Method and system for generating visual elements based on medical concept mapping

Abstract

The present application relates to the technical field of medical data processing and visualization, and particularly relates to a visual element generation method and system based on medical concept mapping. The target medical concept is analyzed in terms of pathological mechanism, a pathological cause-effect chain graph containing cause-effect dependence and time sequence attributes is constructed according to a medical knowledge base, topological structure, time sequence stage and cause-effect strength are converted into spatial layout, gradual change characteristics and connection strength constraint, visual complexity and cognitive load prediction value are calculated, when the prediction value exceeds a threshold value, a target path is identified based on accumulated cause-effect strength and foldable nodes are marked, a simplified chain graph is generated, visual primitives and connection elements are generated according to the simplified chain graph, and an interactive response mechanism is associated with the foldable nodes. The method reduces cognitive load and improves the efficiency of medical concept visualization communication.

Description

Technical Field

[0001] This invention relates to the field of medical data processing and visualization technology, and in particular to a method and system for generating visual elements based on medical concept mapping. Background Technology

[0002] In the field of medical visualization, transforming abstract medical concepts into intuitive visual elements is an important means of assisting clinical decision-making and medical education.

[0003] Conventional approaches generally focus on presenting static relationships between concepts, making it difficult to effectively express the dynamic characteristics of medical concepts in the pathological evolution process. The occurrence and development of diseases often involve causal chains of multiple factors, stages, and pathways. Existing mapping methods usually only retain binary associations between concepts, ignoring the temporal evolutionary relationship between pathogenic factors, pathological processes, and clinical manifestations. This results in generated visual elements that fail to reflect the stage differences in disease progression and are unable to provide observers with a holistic reasoning path from etiology to outcome. When faced with such static maps, clinical users often need to manually integrate scattered relational information to reconstruct the pathological process, significantly increasing their cognitive burden.

[0004] Medical concept networks are typically densely packed with interconnected nodes. Existing generation methods tend to present all mappable visual elements in an attempt to achieve comprehensive information. This approach ignores the differences in users' cognitive capacity across various usage scenarios. When the target users are patients or medical students, highly complex visual outputs can easily lead to information overload, hindering the understanding of core pathological mechanisms. Even for professional physicians, the complex, nested connections reduce the efficiency of identifying critical paths, diminishing the practical value of visualization tools in emergency decision-making. Overall, existing technologies lack dynamic simplification mechanisms based on cognitive load, resulting in fixed and rigid visual element generation that struggles to provide flexible hierarchical display and highlight key points according to user status or interaction needs. Summary of the Invention

[0005] The embodiments of the present invention provide a method and system for generating visual elements based on medical concept mapping, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a method for generating visual elements based on medical concept mapping, comprising:

[0007] The pathological mechanism of the acquired target medical concept to be visualized is analyzed, the pathogenic factor node, pathological process node and clinical manifestation node are identified, the causal dependency relationship between the nodes is established based on the medical knowledge base and the temporal stage attribute is marked to form a pathological causal chain diagram.

[0008] The topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph are respectively converted into spatial layout constraints, gradient feature constraints, and connection strength constraints of visual elements.

[0009] The visual complexity index is calculated based on the number of nodes, connection density, and nesting level of the pathological causal chain diagram, and the cognitive load prediction value is calculated in combination with the cognitive capacity threshold of the target user.

[0010] When the predicted cognitive load value exceeds the preset cognitive load threshold, the target display path set is identified based on the cumulative causal strength value of each path, and the nodes outside of it are marked as collapsible nodes, generating a simplified pathological causal chain diagram.

[0011] Based on spatial layout constraints, gradient feature constraints, and connection strength constraints, visual primitives and connecting elements are generated for each node of the simplified pathological causal chain diagram or pathological causal chain diagram to form composite visual elements.

[0012] The associated interactive response mechanism for the foldable node receives user interaction commands, renders the visual display of the corresponding folded content, and records the interaction behavior data.

[0013] The acquired target medical concepts to be visualized are analyzed for pathological mechanisms, identifying pathogenic factor nodes, pathological process nodes, and clinical manifestation nodes. Based on a medical knowledge base, causal dependencies between nodes are established and temporal stage attributes are labeled, forming a pathological causal chain diagram, including:

[0014] Semantic parsing of the target medical concept is performed to extract core pathological entities;

[0015] Based on the etiological classification relationship of core pathological entities in the medical knowledge base, pathogenic factor nodes are identified; based on the pathological evolution relationship, pathological process nodes are identified; and based on the symptom mapping relationship, clinical manifestation nodes are identified.

[0016] The causal association rules between nodes are extracted from the medical knowledge base. The logical dependency type between the antecedent conditions and the subsequent results in the causal association rules is analyzed. The dependency strength weight is marked according to the directed causal dependency relationship between the pathogenic factor node, the pathological process node and the clinical manifestation node.

[0017] Time series data of pathological evolution are extracted from the medical knowledge base. Based on the time distribution characteristics of the occurrence of each pathological state recorded in the time series data, the temporal position quantification values ​​of each pathogenic factor node, pathological process node and clinical manifestation node in the whole cycle of pathological evolution are calculated as temporal stage attributes.

[0018] The pathological causal chain graph is constructed by using the pathogenic factor node, pathological process node, and clinical manifestation node as graph nodes, the directed causal dependency relationship labeled with the dependency strength weight as graph edges, and the quantized value of the time sequence position as node attribute.

[0019] The topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph are respectively converted into spatial layout constraints, gradation feature constraints, and connection strength constraints of visual elements, including:

[0020] The topological level position of each node is calculated based on the connection relationship between nodes in the pathological causal chain diagram. Vertical hierarchical layout rules are generated according to the topological level position. Horizontal expansion layout rules are generated according to the causal branch relationship of nodes in the same topological level. The vertical hierarchical layout rules and the horizontal expansion layout rules are combined to form spatial layout constraints.

[0021] Extract the temporal stage attributes of each node in the pathological causal chain graph, and map the temporal stage attributes to visual gradient parameters. The visual gradient parameters include color gradient parameters determined according to the relative position of the temporal stage attributes on the time axis, and transparency gradient parameters determined according to the time span of the temporal stage attributes. The visual gradient parameters are used as gradient feature constraints.

[0022] Extract the causal dependency strength corresponding to each edge in the pathological causal chain graph, and map the causal dependency strength to the visual parameters of the connecting line. The visual parameters of the connecting line include the line thickness parameter and the line saturation parameter, both of which are positively correlated with the causal dependency strength. The visual parameters of the connecting line are used as the connection strength constraint.

[0023] Based on the number of nodes, connection density, and nesting level of the pathological causal chain graph, a visual complexity index is calculated. Combined with the cognitive capacity threshold of the target user, a cognitive load prediction value is calculated, including:

[0024] The total number of nodes in the pathological causal chain graph is used as the number of nodes, the ratio of the total number of edges to the total number of nodes is used as the connection density, and the maximum value of the topological level is used as the nesting level.

[0025] Identify the node type identifiers for each node, including pathogenic factor type identifiers, pathological process type identifiers, and clinical manifestation type identifiers. Set cognitive load contribution weights based on the cognitive processing depth of each node type identifier in medical knowledge understanding. Calculate the weighted sum of the number of nodes of each type and their corresponding weights as the node heterogeneity complexity.

[0026] Detect locally high-density connected subgraphs whose connection density exceeds the average connection density of the entire graph, and obtain the connection aggregation complexity by summing the products of the number of nodes and the number of edges in each subgraph.

[0027] The visual complexity index is obtained by exponentially calculating the node heterogeneity complexity, connection aggregation complexity, and nesting level, respectively, and then summing them. The exponent of the exponential calculation is determined based on the marginal contribution rate of each complexity component to the cognitive load.

[0028] The average information processing rate and visual dwell time distribution in the target user's historical interactions are obtained, and then converted into the number of nodes that can be processed per unit time and the attention concentration ability coefficient, respectively, and multiplied to obtain the cognitive capacity threshold.

[0029] The ratio of the visual complexity index to the cognitive capacity threshold is used as a predictor of cognitive load.

[0030] Based on the cumulative causal strength values ​​of each path, the target path set is identified, and nodes outside of it are marked as collapsible nodes, generating a simplified pathological causal chain graph, including:

[0031] When the predicted cognitive load value exceeds the preset cognitive load threshold, a simplified processing procedure is executed.

[0032] Extract the complete causal transmission path from the pathogenic factor node to the clinical manifestation node in the pathological causal chain graph, and perform cumulative causal dependence strength calculation on each side along each complete causal transmission path to obtain the cumulative causal strength value.

[0033] All complete causal transmission paths are sorted in descending order of cumulative causal intensity value. The preceding paths whose sum of cumulative causal intensity values ​​reaches the proportion threshold of the sum of cumulative causal intensity values ​​of all paths are selected to form the target display path set. The proportion threshold is dynamically adjusted according to the difference between the cognitive load prediction value and the preset cognitive load threshold and is negatively correlated.

[0034] Nodes in the pathological causal chain graph that are not included in the target display path set are marked as collapsible nodes, and nodes and edges in the target display path set are retained to generate a simplified pathological causal chain graph.

[0035] Based on spatial layout constraints, gradation feature constraints, and connection strength constraints, visual primitives and connecting elements are generated for each node of the simplified pathological causal chain diagram or pathological causal chain diagram to form composite visual elements, including:

[0036] Extract each node from the simplified pathological causal chain diagram or the pathological causal chain diagram, determine the spatial coordinate position of the visual primitive of each node according to the spatial layout constraints, and determine the color value and transparency value of the visual primitive of each node according to the gradient feature constraints.

[0037] For each node, the number of neighboring nodes and the density of neighboring connections within its neighborhood are statistically analyzed and weighted to obtain the local visual complexity.

[0038] Based on the local visual complexity of each node, a color saturation attenuation coefficient and a visual primitive size scaling coefficient are generated. When the local visual complexity exceeds the preset local complexity threshold, the color saturation of the corresponding node's visual primitive is reduced and the visual primitive size is reduced.

[0039] The color saturation attenuation coefficient is applied to the color value to generate adjusted color parameters, and the visual primitive size scaling coefficient is applied to the spatial coordinate position to generate adjusted spatial parameters. Based on the adjusted color parameters, transparency value and adjusted spatial parameters, node visual primitives with local complexity adjustment are generated.

[0040] The line thickness and saturation of the corresponding connecting elements on each side are determined according to the connection strength constraint. The connecting elements are generated by combining the adjusted spatial parameters of the starting node visual primitives and the ending node visual primitives of each side. The node visual primitives and connecting elements are combined to form a composite visual element.

[0041] The associated interactive response mechanism for the foldable node receives user interaction commands, renders the visual display of the corresponding folded content, and records interaction behavior data, including:

[0042] Generate interactive control elements for the foldable nodes and bind them to the display positions of the corresponding node visual primitives;

[0043] Monitor user trigger operations on the interactive control element, extract the pathological causal subgraph of the pathological causal chain graph corresponding to the trigger operation, calculate the unfolding complexity based on the number of nodes and connection density of the pathological causal subgraph, and determine the layout scaling ratio and superimposed display position of the unfolded visual element based on the unfolding complexity.

[0044] Each node and edge in the pathological causal subgraph is used to generate an expanded visual element according to the spatial layout constraints, gradient feature constraints and connection strength constraints. The expanded visual element is then superimposed onto the display area of ​​the current composite visual element according to the layout scaling ratio and superimposed display position.

[0045] The trigger timestamp of the recorded action and the corresponding collapsible node identifier form an interaction behavior record.

[0046] A second aspect of this invention provides a visual element generation system based on medical concept mapping, comprising:

[0047] The pathology analysis unit is used to analyze the pathological mechanism of the acquired target medical concepts to be visualized, identify pathogenic factor nodes, pathological process nodes and clinical manifestation nodes, establish causal dependencies between nodes based on the medical knowledge base and mark the temporal stage attributes to form a pathological causal chain diagram.

[0048] The constraint transformation unit is used to convert the topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph into spatial layout constraints, gradient feature constraints, and connection strength constraints of visual elements, respectively.

[0049] The load prediction unit is used to calculate the visual complexity index based on the number of nodes, connection density and nesting level of the pathological causal chain graph, and to calculate the cognitive load prediction value in combination with the cognitive capacity threshold of the target user.

[0050] The path simplification unit is used to identify the target display path set based on the cumulative causal strength value of each path when the cognitive load prediction value exceeds the preset cognitive load threshold, mark the nodes outside of it as collapsible nodes, and generate a simplified pathological causal chain diagram.

[0051] The visual generation unit is used to generate visual primitives and connecting elements for each node of the simplified pathological causal chain diagram or pathological causal chain diagram according to spatial layout constraints, gradient feature constraints and connection strength constraints, so as to form composite visual elements.

[0052] The interactive response unit is used to associate the foldable node with an interactive response mechanism, receive user interaction instructions, render the visual display of the corresponding folded content, and record interactive behavior data.

[0053] A third aspect of the present invention provides an electronic device, comprising:

[0054] processor;

[0055] Memory used to store processor-executable instructions;

[0056] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0057] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0058] The beneficial effects of this invention are:

[0059] By analyzing the pathological mechanisms of medical concepts, a pathological causal chain diagram with causal semantics is constructed, fully presenting the evolutionary relationship between pathogenic factors, pathological processes, and clinical manifestations. This transforms abstract medical concepts into traceable and deducible structured knowledge. The causal chain diagram contains causal dependencies and temporal attributes between nodes, providing accurate semantic basis for visualization and applicable to scenarios such as medical education, clinical auxiliary explanation, and scientific research presentation.

[0060] By mapping topological structure, temporal attributes, and causal strength to spatial, gradation, and connectivity constraints, the arrangement of visual elements strictly corresponds to medical logic. Users can intuitively perceive the sequence and primary / secondary causal relationships of disease development, avoiding misalignment between visual expression and pathological semantics. Based on visual complexity indicators and cognitive load predictions, the system pre-assesses the cognitive load of graphics on the target user. When the prediction exceeds a threshold, secondary paths are automatically filtered out, retaining only key display paths. This ensures that the output content matches the user's cognitive capacity, significantly reducing the cognitive burden. The simplified and retained composite visual elements balance overall overview and in-depth local analysis. Collapsible nodes hide redundant information without disrupting the integrity of the main path, allowing users to expand them as needed, balancing macroscopic and microscopic perspectives.

[0061] The recorded behavioral and preference data during the interaction process are continuously fed back, supporting the adaptive adjustment of subsequent visualization solutions. Over long-term use, the matching degree between the generated results and individual cognitive habits is gradually improved. The accumulation of behavioral data provides a closed-loop basis for system optimization, and the visualization effect is continuously refined with the increase in usage frequency. Attached Figure Description

[0062] Figure 1 This is a flowchart illustrating a visual element generation method based on medical concept mapping.

[0063] Figure 2 Pathological causal chain diagram visual constraint transformation flowchart. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0065] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0066] Figure 1 This is a flowchart illustrating the visual element generation method based on medical concept mapping according to an embodiment of the present invention. The visual element generation method based on medical concept mapping includes:

[0067] The pathological mechanism of the acquired target medical concept to be visualized is analyzed, the pathogenic factor node, pathological process node and clinical manifestation node are identified, the causal dependency relationship between the nodes is established based on the medical knowledge base and the temporal stage attribute is marked to form a pathological causal chain diagram.

[0068] The topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph are respectively converted into spatial layout constraints, gradient feature constraints, and connection strength constraints of visual elements.

[0069] The visual complexity index is calculated based on the number of nodes, connection density, and nesting level of the pathological causal chain diagram, and the cognitive load prediction value is calculated in combination with the cognitive capacity threshold of the target user.

[0070] When the predicted cognitive load value exceeds the preset cognitive load threshold, the target display path set is identified based on the cumulative causal strength value of each path, and the nodes outside of it are marked as collapsible nodes, generating a simplified pathological causal chain diagram.

[0071] Based on spatial layout constraints, gradient feature constraints, and connection strength constraints, visual primitives and connecting elements are generated for each node of the simplified pathological causal chain diagram or pathological causal chain diagram to form composite visual elements.

[0072] The associated interactive response mechanism for the foldable node receives user interaction commands, renders the visual display of the corresponding folded content, and records the interaction behavior data.

[0073] The acquired target medical concepts to be visualized are analyzed for pathological mechanisms, identifying pathogenic factor nodes, pathological process nodes, and clinical manifestation nodes. Based on a medical knowledge base, causal dependencies between nodes are established and temporal stage attributes are labeled, forming a pathological causal chain diagram, including:

[0074] Semantic parsing of the target medical concept is performed to extract core pathological entities;

[0075] Based on the etiological classification relationship of core pathological entities in the medical knowledge base, pathogenic factor nodes are identified; based on the pathological evolution relationship, pathological process nodes are identified; and based on the symptom mapping relationship, clinical manifestation nodes are identified.

[0076] The causal association rules between nodes are extracted from the medical knowledge base. The logical dependency type between the antecedent conditions and the subsequent results in the causal association rules is analyzed. The dependency strength weight is marked according to the directed causal dependency relationship between the pathogenic factor node, the pathological process node and the clinical manifestation node.

[0077] Time series data of pathological evolution are extracted from the medical knowledge base. Based on the time distribution characteristics of the occurrence of each pathological state recorded in the time series data, the temporal position quantification values ​​of each pathogenic factor node, pathological process node and clinical manifestation node in the whole cycle of pathological evolution are calculated as temporal stage attributes.

[0078] The pathological causal chain graph is constructed by using the pathogenic factor node, pathological process node, and clinical manifestation node as graph nodes, the directed causal dependency relationship labeled with the dependency strength weight as graph edges, and the quantized value of the time sequence position as node attribute.

[0079] To analyze the pathological mechanisms of a target medical concept, it is first necessary to deconstruct the concept itself at the semantic level. Taking "type 2 diabetes" as an example, semantic analysis is not simply keyword extraction, but rather a structured analysis of the concept's definition text, diagnostic criteria description, and relevant literature abstracts using natural language processing technology. This process identifies entity words that carry core pathological meaning, such as "insulin resistance," "impaired pancreatic β-cell function," and "hyperglycemia." These core pathological entities constitute the semantic anchors for subsequent node identification, determining the coverage and semantic boundaries of the entire pathological causal chain.

[0080] After the core pathological entities are identified, they are classified and assigned based on a medical knowledge base. The medical knowledge base typically organizes knowledge using an ontology structure, including three core relationship types: etiology relation, pathological progression relation, and symptom mapping relation. Based on the etiology relation, "genetic susceptibility," "high-calorie diet," and "sedentary lifestyle" can be identified as pathogenic factor nodes. Based on the pathological progression relation, "impaired insulin signaling pathway," "glucose transport disorder," and "chronic low-grade inflammation" can be identified as pathological process nodes. Based on the symptom mapping relation, "polydipsia and polyuria," "retinopathy," and "peripheral neuropathy" can be identified as clinical manifestation nodes. The identification logic of these three types of nodes is independent, corresponding to the "causal layer," "mechanism layer," and "manifestation layer" of disease occurrence, respectively, together forming the complete structure of the pathological narrative.

[0081] After node identification, causal relationship rules between nodes need to be extracted from the medical knowledge base. These rules are typically stored in the logical form of "if A then B," where A is the antecedent condition and B is the subsequent result. Parsing these rules focuses on determining the type of logical dependency between the antecedent and the subsequent result. Common dependency types include: sufficient condition dependency (the antecedent condition alone is sufficient to trigger the subsequent result), necessary condition dependency (the subsequent result requires the antecedent condition to occur), and co-condition dependency (multiple antecedent conditions work together to trigger the subsequent result). Different logical dependency types correspond to different dependency strength weighting strategies: sufficient condition dependencies are assigned higher weights, necessary condition dependencies are assigned medium weights, and co-condition dependencies are weighted according to the number of co-factors. Dependency strength weights are numerically labeled on the directed causal dependencies, providing a quantitative basis for subsequent path selection and cognitive load calculation.

[0082] The annotation of temporal stage attributes relies on pathological evolution time-series data stored in the medical knowledge base. This data originates from epidemiological cohort studies, clinical follow-up records, and pathological literature, recording the typical time distribution of each pathological state in the natural history of the disease. For each node, the time distribution characteristics of its corresponding pathological state throughout the entire pathological evolution cycle are extracted, including the median time of occurrence, the time interval of occurrence, and its temporal relationship with other nodes. Taking type 2 diabetes as an example, "insulin resistance" usually appears several years before clinical diagnosis, while "retinopathy" often gradually manifests several years after diagnosis. Based on the time distribution characteristics of each node, its temporal position quantification value in the entire pathological evolution cycle is calculated and denoted as... ,in For node indexing. The calculation method is to use nodes median occurrence time Normalized to the full lifecycle time span Above, that is , making The smaller the value, the earlier the node appears in the pathological evolution; the larger the value, the later it appears. As a node attribute, it is attached to the corresponding node and directly reflects its stage affiliation in the time dimension.

[0083] After completing node identification, dependency labeling, and temporal attribute calculation, these elements are integrated into a graph structure. The graph uses pathogenic factor nodes, pathological process nodes, and clinical manifestation nodes as the vertex set, and directed causal dependencies labeled with dependency strength weights as the edge set. The temporal position quantification value is used as the edge set. As attribute fields for each node, a complete pathological causal chain graph is constructed. The directionality of the graph ensures the explicit expression of the causal direction, the weight attribute supports subsequent path strength calculation, and the temporal attribute of the nodes provides a numerical basis for the gradual feature mapping of visual elements. For complex diseases with multiple parallel causal paths, the graph structure naturally supports branching and converging topologies, and can fully express the mechanistic features of multi-factor synergistic pathogenesis.

[0084] In practice, the quality of the medical knowledge base directly affects the accuracy of node identification and the completeness of causal association rules. To improve robustness, causal association rules with confidence levels below a preset threshold can be filtered out, retaining only high-confidence rules supported by sufficient literature for graph construction. For nodes lacking time-series data in the knowledge base, interpolation estimation can be performed using time-series distribution data of similar diseases, and the confidence level of the data source can be marked in the node attributes to differentiate uncertainty during subsequent visualization. The entire process of constructing the pathological causal chain graph forms a reusable workflow applicable to different types of target medical concepts, possessing good versatility and scalability.

[0085] The topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph are respectively converted into spatial layout constraints, gradation feature constraints, and connection strength constraints of visual elements, including:

[0086] The topological level position of each node is calculated based on the connection relationship between nodes in the pathological causal chain diagram. Vertical hierarchical layout rules are generated according to the topological level position. Horizontal expansion layout rules are generated according to the causal branch relationship of nodes in the same topological level. The vertical hierarchical layout rules and the horizontal expansion layout rules are combined to form spatial layout constraints.

[0087] Extract the temporal stage attributes of each node in the pathological causal chain graph, and map the temporal stage attributes to visual gradient parameters. The visual gradient parameters include color gradient parameters determined according to the relative position of the temporal stage attributes on the time axis, and transparency gradient parameters determined according to the time span of the temporal stage attributes. The visual gradient parameters are used as gradient feature constraints.

[0088] Extract the causal dependency strength corresponding to each edge in the pathological causal chain graph, and map the causal dependency strength to the visual parameters of the connecting line. The visual parameters of the connecting line include the line thickness parameter and the line saturation parameter, both of which are positively correlated with the causal dependency strength. The visual parameters of the connecting line are used as the connection strength constraint.

[0089] Combination Figure 2The flowchart illustrating the visual constraint transformation of a pathological causal chain graph illustrates that after the pathological causal chain graph is constructed, its structured information needs to be converted into a set of constraint parameters that can be used by the visual rendering engine. This transformation process consists of three parallel mapping paths, which respectively handle topological structure, temporal attributes, and causal strength, ultimately forming three types of visual rules: spatial layout constraints, gradient feature constraints, and connection strength constraints.

[0090] The calculation of topological hierarchy is based on the directed connections between nodes. For each node in the pathological causal chain graph, a breadth-first traversal is performed starting from the root node (usually the causative factor node), and the shortest path length required to reach that node is calculated. This length is defined as the node's topological hierarchy number. Nodes with the same hierarchy number are grouped into the same horizontal layer, while nodes with different hierarchy numbers are arranged vertically. The vertical layering layout rule is thus generated: the larger the hierarchy number, the lower the node's vertical coordinate in the visual canvas, and a fixed vertical spacing is maintained between adjacent layers to ensure that the causal transmission direction is clearly presented visually from top to bottom. This layout naturally aligns with the temporal direction of pathological evolution, allowing observers to intuitively track disease progression along the vertical axis.

[0091] Within the same topological level, causal branching relationships may exist between nodes, meaning that a higher-level node simultaneously triggers multiple nodes at the same or lower levels. The lateral expansion layout rule addresses this branching structure: taking the branch's starting node as the center, all its direct child nodes are evenly expanded horizontally. The lateral coordinates of the child nodes are determined by adding an offset to the lateral coordinates of the branch's starting node. The offset is calculated based on the number of child nodes and a preset minimum node spacing. When there are multi-level nested branches, the above rule is recursively applied to ensure that the lateral expansion of each level does not result in node overlap. Combining the vertical layered layout rule with the lateral expansion layout rule forms a complete spatial layout constraint. This constraint is expressed as the minimum spacing requirement between the target coordinate range of each node and its adjacent nodes, and is used in the subsequent visual primitive generation stage.

[0092] The mapping of temporal stage attributes uses the relative position of each node on the pathological evolution timeline as the core input. Each node carries information about the time interval of its corresponding pathological state, including the start time, median time, and duration of that state. After normalizing the relative position of the nodes on the timeline, it is mapped to color gradient parameters: nodes with earlier temporal positions (corresponding to the early stages of the disease) are assigned the cooler end of the color gradient sequence, nodes with later temporal positions (corresponding to the late stages of the disease or clinical manifestations) are assigned the warmer end, and intermediate nodes are obtained by linear interpolation to obtain transitional color values. This color coding scheme allows observers to directly perceive the temporal dimension of pathological progress through color, distinguishing between early pathogenic factors and late clinical manifestations without additional illustrations.

[0093] The transparency gradient parameter is determined by the time span of the pathological state corresponding to the node. Nodes with shorter time spans indicate that the pathological state occurs rapidly and lasts for a limited time, corresponding to higher opacity values, appearing as clear entities visually. Nodes with longer time spans indicate that the state persists or evolves slowly, corresponding to lower opacity values, appearing as semi-transparent, suggesting a chronic or background nature. The specific range of transparency values ​​is set between 0.45 and 1.0 to ensure that all nodes are visually recognizable and to avoid information loss due to excessive transparency. The color gradient parameter and the transparency gradient parameter together constitute the gradient feature constraint, stored in a mapping structure with node identifiers as keys and color value-transparency value pairs as values, for direct lookup during visual primitive generation.

[0094] The mapping of causal dependency strength is performed on each directed edge in the pathological causal chain graph. Each edge carries a causal dependency strength value, which is labeled by the medical knowledge base during the construction of the pathological causal chain graph, reflecting the certainty and magnitude of the causal relationship between two nodes. The causal dependency strength value is linearly mapped to the line thickness parameter of the connecting line: edges with higher strength values ​​correspond to thicker connecting lines, and edges with lower strength values ​​correspond to thinner connecting lines. The range of line thickness values ​​is dynamically determined according to the target rendering resolution and canvas size. Typically, the thinnest line is set to 1 pixel unit, the thickest line to 6 pixel units, and intermediate values ​​are interpolated according to the intensity ratio.

[0095] The line saturation parameter also shows a positive correlation with the strength of causal dependence: edges with high intensity values ​​are rendered with highly saturated colors, visually forming distinct connection channels; edges with low intensity values ​​are rendered with low saturation colors (approaching gray), visually receding into background connections and not interfering with the identification of the main causal path. This dual visual encoding method—utilizing both line thickness and color saturation—enhances the perceptibility of causal strength information, allowing observers to quickly identify the dominant causal relationship even in densely connected areas. The line thickness and line saturation parameters together constitute the visual parameters of the connection line, stored with the edge identifier as the key, forming connection strength constraints.

[0096] The generation processes of the three types of constraints are independent and can be executed in parallel. They are ultimately aggregated into a unified constraint data structure, serving as the input specification for the subsequent visual primitive generation stage. Spatial layout constraints determine the position of each node on the canvas, gradient feature constraints determine the color and transparency of each node, and connection strength constraints determine the visual attributes of each edge's lines. Together, these three constitute a complete mapping system from pathological semantics to visual expression. In practical applications, when the target medical concept is diabetic nephropathy, the insulin resistance node, as an early pathogenic factor node, is given a cool color tone and high opacity; the glomerular sclerosis node, as a late-stage pathological process node, is given a warm color tone and medium opacity; and the strong causal edge from insulin resistance to hyperglycemia is rendered with thick lines and high saturation, intuitively presenting the core position of this path in the overall pathological mechanism.

[0097] Based on the number of nodes, connection density, and nesting level of the pathological causal chain graph, a visual complexity index is calculated. Combined with the cognitive capacity threshold of the target user, a cognitive load prediction value is calculated, including:

[0098] The total number of nodes in the pathological causal chain graph is used as the number of nodes, the ratio of the total number of edges to the total number of nodes is used as the connection density, and the maximum value of the topological level is used as the nesting level.

[0099] Identify the node type identifiers for each node, including pathogenic factor type identifiers, pathological process type identifiers, and clinical manifestation type identifiers. Set cognitive load contribution weights based on the cognitive processing depth of each node type identifier in medical knowledge understanding. Calculate the weighted sum of the number of nodes of each type and their corresponding weights as the node heterogeneity complexity.

[0100] Detect locally high-density connected subgraphs whose connection density exceeds the average connection density of the entire graph, and obtain the connection aggregation complexity by summing the products of the number of nodes and the number of edges in each subgraph.

[0101] The visual complexity index is obtained by exponentially calculating the node heterogeneity complexity, connection aggregation complexity, and nesting level, respectively, and then summing them. The exponent of the exponential calculation is determined based on the marginal contribution rate of each complexity component to the cognitive load.

[0102] The average information processing rate and visual dwell time distribution in the target user's historical interactions are obtained, and then converted into the number of nodes that can be processed per unit time and the attention concentration ability coefficient, respectively, and multiplied to obtain the cognitive capacity threshold.

[0103] The ratio of the visual complexity index to the cognitive capacity threshold is used as a predictor of cognitive load.

[0104] Before visualizing the pathological causal chain diagram, it is necessary to quantify and statistically analyze its structural features. This involves traversing all nodes in the pathological causal chain diagram and counting the total number of nodes. Traverse all directed edges and count the total number of edges. The ratio of the total number of edges to the total number of nodes. Connection density reflects the average density of causal relationships between nodes in the graph; a hierarchical traversal of the graph is performed, and the maximum topological depth from the root node to the leaf node is recorded, with this maximum value used as the nesting level. These three basic statistics together describe the structural characteristics of the pathological causal chain graph in terms of scale, density, and hierarchical depth, providing the raw input for subsequent complexity calculations.

[0105] The calculation of node heterogeneity complexity depends on the distinction between node types. Nodes in the pathological causal chain graph are classified into three categories according to medical semantics: pathogenic factor type nodes, pathological process type nodes, and clinical manifestation type nodes. These three types of nodes require significantly different levels of cognitive processing depth in understanding medical knowledge. Pathogenic factor type nodes typically correspond to etiological knowledge, requiring readers to have a certain level of basic medical background; pathological process type nodes involve dynamic physiological and pathological changes, requiring the highest cognitive processing depth; clinical manifestation type nodes correspond to observable symptoms and signs, are relatively intuitive, and require a lower cognitive processing depth. Based on these differences, cognitive load contribution weights are assigned to each of the three types of nodes. , , ,in The weight values ​​can be determined based on domain expert evaluation or user experimental data. Let the number of nodes in the three types be respectively... , , Then the node heterogeneity complexity is:

[0106]

[0107] The computation of connection clustering complexity is applied to locally high-density regions in the graph. The average connection density of the entire graph is calculated instead. Then, the local connectivity density within each connected subgraph or local neighborhood is detected, and the local connectivity density exceeding the specified value is considered. The subgraph is identified as a high-density connected subgraph. Let a total of [number] subgraphs be identified. The high-density subgraph, the first The number of nodes in each subgraph is The number of edges is The complexity of join aggregation is:

[0108]

[0109] The physical meaning of this indicator is that the product of nodes and edges in a local high-density region can simultaneously reflect the scale and density of the region. The combined effect of the two on cognitive load is much higher than the contribution of an equal number of nodes in a sparse region.

[0110] Visual complexity index The result is obtained by summing the three components after performing exponentiation on each of them:

[0111]

[0112] in , , These are the power exponents corresponding to node heterogeneity complexity, connection aggregation complexity, and nesting level, respectively. The power exponents are determined based on the marginal contribution rate of each component to cognitive load: if the influence of a component on cognitive load increases exponentially with increasing value, the corresponding exponent is greater than 1; if the influence tends to saturate, the exponent is less than 1. The marginal contribution rate can be obtained through regression fitting using eye-tracking data, task completion time, or subjective load scores (such as the NASA-TLX scale) from user cognitive experiments. In the initial deployment phase, where experimental data is lacking, all three exponents can be set to 1, i.e., degenerating into a linear summation, and nonlinear calibration can be performed after accumulating sufficient user data.

[0113] Cognitive capacity threshold The calculations are derived from the target user's historical interaction data. Two types of metrics are extracted from the user's past graphical browsing records: average information processing speed. (Unit: nodes / second, obtained by counting the number of nodes successfully understood and operated by the user per unit time) and visual dwell time distribution (i.e., the sequence of user gaze duration at each node). Statistical analysis is performed on the visual dwell time distribution to calculate its central tendency and dispersion: the more concentrated the dwell time distribution, the more stable the user's attention allocation and the higher the attention concentration ability coefficient. The higher the value, the more dispersed the distribution. The lower the value, the better. Specifically, the ratio of the mean to the standard deviation of the duration of stay distribution (i.e., the reciprocal of the coefficient of variation) can be normalized to a certain value. interval as The estimated value. Convert the average information processing rate into the number of nodes that can be processed per unit time. (If the original data is already in nodes / second, use it directly), multiply it by the attention concentration ability coefficient to obtain the cognitive capacity threshold: . The meaning is the number of equivalent nodes that a target user can process per unit of time while maintaining effective cognitive processing, which comprehensively reflects the user's processing speed and attention stability.

[0114] Cognitive Load Predictions Defined as the ratio of visual complexity index to cognitive capacity threshold: .when When the complexity is small, the visual complexity of the diagram is within the user's cognitive capacity, and the complete pathological causal chain diagram can be directly displayed; when Exceeding the preset cognitive load threshold At this point, subsequent path filtering and node folding processes are triggered, generating a simplified pathological causal chain diagram. Preset threshold. It can be stratified according to the characteristics of user groups. For example, a higher threshold can be set for medical professionals and a lower threshold can be set for patients or non-professional users, thereby realizing an adaptive display strategy for different audiences.

[0115] In practical deployments, users' historical interaction data may be insufficient during initial use. To address this cold start issue, a user grouping strategy can be adopted: users are grouped into corresponding user groups based on their professional background (e.g., clinicians, medical students, general users) provided during registration, with the average user count used within each group. and As an initial value, it is gradually replaced with a personalized estimate as interaction data accumulates. This mechanism ensures that the cognitive capacity threshold still has reasonable estimation accuracy during the data sparsity phase, avoiding serious deviations in cognitive load prediction due to insufficient initial data.

[0116] Based on the cumulative causal strength values ​​of each path, the target path set is identified, and nodes outside of it are marked as collapsible nodes, generating a simplified pathological causal chain graph, including:

[0117] When the predicted cognitive load value exceeds the preset cognitive load threshold, a simplified processing procedure is executed.

[0118] Extract the complete causal transmission path from the pathogenic factor node to the clinical manifestation node in the pathological causal chain graph, and perform cumulative causal dependence strength calculation on each side along each complete causal transmission path to obtain the cumulative causal strength value.

[0119] All complete causal transmission paths are sorted in descending order of cumulative causal intensity value. The preceding paths whose sum of cumulative causal intensity values ​​reaches the proportion threshold of the sum of cumulative causal intensity values ​​of all paths are selected to form the target display path set. The proportion threshold is dynamically adjusted according to the difference between the cognitive load prediction value and the preset cognitive load threshold and is negatively correlated.

[0120] Nodes in the pathological causal chain graph that are not included in the target display path set are marked as collapsible nodes, and nodes and edges in the target display path set are retained to generate a simplified pathological causal chain graph.

[0121] When cognitive load prediction value Exceeding the preset cognitive load threshold At this point, a simplified processing flow is triggered. The number of nodes and connections in the pathological causal chain diagram exceeds the cognitive capacity of the target user. It is necessary to retain the main causal transmission trunk with the most medical diagnostic value through a path filtering mechanism, while marking secondary nodes as collapsible to reduce the information density of the visual presentation.

[0122] In the path extraction phase, all complete causal transmission paths originating from pathogenic factor nodes and terminating at clinical manifestation nodes are enumerated from the pathological causal chain graph. Each complete path consists of several directed edges connected sequentially, and each edge carries a causal dependency strength value. This value is labeled by the pathological mechanism analysis phase based on the medical knowledge base, and its range is normalized to a certain value. A higher value indicates a more direct and definite causal relationship between the two nodes. For a complete path, the cumulative causal strength value of the path is obtained by multiplying the causal dependency strength values ​​of all the edges it traverses. The physical meaning of multiplication is that if the causal relationship of any link in a transmission chain is weak, the overall transmission efficiency of the entire path will decrease accordingly. This is consistent with the actual law of the "weakest link effect" in multi-step pathological mechanisms in medicine.

[0123] Let's illustrate the calculation process using a specific scenario. Suppose a pathological causal chain diagram describes the mechanism by which hypertension leads to myocardial infarction, where there exists a path: hypertension → vascular endothelial injury → atherosclerosis → coronary artery stenosis → myocardial ischemia → myocardial infarction. The causal dependence strength values ​​of each edge are as follows: , , , , Then the cumulative causal strength value of this path If another path passes through an intermediate node with a weak causal relationship (such as some rare secondary mechanism), the causal dependency strength of its corresponding edge may only be [value missing]. Then the entire path It will be significantly lower than the main path and will naturally fall into the low priority range in subsequent sorting.

[0124] After calculating the cumulative causal strength value for all complete paths, according to Sort all paths in descending order to form an ordered path list. Let the sum of the cumulative causal strength values ​​of all paths be... The ratio threshold is Then, starting from the top of the ordered list, the paths are accumulated sequentially. until the accumulated value reaches Up to this point, the selected set of preceding paths constitutes the target display path set. (Scale threshold) It is not a fixed constant, but rather based on the predicted value of cognitive load. Compared with the preset cognitive load threshold The difference Dynamically adjusted, and with Negative correlation: when A lower level indicates that the cognitive load is only slightly excessive. Take the higher value (e.g.) to ), retain more paths to maintain content integrity; when A higher value indicates that the cognitive load is severely excessive. Take the lower value (e.g.) to Only the most critical, high-intensity paths are retained, while the displayed content is significantly compressed. This dynamic adjustment mechanism ensures that the degree of simplification is proportional to the actual cognitive load, avoiding oversimplification that could lead to the loss of key pathological information, and also avoiding undersimplification that could result in an excessive cognitive load.

[0125] and The negative correlation between them can be achieved through piecewise linear mapping or a monotonically decreasing function. For example, setting... The upper limit of the reference is ,but:

[0126]

[0127] in This is the upper limit of the proportional threshold. This is the lower limit of the proportional threshold. This is the reference maximum value for the cognitive load difference, when hour Pick .

[0128] Once the target display path set is determined, all nodes in the pathological causal chain graph are traversed, and nodes not included in any path in the target display path set are marked as collapsible nodes. The specific judgment logic is as follows: if a node does not appear in any path in the target display path set, it is marked as collapsible; if a node appears in at least one path in the target display path set, it is retained as a display node. For nodes marked as collapsible, the connecting edges between them and their adjacent nodes are not included in the simplified pathological causal chain graph display layer, but these nodes and their associated complete subgraph structures are fully preserved at the data level to support dynamic expansion rendering during subsequent user interaction.

[0129] The simplified pathological causal chain graph consists of all nodes and edges involved in the target display path set. If multiple paths in the target display path set share some nodes or edges (i.e., there is a convergence or branching structure between paths), the shared nodes and edges are retained only once and not counted repeatedly. The final simplified pathological causal chain graph is a connected subgraph of the original pathological causal chain graph in terms of topological structure. It retains the most important causal transmission framework from pathogenic factors to clinical manifestations, while providing users with an interactive entry point to explore secondary paths on demand through a collapsible node marking mechanism.

[0130] Based on spatial layout constraints, gradation feature constraints, and connection strength constraints, visual primitives and connecting elements are generated for each node of the simplified pathological causal chain diagram or pathological causal chain diagram to form composite visual elements, including:

[0131] Extract each node from the simplified pathological causal chain diagram or the pathological causal chain diagram, determine the spatial coordinate position of the visual primitive of each node according to the spatial layout constraints, and determine the color value and transparency value of the visual primitive of each node according to the gradient feature constraints.

[0132] For each node, the number of neighboring nodes and the density of neighboring connections within its neighborhood are statistically analyzed and weighted to obtain the local visual complexity.

[0133] Based on the local visual complexity of each node, a color saturation attenuation coefficient and a visual primitive size scaling coefficient are generated. When the local visual complexity exceeds the preset local complexity threshold, the color saturation of the corresponding node's visual primitive is reduced and the visual primitive size is reduced.

[0134] The color saturation attenuation coefficient is applied to the color value to generate adjusted color parameters, and the visual primitive size scaling coefficient is applied to the spatial coordinate position to generate adjusted spatial parameters. Based on the adjusted color parameters, transparency value and adjusted spatial parameters, node visual primitives with local complexity adjustment are generated.

[0135] The line thickness and saturation of the corresponding connecting elements on each side are determined according to the connection strength constraint. The connecting elements are generated by combining the adjusted spatial parameters of the starting node visual primitives and the ending node visual primitives of each side. The node visual primitives and connecting elements are combined to form a composite visual element.

[0136] After extracting the nodes from the simplified pathological causal chain diagram, the spatial coordinates of each node's visual primitives are determined based on spatial layout constraints. These spatial layout constraints originate from the topological transformation of the pathological causal chain diagram and are specifically manifested as the coordinate allocation rules for nodes in the two-dimensional or three-dimensional rendering space: pathogenic factor nodes upstream in the causal chain are typically constrained to the starting region of the layout space, pathological process nodes are centrally distributed, and clinical manifestation nodes are constrained to the ending region, thus visually presenting the directionality of pathological evolution. The determination of coordinate positions also needs to consider the hierarchical relationship between nodes; nodes within the same nested level maintain uniform spacing in the vertical or horizontal directions to avoid overlap and occlusion of visual elements.

[0137] While determining the spatial coordinates, initial color and transparency values ​​are assigned to each node's visual primitives based on gradient feature constraints. Gradient feature constraints are derived from the temporal stage attributes of the pathological causal chain diagram. Nodes with earlier temporal positions (corresponding to the early stages of pathological evolution) are typically assigned warm colors with higher saturation, while nodes with later temporal positions gradually transition to cooler colors or low-saturation colors, visually mapping the time dimension of pathological progress through color gradients. The initial setting of transparency values ​​is related to the strength of causal dependence between nodes. Core nodes with higher dependence strength have transparency values ​​close to 1 (completely opaque), while edge nodes have relatively lower transparency values, creating a visual sense of hierarchy.

[0138] After assigning initial color and coordinate values, the number of neighboring nodes and the neighborhood connection density are calculated for each node, and the local visual complexity of that node is obtained through a weighted combination. The neighborhood is typically defined by the topological distance of the graph, including nodes directly connected to the target node or within a specified number of hops. The number of neighboring nodes reflects the information density around the node, while the neighborhood connection density reflects the tightness of connections within the neighborhood. Let the node... The number of neighboring nodes is The neighborhood connection density is Local visual complexity Calculated by weighted combination: .in The weighting coefficient is the number of neighboring nodes. , where is the weighting coefficient for neighborhood connection density. Both are positive real numbers and can be adjusted according to the specific medical visualization scenario.

[0139] Based on the local visual complexity of each node Generate the corresponding color saturation attenuation coefficient. scaling factor with visual primitive size .when Exceeding the preset local complexity threshold When this occurs, a local complexity adjustment mechanism is triggered: the color saturation of the visual primitive of that node is reduced and the size of the visual primitive is reduced to reduce the excessive occupation of the user's attention in that area and alleviate the feeling of local visual crowding. and The calculation method is as follows:

[0140]

[0141]

[0142] in This represents the maximum local visual complexity of all nodes. The saturation attenuation intensity coefficient. ...

[0143] Color saturation attenuation coefficient This is applied to the initial color values ​​to generate adjusted color parameters. Specifically, the initial color values ​​are converted from the RGB color space to the HSL color space, and the saturation components are multiplied by... Then it is converted back to RGB to obtain the adjusted color parameters. This operation preserves the hue and brightness information, and only compresses the saturation, so that nodes in highly complex areas appear to have a softer, more subtle color effect, without affecting the temporal semantics carried by the color.

[0144] Scaling factor of visual primitive size This is applied to spatial coordinates to generate adjusted spatial parameters. The adjustment method is as follows: Using the node's initial coordinates as the center, multiply the rendering size of the visual primitive (such as the radius of a circular node, the side length of a rectangular node) by... Simultaneously, the placement boundaries of nodes in the layout are adjusted accordingly to ensure that the scaled-down visual primitives remain centered at their original coordinates, preventing layout offset. The adjusted spatial parameters include the node center coordinates and updated size information, which are used when generating subsequent connected elements.

[0145] Based on the adjusted color parameters, transparency values, and spatial parameters, the final rendering of the node visual primitives, after local complexity adjustment, is completed. The shape type of each node visual primitive is determined by its node category: pathogenic factor nodes are typically rendered as rhombuses, pathological process nodes as rounded rectangles, and clinical manifestation nodes as ellipses, using shape differences to help users quickly distinguish the semantic categories of nodes. Text labels can be overlaid inside nodes, and the label font size scales proportionally to the size of the visual primitive, ensuring that the labels remain readable even after the size is reduced.

[0146] The generation of connection elements is based on connection strength constraints. These constraints are derived from the causal dependency strength of each edge in the pathological causal chain graph. Edges with higher causal dependency strength correspond to thicker lines and higher line saturation, while edges with lower causal dependency strength correspond to thinner lines and lower line saturation. Let the edges... The strength of causal dependence is Line thickness With line saturation They are determined by the following relationships respectively:

[0147]

[0148]

[0149] in , The lower and upper limits for line thickness (in pixels). , These represent the lower and upper limits of line saturation. The coordinates of the start and end points of the connecting element are taken from the center coordinates of the adjusted spatial parameters of the starting and ending visual primitives, and the endpoints are offset according to the boundaries of the visual primitives, so that the connecting line is tangent to the node boundaries rather than penetrating into the nodes. For directional causal dependencies, an arrow marker is added to the end of the connecting element, with the arrow size proportional to the line thickness.

[0150] All node visual primitives and connecting elements are combined to form composite visual elements. The combination follows a rendering hierarchy: connecting elements are at the bottom layer, node visual primitives at the top layer, and node text labels at the top layer, ensuring that labels and node shapes are always visible and not obscured by connecting lines. For complex graph structures with intersecting connections, the connecting lines at intersections are treated as bridges, using small, visually prominent arcs to distinguish different connections and prevent users from misinterpreting the connections. The final composite visual element fully preserves the topological semantics of the pathological causal chain graph, while actively reducing visual noise in information-dense areas through a local complexity adaptive adjustment mechanism, guiding user attention to the core path of causal relationships.

[0151] The associated interactive response mechanism for the foldable node receives user interaction commands, renders the visual display of the corresponding folded content, and records interaction behavior data, including:

[0152] Generate interactive control elements for the foldable nodes and bind them to the display positions of the corresponding node visual primitives;

[0153] Monitor user trigger operations on the interactive control element, extract the pathological causal subgraph of the pathological causal chain graph corresponding to the trigger operation, calculate the unfolding complexity based on the number of nodes and connection density of the pathological causal subgraph, and determine the layout scaling ratio and superimposed display position of the unfolded visual element based on the unfolding complexity.

[0154] Each node and edge in the pathological causal subgraph is used to generate an expanded visual element according to the spatial layout constraints, gradient feature constraints and connection strength constraints. The expanded visual element is then superimposed onto the display area of ​​the current composite visual element according to the layout scaling ratio and superimposed display position.

[0155] The trigger timestamp of the recorded action and the corresponding collapsible node identifier form an interaction behavior record.

[0156] In simplified pathological causal chain diagrams, the portions marked as collapsible nodes contain compressed pathological details. To allow users to expand this collapsed content as needed, an interactive control element needs to be generated for each collapsible node and bound to the display position of that node's visual primitive. The interactive control element is typically attached to the edge area of ​​the node's visual primitive as a small icon or button. Its display coordinates are synchronized with the anchor point coordinates of the node's visual primitive, ensuring that the control element always moves with the corresponding node when the overall layout is scaled or panned, without positional drift. The visual design of the control element should be consistent with the style of the node's visual primitive, using a color scheme that matches the gradient feature constraints, and distinguished from the background by outlines or shadows, allowing users to intuitively identify its interactive attributes. For collapsible nodes with deep nesting levels, the size of the control element can be adjusted proportionally according to the scaling factor of the node's visual primitive to avoid obscuring key information of adjacent nodes in dense layouts.

[0157] When a user triggers an interactive control element, the pathological causal subgraph corresponding to that collapsible node needs to be extracted from the pathological causal chain graph. The scope of the pathological causal subgraph is determined by the downstream dependencies of that node in the original pathological causal chain graph, and includes all child nodes reachable from that node and the connecting edges between them. After extraction, the number of nodes in the subgraph is counted. With the number of edges Calculate the subgraph connectivity density This leads to the expansion complexity. : . This reflects the visual information density of the sub-image and directly determines the display space required to expand visual elements. When When the content is smaller, it can be displayed with a larger layout scaling ratio, and nodes and connecting elements all maintain a clearly distinguishable size; when When dealing with larger content, it's necessary to appropriately compress the layout scaling ratio to ensure that the expanded content can be fully contained within the available display area without exceeding the boundaries. Layout scaling ratio with expansion complexity The relationship between them is determined by an inverse proportionality: .in The scaling factor is preset based on the overall display area size of the current composite visual element. The overlay display position is determined by the coordinates of the trigger node in the current composite visual element, the estimated size of the expanded content, and the boundary of the available display area. The expanded content is preferentially placed in the blank area adjacent to the trigger node. If there is insufficient space, it is expanded outward or overlaid using a floating layer method.

[0158] After determining the layout scaling ratio and overlay display position, the nodes and edges in the pathological cause-and-effect subgraph are arranged according to the main... Figure 1 The spatial layout constraints, gradient feature constraints, and connection strength constraints generate the unfolded visual elements. The temporal stage attributes of each node in the subgraph determine its vertical arrangement order within the unfolded area, while causal dependencies determine the horizontal hierarchical relationship between nodes, thus forming a spatial layout consistent with the topological logic of the main graph. The visual primitives of each node select corresponding shapes and basic colors based on their node type (pathogenic factors, pathological processes, or clinical manifestations), and then overlay temporal gradient effects according to gradient feature constraints, ensuring that the unfolded content maintains visual consistency with the main graph. Users can understand the meaning of the unfolded content without relearning visual coding rules. The connection elements corresponding to each edge map line thickness and saturation based on the causal dependency strength, perfectly consistent with the visual rules of connection elements in the main graph, ensuring the visual expression of causal relationships within the unfolded area is coherent.

[0159] After the visual elements are generated, scale them according to the calculated layout ratio. The entire sub-image's visual elements are scaled proportionally, and the scaled expanded content is positioned for overlay display, superimposed on the current composite visual element's display area as a floating layer. During overlay, the visual connection line between the expanded content and the triggering node must be preserved to clearly indicate which collapsible node the expanded content belongs to, avoiding user confusion when multiple collapsible nodes are expanded simultaneously. The boundaries of the expanded content are distinguished using a lightweight border or semi-transparent overlay that contrasts sharply with the main image background, allowing users to clearly perceive the extent of the expanded area. If the user triggers the same interactive control element again, the expanded content collapses, the overlay is removed from the display area, and the composite visual element returns to its collapsed state.

[0160] After each triggered operation is completed, record the trigger timestamp of this interaction. This, along with the node identifier of the corresponding collapsible node, forms an interaction behavior record. Trigger timestamp. Accurate to the millisecond level, a unified time base is used to ensure temporal comparability between multiple records. Node identifiers are consistent with the unique codes of nodes in the pathological causal chain diagram, enabling precise association between interaction behavior records and specific pathological concept nodes. Interaction behavior records are appended and stored in chronological order of trigger time, forming a complete user interaction sequence. This data can be used for subsequent analysis of users' focus and cognitive paths when browsing the pathological causal chain diagram, such as which collapsible nodes are frequently expanded and the temporal distribution of expansion operations throughout the browsing process, providing data support for further optimization of visual element collapsing strategies and layout design. In multi-user scenarios, the interaction behavior records of different users are stored independently, grouped by user identifiers to avoid interference between the behavioral data of different users.

[0161] A second aspect of this invention provides a visual element generation system based on medical concept mapping, comprising:

[0162] The pathology analysis unit is used to analyze the pathological mechanism of the acquired target medical concepts to be visualized, identify pathogenic factor nodes, pathological process nodes and clinical manifestation nodes, establish causal dependencies between nodes based on the medical knowledge base and mark the temporal stage attributes to form a pathological causal chain diagram.

[0163] The constraint transformation unit is used to convert the topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph into spatial layout constraints, gradient feature constraints, and connection strength constraints of visual elements, respectively.

[0164] The load prediction unit is used to calculate the visual complexity index based on the number of nodes, connection density and nesting level of the pathological causal chain graph, and to calculate the cognitive load prediction value in combination with the cognitive capacity threshold of the target user.

[0165] The path simplification unit is used to identify the target display path set based on the cumulative causal strength value of each path when the cognitive load prediction value exceeds the preset cognitive load threshold, mark the nodes outside of it as collapsible nodes, and generate a simplified pathological causal chain diagram.

[0166] The visual generation unit is used to generate visual primitives and connecting elements for each node of the simplified pathological causal chain diagram or pathological causal chain diagram according to spatial layout constraints, gradient feature constraints and connection strength constraints, so as to form composite visual elements.

[0167] The interactive response unit is used to associate the foldable node with an interactive response mechanism, receive user interaction instructions, render the visual display of the corresponding folded content, and record interactive behavior data.

[0168] A third aspect of the present invention provides an electronic device, comprising:

[0169] processor;

[0170] Memory used to store processor-executable instructions;

[0171] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0172] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0173] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

Claims

1. A method for generating visual elements based on medical concept mapping, characterized in that, include: The pathological mechanism of the acquired target medical concept to be visualized is analyzed, the pathogenic factor node, pathological process node and clinical manifestation node are identified, the causal dependency relationship between the nodes is established based on the medical knowledge base and the temporal stage attribute is marked to form a pathological causal chain diagram. The topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph are respectively converted into spatial layout constraints, gradient feature constraints, and connection strength constraints of visual elements. The visual complexity index is calculated based on the number of nodes, connection density, and nesting level of the pathological causal chain diagram, and the cognitive load prediction value is calculated in combination with the cognitive capacity threshold of the target user. When the predicted cognitive load value exceeds the preset cognitive load threshold, the target display path set is identified based on the cumulative causal strength value of each path, and the nodes outside of it are marked as collapsible nodes, generating a simplified pathological causal chain diagram. Based on spatial layout constraints, gradient feature constraints, and connection strength constraints, visual primitives and connecting elements are generated for each node of the simplified pathological causal chain diagram or pathological causal chain diagram to form composite visual elements. The associated interactive response mechanism for the foldable node receives user interaction commands, renders the visual display of the corresponding folded content, and records the interaction behavior data.

2. The method according to claim 1, characterized in that, The acquired target medical concepts to be visualized are analyzed for pathological mechanisms, identifying pathogenic factor nodes, pathological process nodes, and clinical manifestation nodes. Based on a medical knowledge base, causal dependencies between nodes are established and temporal stage attributes are labeled, forming a pathological causal chain diagram, including: Semantic parsing of the target medical concept is performed to extract core pathological entities; Based on the etiological classification relationship of core pathological entities in the medical knowledge base, pathogenic factor nodes are identified; based on the pathological evolution relationship, pathological process nodes are identified; and based on the symptom mapping relationship, clinical manifestation nodes are identified. The causal association rules between nodes are extracted from the medical knowledge base. The logical dependency type between the antecedent conditions and the subsequent results in the causal association rules is analyzed. The dependency strength weight is marked according to the directed causal dependency relationship between the pathogenic factor node, the pathological process node and the clinical manifestation node. Time series data of pathological evolution are extracted from the medical knowledge base. Based on the time distribution characteristics of the occurrence of each pathological state recorded in the time series data, the temporal position quantification values ​​of each pathogenic factor node, pathological process node and clinical manifestation node in the whole cycle of pathological evolution are calculated as temporal stage attributes. The pathological causal chain graph is constructed by using the pathogenic factor node, pathological process node, and clinical manifestation node as graph nodes, the directed causal dependency relationship labeled with the dependency strength weight as graph edges, and the quantized value of the time sequence position as node attribute.

3. The method according to claim 1, characterized in that, The topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph are respectively converted into spatial layout constraints, gradation feature constraints, and connection strength constraints of visual elements, including: The topological level position of each node is calculated based on the connection relationship between nodes in the pathological causal chain diagram. Vertical hierarchical layout rules are generated according to the topological level position. Horizontal expansion layout rules are generated according to the causal branch relationship of nodes in the same topological level. The vertical hierarchical layout rules and the horizontal expansion layout rules are combined to form spatial layout constraints. Extract the temporal stage attributes of each node in the pathological causal chain graph, and map the temporal stage attributes to visual gradient parameters. The visual gradient parameters include color gradient parameters determined according to the relative position of the temporal stage attributes on the time axis, and transparency gradient parameters determined according to the time span of the temporal stage attributes. The visual gradient parameters are used as gradient feature constraints. Extract the causal dependency strength corresponding to each edge in the pathological causal chain graph, and map the causal dependency strength to the visual parameters of the connecting line. The visual parameters of the connecting line include the line thickness parameter and the line saturation parameter, both of which are positively correlated with the causal dependency strength. The visual parameters of the connecting line are used as the connection strength constraint.

4. The method according to claim 1, characterized in that, Based on the number of nodes, connection density, and nesting level of the pathological causal chain graph, a visual complexity index is calculated. Combined with the cognitive capacity threshold of the target user, a cognitive load prediction value is calculated, including: The total number of nodes in the pathological causal chain graph is used as the number of nodes, the ratio of the total number of edges to the total number of nodes is used as the connection density, and the maximum value of the topological level is used as the nesting level. Identify the node type identifiers for each node, including pathogenic factor type identifiers, pathological process type identifiers, and clinical manifestation type identifiers. Set cognitive load contribution weights based on the cognitive processing depth of each node type identifier in medical knowledge understanding. Calculate the weighted sum of the number of nodes of each type and their corresponding weights as the node heterogeneity complexity. Detect locally high-density connected subgraphs whose connection density exceeds the average connection density of the entire graph, and obtain the connection aggregation complexity by summing the products of the number of nodes and the number of edges in each subgraph. The visual complexity index is obtained by exponentially calculating the node heterogeneity complexity, connection aggregation complexity, and nesting level, respectively, and then summing them. The exponent of the exponential calculation is determined based on the marginal contribution rate of each complexity component to the cognitive load. The average information processing rate and visual dwell time distribution in the target user's historical interactions are obtained, and then converted into the number of nodes that can be processed per unit time and the attention concentration ability coefficient, respectively, and multiplied to obtain the cognitive capacity threshold. The ratio of the visual complexity index to the cognitive capacity threshold is used as a predictor of cognitive load.

5. The method according to claim 1, characterized in that, Based on the cumulative causal strength values ​​of each path, the target path set is identified, and nodes outside of it are marked as collapsible nodes, generating a simplified pathological causal chain graph, including: When the predicted cognitive load value exceeds the preset cognitive load threshold, a simplified processing procedure is executed. Extract the complete causal transmission path from the pathogenic factor node to the clinical manifestation node in the pathological causal chain graph, and perform cumulative causal dependence strength calculation on each side along each complete causal transmission path to obtain the cumulative causal strength value. All complete causal transmission paths are sorted in descending order of cumulative causal intensity value. The preceding paths whose sum of cumulative causal intensity values ​​reaches the proportion threshold of the sum of cumulative causal intensity values ​​of all paths are selected to form the target display path set. The proportion threshold is dynamically adjusted according to the difference between the cognitive load prediction value and the preset cognitive load threshold and is negatively correlated. Nodes in the pathological causal chain graph that are not included in the target display path set are marked as collapsible nodes, and nodes and edges in the target display path set are retained to generate a simplified pathological causal chain graph.

6. The method according to claim 1, characterized in that, Based on spatial layout constraints, gradation feature constraints, and connection strength constraints, visual primitives and connecting elements are generated for each node of the simplified pathological causal chain diagram or pathological causal chain diagram to form composite visual elements, including: Extract each node from the simplified pathological causal chain diagram or the pathological causal chain diagram, determine the spatial coordinate position of the visual primitive of each node according to the spatial layout constraints, and determine the color value and transparency value of the visual primitive of each node according to the gradient feature constraints. For each node, the number of neighboring nodes and the density of neighboring connections within its neighborhood are statistically analyzed and weighted to obtain the local visual complexity. Based on the local visual complexity of each node, a color saturation attenuation coefficient and a visual primitive size scaling coefficient are generated. When the local visual complexity exceeds the preset local complexity threshold, the color saturation of the corresponding node's visual primitive is reduced and the visual primitive size is reduced. The color saturation attenuation coefficient is applied to the color value to generate adjusted color parameters, and the visual primitive size scaling coefficient is applied to the spatial coordinate position to generate adjusted spatial parameters. Based on the adjusted color parameters, transparency value and adjusted spatial parameters, node visual primitives with local complexity adjustment are generated. The line thickness and saturation of the corresponding connecting elements on each side are determined according to the connection strength constraint. The connecting elements are generated by combining the adjusted spatial parameters of the starting node visual primitives and the ending node visual primitives of each side. The node visual primitives and connecting elements are combined to form a composite visual element.

7. The method according to claim 1, characterized in that, The associated interactive response mechanism for the foldable node receives user interaction commands, renders the visual display of the corresponding folded content, and records interaction behavior data, including: Generate interactive control elements for the foldable nodes and bind them to the display positions of the corresponding node visual primitives; Monitor user trigger operations on the interactive control element, extract the pathological causal subgraph of the pathological causal chain graph corresponding to the trigger operation, calculate the unfolding complexity based on the number of nodes and connection density of the pathological causal subgraph, and determine the layout scaling ratio and superimposed display position of the unfolded visual element based on the unfolding complexity. Each node and edge in the pathological causal subgraph is used to generate an expanded visual element according to the spatial layout constraints, gradient feature constraints and connection strength constraints. The expanded visual element is then superimposed onto the display area of ​​the current composite visual element according to the layout scaling ratio and superimposed display position. The trigger timestamp of the recorded action and the corresponding collapsible node identifier form an interaction behavior record.

8. A visual element generation system based on medical concept mapping, used to implement the method as described in any one of claims 1-7, characterized in that, include: The pathology analysis unit is used to analyze the pathological mechanism of the acquired target medical concepts to be visualized, identify pathogenic factor nodes, pathological process nodes and clinical manifestation nodes, establish causal dependencies between nodes based on the medical knowledge base and mark the temporal stage attributes to form a pathological causal chain diagram. The constraint transformation unit is used to convert the topological structure, temporal stage attributes, and causal dependency strength of the pathological causal chain graph into spatial layout constraints, gradient feature constraints, and connection strength constraints of visual elements, respectively. The load prediction unit is used to calculate the visual complexity index based on the number of nodes, connection density and nesting level of the pathological causal chain graph, and to calculate the cognitive load prediction value in combination with the cognitive capacity threshold of the target user. The path simplification unit is used to identify the target display path set based on the cumulative causal strength value of each path when the cognitive load prediction value exceeds the preset cognitive load threshold, mark the nodes outside of it as collapsible nodes, and generate a simplified pathological causal chain diagram. The visual generation unit is used to generate visual primitives and connecting elements for each node of the simplified pathological causal chain diagram or pathological causal chain diagram according to spatial layout constraints, gradient feature constraints and connection strength constraints, so as to form composite visual elements. The interactive response unit is used to associate the foldable node with an interactive response mechanism, receive user interaction instructions, render the visual display of the corresponding folded content, and record interactive behavior data.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images