and operation duration. However, the naming and formatting of these fields vary across different medical information systems. For example, the surgical time field may be stored as a dual field of "start time + end time" or as "start time + duration in minutes". The computer first converts these formats into a unified "surgical time interval". The notation of “, where and All are expressed in integer minutes, with a minute offset based on 0:00 on the day of surgery.

[0037] For the surgical area description field, the computer maps it to a unified anatomical location code in the following way: First, using the surgical area description string as input, an exact match is performed in the hospital's standard anatomical location vocabulary. If a match is successful, the corresponding code is directly retrieved. Second, if an exact match fails, the edit distance between the surgical area description string and each entry in the vocabulary is calculated, and the entry with the smallest edit distance that does not exceed a threshold is selected. The entry encoding, in this embodiment Third step, if the minimum edit distance exceeds The longest continuous substring describing the surgical area is used as the query condition. The longest common prefix match is used to search the vocabulary, and the entry with the longest matching length is selected. In the fourth step, if no confidence mapping is found in the above three steps, the field is marked as "location to be confirmed," the original text is retained, and the confidence level is recorded as low in the structured surgical data. For example, "right L3-L4 intervertebral space" is mapped to the code "LS-R-0304" through exact matching, and "femoral nerve course area" is mapped to the code "FN-01" through longest common prefix matching, corresponding to the femoral nerve trunk area; both have high confidence levels.

[0038] After completing field extraction and encoding mapping, the computer converts each surgical procedure record into a structured surgical data entry, with fields including: patient identifier, surgical time range, etc. Surgical procedure name, standardized anatomical site code, procedure duration (minutes), and confidence level indicator.

[0039] Next, the postoperative pain records are processed in a structured manner. The computer reads the existing postoperative pain records from the nursing record interface of the medical information system. Each postoperative pain record contains four core fields: recording time, pain location, pain nature, and pain intensity. Pain intensity is usually stored as a NRS numerical score (0-10 integers), while some systems store it as a VAS score (0-100) or a textual rating (mild / moderate / severe). The computer converts all scoring formats other than NRS into equivalent NRS values, divides the VAS by 10 and rounds it down, and assigns NRS values ​​of 1-3 for mild, 4-6 for moderate, and 7-10 for severe, taking the median value of the range.

[0040] The pain location field is converted to a unified anatomical location code using the same four-step mapping process as the surgical procedure record. The pain nature field is converted into five categories: burning (NP-01), stabbing (NP-02), aching (NP-03), electric shock (NP-04), and pressure (NP-05). The computer takes the pain nature description string as input, searches the pain nature category thesaurus for the closest category entry, and uses the keyword set of each entry in the category thesaurus as the matching basis. If the description matches keywords of multiple categories, the highest priority category is selected as the primary category code in the order of electric shock > burning > stabbing > pressure > aching, and the remaining matched categories are stored as secondary nature identifiers. If the pain nature field is empty or cannot be recognized, the primary category code is recorded as "NP-00" (unclassified), and a field missing identifier is marked. The computer converts each postoperative pain record into a structured pain data entry with fields including: patient identifier, record time (converted to minute offset relative to the start time of surgery), uniform anatomical site code, primary pain category code, secondary nature identifier list, NRS equivalent score, and confidence level identifier.

[0041] Then, the postoperative local abnormal event records are processed in a structured manner. The computer reads the existing postoperative local abnormal event records from the abnormal event interface of the medical information system. Each record contains four core fields: the location of the abnormal event, the start time, the end time, and the abnormal status. The location of the abnormal event is converted into a unified anatomical location code using the same four-step mapping process as described above. Both the start time and the end time are converted into minute offsets relative to the start time of the surgery; if the end time field is missing, the computer estimates the duration based on the default duration set according to the abnormal event type, for example: incision abnormality AE-01: 4320 minutes, or 72 hours; inflammatory response AE-02: 2880 minutes, or 48 hours; tissue swelling AE-03: 2880 minutes, or 48 hours; postural traction AE-04: 120 minutes, or 2 hours; dressing change stimulation AE-05: 240 minutes, or 4 hours; drainage stimulation AE-06: 480 minutes, or 8 hours, and the estimated end time is marked as an estimation identifier. The abnormal status field is categorized into six types: incision abnormality (AE-01), inflammatory response (AE-02), tissue swelling (AE-03), postural traction (AE-04), dressing change stimulation (AE-05), and drainage stimulation (AE-06). The mapping method is similar to the pain nature category coding mapping, using keyword matching. The computer converts each record into a structured abnormal event data entry, with fields including: patient identifier, uniform anatomical site code, start time offset, end time offset, end time estimation identifier, abnormal event type code, and confidence level identifier.

[0042] Finally, patient identification and time-based associations are established among the three types of structured data. After generating the three types of structured data, the computer establishes patient identification associations among them: using the patient's unique identifier as the association key, such as the hospital medical record number, the structured surgical data, structured pain data, and structured abnormal event data of the same patient are organized into the same patient data set. The time-based association has been completed in various data processing steps, all converted to minute offsets with the surgery start time as zero. This step further verifies whether the time range of the three types of data is within a reasonable range: the time offset of structured pain data and structured abnormal event data should be greater than the surgery start time offset and within a reasonable postoperative time window. In this embodiment, the reasonable postoperative time window is set to 30 days postoperatively, i.e., 43,200 minutes. Records exceeding this range are marked as time abnormalities but are still retained in the data set, and can be selectively skipped or downgraded in subsequent steps. After completing the above association processing, the structured medical data of each patient forms a triplet (a list of structured surgical data, a list of structured pain data, and a list of structured abnormal event data) indexed by the patient identifier.

[0043] To ensure the reproducibility of the above structured processing, the hospital's standard anatomical site thesaurus should include at least the following fields: site name, synonyms, unified anatomical site code, region, left / right side identifier, superior / inferior site code, and confidence level; the pain nature category thesaurus should include at least the following fields: pain nature category code, keyword set, priority, corresponding description, and missing information handling identifier.

[0044] The anatomical location adjacency table uses the unified anatomical location code as the node identifier and records the adjacent location code, adjacency direction, adjacency distance, and applicable surgical area; the anatomical region to nerve innervation region mapping table uses the unified anatomical location code as the query key and records the nerve root or nerve trunk identifier, nerve innervation region code, mapping confidence, and mapping source; the nerve innervation region adjacency table records the codes of adjacent nerve innervation regions and the adjacency relationship type.

[0045] In this embodiment, the parameters are stored in a configurable manner in a parameter configuration table. The parameter configuration table includes at least the parameter name, default value, applicable data type, value range, value source, and adjustment condition fields. , The reasonable time window, the matching time window after surgery, the migration continuous time threshold, the number of neural adjacency expansion steps, the estimated boundary tolerance time, the time difference of merging intervals to be reviewed, and the individual writing duration are all read through this parameter configuration table.

[0046] In one example configuration, Edit distance tolerance for short text part descriptions To limit excessive expansion of cross-site pathways, the reasonable postoperative time window is 43,200 minutes, the migration continuity time threshold is 1,440 minutes, and the postoperative matching time window is 4,320 minutes. When the recording frequency, surgical type, or departmental rules of the target medical information system change, the above parameters are adjusted through the parameter configuration table, and the parameter version identifier used is recorded in the output data. Furthermore, the surgical area segmentation time threshold is 10 minutes, the surgical area anatomical adjacency threshold is 2 steps, the neural adjacency expansion step count is 3 steps, the estimated boundary tolerance time is 60 minutes, the time difference for merging intervals to be reviewed is 120 minutes, and the individual write duration is 180 minutes. All of these are recorded as independent parameters in the parameter configuration table, recording their default values, value ranges, value sources, and adjustment conditions.

[0047] Step S2: Construct a surgical area perturbation map based on the structured medical data, which includes surgical area operation nodes, adjacent neural nodes, and node relationships.

[0048] The purpose of step S2 is to convert the structured surgical data into a directed graph data structure, namely the surgical area perturbation graph. This graph contains surgical area operation nodes, neighboring neural nodes, and three types of edges, specifically directed operation edges, region mapping edges, and neural adjacency edges, which are used to express the searchable topological relationship between surgical area operation information and neighboring neural regions at the computer level.

[0049] Generate surgical area operation nodes. The computer takes a list of structured surgical data from the patient dataset as input and processes the data according to the start point of the surgical time interval. Sort the records in ascending order to obtain an ordered sequence of operation records. For two adjacent operation records, insert a split point between them if either of the following conditions is met: 1) the time interval between the operation time intervals of the two adjacent records exceeds 10 minutes; 2) the operation names of the two adjacent records are different, and the strings are completely different, so no fuzzy matching is performed. Each continuous operation segment defined by the split point is converted into a surgical area operation node. The node attributes include: node identifier, operation time interval, unified anatomical location code set, and main operation name; where the node identifier consists of the patient identifier, the first operation record of the segment, and the main operation name. It is composed of unified anatomical site codes, and the surgical time interval is the total number of operations recorded in that segment. Minimum value to The maximum value is the unified anatomical location code set, which is the set of duplicate anatomical location codes for all operation records in this segment. The main operation name is the name of the surgical operation that appears most frequently in this segment. If the same patient has multiple surgical stages, such as laparoscopic exploration followed by open surgery, then operation segments for each stage generate corresponding surgical area operation nodes, and each node carries the identifier of the surgical stage to which it belongs.

[0050] Generate directed operation edges. The computer traverses the generated sequence of surgical area operation nodes in chronological order. For two adjacent nodes, such as node A and node B, node A is performed before node B in time, and the following judgment is made: Any code from the anatomical site code set of node A is selected. And any code in the anatomical part coding set of node B. Query the adjacency list of anatomical sites and The shortest adjacent path length d between them; if there exists at least one pair Make d not exceed 2, that is and If an anatomical site is reachable in at most two steps from its adjacency list, then a directed edge is generated between node A and node B, with the direction A→B. The edge attributes include: source node identifier, target node identifier, and time interval (node ​​B's...). Subtract node A (Unit: minutes), anatomical adjacency distance d. If the adjacency condition is not met, no directed operation edge is generated, but the two nodes remain in the surgical area perturbation graph and will not be connected by the operation edge during subsequent path retrieval.

[0051] The computer generates neighboring nerve nodes and region mapping edges. It queries a mapping table of anatomical regions and their innervation regions, using a unified anatomical location code as the query key. This table returns a list of innervation region codes corresponding to the given anatomical region. Each innervation region code consists of a nerve root / trunk identifier and a dermatome segment number; for example, "L3-dermatome" indicates a dermatome innervated by the 3rd lumbar nerve root. For each surgical operation node, the computer iterates through its anatomical location code set, querying each code in the mapping table to obtain a complete set of innervation region codes associated with that node. For each unique innervation region code in the complete set, a neighboring nerve node is generated. Node attributes include: node identifier (innervation region code), nerve type (nerve root / trunk / branch), and a list of corresponding anatomical region codes. Subsequently, the computer generates a region mapping edge between the corresponding surgical operation node and each neighboring nerve node. Edge attributes include: source node identifier (surgical operation node), target node identifier (neighboring nerve node), source anatomical location code, and mapping confidence, where the mapping confidence is inherited from the confidence identifier of the anatomical location code in step S1.

[0052] Generate nerve adjacency edges. The computer queries the adjacency table of nerve innervation regions. This adjacency table uses the nerve innervation region code as the query key and returns a list of other nerve innervation region codes that have a direct anatomical adjacency relationship with it. Adjacency relationship types include adjacent dermatomes of the same nerve root, such as between L3 and L4 dermatomes, and adjacent branches of the same nerve trunk, such as between the common fibular branch and the tibial branch of the sciatic nerve. For all the generated adjacent nerve nodes, pair them up and check whether the nerve innervation region codes of the two nodes have a direct adjacency record in the nerve innervation region adjacency table; if so, generate an undirected nerve adjacency edge between the two adjacent nerve nodes. The edge attributes include: identifiers of the two endpoint nodes and the adjacency relationship type. At this point, the surgical area perturbation graph is completed, containing all surgical area operation nodes, adjacent nerve nodes, directed operation edges, region mapping edges, and nerve adjacency edges. This graph is stored in the form of an adjacency list, supporting breadth-first or depth-first traversal from any node along edges of a specified type.

[0053] Step S3: Generate pain migration units based on adjacent postoperative pain records, and connect multiple pain migration units into a pain migration path.

[0054] The purpose of step S3 is to convert discrete, time-ordered structured pain data into a path-based pain migration path data structure so that node matching and continuity retrieval can be performed with the surgical area perturbation map in step S4.

[0055] Generate pain transfer units. The computer takes a structured list of pain data from the patient dataset as input, sorts it by record time offset in ascending order, and obtains an ordered sequence of n pain records, denoted as . The computer sequentially retrieves two adjacent pain records. and The i-th pain record pair is formed, and the i-th pain transfer unit is generated as follows: .

[0056] Migration start point code: Take The unified anatomical location code is denoted as .

[0057] Migration endpoint code: Take The unified anatomical location code is denoted as .

[0058] Partial adjacency sequence generation: If and If they are the same, then the adjacency sequence of the parts is a single-element sequence. If they differ, the computer will... Starting point Using the endpoint as the starting point, perform a breadth-first search in the undirected adjacency graph of the anatomical site adjacency data to find the shortest path; if the shortest path length does not exceed the upper limit... In this embodiment If the shortest path length exceeds a certain threshold, then all anatomical site codes (including start and end points) along the shortest path are sequentially written into the site adjacency sequence field; if the shortest path length exceeds a certain threshold... If no connected path exists, then the part adjacency sequence field should be set to contain only... and Discontinuous sequences are marked with "location span too large" or "location unreachable".

[0059] Pain state feature generation: The computer extracts the following three types of state change information from pain record pairs: First, the direction of change of the pain nature category encoding, i.e., comparison. and The primary category code is used; if the primary category codes are different, it is recorded as "change in nature"; if they are the same, it is recorded as "unchanged nature". Secondly, the change in pain intensity is taken as... NRS equivalent value minus The NRS equivalent value is used, with positive values ​​indicating increased intensity, negative values ​​indicating decreased intensity, and 0 indicating no change. Thirdly, the normalized labeling of the direction of change in nature is used: if the main category coding sequence follows a "stable or escalating (burning > stinging > shock)" pattern, it is labeled "progressive"; if it follows a "deteriorating" pattern, it is labeled "relieved"; and irregular changes are labeled "fluctuating". These three types of information are combined into a pain state feature field and written into... .

[0060] Complete attributes of the pain migration unit: Unit identifier, migration start code Migration endpoint coding The elements include: adjacent sequence of locations, pain record pair identifiers, time intervals, pain state characteristics, and confidence level identifiers. The unit identifier consists of the patient identifier, i, and... It is pieced together; pain records are used as markers. and The record identifier; the time interval refers to Recording time offset to The recording time offset.

[0061] The computer generates migration continuity identifiers and connects them into pain migration paths. The computer then analyzes adjacent pain migration units. and Perform migration continuity determination and generate migration continuity identifiers. The decision logic is as follows: Condition A (Time Continuity): The start point of the time interval and The difference between the endpoints of the time intervals shall not exceed 1440 minutes (24 hours). Condition B (Local Continuity): migration start point encoding and The migration endpoint codes are the same, or the shortest path length between the two in the anatomical site adjacency data does not exceed 1 step; Condition C (State Continuity): and The normalized indicators for the direction of change in the nature of pain state characteristics are the same: all are progressive, all are remission, or all are fluctuating, or both are "uniform in nature".

[0062] When conditions A, B, and C are all satisfied simultaneously Set to "continuous"; when condition A is not satisfied, Set to "Interruption - Time Exceeded"; when condition A is met but condition B is not met, Set to "Interruption-Part Jump"; when conditions A and B are both met but condition C is not met, Set to "Interrupt-State Reversal". Different interrupt types provide additional reference information for subsequent steps. Based on the migration continuity identifier, the computer will... Adjacent pain migration units are connected end-to-end in a "continuous" manner to form a pain migration path; in The current pain migration path is terminated at any "interruption" location, and a new pain migration path is started. Each pain migration path contains a path identifier, an ordered list of pain migration units, a path start time offset, and a path end time offset. The path identifier consists of a patient identifier and a path sequence number.

[0063] Step S4: Perform node matching and continuity retrieval between the pain migration path and the surgical area perturbation map to extract candidate neural association paths.

[0064] The pain migration pathway was matched with the surgical area perturbation map for node matching and continuity retrieval to extract the processing relationships of candidate neural pathways, such as... Figure 2 As shown.

[0065] The purpose of step S4 is to correlate and match the pain migration path generated in step S3 with the surgical area perturbation map generated in step S2 to extract candidate neural correlation paths. This step is divided into two stages: node matching and continuous retrieval, to output candidate correlation data.

[0066] First, in the node matching phase, the computer traverses each pain migration unit in each pain migration path. Search for matching surgical area operation nodes in the surgical area perturbation map. The matching conditions must satisfy both of the following conditions.

[0067] Part code matching: At least one anatomical location code exists in the adjacent sequence of the surgical site, which is identical to a code in the set of anatomical location codes for the surgical site operation node. That is, suppose... The adjacency sequence of the part is surgical area operation nodes The anatomical part coding set is The matching condition is .

[0068] Time interval matching: Time offset of the last record in the time interval satisfy: surgical time range ,and of This means the pain record occurred within 72 hours after the surgery. This condition indicates that the pain record occurred within 72 hours after the operation in that surgical area, and thus has a reasonable temporal causal relationship.

[0069] for All surgical operation nodes that satisfy the above two matching conditions form a matching node set. .like If it is an empty set, then the Nodes marked as "no matching node" are skipped in subsequent sequential searches but retained in the path data as candidate locations for path breakpoints. If If there are multiple nodes, such as multiple operations involving multiple adjacent sites during surgery, all matching nodes are retained. Subsequent continuous searches are carried out starting from each matching node, and path segment identifiers are retained for different branches.

[0070] Then it proceeds to the continuous retrieval phase. (Regarding...) Each matching node in Computers Starting from this point, generate as follows: Corresponding neighboring neural node sequences .

[0071] The first step is to read along the region mapping edge: from Starting from, read along the region mapping edge and... The set of all directly related neighboring neural nodes ,Right now .

[0072] The second step is to extend along the adjacent edges of the nerve: Starting from the set, along the adjacent edges of the nerves... The pain migration direction is progressively expanded, that is, along the direction vector corresponding to the migration origin code to the migration endpoint code in the anatomical adjacency data, such as "from proximal to distal" or "from dorsal to ventral". Each expansion step selects nerve adjacency edges consistent with the pain migration direction, with a maximum of 3 expansion steps; the newly added neighboring nerve nodes in each expansion step are added to... At the end of the sequence. If different matching nodes produce the same neighboring neural node sequence prefix at a certain expansion step, they are merged into the same sequence; if a bifurcation occurs, that is, the same pain migration unit corresponds to multiple candidate neural sequences, they are recorded separately and independent candidate neural association path segments are generated for each bifurcation sequence in subsequent processing.

[0073] Complete each Neighboring neural node sequences After generation, the computer analyzes the adjacent sequences in the pain migration path. and Perform neural continuity determination: take End node and starting node Check the adjacency table of the nerve innervation area for presence. arrive Neural adjacent edges; if they exist, then neural continuity markers. It is "neural continuity"; if there are no adjacent nerve edges but both belong to the same nerve root, that is, the nerve root field encoded by the nerve innervation region is the same, then "Same root, not adjacent"; other cases This is referred to as "neural interruption".

[0074] Based on neural continuity markers, the computer will For "neural continuity" adjacent and And their corresponding NNS combinations are written into the same candidate neural association path segment; in For locations with "non-adjacent roots", the same candidate neural pathway segment is preserved and written with a common root gap identifier. Write a path segmentation identifier at the location of the "nerve interruption," terminate the current candidate neural connection path segment, and start a new candidate neural connection path segment. In the final output set of candidate neural connection paths, each candidate neural connection path contains a path segment identifier, an ordered list of pain migration units, a sequence of neighboring neural nodes corresponding to each unit, a path time range, a common root gap identifier, and a path segmentation identifier.

[0075] Step S5: Generate abnormal event coverage data based on the postoperative local abnormal event records, and use the abnormal event coverage data to segment the pain migration units in the candidate nerve association pathways to obtain fully covered units, uncovered units, and partially covered units.

[0076] Step S5 consists of two sub-steps: First, abnormal event coverage data is generated from structured abnormal event data; second, the abnormal event coverage data is used to segment the pain migration units in candidate neural pathways to obtain fully covered units, uncovered units, and partially covered units. The processing relationship of abnormal event coverage data for segmenting pain migration units in candidate neural pathways is as follows: Figure 3 As shown.

[0077] First, the computer iterates through the patient's list of structured abnormal event data, processing each structured abnormal event data entry. Perform the following processes sequentially to generate three types of coverage areas.

[0078] Generation of part coverage area: using Unified anatomical site coding Using the central node, perform a breadth-first search in the undirected adjacency graph of the anatomical region's adjacency data to retrieve nodes that are adjacent to the central node. The set of codes for all directly adjacent anatomical sites within a distance of no more than one step is denoted as . .Will itself and The union of sets is defined as follows: Part coverage area This design is based on the fact that local abnormal events, such as inflammatory responses and tissue swelling, often affect not only the site of occurrence but also directly adjacent anatomical areas. For more precise control over coverage, the extension radius can be adjusted to 0 steps (center only) or 2 steps (two adjacent steps), depending on the system configuration parameters.

[0079] Generation of time coverage: with start time offset Lower bound, end time offset As the upper bound, generate a closed interval. If the end time offset is an estimated value, then in The coverage data records time boundary estimation identifiers, indicating that the upper limit of the time coverage area is an estimated value. In subsequent coverage vector calculations, time points that fall near the estimated upper limit and whose difference from the estimated upper limit is no more than 60 minutes are marked as "time boundary uncertain".

[0080] State coverage generation: based on Exception event type encoding The state coverage area is determined according to the following fixed mapping relationship. : Incision abnormality (AE-01) Inflammatory response (AE-02) corresponds to Tissue swelling (AE-03) corresponds to Postural stretching (AE-04) corresponds to ; Medication change stimulus (AE-05) corresponds to ; Drainage stimulation (AE-06) corresponds to The above mapping relationships reflect the coverage patterns that different types of local abnormal events may typically produce in the manifestation of pain. If the type encoding is "unrecognized", then That is, no state overwriting occurs.

[0081] A fixed mapping relationship between abnormal event types and pain nature categories is stored in an abnormal event status coverage mapping table. This table includes at least the following fields: abnormal event type code, abnormal event name, corresponding pain nature category code set, mapping basis, applicable conditions, and version identifier. For example, inflammatory response AE-02 corresponds to burning pain NP-01 and aching pain NP-03, and the mapping basis is recorded as in-hospital nursing record keyword statistics or expert rule configuration. The mapping basis field also records the in-hospital nursing record keyword statistics sample range, statistical time window, expert review role, activation date, and version number. When the same abnormal event can correspond to multiple pain nature categories, the status coverage is determined according to the priority of applicable conditions in the mapping table and the expert review results.

[0082] When the exception event type is coded as unrecognized, the state coverage is... The set is empty, and a state coverage missing identifier is written in the coverage data record; subsequent coverage vector calculations do not use this abnormal event as the source of state dimension coverage, but still retain its location coverage range and time coverage range for tracing the candidate neural association interval to be reviewed.

[0083] The same of , , Associated storage, generation Coverage data records The attributes include: overlay data identifier , Location Coverage (Encoding set), time coverage (Closed interval), state coverage area (Encoding set), time boundary estimation identifier, source anomaly event identifier ( (A unique identifier).

[0084] For areas where coverage overlaps (i.e.) And the time coverage areas overlap (i.e.) Two coverage data points and Instead of merging them, the computer retains each coverage data record separately and assigns a unique event source identifier to ensure that subsequent steps can distinguish the independent coverage contributions of different anomalous events to the pain migration unit. The computer maintains a coverage data index structure to support subsequent rapid retrieval of the set of coverage data records corresponding to a given anatomical site code and time point.

[0085] Then, the candidate neural pathways are segmented for coverage. The computer traverses each pain transfer unit in the set of candidate neural pathways generated in step S4. For all covered data records Calculate the three types of covering vectors.

[0086] Calculation of part coverage vector: using Adjacency sequence of parts For each pain migration unit, the set of locations is used to cover the data. ,judge Is it not empty? If not empty, record " In terms of location "Coverage", that is, the location coverage vector. The component is 1 if it is not 0 otherwise. If it contains the "location inaccessible" label, then only... and The two endpoint codes are used to determine the intersection of the sets.

[0087] Calculation of the time coverage vector: The time offset of the next pain record For each piece of covered data, a time checkpoint is used. ,judge Whether it falls into If it falls into the range, record " In the time dimension "Coverage", time coverage vector The component is 1 if it is not 0 otherwise. Falling The upper bound of the estimate is near, i.e. and For the estimated value, the time coverage vector is... The components are recorded as having "undefined time boundaries" rather than simply being 0 or 1.

[0088] Calculation of the state coverage vector: The main category code of the next pain record For each piece of covered data, the inspection target is... ,judge Does it belong to If it belongs to the category, record " In the state dimension "Coverage", state coverage vector The component is 1 if it is not 0 otherwise. If the value is "NP-00" (unclassified), then the state coverage vector is... Set the component to 0 and record the missing status field identifier.

[0089] After calculating the three types of covering vectors, the computer then... Segment and categorize.

[0090] Determination of a fully covered cell: If a certain line exists This makes the part coverage vector, time coverage vector, and state coverage vector... If all components are 1, meaning all three dimensions point to the same covered data, then... It is classified into a set of fully covered units, and the associated ones are recorded. (Event source identifier) ​​and The sequence identifier of the neighboring neural node in the candidate neural associated path segment (neighboring neural node identifier).

[0091] Determination of uncovered cells: If there are no uncovered cells... Coverage in any dimension: location, time, or state. That is, there is no coverage data where the component of the location coverage vector, time coverage vector, or state coverage vector is 1. It is classified into the uncovered cell set. This decision is mutually exclusive with fully covered cells and partially covered cells, ensuring that the uncovered cell set contains only pain migration cells that cannot be interpreted by data covered by anomalous events.

[0092] Determination of partial coverage units: If The complete coverage cell criterion is not met, and at least one condition exists. If one or two-dimensional coverage vector components in the three dimensions of location, time, and state are all equal to 1, then They are categorized into a set of partially covered units. The segmentation process for partially covered units is as follows: the computer prioritizes the temporal coverage boundary as the segmentation dimension, i.e., if... time interval and There is a partial intersection, that is or Then at the intersection boundary, Divided into two sub-units: Covered sub-units and uncovered The covered sub-units; if the time dimension is fully covered but the location or state dimension is only partially satisfied, it is classified as "overall partial coverage" and no further sub-unit division is performed, and it is directly included. Write the partially covered cell set and attach the corresponding dimension coverage tag. The computer associates fully covered cells, uncovered cells, and partially covered cells with the event source identifier. The neighboring neural node identifiers are associated and stored to form a data structure for covering segmented results.

[0093] Step S6: Based on the distribution of the uncovered and partially covered units in the surgical area perturbation map, generate candidate neural association intervals and output intermediate result data containing the candidate neural association intervals and their data source identifiers.

[0094] Step S6, based on the coverage segmentation results, merges and adds labels to the candidate neural connection intervals, ultimately outputting intermediate result data including data source labels. This step consists of two sub-steps: interval merging and result output.

[0095] Interval merging steps. The computer reads the coverage segmentation results forward along the time axis of each candidate neural connection path and performs the following two types of interval merging operations.

[0096] The first type of merging generates primary candidate neural association regions: Computer retrieval of consecutively arranged uncovered unit sequences. The criterion for consecutive arrangement is: two adjacent uncovered units... and Neural continuity markers The insertion of a unit is defined as either "neurally continuous" or "non-adjacent to the same root," and neither of these segments has other segmentation types in the candidate neural association path; that is, a unit that fully covers or partially covers a unit. If the conditions are met, the computer reads the time range covered by the continuous uncovered unit sequence, from the time offset of the pain record preceding the first PMU in the sequence to the time offset of the pain record following the last PMU, and the corresponding adjacent neural node sequences, i.e., all the units in the sequence. The ordered concatenation of nodes, with repeated nodes taken once, combines the above time range and the sequences of neighboring neural nodes to form a primary candidate neural association interval, and assigns a unique interval identifier to this interval. If there are consecutive uncovered unit sequences For locations marked as "non-adjacent within the same root", the computer sets that location as an internal breakpoint within the interval and labels the interval as "interval within the same root" to indicate that the interval contains an internal gap within the nerve root.

[0097] The second type of merging generates candidate neural network regions to be reviewed: the computer traverses the set of partial coverage units and extracts each partial coverage unit. Data not covered The covered time segment, i.e. Not belonging to any time interval The portion that is not covered is denoted as the uncovered subfield. For adjacent partial coverage units and If the uncovered sub-segments are continuous in time, that is of and of If the difference is no more than 120 minutes, and the corresponding neighboring neural node sequences have "neural continuity" or "non-adjacent to the same root" continuity in terms of neural adjacency, then they are merged into a single candidate neural association interval to be reviewed, and assigned a unique interval identifier. The "Pending Review" category is also marked. If there is only a single partially covered unit with an uncovered segment and the segment duration exceeds 180 minutes, it is also written as a separate candidate neural association interval pending review.

[0098] Results Output Steps. The computer adds the following data source identifiers to each major candidate neural association interval: firstly, surgical area operation node identifiers, recording the set of all surgical area operation node identifiers within the interval's coverage area that have site and time matching with each uncovered unit; secondly, pain record identifiers, recording all uncovered units constituting the interval. The third is the identifier of the pain record being referenced; the fourth is the identifier of the neighboring nerve nodes, which records the unique identifier of each node in the sequence of neighboring nerve nodes corresponding to the interval.

[0099] The computer adds an abnormal event source identifier to each candidate neural network region to be reviewed, in addition to the three types of identifiers mentioned above, and records the set of abnormal event coverage data identifiers that generated the corresponding partial coverage units. This is to trace which abnormal events covered certain time periods or parts of the interval.

[0100] The computer also writes each fully covered unit and its data source identifier into the output data to completely preserve the processing results of all segments in the candidate neural pathway. The data source identifier of a fully covered unit includes the corresponding... Adjacent nerve node markers and pain record markers.

[0101] Finally, the computer organizes the list of primary candidate neural association intervals, the list of candidate neural association intervals to be reviewed, and the list of fully covered units into intermediate result data, which is then written into the pending data queue of the medical information system or a designated storage path. Each data object in the intermediate result data retains an independent interval identifier and data source identifier, facilitating subsequent cross-tracing and review by users with appropriate permissions according to pain records, surgical procedures, or abnormal events. The intermediate result data does not contain disease diagnosis conclusions or treatment recommendations; it only represents the structured processing results of the computer on existing medical records for subsequent medical judgment.

[0102] This embodiment provides a neuropathic pain assessment system based on medical big data, such as... Figure 4 As shown, corresponding to the neuropathic pain assessment method based on medical big data, it consists of the following functional modules, each of which is implemented in software form on a computer, and works together to complete the structured processing of existing medical records and the generation of intermediate result data.

[0103] The data acquisition module is responsible for reading existing surgical procedure records, postoperative pain records, and postoperative local abnormal event records from the medical information system's interface, and then transmitting the raw record data to downstream modules. This module supports batch reading of multiple records using patient identifiers as query criteria, and generates a raw data set grouped by patient identifier after the reading is complete. The data acquisition module does not modify the data content; it only performs reading and grouping operations, ensuring that subsequent processing can be traced back to the original records.

[0104] The structured processing module receives the raw records from the data acquisition module and performs the following processing on the three types of records respectively.

[0105] For surgical procedure records: the module uniformly converts the surgical time field to... The data is formatted in minute intervals; a four-step anatomical location coding mapping is performed on the surgical area description field: exact matching → edit distance matching → longest common prefix matching → “location to be confirmed” marker, generating a unified anatomical location code and confidence level identifier; the processing results are written into the structured surgical data entry.

[0106] For postoperative pain records: the module uniformly converts pain intensity into NRS equivalent scores; performs the same four-step encoding mapping on the pain location field; performs category encoding mapping on the pain nature field according to the priority of electric shock-like > burning-like > stabbing-like > pressure-like > aching-like, with the primary category code written into the primary field and the remaining matching categories written into the secondary nature identifier list; converts the recording time into a minute offset relative to the surgery start time; and writes the processing results into the structured pain data entry.

[0107] For postoperative local abnormal event records: the module performs a four-step encoding mapping on the occurrence site; converts the start time and end time into minute offsets; estimates the duration of records with missing end time fields according to the event type (AE-01 to AE-06) by setting a default duration and marking the estimation identifier; performs encoding mapping on the abnormal status field for six types of abnormal events; and writes the processing results into structured abnormal event data entries.

[0108] After the structured processing module completes the processing of the three types of data, it establishes a triplet (structured surgical data list, structured pain data list, and structured abnormal event data list) with the patient identifier as the association key, and verifies whether the time offset of each type of data is within the reasonable postoperative time window. Records that exceed the range are marked with a time abnormality indicator. The reasonable postoperative time window is 43,200 minutes.

[0109] The perturbation map construction module receives the structured surgical data output by the structured processing module, executes the four sub-steps described in step S2, and generates a perturbation map of the surgical area.

[0110] This module internally maintains three types of static data structures: an anatomical site adjacency table, a mapping table between anatomical regions and nerve innervation regions, and a nerve innervation region adjacency table. When generating surgical area operation nodes, a time interval exceeding 10 minutes or different surgical operation names are used as segmentation conditions; when generating directed operation edges, an anatomical adjacency distance of no more than 2 steps is used as a connection condition; when generating adjacent nerve nodes, the corresponding nerve innervation region code for each surgical area operation node is queried through the mapping table; when generating nerve adjacency edges, the same nerve root or nerve trunk is used as the adjacency determination criterion. The surgical area perturbation graph is stored in the form of an adjacency table, supporting subsequent modules to efficiently retrieve associated nodes by node identifier.

[0111] The pain path generation module receives the structured pain data output by the structured processing module, executes the two sub-steps described in step S3, and generates a pain migration path.

[0112] This module pairs two adjacent pain records together, and performs a breadth-first search (maximum path length) on the anatomical adjacency data based on the anatomical location encoding of the migration start and end points. Generate site adjacency sequences; generate pain state characteristics based on changes in pain nature category coding and NRS equivalent score changes, including the direction of nature change, the amount of intensity change, and the progression type identifier; generate migration continuity identifiers (continuous / interrupted - time limit exceeded / interrupted - site jump / interrupted - state reversal) based on three conditions: time continuity (≤1440 minutes), site continuity (adjacency distance ≤1 step), and state continuity (consistent direction of nature change); connect pain migration units into pain migration paths based on continuity identifiers.

[0113] The association path extraction module receives the pain migration path output by the pain path generation module and the surgical area perturbation map output by the perturbation map construction module, and performs the two stages described in step S4 to generate candidate neural association paths.

[0114] Node matching phase: The module matches each pain migration unit. The matching node set is retrieved based on the condition that the intersection of the site coding sets is non-empty and the time offset of the subsequent pain record is within 4320 minutes after the operation in the corresponding surgical area. Continuous retrieval phase: Starting from the matching node, generate a sequence of neighboring neural nodes consistent with the pain migration direction along the region mapping edge and neural adjacency edge (maximum extension of 3 steps); for adjacent... For each neural continuity determination (neural continuity / non-adjacent along the same root / neural interruption), the candidate neural associated path segment or path segment identifier is written according to the determination result. Each path in the output candidate neural associated path set retains a complete path segment identifier and path segment identifier.

[0115] The exception coverage generation module receives the structured exception event data output by the structured processing module, executes the processing described in the first sub-step of step S5, and generates exception event coverage data.

[0116] This module takes each structured exception event data as an example. As input, the part coverage area is generated through breadth-first search (expanding the radius by 1 step). Time coverage range is generated by time offset interval. (Add time boundary estimation identifiers to the estimated end time), and generate state coverage using a fixed mapping table. (AE-01 to AE-06 each correspond to a specific set of pain nature categories). For multiple coverage data entries that overlap in both location and time coverage, each entry is retained and assigned a unique event source identifier. The module maintains a coverage data index to support subsequent rapid retrieval.

[0117] The hierarchical processing module receives candidate neural association paths output by the association path extraction module and abnormal event coverage data output by the abnormal coverage generation module, and executes the coverage segmentation processing described in the second sub-step of step S5, segmenting each pain migration unit. They are classified as fully covered units, uncovered units, or partially covered units.

[0118] This module is for each Calculate the part coverage vector (in) and The criterion is whether the intersection is non-empty, and the time coverage vector (based on) Whether it falls into As a basis for judgment, estimate the uncertain states near the boundary and the state coverage vector (in order to determine the state). Does it belong to (As a basis for judgment) Three types of covering vectors; pointing to the same thing in three dimensions simultaneously. The condition for determining complete coverage is that there is no coverage in any dimension. The conditions for determining uncovered conditions are: there is no complete coverage but there is one-dimensional or two-dimensional coverage. To determine the conditions for partial coverage, sub-unit segmentation is performed according to the time coverage boundary. The classification results are stored in association with the event source identifier and the identifier of the neighboring neural node, forming the coverage segmentation result data structure.

[0119] The result output module receives the coverage segmentation results output by the hierarchical processing module, executes the interval merging and result output processing described in step S6, and generates intermediate result data.

[0120] This module reads the coverage segmentation results along the time axis in a forward direction: merging consecutive uncovered unit sequences (identified by neural continuity). The primary candidate neural association intervals are defined as either "neural continuity" or "non-adjacent to the same root" as the continuity criteria. Uncovered segments from partially covered units are extracted and adjacent segments (with a time difference not exceeding 120 minutes and neural continuity marked as "neural continuity" or "non-adjacent to the same root") are merged to form candidate neural association intervals for review (segments exceeding 180 minutes are also written separately). Surgical area operation node identifiers, pain record identifiers, and adjacent nerve node identifiers are added to the primary candidate neural association intervals. Abnormal event source identifiers are additionally added to the candidate neural association intervals for review. Fully covered units, along with their data source identifiers, are written into the output data. Finally, the lists of primary candidate neural association intervals, candidate neural association intervals for review, and fully covered units are organized into intermediate result data and written into the output queue. Intermediate result data does not include disease diagnosis conclusions or treatment recommendations and is intended for subsequent medical judgment.

[0121] This embodiment provides an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor.

[0122] The processor may be a central processing unit (CPU) or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The memory may include mass storage for data or instructions, such as a hard disk drive (HDD), a solid-state drive (SSD), random access memory (RAM), or read-only memory (ROM).

[0123] When the computer program is executed by the processor, it implements all the steps of the neuropathic pain assessment method based on medical big data described in this invention, including: obtaining the formed medical records from the medical information system and performing structured encoding, constructing a surgical area perturbation map, generating pain migration paths and pain migration units, extracting candidate nerve association paths, generating abnormal event coverage data and performing coverage segmentation on the candidate nerve association paths, generating candidate nerve association intervals and outputting intermediate result data.

[0124] The electronic device can be a server, workstation, or other device with computing capabilities. It communicates with the data interface of the medical information system to read the formed medical records and write the intermediate result data of the processing to the pending data queue or a specified storage path of the medical information system.

[0125] The foregoing has provided a detailed description of the neuropathic pain assessment system and method based on medical big data provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and core ideas of this invention. It should be noted that those skilled in the art can make various improvements and modifications to this invention without departing from its principles, and these improvements and modifications also fall within the protection scope of this invention.

Claims

1. A method for assessing neuropathic pain based on medical big data, characterized in that, The method is executed by a computer and includes: The system retrieves postoperative patient surgical procedure records, postoperative pain records, and postoperative local abnormal event records from the medical information system, and performs time stamping, anatomical location coding, and status feature coding to obtain structured medical data. Based on the structured medical data, a surgical area perturbation map is constructed, which includes surgical area operation nodes, adjacent neural nodes, and node relationships. Pain migration units are generated based on adjacent postoperative pain records, and multiple pain migration units are connected to form a pain migration path. The pain migration path is matched with the surgical area perturbation map for node matching and continuity retrieval to extract candidate neural association paths; Abnormal event coverage data is generated based on the postoperative local abnormal event records, and the abnormal event coverage data is used to segment the pain migration units in the candidate neural association pathways to obtain fully covered units, uncovered units, and partially covered units. Based on the distribution of the uncovered and partially covered units in the surgical area perturbation map, candidate neural association intervals are generated, and intermediate result data containing the candidate neural association intervals and their data source identifiers are output.

2. The method for assessing neuropathic pain based on medical big data according to claim 1, characterized in that, The process involves retrieving pre-existing postoperative patient surgical procedure records, postoperative pain records, and postoperative local abnormal event records from the medical information system, and then performing time stamping, anatomical location coding, and state feature coding to obtain structured medical data, including: The surgical time, surgical procedure name, surgical area description, and operation duration are extracted from the surgical procedure record. The surgical area description is mapped to a unified anatomical site code to obtain structured surgical data. The recording time, pain location, pain nature, and pain intensity are extracted from the postoperative pain records. The pain location is mapped to the unified anatomical site code, and the pain nature is converted into a pain nature category code to obtain structured pain data. The location, start time, end time and abnormal state of the abnormal event are extracted from the postoperative local abnormal event record, and the location of the abnormal event is mapped to the unified anatomical site code to obtain structured abnormal event data. Establish patient identification associations and time-based associations among the structured surgical data, structured pain data, and structured abnormal event data.

3. The method for assessing neuropathic pain based on medical big data according to claim 1, characterized in that, The construction of a surgical area perturbation map based on the structured medical data, including surgical area operation nodes, adjacent neural nodes, and node relationships, includes: The structured surgical data was divided into multiple operation segments according to the operation time, and each operation segment was converted into a surgical area operation node. Based on the time interval between adjacent surgical area operation nodes and the adjacency relationship of anatomical sites, a directed operation edge is generated between surgical area operation nodes. Query the mapping table between anatomical regions and nerve innervation regions, generate neighboring nerve nodes for each surgical region operation node, and generate region mapping edges between the surgical region operation node and neighboring nerve nodes. Based on the anatomical adjacency relationship between the nerve innervation areas corresponding to adjacent nerve nodes, nerve adjacency edges between adjacent nerve nodes are generated to form the surgical area perturbation map.

4. The method for assessing neuropathic pain based on medical big data according to claim 1, characterized in that, The step of generating pain migration units based on adjacent postoperative pain records and connecting multiple pain migration units into a pain migration path includes: Two postoperative pain records with adjacent times are combined into a pain record pair. The code of the previous pain location in the pain record pair is used as the migration starting point, and the code of the next pain location is used as the migration ending point. Query the anatomical site adjacency data, generate the site adjacency sequence between the migration start point and the migration end point, and write the site adjacency sequence into the pain migration unit; Pain state features are generated based on changes in pain nature category coding and changes in pain intensity parameters in the pain record pairs; Based on the time interval between adjacent pain migration units, the location adjacency relationship, and the pain state characteristics, a migration continuity identifier is generated, and multiple pain migration units are connected into a pain migration path according to the migration continuity identifier.

5. The method for assessing neuropathic pain based on medical big data according to claim 1, characterized in that, The step of performing node matching and continuity retrieval between the pain migration path and the surgical area perturbation map to extract candidate neural pathways includes: The pain migration units in the pain migration path are matched with the surgical area operation nodes in the surgical area perturbation map by location encoding and time interval matching to obtain the matching nodes. Using the matching node as the starting point for retrieval, the corresponding neighboring neural nodes are read along the region mapping edge, and a sequence of neighboring neural nodes corresponding to the pain migration direction is generated along the neural adjacency edge. Generate neural continuity identifiers based on the sequence of neighboring neural nodes corresponding to continuous pain migration units; According to the neural continuity identifier, continuous pain migration units and their corresponding neighboring neural node sequences are written into the candidate neural association path, and path break identifiers are written at the locations where the neural continuity identifier is interrupted.

6. The method for assessing neuropathic pain based on medical big data according to claim 1, characterized in that, The generation of abnormal event coverage data based on postoperative local abnormal event records includes: Based on the location of the abnormal event location code in the structured abnormal event data in the adjacent data of the anatomical region, generate the location coverage range; Generate a time coverage range based on the start and end times of the abnormal event; Based on the mapping relationship between abnormal event types and pain nature categories, generate state coverage; The location coverage, time coverage, and status coverage of the same abnormal event are associated and stored together to form abnormal event coverage data; For multiple abnormal events with overlapping coverage areas and overlapping time coverage areas, separate event source identifiers are set.

7. The method for assessing neuropathic pain based on medical big data according to claim 1, characterized in that, The step of using the abnormal event coverage data to segment the pain migration units in the candidate neural pathways to obtain fully covered units, uncovered units, and partially covered units includes: For pain migration units in candidate neural pathways, generate location coverage vectors, time coverage vectors, and state coverage vectors between them and anomalous event coverage data; Write the pain migration unit of the complete coverage unit set into which the location coverage vector, time coverage vector and state coverage vector all point to the same abnormal event coverage data. Write the pain migration units that do not have any abnormal event coverage data in any dimension of location, time or state into the set of uncovered units; Pain migration units that do not belong to the complete coverage unit but have one-dimensional or two-dimensional coverage are segmented according to the coverage boundary and written into the partial coverage unit set. Fully covered units, uncovered units, and partially covered units are associated with and stored as anomaly source identifier and neighboring neural node identifier, respectively.

8. The method for assessing neuropathic pain based on medical big data according to claim 1, characterized in that, The step involves generating candidate neural association regions based on the distribution of uncovered and partially covered units in the surgical area perturbation map, and outputting intermediate result data containing the candidate neural association regions and their data source identifiers, including: Read continuously arranged uncovered units along the candidate neural association path, and merge them into the first association interval based on the continuity relationship of adjacent neural nodes and the direction of pain migration; Read the segments in the partial coverage unit that are not covered by abnormal events, and merge them into a second associated interval according to the connection relationship of the corresponding neighboring neural nodes; The first association interval was identified as the primary candidate neural association interval, and the second association interval was identified as the candidate neural association interval to be reviewed. Add surgical site operation node identifiers, pain record identifiers, and adjacent nerve node identifiers to the main candidate nerve association intervals, and further add abnormal event source identifiers to the candidate nerve association intervals to be reviewed; The output includes intermediate result data containing the main candidate neural association intervals, candidate neural association intervals to be reviewed, fully covered units, and their data source identifiers.

9. A neuropathic pain assessment system based on medical big data, characterized in that, include: The data acquisition module is used to retrieve existing postoperative patient surgical procedure records, postoperative pain records, and postoperative local abnormal event records from the medical information system; The structured processing module is used to perform time stamping, anatomical location encoding, and state feature encoding on the surgical procedure record, postoperative pain record, and postoperative local abnormal event record to obtain structured medical data. The perturbation graph construction module is used to construct a perturbation graph of the surgical area based on the structured medical data, which includes surgical area operation nodes, adjacent nerve nodes and node relationships. The pain path generation module is used to generate pain migration units based on adjacent postoperative pain records and connect multiple pain migration units into a pain migration path. The association path extraction module is used to perform node matching and continuity retrieval between the pain migration path and the surgical area perturbation map to extract candidate neural association paths; An abnormal coverage generation module is used to generate abnormal event coverage data based on postoperative local abnormal event records; a hierarchical processing module is used to use the abnormal event coverage data to segment the pain migration units in the candidate neural association pathways to obtain fully covered units, uncovered units, and partially covered units. The result output module is used to generate candidate neural association intervals based on the distribution of the uncovered and partially covered units in the surgical area perturbation map, and output intermediate result data containing the candidate neural association intervals and their data source identifiers.

10. An electronic device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it implements the neuropathic pain assessment method based on medical big data as described in any one of claims 1 to 8.

Images