Airway stenosis treatment data standardization storage method

Abstract

This invention relates to the field of medical informatics and medical data processing technology, and discloses a standardized storage method for airway stenosis treatment data. By converting a three-dimensional medical image of the airway into a topological graph, and sequentially constructing an adjacency matrix, a node attribute matrix, and an edge attribute matrix, the multidimensional structural features of each node are calculated, and a standardized permutation matrix is ​​generated through normalized sorting. Then, a unified standardized structural encoding vector is formed through rearrangement and vectorization concatenation. Standardized storage records are generated by combining patient identification and time, and fingerprint verification is performed. This solves the problems of incomplete preservation, difficulty in retrieval, and easy loss of context in traditional clinical data storage. This method restores the structure without secondary reconstruction through matrix and vectorization processing; unified sorting eliminates differences in node order among different cases; the high-dimensional encoding after vectorization saves storage space and is compatible with vector retrieval libraries; and fingerprint verification ensures data integrity and security.

Description

Technical Field

[0001] This invention relates to the field of medical informatics and medical data processing technology, specifically a standardized storage method for airway stenosis treatment data. Background Technology

[0002] In the current diagnosis and treatment of airway stenosis, examinations, assessments, and interventional procedures generate heterogeneous data from multiple sources, including 3D imaging, bronchoscopic findings, stenosis location and grading, treatment device parameters, and follow-up results. Existing technologies typically employ parallel storage of image files and text records: the image side primarily uses DICOM and its derivatives, while the text side mainly uses structured forms or medical records, with some systems using generic codes or template fields for archiving. While this approach facilitates data collection and retrospection, it lacks a unified and reproducible standardized representation of the crucial airway anatomy, making it difficult to consistently compare and reliably correlate data from different time points, different devices, or different institutions for the same patient.

[0003] In existing technologies, airway structural information is often fixed using methods such as "segment name + description" or "screenshot annotation + text location," or stored separately as 3D reconstruction model files. The former relies on the operator's subjective description and naming habits, easily leading to problems such as different segments with the same name, different names for the same segment, and inconsistent recording granularity. The latter, while preserving geometric details, suffers from large model file sizes, diverse formats, and a lack of unified indexing. Furthermore, it is often difficult to ensure that node numbers and topological order are consistent during cross-reconstruction, resulting in difficulties in structural alignment. In addition, the attributes of edges and nodes (such as stenosis-related measurements, lumen morphology parameters, bifurcation characteristics, etc.) are often scattered across different tables or report paragraphs, lacking a stable binding relationship with anatomical structures. This leads to a large amount of manual secondary processing required for subsequent statistical analysis, efficacy comparison, device selection, and pathway planning. Regarding data consistency verification, existing systems mostly rely on file check codes or database primary key constraints, which can only ensure that files have not been tampered with or records have not been inserted repeatedly. It is difficult to determine whether "the same structure is equivalently represented" or "whether there are structural coding alignment deviations at different time points for the same patient." In terms of data retrieval, existing technologies mostly use fields such as patient ID and examination time for indexing, which makes it difficult to support rapid retrieval and cross-record comparison based on anatomical structure and stenosis topology.

[0004] To this end, this case aims to propose the following: First, construct a graph structure reflecting the anatomical topology of the airway based on 3D medical images, and assign a stable original number to each anatomical node in the graph. Then, transform the graph into an adjacency matrix, a node attribute matrix, and an edge attribute matrix. Next, calculate the multidimensional structural features of each node based on the matrix content and organize them into a feature matrix to determine the standard order of the nodes. A permutation matrix is ​​used to achieve a unified rearrangement of the matrices. Then, the three types of rearranged matrices are vectorized and concatenated to form a unified high-dimensional structural encoding vector, and standardized storage records are constructed by combining patient identifiers, record identifiers, and relative time information. Finally, fingerprint calculation and joint indexing are used to ensure data integrity, uniqueness, and rapid retrieval. Summary of the Invention

[0005] This invention provides a standardized storage method for airway stenosis treatment data, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for standardized storage of airway stenosis treatment data, comprising:

[0007] An airway structure map representing the anatomical topology of the airway is constructed based on airway medical images, and original node numbers are assigned to the airway anatomical nodes.

[0008] Based on the airway structure diagram, an original adjacency matrix representing the node connection relationship is constructed, and a node attribute matrix recording node attribute information and an edge attribute matrix recording edge attribute information are constructed respectively.

[0009] The structural features of each node are calculated based on the original adjacency matrix and the node attribute matrix, and the structural features of each node are combined into a structural feature matrix.

[0010] Perform a canonical sort on all nodes based on the structural feature matrix, and construct a permutation matrix that reflects the canonical sort result;

[0011] The original adjacency matrix, node attribute matrix, and edge attribute matrix are normalized and rearranged using a permutation matrix to obtain a normalized adjacency matrix, normalized node attribute matrix, and normalized edge attribute matrix with uniform node order.

[0012] The standardized adjacency matrix, standardized node attribute matrix, and standardized edge attribute matrix are vectorized and encoded, and the resulting multiple encoded vectors are concatenated into a unified standardized structure encoded vector according to a preset order.

[0013] The standardized structure encoding vector is combined with the patient's unique ID, the record's unique ID, and the time interval relative to the patient's first record to form a standardized storage record, and a composite index key is established for the standardized storage record;

[0014] The standardized fingerprint value is calculated based on the standardized structure encoding vector. The encoding integrity is verified on the standardized storage record. The standardized structure encoding vector is restored into a standardized adjacency matrix, a standardized node attribute matrix, and a standardized edge attribute matrix according to the preset decoding rules.

[0015] Optionally, the step of constructing an airway structure map representing the anatomical topology of the airway based on airway medical images, and assigning original node numbers to airway anatomical nodes, specifically includes:

[0016] All airway anatomical nodes are obtained from the three-dimensional medical images of the airway, and a unique original node number is assigned to each airway anatomical node in a preset order.

[0017] All airway anatomical nodes are compiled into a node set according to the original node number, so that each element in the node set corresponds one-to-one with an airway anatomical node.

[0018] Select any two airway anatomical nodes from the node set. If there is a direct anatomical connection, establish a directed edge between the two nodes from the starting node to the ending node, and summarize all directed edges into an edge set.

[0019] The set of nodes and the set of edges are combined to form an airway structure diagram. In the airway structure diagram, the original node number is used to identify the airway anatomical nodes, and the original starting node number and the original ending node number of the directed edges are used to identify the airway anatomical connection relationship.

[0020] Optionally, the step of constructing an original adjacency matrix representing node connections based on the airway structure diagram, and constructing a node attribute matrix recording node attribute information and an edge attribute matrix recording edge attribute information, specifically includes:

[0021] In the airway structure diagram, the nodes in the node set are used as the row index and column index of the original adjacency relation matrix to construct the original adjacency relation matrix. When there is a directed edge from the corresponding row node to the corresponding column node, the matrix element is recorded as connected; when there is no directed edge, the matrix element is recorded as not connected.

[0022] For each airway anatomical node, the attribute information of each airway anatomical node is obtained, and the attributes are arranged in a fixed order to form a node attribute vector, which contains multiple attribute components.

[0023] Arrange the node attribute vectors of all nodes in the order of their original node numbers into rows of the node attribute matrix, so that each column of the node attribute matrix corresponds one-to-one with a node attribute component.

[0024] Select all matrix elements in the original adjacency matrix that indicate the existence of directed edges, and arrange them in alphabetical order according to row index from smallest to largest and column index from smallest to largest within the same row. During the arrangement process, count the total number of directed edges, and assign a unique original edge number to each pair of original start node number and original end node number that have directed edges.

[0025] Record the original starting node number and the original ending node number corresponding to each original edge number to form a one-to-one correspondence between the original edge number and the original starting node number and the original ending node number.

[0026] For each directed edge, obtain the attribute information of the corresponding directed edge, arrange the edge attributes in a fixed order to form an edge attribute vector, and arrange all edge attribute vectors in the order of the original edge numbers to form the rows of the edge attribute matrix.

[0027] Optionally, the step of calculating the structural features of each node based on the original adjacency matrix and node attribute matrix, and combining the structural features of each node into a structural feature matrix, specifically includes:

[0028] For each node, select all matrix elements of the corresponding row in the original adjacency matrix, count the number of matrix elements that indicate the existence of directed edges, and use the counted number of matrix elements as the degree of the corresponding node.

[0029] For each node, based on the original adjacency matrix, identify all matrix elements in the corresponding node's row that represent the existence of directed edges. Use these matrix elements to determine all adjacent nodes starting from the corresponding node, obtain the degree of each adjacent node, and sum the degrees of all adjacent nodes to obtain the total degree of the adjacent nodes of the corresponding node.

[0030] For each node, based on the original adjacency relation matrix, identify all matrix elements in the corresponding node's row that indicate the existence of directed edges. Use these matrix elements to determine all adjacent nodes starting from the corresponding node. Sum the node attribute vector of each adjacent node according to each attribute component, and combine the summation results of each attribute component into the adjacent node attribute summation vector of the corresponding node.

[0031] For each node, the node degree, the sum of the degrees of adjacent nodes, and the sum of the attributes of adjacent nodes are combined in a fixed order to form the structural feature vector of the corresponding node, so that each component of the structural feature vector corresponds to the node degree, the sum of the degrees of adjacent nodes, and the sum of the attributes of adjacent nodes, respectively.

[0032] Following the original node index order, the structural feature vectors of all nodes are arranged sequentially into rows of the structural feature matrix, so that each row of the structural feature matrix corresponds to the structural feature of a node, and each column corresponds to a structural feature component.

[0033] Optionally, the step of performing a canonical sort on all nodes based on the structural feature matrix and constructing a permutation matrix reflecting the canonical sort result specifically includes:

[0034] Based on the structural feature matrix, a sorting operation is performed on the structural feature vectors of all nodes. During sorting, the first structural feature component in the structural feature vector is compared first. If the first structural feature components are the same, the second structural feature component is compared. If the second structural feature components are the same, the third structural feature component is compared. If all structural feature components are the same, they are arranged in ascending order according to the original node number to obtain the normal sorting order of the nodes.

[0035] A node sorting index vector is constructed according to the standard sorting order. In the node sorting index vector, each position records the original node number corresponding to each sorting position, so that the node sorting index vector fully reflects the standard sorting result.

[0036] Construct a permutation matrix based on the node sorting index vector. The permutation matrix is ​​a square matrix. Each row of the matrix corresponds to the normal sorting position, and each column corresponds to the original node index. The matrix element is assigned a value of 1 at the position where the normal sorting position matches the original node index, and a value of 0 at the other positions, so that there is only one matrix element with a value of 1 in each row and each column of the permutation matrix.

[0037] Optionally, the step of normalizing and rearranging the original adjacency relation matrix, node attribute matrix, and edge attribute matrix using a permutation matrix to obtain a normalized adjacency relation matrix, normalized node attribute matrix, and normalized edge attribute matrix with unified node order specifically includes:

[0038] Based on the permutation matrix, the rows and columns of the original adjacency relation matrix are rearranged simultaneously according to the normal sort order to obtain the normal adjacency relation matrix, so that the row index and column index of the normal adjacency relation matrix correspond to the normal sort order;

[0039] Based on the permutation matrix, the rows of the node attribute matrix are rearranged according to the normal sorting order to obtain the normal node attribute matrix, so that each row of the normal node attribute matrix corresponds one-to-one with a node after normal sorting.

[0040] The normal adjacency matrix is ​​scanned sequentially in ascending order of row index and column index within the same row. When a matrix element is found to indicate the existence of a directed edge, a normal edge number is assigned to each detected directed edge in the order of scanning. At the same time, the number of normal edge numbers assigned is counted during the scanning process.

[0041] For each canonical edge number, obtain the canonical sorting position of the starting node and the canonical sorting position of the ending node corresponding to each canonical edge according to the canonical sorting order, and find the original starting node number and the original ending node number corresponding to them through the node sorting index vector.

[0042] In the original adjacency matrix, all matrix elements are scanned in ascending order of row index and ascending order of column index within the same row. The number of matrix elements that indicate the existence of directed edges before reaching the corresponding positions of the original starting node number and the original ending node number is counted. The counted number is incremented by one to obtain the original edge number of the corresponding directed edge. The corresponding edge attribute vector is then searched in the edge attribute matrix based on the original edge number.

[0043] In the normalized edge attribute matrix, the normalized edge number is selected as the row index, and the edge attribute vector corresponding to each normalized edge number is written into the corresponding row of the normalized edge attribute matrix, so that the arrangement order of each row in the normalized edge attribute matrix is ​​consistent with the lexicographical order of the edges in the normalized adjacency relation matrix.

[0044] Optionally, the step of vectorizing the standardized adjacency matrix, standardized node attribute matrix, and standardized edge attribute matrix, and concatenating the resulting multiple encoded vectors into a unified standardized structure encoded vector according to a preset order, specifically includes:

[0045] The matrix elements of the normal adjacency relation matrix are read sequentially in column index priority order. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form an adjacency relation encoding vector.

[0046] The matrix elements of the standard node attribute matrix are read sequentially according to the column index priority. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form a node attribute encoding vector.

[0047] The matrix elements of the standard edge attribute matrix are read sequentially according to column index priority. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form the edge attribute encoding vector.

[0048] The adjacency relationship encoding vector, node attribute encoding vector, and edge attribute encoding vector are concatenated in a preset order to obtain a unified standardized structure encoding vector.

[0049] Optionally, the step of combining the standardized structure encoding vector with the patient's unique ID, the record's unique ID, and the time interval relative to the patient's first record to form a standardized storage record, and establishing a composite index key for the standardized storage record, specifically includes:

[0050] For each examination or treatment record related to airway stenosis, obtain the patient's unique ID, the record's unique ID, the time interval relative to the patient's first record, and the standardized structure encoding vector of the corresponding record, and combine these four pieces of information into a standardized storage record.

[0051] A composite index key is constructed using the patient's unique ID and the time interval relative to the patient's first record. The composite index key is then used to establish an index relationship with the standardized storage record, and the standardized storage record is written into the storage medium. This allows the corresponding standardized storage record to be retrieved using the patient's unique ID and the time interval relative to the patient's first record.

[0052] Optionally, the step of calculating the standardized fingerprint value based on the standardized structure encoding vector, performing encoding integrity verification on the standardized storage record, and restoring the standardized structure encoding vector into a standardized adjacency matrix, a standardized node attribute matrix, and a standardized edge attribute matrix according to a preset decoding rule specifically includes:

[0053] For each standardized structure encoding vector in a standardized storage record, the value of each component is obtained sequentially according to the position order of each component in the vector. A fixed weight coefficient and the position number of each component in the encoding vector are assigned to each component. The value of each component is multiplied by the corresponding weight coefficient and position number, and all products are summed to obtain the standardized fingerprint value of the corresponding encoding vector.

[0054] When the standardized storage process is executed again on the same examination record or treatment record, a new standardized structure encoding vector corresponding to the same examination record or treatment record is generated. The value of each component is obtained in the order of its position in the vector. A fixed weight coefficient and the position number of each component in the encoding vector are assigned to each component. The value of each component is multiplied by the corresponding weight coefficient and position number, and all products are summed to obtain the recalculated fingerprint value. The recalculated fingerprint value is compared with the standardized fingerprint value when it was first written. If the two standardized fingerprint values ​​are equal, the same examination record or treatment record is regarded as a record that has been written before. Only the original record content is updated and no new standardized storage record is added.

[0055] When the recalculated fingerprint value obtained by recalculating a certain standardized storage record is not equal to the original standardized fingerprint value of the corresponding standardized storage record, the standardized storage record is marked as an illegal record, and write and update operations are refused to be performed on the standardized storage record.

[0056] When two examination records or treatment records from different sources yield the same standardized fingerprint value after calculating the standardized fingerprint value, these two examination records or treatment records are marked as structurally equivalent records. The structurally equivalent marking indicates that the two records have the same airway structure map information at the level of standardized structural coding vector.

[0057] During the standard decoding process, the standardized structure encoding vector is divided according to the number of rows and columns of the standard adjacency matrix, the number of rows and columns of the standard node attribute matrix, and the number of rows and columns of the standard edge attribute matrix. The encoding vector is divided into three segments in a pre-defined order: adjacency encoding vector, node attribute encoding vector, and edge attribute encoding vector.

[0058] The adjacency relationship encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal adjacency relationship matrix; the node attribute encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal node attribute matrix; the edge attribute encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal edge attribute matrix, thus completing the reverse reconstruction of the normal airway structure diagram.

[0059] The present invention has the following beneficial effects:

[0060] 1. First, all anatomical nodes are extracted from 3D airway medical images and assigned stable original node numbers, achieving precise digitization of spatial structural information. Unlike traditional methods that rely solely on planar slicing or 2D skeletonization, this approach preserves the complete topological relationships, enabling distortion-free reflection of airway branch connections and hierarchical structures in subsequent analyses. Integrating graphic construction and automatic node numbering reduces manual intervention and improves data processing consistency. A standardized airway structure diagram is generated for each patient, facilitating cross-case comparisons and retrospective analysis; it also lays a reliable foundation for subsequent matrixing and coding processes.

[0061] 2. After constructing the graph structure, this scheme builds an adjacency matrix representing the connections between nodes, a node attribute matrix recording various node attributes, and an edge attribute matrix recording edge attributes. Edge attribute information is independently abstracted into a matrix dimension and processed synchronously with node attributes. Each directed edge is assigned a unique original edge number according to a unified lexicographical order, ensuring the traceability of edge attributes in the matrix and subsequent vectorization stages. Compared to traditional methods that typically only focus on adjacency relationships or node attributes, this scheme, through parallel management of three matrices, not only retains richer anatomical and physical detection information but also avoids information loss in subsequent encoding stages. When querying, comparing, or replaying the entire airway map, the original image does not need to be accessed again; the complete structure can be reconstructed through the matrices and their numbers, improving storage and retrieval efficiency.

[0062] 3. For the original adjacency relation and node attribute matrix, this scheme calculates a multi-dimensional structural feature vector, including node degree, the sum of the degrees of adjacent nodes, and the sum of the attributes of adjacent nodes. Based on this, a multi-level lexicographical sort is performed on all nodes, ultimately generating a permutation matrix. It is not limited to a single topological indicator but combines attribute information to form a composite vector, achieving a more discriminative node sorting. The sorting process employs multi-level comparisons, ensuring the stability and repeatability of the sorting results, allowing nodes from different cases with the same structural features to obtain consistent sorting positions. This provides a clear and unambiguous index for subsequent unified matrix rearrangement, enabling any node or edge to be accurately located according to the standardized sorting position in subsequent encoding and decoding processes, reducing the complexity of structure reconstruction.

[0063] 4. Based on the permutation matrix, this scheme simultaneously normalizes and rearranges the original adjacency relation matrix, node attribute matrix, and edge attribute matrix, unifying the row and column indices of the three types of matrices into the same sorting system. Subsequently, the rearranged matrices are vectorized in column-major order and concatenated into a unified structure encoding vector according to a preset order. This integrated processing of the three types of matrices avoids the problem of synchronizing node and edge information after traditional step-by-step vectorization; and the reversible vectorization-reconstruction design allows the vector to be directly restored to a three-matrix structure. It simplifies storage dimensions, compressing complex graph data into an efficient one-dimensional vector form; it also supports accurate restoration, improving the flexibility of database retrieval and model analysis. Compared to existing methods that only store adjacency matrices or graph embedding, this scheme balances attribute information and reversibility.

[0064] 5. After obtaining the structural encoding vector, this solution combines it with the patient's unique ID, the record's unique ID, and relative time information to construct a standardized storage record. A standardized fingerprint is generated based on the product of the encoding vector's position and weight coefficients, used for integrity verification of each record. Structural information is integrated with business metadata to form a jointly indexable storage unit. A fingerprint verification mechanism is introduced to ensure data consistency and tamper-proof capabilities during writing, updating, or recalculation. It supports idempotent writing to the same record, avoiding duplicate storage; it can detect and reject illegal or corrupted records; and it can identify structurally equivalent records, facilitating deduplication and fusion of multi-source heterogeneous inputs. Unlike traditional methods relying solely on primary key constraints or transaction logs, this solution provides a lighter and more efficient data integrity guarantee through internal vector fingerprints.

[0065] 6. Compared to common graph database solutions or pure vector embedding methods in existing technologies, this solution, through an integrated "graph-matrix-vector" process, not only preserves complete anatomical topology and attribute information but also achieves efficient and reversible structural encoding and retrieval. This is particularly crucial in large-scale clinical data management and AI-assisted diagnosis: it reduces storage costs while ensuring accurate playback of structural details, supporting rapid comparison, quality control, and automated analysis. Overall, this solution effectively solves the challenge of heterogeneous fusion of multi-source medical image data, providing systematic and scalable technical support for graph-based clinical decision-making and research. Attached Figure Description

[0066] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0067] The technical solutions of 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.

[0068] Example, refer to Figure 1 A standardized storage method for airway stenosis treatment data, comprising:

[0069] An airway structure map representing the anatomical topology of the airway is constructed based on airway medical images, and original node numbers are assigned to the airway anatomical nodes.

[0070] Based on the airway structure diagram, an original adjacency matrix representing the node connection relationship is constructed, and a node attribute matrix recording node attribute information and an edge attribute matrix recording edge attribute information are constructed respectively.

[0071] The structural features of each node are calculated based on the original adjacency matrix and the node attribute matrix, and the structural features of each node are combined into a structural feature matrix.

[0072] Perform a canonical sort on all nodes based on the structural feature matrix, and construct a permutation matrix that reflects the canonical sort result;

[0073] The original adjacency matrix, node attribute matrix, and edge attribute matrix are normalized and rearranged using a permutation matrix to obtain a normalized adjacency matrix, normalized node attribute matrix, and normalized edge attribute matrix with uniform node order.

[0074] The standardized adjacency matrix, standardized node attribute matrix, and standardized edge attribute matrix are vectorized and encoded, and the resulting multiple encoded vectors are concatenated into a unified standardized structure encoded vector according to a preset order.

[0075] The standardized structure encoding vector is combined with the patient's unique ID, the record's unique ID, and the time interval relative to the patient's first record to form a standardized storage record, and a composite index key is established for the standardized storage record;

[0076] The standardized fingerprint value is calculated based on the standardized structure encoding vector. The encoding integrity is verified on the standardized storage record. The standardized structure encoding vector is restored into a standardized adjacency matrix, a standardized node attribute matrix, and a standardized edge attribute matrix according to the preset decoding rules.

[0077] This method directly converts 3D medical images of the airway into a topological graph, then sequentially constructs adjacency matrices, node attribute matrices, and edge attribute matrices. Next, it calculates the multidimensional structural features of nodes and generates a permutation matrix by sorting them lexicographically. The three matrices are then uniformly rearranged and quantized and concatenated into a long encoded vector. Finally, standardized storage records are generated by combining patient identifiers and timestamps, and fingerprint verification is performed. This method solves the problems of incomplete preservation, difficulty in retrieval, and easy loss of context in traditional clinical data storage. First, graph structure and attributes are matrixed and vectorized with equal importance, allowing structure recovery without secondary graph reconstruction. Second, unified sorting and rearrangement eliminate storage heterogeneity caused by differences in node order among different cases. Third, the high-dimensional encoding after vectorization reduces storage space and is compatible with existing vector retrieval libraries. Finally, fingerprint verification ensures the idempotency and integrity of the write and update processes, automatically detecting and rejecting illegal or duplicate writes, greatly improving data quality and security. Unlike existing methods that only store adjacency lists or perform secondary graph construction, this solution integrates graph, matrix, and vector processing, preserving complete structural information while supporting efficient indexing and restoration.

[0078] The construction of an airway structure map representing the anatomical topology of the airway based on airway medical images, and the assignment of original node numbers to airway anatomical nodes, specifically includes:

[0079] All airway anatomical nodes are obtained from the three-dimensional medical images of the airway, and a unique original node number is assigned to each airway anatomical node in a preset order.

[0080] All airway anatomical nodes are compiled into a node set according to the original node number, so that each element in the node set corresponds one-to-one with an airway anatomical node.

[0081] Select any two airway anatomical nodes from the node set. If there is a direct anatomical connection, establish a directed edge between the two nodes from the starting node to the ending node, and summarize all directed edges into an edge set.

[0082] The set of nodes and the set of edges are combined to form an airway structure diagram. In the airway structure diagram, the original node number is used to identify the airway anatomical nodes, and the original starting node number and the original ending node number of the directed edges are used to identify the airway anatomical connection relationship.

[0083] Import anatomical nodes from the 3D medical image of the airway, and record the total number of anatomical nodes as follows: , will the The node number is ;in, This is the index of the node in the node set; For the first One airway anatomical node;

[0084] Construct a set of nodes as ;

[0085] At any two nodes , If there is a direct anatomical connection between them, then they are constructed as directed edges. , forming an edge set for:

[0086] ;in, For node sequence number; For the first One airway anatomical node; For the node Pointing to node The directed edges;

[0087] The diagram of the airway structure is as follows .

[0088] The step of constructing an original adjacency matrix representing node connections based on the airway structure diagram, and constructing a node attribute matrix recording node attribute information and an edge attribute matrix recording edge attribute information, specifically includes:

[0089] In the airway structure diagram, the nodes in the node set are used as the row index and column index of the original adjacency relation matrix to construct the original adjacency relation matrix. When there is a directed edge from the corresponding row node to the corresponding column node, the matrix element is recorded as connected; when there is no directed edge, the matrix element is recorded as not connected.

[0090] For each airway anatomical node, the attribute information of each airway anatomical node is obtained, and the attributes are arranged in a fixed order to form a node attribute vector, which contains multiple attribute components.

[0091] Arrange the node attribute vectors of all nodes in the order of their original node numbers into rows of the node attribute matrix, so that each column of the node attribute matrix corresponds one-to-one with a node attribute component.

[0092] Select all matrix elements in the original adjacency matrix that indicate the existence of directed edges, and arrange them in alphabetical order according to row index from smallest to largest and column index from smallest to largest within the same row. During the arrangement process, count the total number of directed edges, and assign a unique original edge number to each pair of original start node number and original end node number that have directed edges.

[0093] Record the original starting node number and the original ending node number corresponding to each original edge number to form a one-to-one correspondence between the original edge number and the original starting node number and the original ending node number.

[0094] For each directed edge, obtain the attribute information of the corresponding directed edge, arrange the edge attributes in a fixed order to form an edge attribute vector, and arrange all edge attribute vectors in the order of the original edge numbers to form the rows of the edge attribute matrix.

[0095] Constructing an adjacency matrix Its elements are specifically:

[0096] ;in, For the node To the node Does an edge exist?

[0097] Construct the attribute vector for each node. ;in, For the first The attribute vector of each node; for 3D real vector; The number of components in the attribute vector for each node;

[0098] Each node attribute vector It is a vector containing the features of the node, representing the node. A quantitative description of various structural or biological shapes, these attributes may come from the geometric information of the airway structure, pathological examination results or other quantitative medical indicators; for example: the coordinates of the node location, indicating the specific location of the node in the airway; the geometric features of the node such as diameter, shape, and size; the connectivity features of adjacent nodes; and clinical data-related attributes such as airway stenosis, blood flow velocity, and clinical symptom indicators.

[0099] Constructing the node attribute matrix The node attribute matrix Behavior : , ; ;in, Number the components of the node attribute vector; For the first The node of the first Each attribute value; For nodes attribute vector The One component;

[0100] For all satisfied directed edges Node index First press From childhood to adulthood, then by Number them in ascending order;

[0101] The total number of edges is calculated as follows ;in, , This is an index variable used to iterate through the elements of the adjacency matrix; For indicator functions;

[0102] For each group, satisfy Node index pairs The edge sequence number is:

[0103] ;in, In the original adjacency matrix In the middle, satisfy Edge-to-node index Deterministic ordinal number in lexicographical order;

[0104] The lexicographical order includes node indexes. First press From childhood to adulthood, then by Number them in ascending order;

[0105] And number the edges ,side satisfy: , ;in, Index the edge numbers; For the number The edge; For the first The starting node number of the edge; For the first The index of the endpoint node of the edge;

[0106] For each edge Constructing edge attribute vectors Constructing the edge attribute matrix The elements in the attribute matrix are: , , ;in, For the first The edge attribute vector of a single edge; The component index of the edge attribute vector; For the first The first edge Each attribute value; for The Each component.

[0107] edge attribute vector It represents an edge The feature vector of each edge Connecting two nodes, edge attribute vector Various related attributes are used to describe this edge. For example, edge attributes may include: edge length, which represents the physical distance between two connected nodes, or the path length between the two nodes; flow rate, which represents the speed at which gas flows within the airway, and may be related to clinical measurements or image reconstruction data; pressure difference, if the edge represents a narrow region of the airway, the edge attributes can describe the pressure difference or other hydrodynamic characteristics of that region; and other clinical features, such as the width, thickness, or relevance to a specific treatment method of the edge.

[0108] The step of calculating the structural features of each node based on the original adjacency matrix and node attribute matrix, and combining the structural features of each node into a structural feature matrix, specifically includes:

[0109] For each node, select all matrix elements of the corresponding row in the original adjacency matrix, count the number of matrix elements that indicate the existence of directed edges, and use the counted number of matrix elements as the degree of the corresponding node.

[0110] For each node, based on the original adjacency matrix, identify all matrix elements in the corresponding node's row that represent the existence of directed edges. Use these matrix elements to determine all adjacent nodes starting from the corresponding node, obtain the degree of each adjacent node, and sum the degrees of all adjacent nodes to obtain the total degree of the adjacent nodes of the corresponding node.

[0111] For each node, based on the original adjacency relation matrix, identify all matrix elements in the corresponding node's row that indicate the existence of directed edges. Use these matrix elements to determine all adjacent nodes starting from the corresponding node. Sum the node attribute vector of each adjacent node according to each attribute component, and combine the summation results of each attribute component into the adjacent node attribute summation vector of the corresponding node.

[0112] For each node, the node degree, the sum of the degrees of adjacent nodes, and the sum of the attributes of adjacent nodes are combined in a fixed order to form the structural feature vector of the corresponding node, so that each component of the structural feature vector corresponds to the node degree, the sum of the degrees of adjacent nodes, and the sum of the attributes of adjacent nodes, respectively.

[0113] Following the original node index order, the structural feature vectors of all nodes are arranged sequentially into rows of the structural feature matrix, so that each row of the structural feature matrix corresponds to the structural feature of a node, and each column corresponds to a structural feature component.

[0114] For each node Perform steps S301 to S303 to construct the first... The structural feature vector of each node Specifically:

[0115] S301, Node Degree: ;in, For nodes The first structural feature component;

[0116] S302, Sum of the degrees of adjacent nodes: ;in, For nodes The second structural feature component; This is an index variable used to iterate through the first... Row adjacency;

[0117] S303, Adjacent Node Attributes and: ;in, For nodes The third structural feature component;

[0118] Constructing the structural feature matrix The first structural feature matrix row element is .

[0119] The step of performing a canonical sort on all nodes based on the structural feature matrix and constructing a permutation matrix reflecting the canonical sort result specifically includes:

[0120] Based on the structural feature matrix, a sorting operation is performed on the structural feature vectors of all nodes. During sorting, the first structural feature component in the structural feature vector is compared first. If the first structural feature components are the same, the second structural feature component is compared. If the second structural feature components are the same, the third structural feature component is compared. If all structural feature components are the same, they are arranged in ascending order according to the original node number to obtain the normal sorting order of the nodes.

[0121] A node sorting index vector is constructed according to the standard sorting order. In the node sorting index vector, each position records the original node number corresponding to each sorting position, so that the node sorting index vector fully reflects the standard sorting result.

[0122] Construct a permutation matrix based on the node sorting index vector. The permutation matrix is ​​a square matrix. Each row of the matrix corresponds to the normal sorting position, and each column corresponds to the original node index. The matrix element is assigned a value of 1 at the position where the normal sorting position matches the original node index, and a value of 0 at the other positions, so that there is only one matrix element with a value of 1 in each row and each column of the permutation matrix.

[0123] Construct the node sorting index vector as follows ;in, For the sorted number The original node index corresponding to the bit; For sorting position index;

[0124] The sorting is based on lexicographical comparison: When there is and At that time, the index Ranked in the index Previously; among them, This is a lexicographical comparison symbol used to compare three-dimensional vectors. The size of the components is compared sequentially, with the order determined when the first different component is encountered.

[0125] Constructing the permutation matrix ,satisfy:

[0126] , , ;in, The elements of the permutation matrix represent the original node indices. Position after normal sorting Is there a corresponding relationship between them?

[0127] The process of normalizing and rearranging the original adjacency relation matrix, node attribute matrix, and edge attribute matrix using a permutation matrix to obtain a normalized adjacency relation matrix, normalized node attribute matrix, and normalized edge attribute matrix with unified node order specifically includes:

[0128] Based on the permutation matrix, the rows and columns of the original adjacency relation matrix are rearranged simultaneously according to the normal sort order to obtain the normal adjacency relation matrix, so that the row index and column index of the normal adjacency relation matrix correspond to the normal sort order;

[0129] Based on the permutation matrix, the rows of the node attribute matrix are rearranged according to the normal sorting order to obtain the normal node attribute matrix, so that each row of the normal node attribute matrix corresponds one-to-one with a node after normal sorting.

[0130] The normal adjacency matrix is ​​scanned sequentially in ascending order of row index and column index within the same row. When a matrix element is found to indicate the existence of a directed edge, a normal edge number is assigned to each detected directed edge in the order of scanning. At the same time, the number of normal edge numbers assigned is counted during the scanning process.

[0131] For each canonical edge number, obtain the canonical sorting position of the starting node and the canonical sorting position of the ending node corresponding to each canonical edge according to the canonical sorting order, and find the original starting node number and the original ending node number corresponding to them through the node sorting index vector.

[0132] In the original adjacency matrix, all matrix elements are scanned in ascending order of row index and ascending order of column index within the same row. The number of matrix elements that indicate the existence of directed edges before reaching the corresponding positions of the original starting node number and the original ending node number is counted. The counted number is incremented by one to obtain the original edge number of the corresponding directed edge. The corresponding edge attribute vector is then searched in the edge attribute matrix based on the original edge number.

[0133] In the normalized edge attribute matrix, the normalized edge number is selected as the row index, and the edge attribute vector corresponding to each normalized edge number is written into the corresponding row of the normalized edge attribute matrix, so that the arrangement order of each row in the normalized edge attribute matrix is ​​consistent with the lexicographical order of the edges in the normalized adjacency relation matrix.

[0134] The adjacency matrix is ​​normalized using a permutation matrix to: ;in, The adjacency matrix is ​​normalized.

[0135] Standardize the node attribute matrix as follows: ;in, A normalized node attribute matrix;

[0136] Side attribute matrix Rearrange the edges to obtain the standardized edge attribute matrix. Specifically:

[0137] For all and ,according to From childhood to adulthood, in the same Press down Process in ascending order, if and only if At that time, execute steps S501 to S503:

[0138] S501, ;in, , This serves as the row and column index for the normalized node; To normalize the elements of the adjacency matrix; To normalize the adjacency matrix Middle Lexicographical edge numbering; , This is an index variable used for iteration. ;

[0139] S502, Order: , ;in, To standardize the position The corresponding original node index; To standardize the position The corresponding original node index;

[0140] S503, Calculation ;in, In the original adjacency matrix Middle Lexicographical edge numbering;

[0141] For all Assignment: ;in, To normalize the elements of the edge attribute matrix; These are the elements of the original edge attribute matrix.

[0142] The step of vectorizing the standardized adjacency matrix, standardized node attribute matrix, and standardized edge attribute matrix, and then concatenating the resulting multiple encoded vectors into a unified standardized structure encoded vector according to a preset order, specifically includes:

[0143] The matrix elements of the normal adjacency relation matrix are read sequentially in column index priority order. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form an adjacency relation encoding vector.

[0144] The matrix elements of the standard node attribute matrix are read sequentially according to the column index priority. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form a node attribute encoding vector.

[0145] The matrix elements of the standard edge attribute matrix are read sequentially according to column index priority. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form the edge attribute encoding vector.

[0146] The adjacency relationship encoding vector, node attribute encoding vector, and edge attribute encoding vector are concatenated in a preset order to obtain a unified standardized structure encoding vector.

[0147] Vectorizing the adjacency matrix into a column-major sequence is as follows: ;in, To be The vector obtained by expanding according to the column precedence rule; Vectorized functions;

[0148] Concatenate the node attribute matrix vector column by column, specifically as follows: ;in, To be The vector obtained by expanding according to the column precedence rule;

[0149] Vectorize the edge attribute matrix into column concatenation, specifically as follows: ;in, To be The vector obtained by expanding according to the column precedence rule;

[0150] Concatenate them into a unified structure encoding vector, specifically: ;in, This is the final standardized structure encoding vector; The vector concatenation operator concatenates multiple vectors into a single long vector in sequence.

[0151] The process of combining the standardized structure encoding vector with the patient's unique ID, the record's unique ID, and the time interval relative to the patient's first record to form a standardized storage record, and establishing a composite index key for the standardized storage record, specifically includes:

[0152] For each examination or treatment record related to airway stenosis, obtain the patient's unique ID, the record's unique ID, the time interval relative to the patient's first record, and the standardized structure encoding vector of the corresponding record, and combine these four pieces of information into a standardized storage record.

[0153] A composite index key is constructed using the patient's unique ID and the time interval relative to the patient's first record. The composite index key is then used to establish an index relationship with the standardized storage record, and the standardized storage record is written into the storage medium. This allows the corresponding standardized storage record to be retrieved using the patient's unique ID and the time interval relative to the patient's first record.

[0154] For each examination or treatment record, construct and store the record tuple as follows:

[0155] ;in, This is a standardized storage record; Each patient is assigned a unique identification number; Used as a unique number for examination or treatment records; This refers to the time elapsed since the first record was made;

[0156] Construct a composite index key as .

[0157] The process of calculating standardized fingerprint values ​​based on standardized structure encoding vectors, performing encoding integrity verification on standardized storage records, and restoring the standardized structure encoding vectors to a standardized adjacency matrix, a standardized node attribute matrix, and a standardized edge attribute matrix according to preset decoding rules specifically includes:

[0158] For each standardized structure encoding vector in a standardized storage record, the value of each component is obtained sequentially according to the position order of each component in the vector. A fixed weight coefficient and the position number of each component in the encoding vector are assigned to each component. The value of each component is multiplied by the corresponding weight coefficient and position number, and all products are summed to obtain the standardized fingerprint value of the corresponding encoding vector.

[0159] When the standardized storage process is executed again on the same examination record or treatment record, a new standardized structure encoding vector corresponding to the same examination record or treatment record is generated. The value of each component is obtained in the order of its position in the vector. A fixed weight coefficient and the position number of each component in the encoding vector are assigned to each component. The value of each component is multiplied by the corresponding weight coefficient and position number, and all products are summed to obtain the recalculated fingerprint value. The recalculated fingerprint value is compared with the standardized fingerprint value when it was first written. If the two standardized fingerprint values ​​are equal, the same examination record or treatment record is regarded as a record that has been written before. Only the original record content is updated and no new standardized storage record is added.

[0160] When the recalculated fingerprint value obtained by recalculating a certain standardized storage record is not equal to the original standardized fingerprint value of the corresponding standardized storage record, the standardized storage record is marked as an illegal record, and write and update operations are refused to be performed on the standardized storage record.

[0161] When two examination records or treatment records from different sources yield the same standardized fingerprint value after calculating the standardized fingerprint value, these two examination records or treatment records are marked as structurally equivalent records. The structurally equivalent marking indicates that the two records have the same airway structure map information at the level of standardized structural coding vector.

[0162] During the standard decoding process, the standardized structure encoding vector is divided according to the number of rows and columns of the standard adjacency matrix, the number of rows and columns of the standard node attribute matrix, and the number of rows and columns of the standard edge attribute matrix. The encoding vector is divided into three segments in a pre-defined order: adjacency encoding vector, node attribute encoding vector, and edge attribute encoding vector.

[0163] The adjacency relationship encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal adjacency relationship matrix; the node attribute encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal node attribute matrix; the edge attribute encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal edge attribute matrix, thus completing the reverse reconstruction of the normal airway structure diagram.

[0164] For the encoding vector Calculate the standardized fingerprint:

[0165] ;in, Encoding vector The total length; Encoding vector The One component; The position index of the encoded vector component;

[0166] The encoded vector obtained during recalculation of the same record is denoted as:

[0167] ;in, This is a recalculated encoding vector for the same business record; For complex encoding vectors The One component;

[0168] Its corresponding recalculated fingerprint value is ;

[0169] The verification criteria are as follows:

[0170] S801. When performing a write operation on the same record again, if the fingerprint value is recalculated... Compared with the original fingerprint value If they are equal, the records are considered the same record and no new record is added for storage.

[0171] S802, The fingerprint value obtained by recalculating a certain record. Compared with the original fingerprint value If they are not equal, mark the record as an illegal record and prohibit writing to or updating the corresponding stored record;

[0172] S803, Two records from different sources yield the same fingerprint value after calculation. At that time, mark the two records as structurally equivalent records;

[0173] The standard decoding process is as follows:

[0174] S804, from the encoded vector The vector is divided into three segments, specifically:

[0175] , , ;

[0176] S805, and then recovers the standard graph structure through vectorized inverse transform, specifically:

[0177] Pick , Pick , Pick ;in, The number of rows is a parameter; The number of columns is a parameter.

[0178] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0179] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for standardized storage of airway stenosis treatment data, characterized in that, include: An airway structure map representing the anatomical topology of the airway is constructed based on airway medical images, and original node numbers are assigned to the airway anatomical nodes. Based on the airway structure diagram, an original adjacency matrix representing the node connection relationship is constructed, and a node attribute matrix recording node attribute information and an edge attribute matrix recording edge attribute information are constructed respectively. The structural features of each node are calculated based on the original adjacency matrix and the node attribute matrix, and the structural features of each node are combined into a structural feature matrix. Perform a canonical sort on all nodes based on the structural feature matrix, and construct a permutation matrix that reflects the canonical sort result; The original adjacency matrix, node attribute matrix, and edge attribute matrix are normalized and rearranged using a permutation matrix to obtain a normalized adjacency matrix, normalized node attribute matrix, and normalized edge attribute matrix with uniform node order. The standardized adjacency matrix, standardized node attribute matrix, and standardized edge attribute matrix are vectorized and encoded, and the resulting multiple encoded vectors are concatenated into a unified standardized structure encoded vector according to a preset order. The standardized structure encoding vector is combined with the patient's unique ID, the record's unique ID, and the time interval relative to the patient's first record to form a standardized storage record, and a composite index key is established for the standardized storage record; The standardized fingerprint value is calculated based on the standardized structure encoding vector. The encoding integrity is checked on the standardized storage record. The standardized structure encoding vector is restored into a standardized adjacency matrix, a standardized node attribute matrix, and a standardized edge attribute matrix according to the preset decoding rules. The calculation of the standardized fingerprint value based on the standardized structure encoding vector specifically includes: for each standardized storage record, the standardized structure encoding vector is used to obtain the value of each component in the order of its position in the vector, a fixed weight coefficient and a position number of each component in the encoding vector are assigned to each component, the value of each component is multiplied by the corresponding weight coefficient and position number, and all products are summed to obtain the standardized fingerprint value of the corresponding encoding vector. The process of performing encoding integrity verification on standardized storage records specifically includes: when the standardized storage process is performed again on the same examination record or treatment record, a new standardized structure encoding vector corresponding to the same examination record or treatment record is generated; the value of each component is obtained sequentially according to the position order of each component in the encoding vector; a fixed weight coefficient and the position number of each component in the encoding vector are assigned to each component; the value of each component is multiplied by the corresponding weight coefficient and position number; and all products are summed to obtain the recalculated fingerprint value; the recalculated fingerprint value is compared with the standardized fingerprint value when it is first written; when the two standardized fingerprint values ​​are equal, the same examination record or treatment record is regarded as a record that has been written before, and only the original record content is updated without adding a new standardized storage record. When the recalculated fingerprint value obtained by recalculating a standardized storage record is not equal to the original standardized fingerprint value of the corresponding standardized storage record, the standardized storage record is marked as an illegal record, and write and update operations on the standardized storage record are refused.

2. The method for standardized storage of airway stenosis treatment data according to claim 1, characterized in that, The construction of an airway structure map representing the anatomical topology of the airway based on airway medical images, and the assignment of original node numbers to airway anatomical nodes, specifically includes: All airway anatomical nodes are obtained from the three-dimensional medical images of the airway, and a unique original node number is assigned to each airway anatomical node in a preset order. All airway anatomical nodes are compiled into a node set according to the original node number, so that each element in the node set corresponds one-to-one with an airway anatomical node. Select any two airway anatomical nodes from the node set. If there is a direct anatomical connection, establish a directed edge between the two nodes from the starting node to the ending node, and summarize all directed edges into an edge set. The set of nodes and the set of edges are combined to form an airway structure diagram. In the airway structure diagram, the original node number is used to identify the airway anatomical nodes, and the original starting node number and the original ending node number of the directed edges are used to identify the airway anatomical connection relationship.

3. The method for standardized storage of airway stenosis treatment data according to claim 2, characterized in that, The step of constructing an original adjacency matrix representing node connections based on the airway structure diagram, and constructing a node attribute matrix recording node attribute information and an edge attribute matrix recording edge attribute information, specifically includes: In the airway structure diagram, the nodes in the node set are used as the row index and column index of the original adjacency relation matrix to construct the original adjacency relation matrix. When there is a directed edge from the corresponding row node to the corresponding column node, the matrix element is recorded as connected; when there is no directed edge, the matrix element is recorded as not connected. For each airway anatomical node, the attribute information of each airway anatomical node is obtained, and the attributes are arranged in a fixed order to form a node attribute vector, which contains multiple attribute components. Arrange the node attribute vectors of all nodes in the order of their original node numbers to form rows of the node attribute matrix, so that each column of the node attribute matrix corresponds one-to-one with a node attribute component. Select all matrix elements in the original adjacency matrix that indicate the existence of directed edges, and arrange them in alphabetical order according to row index from smallest to largest and column index from smallest to largest within the same row. During the arrangement process, count the total number of directed edges, and assign a unique original edge number to each pair of original start node number and original end node number that have directed edges. Record the original starting node number and the original ending node number corresponding to each original edge number to form a one-to-one correspondence between the original edge number and the original starting node number and the original ending node number. For each directed edge, obtain the attribute information of the corresponding directed edge, arrange the edge attributes in a fixed order to form an edge attribute vector, and arrange all edge attribute vectors in the order of the original edge numbers to form the rows of the edge attribute matrix.

4. The method for standardized storage of airway stenosis treatment data according to claim 3, characterized in that, The step of calculating the structural features of each node based on the original adjacency matrix and node attribute matrix, and combining the structural features of each node into a structural feature matrix, specifically includes: For each node, select all matrix elements of the corresponding row in the original adjacency matrix, count the number of matrix elements that indicate the existence of directed edges, and use the counted number of matrix elements as the degree of the corresponding node. For each node, based on the original adjacency matrix, identify all matrix elements in the corresponding node's row that represent the existence of directed edges. Use these matrix elements to determine all adjacent nodes starting from the corresponding node, obtain the degree of each adjacent node, and sum the degrees of all adjacent nodes to obtain the total degree of the adjacent nodes of the corresponding node. For each node, based on the original adjacency relation matrix, identify all matrix elements in the corresponding node's row that indicate the existence of directed edges. Use these matrix elements to determine all adjacent nodes starting from the corresponding node. Sum the node attribute vector of each adjacent node according to each attribute component, and combine the summation results of each attribute component into the adjacent node attribute summation vector of the corresponding node. For each node, the node degree, the sum of the degrees of adjacent nodes, and the sum of the attributes of adjacent nodes are combined in a fixed order to form the structural feature vector of the corresponding node, so that each component of the structural feature vector corresponds to the node degree, the sum of the degrees of adjacent nodes, and the sum of the attributes of adjacent nodes, respectively. Following the original node index order, the structural feature vectors of all nodes are arranged sequentially into rows of the structural feature matrix, so that each row of the structural feature matrix corresponds to the structural feature of a node, and each column corresponds to a structural feature component.

5. The method for standardized storage of airway stenosis treatment data according to claim 4, characterized in that, The step of performing a canonical sort on all nodes based on the structural feature matrix and constructing a permutation matrix reflecting the canonical sort result specifically includes: Based on the structural feature matrix, a sorting operation is performed on the structural feature vectors of all nodes. During sorting, the first structural feature component in the structural feature vector is compared first. If the first structural feature components are the same, the second structural feature component is compared. If the second structural feature components are the same, the third structural feature component is compared. If all structural feature components are the same, they are arranged in ascending order according to the original node number to obtain the normal sorting order of the nodes. A node sorting index vector is constructed according to the standard sorting order. In the node sorting index vector, each position records the original node number corresponding to each sorting position, so that the node sorting index vector fully reflects the standard sorting result. Construct a permutation matrix based on the node sorting index vector. The permutation matrix is ​​a square matrix. Each row of the matrix corresponds to the normal sorting position, and each column corresponds to the original node index. The matrix element is assigned a value of 1 at the position where the normal sorting position matches the original node index, and a value of 0 at the other positions, so that there is only one matrix element with a value of 1 in each row and each column of the permutation matrix.

6. The method for standardized storage of airway stenosis treatment data according to claim 5, characterized in that, The process of normalizing and rearranging the original adjacency relation matrix, node attribute matrix, and edge attribute matrix using a permutation matrix to obtain a normalized adjacency relation matrix, normalized node attribute matrix, and normalized edge attribute matrix with unified node order specifically includes: Based on the permutation matrix, the rows and columns of the original adjacency relation matrix are rearranged simultaneously according to the normal sort order to obtain the normal adjacency relation matrix, so that the row index and column index of the normal adjacency relation matrix correspond to the normal sort order; Based on the permutation matrix, the rows of the node attribute matrix are rearranged according to the normal sorting order to obtain the normal node attribute matrix, so that each row of the normal node attribute matrix corresponds one-to-one with a node after normal sorting. The normal adjacency matrix is ​​scanned sequentially in ascending order of row index and column index within the same row. When a matrix element is found to indicate the existence of a directed edge, a normal edge number is assigned to each detected directed edge in the order of scanning. At the same time, the number of normal edge numbers assigned is counted during the scanning process. For each canonical edge number, obtain the canonical sorting position of the starting node and the canonical sorting position of the ending node corresponding to each canonical edge according to the canonical sorting order, and find the original starting node number and the original ending node number corresponding to them through the node sorting index vector. In the original adjacency matrix, all matrix elements are scanned in ascending order of row index and ascending order of column index within the same row. The number of matrix elements that indicate the existence of directed edges before reaching the corresponding positions of the original starting node number and the original ending node number is counted. The counted number is incremented by one to obtain the original edge number of the corresponding directed edge. The corresponding edge attribute vector is then searched in the edge attribute matrix based on the original edge number. In the normalized edge attribute matrix, the normalized edge number is selected as the row index, and the edge attribute vector corresponding to each normalized edge number is written into the corresponding row of the normalized edge attribute matrix, so that the arrangement order of each row in the normalized edge attribute matrix is ​​consistent with the lexicographical order of the edges in the normalized adjacency relation matrix.

7. The method for standardized storage of airway stenosis treatment data according to claim 6, characterized in that, The step of vectorizing the standardized adjacency matrix, standardized node attribute matrix, and standardized edge attribute matrix, and then concatenating the resulting multiple encoded vectors into a unified standardized structure encoded vector according to a preset order, specifically includes: The matrix elements of the normal adjacency relation matrix are read sequentially in column index priority order. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form an adjacency relation encoding vector. The matrix elements of the standard node attribute matrix are read sequentially according to the column index priority. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form a node attribute encoding vector. The matrix elements of the standard edge attribute matrix are read sequentially according to column index priority. The matrix elements in each column are arranged from top to bottom, and the columns are connected sequentially to form the edge attribute encoding vector. The adjacency relationship encoding vector, node attribute encoding vector, and edge attribute encoding vector are concatenated in a preset order to obtain a unified standardized structure encoding vector.

8. The method for standardized storage of airway stenosis treatment data according to claim 7, characterized in that, The process of combining the standardized structure encoding vector with the patient's unique ID, the record's unique ID, and the time interval relative to the patient's first record to form a standardized storage record, and establishing a composite index key for the standardized storage record, specifically includes: For each examination or treatment record related to airway stenosis, obtain the patient's unique ID, the record's unique ID, the time interval relative to the patient's first record, and the standardized structure encoding vector of the corresponding record, and combine these four pieces of information into a standardized storage record. A composite index key is constructed using the patient's unique ID and the time interval relative to the patient's first record. The composite index key is then used to establish an index relationship with the standardized storage record, and the standardized storage record is written into the storage medium. This allows the corresponding standardized storage record to be retrieved using the patient's unique ID and the time interval relative to the patient's first record.

9. A method for standardized storage of airway stenosis treatment data according to claim 8, characterized in that, The process of calculating standardized fingerprint values ​​based on standardized structure encoding vectors, performing encoding integrity verification on standardized storage records, and restoring the standardized structure encoding vectors to a standardized adjacency matrix, a standardized node attribute matrix, and a standardized edge attribute matrix according to preset decoding rules specifically includes: When two examination records or treatment records from different sources yield the same standardized fingerprint value after calculating the standardized fingerprint value, these two examination records or treatment records are marked as structurally equivalent records. The structurally equivalent marking indicates that the two records have the same airway structure map information at the level of standardized structural coding vector. During the standard decoding process, the standardized structure encoding vector is divided according to the number of rows and columns of the standard adjacency matrix, the number of rows and columns of the standard node attribute matrix, and the number of rows and columns of the standard edge attribute matrix. The encoding vector is divided into three segments in a pre-defined order: adjacency encoding vector, node attribute encoding vector, and edge attribute encoding vector. The adjacency relationship encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal adjacency relationship matrix; the node attribute encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal node attribute matrix; the edge attribute encoding vector is filled with matrix elements in a column index-first, top-down order to recover the normal edge attribute matrix, thus completing the reverse reconstruction of the normal airway structure diagram.

Images