Multi-modal medical image-based plastic surgery auxiliary strategy generation method and system

Abstract

The invention discloses a plastic surgery auxiliary strategy generation method and system based on a multi-modal medical image, and relates to the technical field of auxiliary strategy generation. After the multi-modal medical image is collected, multilevel organization structure information is extracted, and a facial structure function linkage graph is constructed; identifying a structure coupling region with a correlation influence relationship in the function linkage graph as an intervention region; reversely querying an operation strategy knowledge base based on the intervention area to obtain a candidate operation fragment set; according to the structure conflict relation matrix and the recovery time dependence matrix between the candidate operation segments, constructing an initial operation path diagram; according to aesthetic preference parameters set by a user, generating an aesthetic offset information field, and acting on the initial operation path diagram to form a structural deformation map; and based on the structural deformation map, injecting control constraints, generating a reversible control chart, and outputting a personalized operation execution path. Personalized operation path generation is realized, and dissection safety and aesthetic requirements are considered.

Description

Technical Field

[0001] The present invention relates to the technical field of auxiliary strategy generation, and in particular to a method and system for generating auxiliary strategies for plastic surgery based on multimodal medical images. Background Art

[0002] With the continuous advancement of medical image acquisition and intelligent processing technologies, multimodal images (such as three-dimensional CT, MRI, ultrasound, and stereoscopic photography) are increasingly being used in plastic surgery. Facial plastic surgery not only involves precise spatial manipulation of anatomical structures but also requires achieving personalized aesthetic goals while ensuring structural stability. However, existing plastic surgery assistance strategies mostly rely on the surgeon's experience and lack systematic modeling and intelligent planning of complex structural coupling relationships, individual aesthetic differences, and postoperative recovery processes, making it difficult to meet the needs of precise and customized interventions.

[0003] Some studies have attempted to introduce graph structures to model organizational relationships, but these generally lack mechanisms for integrating structure and function, failing to reflect the dynamic coupling between different organizational hierarchies. Furthermore, current path planning methods are often based on static processes, ignoring factors such as recovery dependencies, structural conflicts, and aesthetic regulation between surgical segments. This makes it difficult to dynamically adjust surgical sequences or support personalized adjustment strategies. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing plastic surgery strategies lack structural hierarchical understanding, functional coupling perception and integration of aesthetic preferences, and cannot simultaneously meet clinical anatomical safety and individualized aesthetic goals. In addition, the path generation process lacks dynamic adjustability, making it difficult to achieve accurate planning and flexible adjustment of complex surgical processes.

[0006] To solve the above technical problems, the present invention provides the following technical solution: a method for generating a plastic surgery auxiliary strategy based on multimodal medical images, comprising:

[0007] After acquiring multimodal medical images, multi-level tissue structure information is extracted and a facial structure-function linkage map is constructed;

[0008] Identifying structural coupling areas with associated impact relationships in the functional linkage diagram as intervention areas;

[0009] Reversely query the surgical strategy knowledge base based on the intervention area to obtain a set of candidate surgical fragments;

[0010] Construct an initial surgical pathway diagram based on the structural conflict relationship matrix and the recovery time dependency matrix between candidate surgical segments;

[0011] According to the aesthetic preference parameters set by the user, an aesthetic deviation information field is generated and applied to the initial surgical path map to form a structural deformation map;

[0012] Based on the structural deformation map, control constraints are injected, a reversible control map is generated, and a personalized surgical execution path is output.

[0013] As a preferred embodiment of the method for generating a plastic surgery auxiliary strategy based on multimodal medical images of the present invention, the multimodal medical images include three-dimensional CT images, MRI images, ultrasound images, and three-dimensional stereoscopic images;

[0014] The multi-level tissue structure information includes bone feature information, muscle feature information, soft tissue feature information and surface skin feature information.

[0015] As a preferred embodiment of the method for generating a plastic surgery auxiliary strategy based on multimodal medical images described in the present invention, the facial structure-function linkage map construction process is specifically as follows: the facial structure-function linkage map construction process is specifically as follows: using an inter-modal cross-attention mechanism to extract the multi-level tissue structure information from the spatially aligned multimodal medical image; using a cross-modal feature aggregation mechanism to aggregate the multi-level tissue structure information into three-dimensional spatial coordinates to obtain all structural units as a node set for constructing the structure-function linkage map;

[0016] A priori coupling factors between different tissue levels are preset; the product of the prior coupling factor and the functional conduction strength is used as the positive factor of coupling, and the positive factor is normalized with the Euclidean distance between nodes as the denominator to serve as the coupling weight between nodes; among all node pairs, node pairs with coupling weights greater than zero are screened out to establish edge connections and included in the edge set, and a two-dimensional symmetrical edge weight matrix is ​​generated at the same time, storing the coupling weight value between each pair of nodes in numerical form; the structural function linkage diagram is composed of the node set, the edge set and the edge weight matrix; the tissue level labels include bone, muscle, soft tissue and surface skin.

[0017] As a preferred embodiment of the method for generating an auxiliary strategy for plastic surgery based on multimodal medical images described in the present invention, the method further comprises: identifying the structural coupling region, extracting all adjacent nodes that have a structural-functional coupling relationship with the current node as a first-order adjacent node set; adding the edge weights between the first-order adjacent node pairs and taking the average value as the local edge density index of the current node; determining whether the tissue hierarchy labels between the first-order adjacent node pairs are consistent, and using the average value as the semantic consistency score; and converting the Euclidean distance between the first-order adjacent node pairs into a spatial consistency score through exponential decay.

[0018] The semantic consistency score and spatial consistency score between each pair of first-order adjacent nodes are multiplied and then weighted averaged to obtain the functional aggregation index. The local edge density index and the functional aggregation index are concatenated into a set of coupling feature vectors, which are then divided into multiple preliminary structural coupling regions using a density-aware clustering algorithm. The edge weights between all node pairs within each preliminary structural coupling region are summed and averaged as the regional coupling score. All coupling regions with regional coupling scores exceeding the screening threshold are screened to form a set of intervention areas.

[0019] As a preferred embodiment of the method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to the present invention, the step of obtaining a candidate surgical segment set includes counting the proportion of each type of node in the node set of the facial structure-function linkage diagram according to the organizational hierarchy label, and using the result as the structural hierarchy distribution;

[0020] Based on the node coordinate distribution, the geometric center, average curvature and boundary convexity are calculated to form the spatial configuration index; the average weight, maximum weight gradient and minimum weight gradient of the edge are counted to form the structural coupling strength characteristic; the structural hierarchy distribution, spatial configuration index and structural coupling strength characteristic are spliced ​​into a structural comprehensive vector;

[0021] In the surgical strategy knowledge base, the structural hierarchical distribution and the predefined hierarchical distribution template are subjected to cosine similarity calculation. When the label similarity threshold is reached, the label group is consistent; the structural coupling intensity gradient constructs a histogram and calculates the cross entropy with the predefined topological template histogram. When the cross entropy value is When the topological similarity threshold is set, it is topologically similar; only candidate surgical segments that meet both label group consistency and topological similarity are retained and included in the candidate surgical segment set.

[0022] As a preferred embodiment of the method for generating auxiliary strategies for plastic surgery based on multimodal medical images described in the present invention, the construction of the initial surgical path map includes constructing a structural conflict relationship matrix and a recovery time dependency matrix using candidate surgical segments as nodes of the initial surgical path map; the construction of the structural conflict relationship matrix includes calculating the spatial overlap between the node sets of the facial structure-function linkage diagram within any two candidate surgical segments. When the overlap ratio of the spatial overlap exceeds a set threshold, a structural conflict exists, which is recorded as 1, otherwise it is 0; the construction of the recovery time dependency matrix includes counting the expected recovery time of each surgical segment. When the surgical segments have an execution sequence relationship, dependency edges are established between the corresponding nodes of the surgical segments; at the same time, for segments without structural conflict and time dependency, bidirectional structural edges are established; the edge set of the initial surgical path map includes: bidirectional structural edges and dependency edges.

[0023] As a preferred solution of the method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to the present invention, wherein: the aesthetic preference parameter is the user's preference for each facial structure area. The enhancement or softening intention of each structural area; constructing an aesthetic preference parameter vector based on the aesthetic preference parameters , where each dimension Respectively represent the the intention to enhance or soften a structural area; Indicates index; n indicates Dimensions;

[0024] Generating the aesthetic offset information field includes: constructing a continuous aesthetic offset information field covering the three-dimensional facial space using a Gaussian function based on the spatial position of each structural unit in the three-dimensional facial structure model and the hierarchical label index to which it belongs, according to the weight of each dimension in the aesthetic preference parameter vector;

[0025] The structure deformation map comprises calculating the average response value of the intervention region structure node corresponding to each candidate surgical segment in the aesthetic deviation information field as the aesthetic response intensity of the candidate surgical segment; when the aesthetic response intensity When the jump threshold is set, the jump candidate label is assigned to the candidate surgical segment stage; the candidate surgical segments with all jump candidate labels are used to form a label set; the label set is combined with the initial surgical path map to form a structural deformation map;

[0026] After the structural deformation map is constructed, control constraints are introduced to generate a reversible control graph. The control constraints include: setting the weights of all incoming edges of candidate surgical segments in the label set to 0, and retaining only outgoing edges; for segments with a recovery time dependency, when the aesthetic response strength of the subsequent segment is lower than that of the preceding segment, the execution order of the segment pair is swapped, and the edge direction in the recovery dependency matrix is ​​adjusted at the same time; when the aesthetic response strengths of three candidate surgical segments in the path all exceed a multiple threshold of the aesthetic average value, the three candidate surgical segments are merged into one candidate surgical segment node, replacing the original path segment; when the Euclidean distance between the candidate surgical segment pairs is less than the nearest neighbor threshold and there is no structural conflict relationship, a bidirectional connecting edge is added between the candidate surgical segment pairs;

[0027] The process of outputting the personalized surgery execution path is as follows:

[0028] The set of paths in the reversible control diagram that meets the following conditions is screened: the aesthetic response intensity of the segments in the label set is not less than the jump threshold, the order of all segments with recovery time dependencies is not destroyed, and the sum of the expected execution time and recovery period of all segments in the path does not exceed the maximum duration set by the user; the sum of the aesthetic response intensities of all segments in each screened path is calculated as the aesthetic benefit, and the total number of path segments multiplied by the step penalty coefficient is used as the step cost, and the difference between the aesthetic benefit and the step cost is used as the comprehensive score; the path with the highest comprehensive score is selected as the output of the personalized surgical execution path.

[0029] As a preferred solution of the plastic surgery auxiliary strategy generation system based on multimodal medical images described in the present invention, wherein: the facial structure-function linkage map module, after acquiring multimodal medical images, extracts multi-level tissue structure information and constructs a facial structure-function linkage map;

[0030] An intervention area module, which identifies a structural coupling area with an associated influence relationship in the functional linkage diagram as an intervention area;

[0031] A query module, which reversely queries a surgical strategy knowledge base based on the intervention area to obtain a set of candidate surgical fragments;

[0032] The initial path map module constructs the initial surgical path map based on the structural conflict relationship matrix and the recovery time dependency matrix between candidate surgical segments;

[0033] The structural deformation atlas module generates an aesthetic deviation information field based on the aesthetic preference parameters set by the user, and acts on the initial surgical path map to form a structural deformation atlas;

[0034] The execution module injects control constraints based on the structural deformation map, generates a reversible control map, and outputs a personalized surgical execution path.

[0035] A computer device includes a memory and a processor, wherein the memory stores a computer program and the processor executes the computer program to implement the steps of a method for generating a plastic surgery auxiliary strategy based on multimodal medical images.

[0036] A computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of a method for generating a plastic surgery auxiliary strategy based on multimodal medical images.

[0037] Beneficial effects of the present invention: The method for generating auxiliary strategies for plastic surgery based on multimodal medical images provided by the present invention integrates three-dimensional CT, MRI, ultrasound, and stereoscopic photography images to construct a structure-function linkage diagram with anatomical structure and function relationships, comprehensively expressing the coupling characteristics between multi-level tissues such as bone, muscle, soft tissue, and skin. High-function aggregation areas are identified as intervention areas through indicators such as edge density, semantic consistency, and spatial consistency, and a set of candidate intervention areas is formed using coupled eigenvector density clustering. The initial surgical path diagram is generated by combining the structural conflict relationship matrix and the recovery time dependency matrix to ensure the structural rationality of the strategy and the matching of the recovery timing. An aesthetic offset information field is constructed based on the user's aesthetic preferences, mapped to the path diagram and formed into a structural deformation map, and candidate segments for stage jumps are marked. Priority jump, sequence adjustment, aesthetic continuous merging, and structure preservation constraints are constructed and injected into the structural deformation map to form a reversible control diagram. In the reversible control diagram, the optimal execution path that simultaneously satisfies the constraints such as jump, dependency and recovery time is screened to realize multi-stage surgical strategy planning based on individualized aesthetic intentions and clinical implementation constraints, thereby improving the safety, adjustability and aesthetic matching of the intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 This is an overall flow chart of the method for generating a plastic surgery auxiliary strategy based on multimodal medical images provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0040] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0041] Example 1, reference Figure 1 , as one embodiment of the present invention, provides a method for generating a plastic surgery auxiliary strategy based on multimodal medical images, comprising:

[0042] S1: Collect multimodal medical images and fuse multi-level tissue structure vectors to construct a facial structure-function linkage map; identify structural coupling areas with associated influence relationships in the functional linkage map as intervention areas.

[0043] Acquire 3D CT images of bony structures, MRI or ultrasound images of muscle and vascular structures, and 3D stereographic images of the epidermis and soft tissue morphology. After spatial alignment, these image data are fed into a Transformer-based multimodal embedding network, which extracts multi-level tissue structural information through a cross-attention mechanism between modalities.

[0044] The fused multi-level structural feature set includes: bony feature information, muscle feature information, soft tissue feature information, and surface skin feature information. For each type of feature information, the system constructs its corresponding local feature representation and performs linear transformations on each, uniformly mapping the feature representations of different modalities to a latent feature space of the same dimension to obtain comparable feature embedding representations. The embedded features of each modality are fed into a multi-head cross-attention layer to calculate the attention guidance weights between different modalities and generate a fused representation:

[0045]

[0046]

[0047] Among them, MultiHeadAttn(·) represents the multi-head cross attention mechanism; Represents structural unit Embedded features in osseous modalities; Represents structural unit Embedded features in muscle modality; Represents structural unit embedded features in soft tissue modalities; Represents structural unit Embedded features in surface skin modalities. Represents the set of all nodes in the structure-function linkage diagram; Indicates the A node is a triple structure; Represents structural unit The fusion feature vector of Represents structural unit The tissue level label includes four categories: "bone", "muscle", "soft tissue" and "surface skin"; Represents structural unit The position coordinates in three-dimensional space.

[0048] Defining coupling weights between structural units based on structure-function correlation , which is used to quantify the coupling strength between any two nodes:

[0049]

[0050] in, Represents structural unit and The a priori coupling factors between the layers are set by the anatomical knowledge base (e.g., 1.0 between bony and muscle, 0.6 between muscle and soft tissue, and 0.2 between bony and skin). Represents structural unit and Functional conduction strength in historical models; Structural unit and Euclidean distance in three-dimensional space.

[0051] when , it is determined that there is a structural coupling relationship between the two nodes.

[0052] In particular, for cross-layer node pairs, the system only sets the , ensuring that the coupling logic is anatomically reasonable.

[0053] After completing the coupling weight calculation, the edge set of the structure-function linkage graph is constructed according to the following constraints Including: For any node pair ,like , then create an edge , and add it to the edge set Each edge Contains: start and end node identifiers, edge weights and meta-attributes of connection level information (same layer / cross layer). Edge weight matrix Synchronous generation:

[0054]

[0055] The edge weight matrix is ​​a sparse symmetric structure, which only retains node pairs with functional coupling, effectively reducing the computational complexity of subsequent graph operations. The final structure-function linkage graph is represented as:

[0056]

[0057] in, Represents a collection of nodes, represents the edge set, represents the edge weight matrix.

[0058] In the structure-function linkage diagram After the construction is completed, the coupling degree between the structural units in the graph is quantified based on the edge weight matrix and the connection relationship between the nodes, and the structural coupling area is extracted. The specific process includes:

[0059] For any node in the structure-function linkage diagram , based on the edge weight matrix Extract its first-order adjacent node set:

[0060]

[0061] That is, only when the node With node When there is a functional coupling relationship between Considered as On this basis, calculate the local edge density index , representing the structural unit The coupling strength level within its first-order neighboring node set is calculated as follows:

[0062]

[0063] in, Representation node and adjacent nodes The coupling weight between . Represents a node in a structure diagram The set of first-order adjacent nodes of ; Indicates the first nodes; Represents any possible connected nodes; Representation node With node The coupling weight between them (if it is 0, it means there is no connection); Indicates that only when there is a valid coupling relationship between two nodes (that is, there is an edge connected), Inclusion of .

[0064] Secondly, construct the functional aggregation index , metric structural unit The aggregation characteristics of the organizational hierarchical labels and spatial distribution within the set of its first-order neighboring nodes are calculated as follows:

[0065] Determine the structural unit and Organizational level labels If they are the same, it is recorded as 1, otherwise it is 0, and the semantic consistency function is obtained:

[0066]

[0067] Calculate the Euclidean distance between structural units , and generate the spatial consistency factor:

[0068]

[0069] in, Represents structural unit and adjacent nodes Semantic consistency function on the organizational level label dimension; Represents structural unit Organizational level labels; Represents structural unit The organizational level label of When , it indicates that the structural units belong to the same organizational level, and the function value is 1; otherwise, the function value is 0, indicating that the structural units belong to different levels and do not have semantic consistency. Represents structural unit Its adjacent structural units Consistency weight in space; Represents the distance attenuation factor, which is a constant greater than 0 and controls the degree of influence of distance changes on weights. Represents the exponential function.

[0070] Combining the two, we get the functional aggregation index:

[0071]

[0072] Each structural unit The local edge density index and the functional aggregation index are combined to form a coupling feature vector: .

[0073] in, Represents structural unit Functional aggregation index; Represents structural unit The coupled eigenvector of represents the local edge density index; Represents the functional aggregation index.

[0074] Based on coupled eigenvector sets , use density-aware clustering algorithms (such as DBSCAN and HDBSCAN) to perform region division, retain the spatial topological relationship, and output multiple structural coupling regions:

[0075]

[0076] in, Indicates the A clustering region is composed of multiple structural units. represents the set of structural coupling regions; Indicates the structural coupling regions; Indicates the total number of coupled regions divided

[0077] To screen the key areas with high coupling degree, the system calculates each cluster area The average value of the edge weights between all node pairs in the region is defined as the regional coupling score:

[0078]

[0079] Setting scoring thresholds , filter out the set of regions where the regional coupling score exceeds the threshold:

[0080]

[0081] in, Indicates area The coupling score is used to measure the average coupling strength between structural units in a region; Indicates area The number of structural units (nodes) contained in; It represents the final structural coupling area screened out in the structure-function linkage diagram as the intervention area; Indicates the screening threshold.

[0082] By fusing multimodal medical image data, a structure-function linkage diagram covering tissue layers such as bone, muscle, soft tissue, and surface skin is constructed, enabling systematic modeling of the structural-functional relationship of the face. A modality fusion strategy based on a multi-head cross-attention mechanism is introduced to fully preserve the complementary characteristics of different modalities in spatial perception and tissue expression, improving the extraction depth and semantic accuracy of multi-level tissue information. This addresses the problem of traditional images being limited to single-modality modeling and lacking the ability to interactively understand structural hierarchies.

[0083] A structural coupling feature system based on coupling weight calculation and functional aggregation index is proposed. Combining local edge density with semantic consistency and spatial consistency, a coupling region identification mechanism is established that can distinguish the interaction relationship between different tissue levels. This breaks through the bottleneck that existing structural recognition technology cannot simultaneously take into account the fusion of spatial topology and semantic hierarchy. On the basis of maintaining spatial continuity, it achieves accurate mapping of anatomical rationality and physiological functional pathways, providing a structural basis for subsequent intervention area screening.

[0084] Through the density-aware clustering algorithm, combined with the coupling eigenvector, adaptive division and screening of structural coupling areas are achieved. The regional coupling score is introduced as a measure of structural aggregation strength to achieve effective extraction of high-density, high-aggregation functional areas. This breaks through the problem of the intervention area relying on subjective point selection and lack of structural support in existing methods, and significantly improves the accuracy, operability and personalized adaptability of the plastic surgery auxiliary strategy.

[0085] S2: Reversely query the surgical strategy knowledge base based on the intervention area to obtain a set of candidate surgical segments.

[0086] Based on the intervention area identified in the structure-function linkage diagram, reverse query the surgical strategy knowledge base to obtain the candidate surgical fragment set. , extract the corresponding structure node set , from which the following three types of structural features are calculated:

[0087] The structural hierarchical distribution characteristics of the intervention area refer to the classification of the structural nodes in the intervention area according to the tissue level labels (bone, muscle, soft tissue, surface skin) corresponding to the nodes, and the statistical ratio of the number of nodes of each type of label to the total number of nodes in the intervention area to form the structural hierarchical distribution vector , and the sum of all proportions is 1.

[0088] The spatial configuration index of the intervention area is calculated based on the coordinate distribution of the nodes in three-dimensional space, and the geometric center, average curvature and boundary convexity of the area are used to form the spatial configuration index vector .

[0089] The structural coupling strength characteristics of the intervention area are statistically analyzed. When there are edge connections between any two structural nodes in the intervention area, the average weight value, maximum weight gradient change and minimum weight gradient change of these edges are extracted to form the structural coupling strength index vector Among them, the weight gradient of the edge is defined as the difference between the weights of adjacent edges, that is, .in, Represents a slave node To Node The edge weight change gradient is used to characterize the weight difference between the current edge and the reference edge. Represents the slave node in the structure diagram Pointing to a node The current weight value of the edge. Represents a slave node To Node The reference weight value of the edge.

[0090] The above three types of features are spliced ​​into a structural comprehensive vector The surgical strategy knowledge base predefines the mapping relationships between structural label groups, topological relationship templates, and corresponding surgical operation fragments. This knowledge base is modeled by clinical experts and historical surgical data, and has medical interpretability and clinical traceability. The system performs reverse queries based on the following two matching criteria:

[0091] Structural tag group consistency: and the knowledge base Structural label distribution vector of records Calculate the cosine similarity. If the similarity Label similarity threshold , then the label groups are determined to be consistent.

[0092] Topological similarity: extraction The edge weight gradient distribution collection in is organized into an edge weight gradient histogram , and then with the knowledge base Histogram of the standard topology templates in the records Perform cross entropy calculation to obtain the topological structure similarity index. If the cross entropy value Topological similarity threshold , they are considered to be topologically similar.

[0093] Finally, only the surgical fragments corresponding to the templates that meet both label group consistency and topological similarity are retained. , included in the candidate surgical fragment set .in, Represents the total number of candidate surgical fragments. Indicates the The structural synthesis vector of candidate segments; Indicates the The structural level distribution vector of candidate segments; Indicates the The spatial configuration index vector corresponding to each fragment; Indicates the The structural coupling strength index vector of candidate fragments. Represents a vector concatenation operation, which connects multiple feature vectors into an overall comprehensive vector. Indicates the first The structure label group distribution vector of the historical surgical fragment records; represents the label similarity threshold; Indicates that the The edge weight distribution of candidate segments is mapped to the edge weight histogram The histogram image formed after Indicates the first The edge weight histogram of the standard topology template corresponding to the historical records; Indicates the threshold of topological structure similarity.

[0094] By introducing structural hierarchical distribution vectors, spatial configuration indicators and structural coupling strength characteristics, the structural tissue state, three-dimensional spatial morphology and functional coupling relationship of the intervention area can be comprehensively characterized, and high-dimensional structural coding expression of the surgical target area can be achieved. This makes the surgical strategy matching no longer rely on a single tissue label or rough area annotation, breaking through the existing strategy acquisition method based on two-dimensional image position or manual label matching, and significantly improving the accuracy of strategy generation and the clinical rationality of area matching.

[0095] A dual constraint mechanism of structural label group consistency and topological structure similarity is adopted to ensure that the matching of surgical segments not only considers the consistency of structural distribution ratio, but also introduces edge weight gradient histogram and cross entropy indicators to evaluate the topological coupling trend. This solves the problem that the existing technology only performs strategy matching based on label similarity or template labeling, which cannot accurately identify topological structure differences and lead to surgical path mismatch, and improves the semantic alignment and execution feasibility of candidate surgical segments.

[0096] By combining the comprehensive structural vector of the intervention area with the label template and topology template in the expert knowledge base, a structure-surgical fragment mapping mechanism is constructed to achieve traceable matching between historical surgical data and current anatomical structure. This breaks through the traditional single mode of reliance on manual experience or image template matching, has stronger adaptability and medical interpretability, and improves the universality and safety of intelligent assistance systems in multiple anatomical types.

[0097] S3: Construct an initial surgical path map based on the structural conflict relationship matrix and recovery time dependency matrix between candidate surgical segments.

[0098] After obtaining a set of candidate surgical fragments After that, the system takes each fragment As the node of the path diagram, the initial surgical path diagram is constructed by comprehensively considering the structural conflict relationship matrix and the postoperative recovery time dependency constraint.

[0099] The structural conflict relationship matrix is ​​constructed by extracting the corresponding structural node sets for any two candidate surgical segments and determining whether the two sets overlap excessively in three-dimensional space. If the ratio of the number of nodes in the intersection of the two sets to the total number of nodes in either set exceeds a preset threshold, the two segments are considered to have a structural conflict and cannot be executed simultaneously. The corresponding position in the structural conflict relationship matrix is ​​marked as 1; otherwise, it is marked as 0.

[0100] The recovery time dependency matrix is ​​constructed by estimating the postoperative recovery time for each candidate surgical segment. If the execution of a segment requires the completion of another segment's recovery, the corresponding position in the recovery time dependency matrix is ​​marked as 1; if there is no such dependency, it is marked as 0.

[0101] The edge set of the path graph is determined based on the above two matrices. No connecting edges are established for fragment pairs with structural conflict relationships; dependent edges are established for fragment pairs with recovery time dependencies; and fragment pairs without both structural conflict and time dependencies are considered structurally compatible and bidirectional structural edges are established.

[0102] The initial surgical path graph uses candidate surgical segments as nodes, and its edge set includes dependency edges and structural compatibility edges, which are used to support subsequent surgical path optimization and control graph generation.

[0103] A structural conflict relationship matrix is ​​constructed to enable the surgical path generation process to have the ability to recognize spatial overlaps and accurately identify regional combinations that cannot be executed in parallel between candidate segments. This fundamentally solves the problem that the existing system cannot avoid intraoperative operation interference and causes multi-region surgical conflicts, ensuring the physical feasibility of the plastic surgery path and the independence of regional operations.

[0104] The introduction of the recovery time dependency matrix enables the execution order between different candidate surgical segments to be automatically inferred based on the recovery logic, and establishes postoperative dependency paths between segments. This effectively circumvents the limitations of traditional path diagram construction that ignores the recovery period or relies on manual experience, thereby improving the physiological rationality of path generation and the intelligent level of surgical rhythm regulation.

[0105] By combining structural conflict relationships and recovery time dependency constraints to construct the initial surgical pathway map, the pathway map's dual expression capabilities of spatial coordination and temporal feasibility are significantly enhanced. This overcomes the problem of existing solutions that only plan pathways based on structural sequence or fixed templates, and establishes a personalized pathway modeling mechanism that can adapt to different surgical goals and patient structural characteristics.

[0106] S4: Generate an aesthetic offset information field based on the aesthetic preference parameters set by the user, and act on the initial surgical path map to form a structural deformation map; based on the structural deformation map, inject control constraints, generate a reversible control map, and output a personalized surgical execution path.

[0107] After completing the initial surgical path map, the aesthetic preference parameter vector consisting of the aesthetic preference parameters set by the user is introduced:

[0108]

[0109] in, represents the aesthetic preference parameter vector set by the user; Represents the aesthetic preference vector No. Dimension, its value range is , positive values ​​indicate enhancing the area, and negative values ​​indicate softening the area; Represents the aesthetic preference parameter vector Dimensions, Represents an index.

[0110] Based on the spatial position of each structural unit in the three-dimensional facial structure model and its level label index , define a continuous aesthetic offset information field:

[0111]

[0112] in, Represents the total number of structural units in the three-dimensional facial structure model; Indicates the The coordinate position of each structural unit in three-dimensional space; Indicates the The organizational level label to which each structural unit belongs; Indicates position in space The aesthetic offset information field value at ; Represents a point in any three-dimensional space; Indicates the The spatial position of each structural unit; Indicates a point and The Euclidean distance between Represents the width parameter of the Gaussian kernel function; represents the natural exponential function. Indicates the The spatial position of each structural unit; 、 、 They respectively represent the horizontal (X), vertical (Y) and depth (Z) position components of the structural unit in the three-dimensional coordinate system.

[0113] For any candidate surgical segment in the initial path graph , the set of structural unit nodes corresponding to the intervention area is recorded as The aesthetic response intensity of this segment is defined as:

[0114]

[0115] in, Indicates the The aesthetic response intensity of each candidate surgical clip; Indicates surgical footage The set of structural nodes corresponding to the intervention area; Representing a collection The total number of midpoints; Represents an arbitrary point in space selected from this node set.

[0116] Setting the jump threshold , for satisfying The surgical segment nodes are marked as “stage jump candidates” in the generated structural deformation graph to form a label set :

[0117]

[0118] Will , combined with the initial surgical path map, together constitute the structural deformation map:

[0119]

[0120] like and and , then the execution fragment can be jumped on the graph .

[0121] Keep consistent with the original path graph to ensure the constraints of the deformation graph on the structural relationship and recovery order; among them, the label set Indicates the jump node at the key stage when the aesthetic preference and risk tolerance conditions are met. represents a set of candidate surgical fragments; Represents a tag set.

[0122] in, Indicates intervention area Generated structural deformation map; 、 Respectively represent the first and structural units; Indicates surgical footage and Between the set of path graph edges There is an edge in .

[0123] After completing the structural deformation map After construction, a set of control constraints are introduced , further optimizing the path planning strategy. These constraints are injected into the graph structure one by one to generate a reversible control graph:

[0124]

[0125] in, Indicates area Reversible control chart of Represents a control constraint.

[0126] Control constraints Specifically, there are four control constraints:

[0127] Prioritized jump constraints: for segments marked in the label set , setting the weights of all its incoming edges to zero and only retaining its outgoing edges, thereby guiding the search algorithm to preferentially select such fragments.

[0128] Order adjustment constraint: If there is a fragment pair in the path that is dependent on the recovery ,and , that is, the aesthetic strength of the preceding fragment is better than that of the subsequent fragment, then the execution order of the two is exchanged and the direction of the recovery dependency matrix is ​​updated.

[0129] Aesthetic continuous merging constraint: When there are three consecutive surgical segments in the path, the aesthetic response intensity of each segment exceeds the multiple threshold of the average value. ,The system merges these three fragments into a composite node, replacing the original path segment, and retains its original front-to-back connection relationship, facilitating subsequent overall rollback or fine-tuning.

[0130] Structure preservation constraint: If the Euclidean distance between a pair of segments is less than the nearest neighbor threshold , and there is no structural conflict relationship, a bidirectional connecting edge is added to them to enhance their structural stability during the planning process.

[0131] After the above constraint adjustment, the original structure diagram is reconstructed into a reversible control diagram with dynamic scheduling capability. , its structure is more adapted to the balance needs between the user's aesthetic goals and postoperative recovery logic.

[0132] In the reversible control chart In the example, all possible paths are enumerated and only those paths that satisfy the following three constraints are retained:

[0133] Jump constraint: If the path contains a jump segment in the label set, its aesthetic response strength must meet .

[0134] Dependency constraint: All fragments with recovery dependencies in the path must ensure that their order is not destroyed.

[0135] Time window constraint: The sum of the estimated execution time and the corresponding recovery period of all segments in the path must not exceed the maximum duration set by the user .

[0136] Each candidate path that satisfies the constraint is , the system will calculate two quantitative indicators for each path:

[0137] Total aesthetic benefits:

[0138]

[0139] Step cost:

[0140]

[0141] Finally, the comprehensive score is:

[0142]

[0143] The system will score the highest path The output is a personalized surgical execution path for doctors to preview and fine-tune in the preoperative visual interface.

[0144] in, Represents candidate paths The sum of aesthetic benefits; Indicates the candidate paths; represents the strength of the aesthetic response of each segment in the path; represents the cost of the steps in the path; Represents the step penalty weight; Indicates the number of segments in the path. represents the preset step penalty coefficient, Represents the comprehensive score of the path.

[0145] By constructing an aesthetic deviation information field driven by aesthetic preference parameters, this approach introduces a spatially continuous aesthetic control mechanism into facial plastic surgery path planning, achieving a trend-based deformation expression of structural regions. This addresses the existing problem of inability to precisely guide structural deformation based on individual subjective preferences. This deviation information field combines 3D structural coordinates with tissue-level label distribution, providing regional response mapping capabilities tailored to the user's aesthetic goals.

[0146] A structural deformation map is constructed and control constraints including priority jump, sequence adjustment, aesthetic continuous merging and structural maintenance are injected into it. A closed-loop mechanism from "aesthetic intention → path structure adjustment → control guidance" is established. This breaks through the limitation of traditional path diagrams that can only express one-way physical dependency relationships, and realizes dynamic reconstruction and intelligent correction of path structure under aesthetic drive, providing goal-oriented flexible control capabilities for multi-stage surgical paths.

[0147] Under the constraints of recovery dependency, execution sequence, and time window, the reversible control chart systematically integrates the aesthetic response intensity and step cost evaluation mechanism, selects the path with the best comprehensive score from the candidate paths, and realizes the automatic generation of personalized aesthetic-driven surgical execution strategies. This solves the technical shortcoming of traditional surgical planning, which lacks the ability to integrate the dual optimization goals of structure and aesthetics, and improves the aesthetic accuracy of the surgical plan and the rationality of path execution.

[0148] Example 2 is an embodiment of the present invention, which provides a method for generating plastic surgery auxiliary strategies based on multimodal medical images. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0149] The plastic surgery department of a tertiary hospital introduced this method to conduct modeling and surgical path generation experiments using clinical image data from 12 patients undergoing facial contouring surgery. The experiment first collected 3D CT, MRI, ultrasound, and 3D stereoscopic images of each patient. Using an intermodal cross-attention mechanism, the multimodal images were spatially aligned and fused to extract structural units at four levels: bone, muscle, soft tissue, and skin. An average of 412.58 structural units were identified per patient.

[0150] Subsequently, a facial structure-function linkage map was constructed based on the anatomical prior coupling factors and functional conduction strength. The number of nodes and edges in the linkage map was 412.58 and 9,763.42, respectively, and all edge weights were normalized. Next, a density-aware clustering algorithm (based on DBSCAN, with parameters ε = 0.25 and MinPts = 8) was used to identify clusters within the space of local edge density and functional aggregation. An average of 4.33 significantly coupled regions were identified, and regions with a coupling score above 0.67 were selected as intervention areas.

[0151] The system constructs structural hierarchical distribution, spatial configuration indicators and structural coupling strength characteristics for each intervention area, and splices them into a structural comprehensive vector; it then matches them with 250 sets of expert-defined templates in the surgical strategy knowledge base, and screens out matching candidate surgical fragments based on dual judgment of cosine similarity and cross entropy. On average, 8.17 candidate surgical fragments are extracted for each patient.

[0152] A structural conflict relationship matrix and a recovery time dependency matrix were then constructed. Conflict relationships were flagged when the spatial overlap ratio exceeded 30%. The average number of structural conflict segments was 5.42 pairs, and the average number of recovery time dependency segments was 3.25 pairs. An initial surgical path map was constructed, using candidate surgical segments as nodes and conflict and dependency relationships as edges.

[0153] Based on the user's aesthetic preference vector (e.g., enhancing the cheekbones by +0.8, softening the jawline by -0.5), a continuous aesthetic shift information field covering the three-dimensional facial space is constructed. The aesthetic response intensity of each candidate segment in the shift field is then calculated, with an average value of 0.72. A jump threshold of 0.65 is used to identify the transition nodes, and a structural deformation graph is formed by combining the original path graph. The system then injects control constraints A, consisting of four constraints, to adjust the graph structure to form a reversible control graph. The set of paths that satisfy the constraints is then enumerated. The path with the highest overall score (sum of aesthetic responses minus step penalties) is ultimately selected as the output path.

[0154] The average total number of multimodal images collected was approximately 4 groups per patient; the average number of identified structural units was 412.58; the average number of edges in the facial structure linkage map was 9,763.42; the average number of significantly coupled regions was 4.33; the number of matching candidate surgical segments was 8.17; the number of structural conflict relationship segment pairs was 5.42 pairs; the number of recovery dependency segment pairs was 3.25 pairs; the average aesthetic response intensity was 0.72 (maximum 0.93, minimum 0.48); the average number of steps in the output path was 6.08; the average improvement rate of the aesthetic score was 22.36%; the average path screening time during implementation was 2.91 seconds.

[0155] Experimental data demonstrate that the introduction of a multimodal image fusion mechanism and structural-functional linkage modeling not only enables deep correlation mining between different tissue layers but also successfully avoids misleading diagnoses caused by insufficient single-modality information. During the intervention zone identification phase, the introduction of structural semantics and spatial consistency metrics, combined with a density-aware clustering algorithm, significantly improves the accuracy and interpretability of coupled region identification, effectively avoiding the irrational regional divisions and clinical mismatches common in existing technologies.

[0156] During the surgical segment matching process, a dual matching mechanism combining structural hierarchical distribution and topological similarity enables candidate surgical segments to have greater local adaptability and implementability, avoiding the over-reliance on templates and neglect of individual differences found in existing image planning schemes. Regarding path generation, by constructing a structural conflict and recovery dependency matrix, this fundamentally overcomes the bottleneck of conflict-free mapping, which previously led to logical confusion in the execution order of segments, thereby improving the stability of the execution path and the efficiency of recovery.

[0157] In the experiment, after the user set their aesthetic preferences, the system automatically identified stage jump nodes and flexibly adjusted the path structure using four types of control constraints. The resulting personalized path achieved an average improvement of 22.36% in aesthetic response compared to traditional paths, fully demonstrating the technology's high adaptability to the user's aesthetic goals while ensuring surgical safety. Furthermore, the entire path selection process took less than 3 seconds, demonstrating its excellent responsiveness and engineering feasibility in real-world clinical settings.

[0158] Embodiment 3, an embodiment of the present invention, provides a plastic surgery auxiliary strategy generation system based on multimodal medical images, including:

[0159] The facial structure-function linkage map module collects multimodal medical images, extracts multi-level tissue structure information, and constructs a facial structure-function linkage map.

[0160] The intervention area module identifies a structural coupling area with an associated influence relationship in the functional linkage diagram as an intervention area.

[0161] The query module reversely queries the surgical strategy knowledge base based on the intervention area to obtain a set of candidate surgical fragments.

[0162] The initial path map module constructs the initial surgical path map based on the structural conflict relationship matrix and the recovery time dependency matrix between candidate surgical segments.

[0163] The structural deformation atlas module generates an aesthetic offset information field based on the aesthetic preference parameters set by the user, and acts on the initial surgical path map to form a structural deformation atlas.

[0164] The execution module injects control constraints based on the structural deformation map, generates a reversible control map, and outputs a personalized surgical execution path.

[0165] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0166] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0167] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0168] It should be understood that various aspects of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the aforementioned embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, implementation may be achieved using any combination of the following technologies known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application-specific integrated circuits having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), and the like. It should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. While the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will appreciate that modifications or equivalent substitutions may be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and such modifications are intended to be encompassed by the claims of the present invention.

[0169] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for generating auxiliary strategies for plastic surgery based on multimodal medical images, characterized in that: include: After acquiring multimodal medical images, multi-level tissue structure information is extracted and a facial structure-function linkage map is constructed; Identifying structural coupling areas with associated impact relationships in the functional linkage diagram as intervention areas; Reversely query the surgical strategy knowledge base based on the intervention area to obtain a set of candidate surgical fragments; Construct an initial surgical pathway diagram based on the structural conflict relationship matrix and the recovery time dependency matrix between candidate surgical segments; According to the aesthetic preference parameters set by the user, an aesthetic deviation information field is generated and applied to the initial surgical path map to form a structural deformation map; Based on the structural deformation map, control constraints are injected, a reversible control map is generated, and a personalized surgical execution path is output.

2. The method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to claim 1, wherein: The multimodal medical images include three-dimensional CT images, MRI images, ultrasound images and three-dimensional stereoscopic images; The multi-level tissue structure information includes bone feature information, muscle feature information, soft tissue feature information and surface skin feature information.

3. The method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to claim 2, wherein: The facial structure-function linkage map construction process specifically includes: extracting the multi-level organizational structure information from the spatially aligned multimodal medical image using an inter-modal cross-attention mechanism; aggregating the multi-level organizational structure information into three-dimensional spatial coordinates using a cross-modal feature aggregation mechanism to obtain all structural units as a node set for constructing the structure-function linkage map; Preset a priori coupling factors between different organizational levels; The product of the prior coupling factor and the functional conduction strength is used as the positive factor of coupling, and the positive factor is normalized with the Euclidean distance between nodes as the denominator to serve as the coupling weight between nodes; Among all node pairs, node pairs with coupling weights greater than zero are screened out to establish edge connections and included in the edge set. At the same time, a two-dimensional symmetric edge weight matrix is ​​generated to store the coupling weight value between each pair of nodes in numerical form; the structure-function linkage diagram is composed of the node set, the edge set and the edge weight matrix; the tissue hierarchy labels include bone, muscle, soft tissue and surface skin.

4. The method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to claim 3, wherein: Identifying the structural coupling region includes extracting all adjacent nodes that have a structural functional coupling relationship with the current node as a first-order adjacent node set; The edge weights between first-order adjacent node pairs are summed and averaged as the local edge density index of the current node; the consistency of the organizational hierarchy labels between first-order adjacent node pairs is determined as the semantic consistency score; the Euclidean distance between first-order adjacent node pairs is converted into a spatial consistency score through exponential decay; The semantic consistency score and spatial consistency score between each pair of first-order adjacent nodes are multiplied and then weighted averaged to obtain the functional aggregation index. The local edge density index and the functional aggregation index are concatenated into a set of coupling feature vectors, which are then divided into multiple preliminary structural coupling regions using a density-aware clustering algorithm. The edge weights between all node pairs within each preliminary structural coupling region are summed and averaged as the regional coupling score. All coupling regions with regional coupling scores exceeding the screening threshold are screened to form a set of intervention areas.

5. The method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to claim 4, wherein: The acquiring of the candidate surgical fragment set includes counting the proportion of each type of node in the node set of the facial structure-function linkage diagram according to the organizational level label as a structural level distribution; Based on the node coordinate distribution, the geometric center, average curvature and boundary convexity are calculated to form the spatial configuration index; the average weight, maximum weight gradient and minimum weight gradient of the edge are counted to form the structural coupling strength characteristic; the structural hierarchy distribution, spatial configuration index and structural coupling strength characteristic are spliced ​​into a structural comprehensive vector; In the surgical strategy knowledge base, the structural hierarchical distribution and the predefined hierarchical distribution template are subjected to cosine similarity calculation. When the label similarity threshold is reached, the label group is consistent; The structural coupling intensity gradient constructs a histogram and calculates the cross entropy with the predefined topology template histogram. When the topological similarity threshold is set, it is topologically similar; only candidate surgical segments that meet both label group consistency and topological similarity are retained and included in the candidate surgical segment set.

6. The method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to claim 5, wherein: The constructing of the initial surgical path map includes taking the candidate surgical segments as nodes of the initial surgical path map, and constructing a structural conflict relationship matrix and a recovery time dependency matrix; The construction of the structural conflict relationship matrix includes calculating the spatial overlap between the node sets of the facial structure-function linkage graph within any two candidate surgical segments. When the overlap ratio of the spatial overlap exceeds a set threshold, a structural conflict exists, which is recorded as 1; otherwise, it is 0. The construction of the recovery time dependency matrix includes calculating the expected recovery time of each surgical segment. When the surgical segments have an execution order relationship, dependency edges are established between the corresponding nodes of the surgical segments. For segments that do not have both structural conflict and time dependency relationships, bidirectional structural edges are established. The edge set of the initial surgical path graph includes: bidirectional structural edges and dependency edges.

7. The method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to claim 6, wherein: The aesthetic preference parameters are the user's preferences for each facial structure area. the intention to enhance or soften a structural area; Constructing aesthetic preference parameter vector based on aesthetic preference parameters , where each dimension Respectively represent the the intention to enhance or soften a structural area; Indicates index; n indicates Dimensions; Generating the aesthetic offset information field includes: constructing a continuous aesthetic offset information field covering the three-dimensional facial space using a Gaussian function based on the spatial position of each structural unit in the three-dimensional facial structure model and the hierarchical label index to which it belongs, according to the weight of each dimension in the aesthetic preference parameter vector; The structure deformation map comprises calculating the average response value of the intervention region structure node corresponding to each candidate surgical segment in the aesthetic deviation information field as the aesthetic response intensity of the candidate surgical segment; when the aesthetic response intensity When the jump threshold is set, the jump candidate label is assigned to the candidate surgical segment stage; the candidate surgical segments with all jump candidate labels are used to form a label set; the label set is combined with the initial surgical path map to form a structural deformation map; After the structural deformation map is constructed, control constraints are introduced to generate a reversible control graph. The control constraints include: setting the weights of all incoming edges of candidate surgical segments in the label set to 0, and retaining only outgoing edges; for segments with a recovery time dependency, when the aesthetic response strength of the subsequent segment is lower than that of the preceding segment, the execution order of the segment pair is swapped, and the edge direction in the recovery dependency matrix is ​​adjusted at the same time; when the aesthetic response strengths of three candidate surgical segments in the path all exceed a multiple threshold of the aesthetic average value, the three candidate surgical segments are merged into one candidate surgical segment node, replacing the original path segment; when the Euclidean distance between the candidate surgical segment pairs is less than the nearest neighbor threshold and there is no structural conflict relationship, a bidirectional connecting edge is added between the candidate surgical segment pairs; The process of outputting the personalized surgery execution path is as follows: Filter the path set in the reversible control graph that meets the following conditions: the aesthetic response intensity of the segments in the label set is not less than the jump threshold, the order of all segments with recovery time dependencies is not destroyed, and the sum of the estimated execution time and recovery period of all segments in the path does not exceed the maximum duration limit set by the user; The sum of the aesthetic response intensities of all segments in each path obtained by screening was calculated as the aesthetic benefit, and the total number of path segments multiplied by the step penalty coefficient was used as the step cost. The difference between the aesthetic benefit and the step cost was used as the comprehensive score. The pathway with the highest comprehensive score is selected as the output of the personalized surgical execution pathway.

8. A system using the method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to any one of claims 1 to 7, characterized in that: The facial structure-function linkage map module collects multimodal medical images, extracts multi-level tissue structure information, and constructs a facial structure-function linkage map; An intervention area module, which identifies a structural coupling area with an associated influence relationship in the functional linkage diagram as an intervention area; A query module, which reversely queries a surgical strategy knowledge base based on the intervention area to obtain a set of candidate surgical fragments; The initial path map module constructs the initial surgical path map based on the structural conflict relationship matrix and the recovery time dependency matrix between candidate surgical segments; The structural deformation atlas module generates an aesthetic deviation information field based on the aesthetic preference parameters set by the user, and acts on the initial surgical path map to form a structural deformation atlas; The execution module injects control constraints based on the structural deformation map, generates a reversible control map, and outputs a personalized surgical execution path.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the processor implements the steps of the method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for generating a plastic surgery auxiliary strategy based on multimodal medical images according to any one of claims 1 to 7 are implemented.

Images