Digestive department data-based analysis method and system

Abstract

The invention provides an analysis method and system based on gastroenterology department data, and the method comprises the steps: carrying out the triple semantic analysis of the medical concepts of the gastroenterology department, obtaining an anatomical space vector, a pathological space vector and a clinical space vector, forming a semantic vector of the gastroenterology department, obtaining a concept evolution relation between the medical concepts of the gastroenterology department, and carrying out the analysis of the medical concepts of the gastroenterology department; quantifying the cognitive distance weight of the concept evolution relationship, constructing a concept evolution relationship network through the evolution relationship and the cognitive distance weight, constructing a learning model, obtaining concepts mastered by intern doctors, learning behaviors and target learning concepts as input of the learning model, and outputting a learning path; the method comprises the following steps: constructing a semantic map database according to a digestive department semantic vector, an evolution relation network and learning model data, and generating a personalized learning scheme through the semantic map database and a learning path. Therefore, the accuracy and the suitability of the learning scheme in the training of the intern doctor in the digestive department are improved.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and more specifically, to a method and system for analyzing data from the gastroenterology department. Background Technology

[0002] In the field of medical education, especially in the training of interns in internal medicine subspecialties such as gastroenterology, the development of personalized and efficient learning programs has become a key direction for improving the quality of talent training, shortening the cycle of skills transfer, and aligning with the needs of clinical positions. However, existing learning programs for gastroenterology interns often have some problems. Traditional experience-driven teaching relies on standardized textbooks and fixed course arrangements, which are not suitable for interns with different knowledge bases and learning needs. This can easily lead to problems such as a mismatch between learning content and individual abilities, and a lack of logical connection in knowledge transfer, resulting in poor practicality of the learning program. Secondly, there is a problem of poor quantitative coordination. Either it only focuses on the coverage of theoretical knowledge or the uniformity of learning progress while ignoring individual needs, or it lacks the structured integration of the knowledge system and dynamic adaptation to the learner's status, resulting in a waste of learning resources and insufficient training efficiency. Therefore, there is an urgent need for a technical solution with multi-dimensional semantic parsing and personalized path optimization capabilities to improve the accuracy and adaptability of learning programs in the training of gastroenterology interns. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an analysis method based on gastroenterology data, the method comprising: A triple semantic analysis of gastroenterology medical concepts is performed to obtain anatomical space vectors, pathological space vectors, and clinical space vectors, which are then combined to form a gastroenterology semantic vector. The concept evolution relationships between gastroenterological medical concepts are obtained, including premise relationships, refinement relationships, comparison relationships, and application relationships. The cognitive distance weights of the concept evolution relationships are quantified, and a concept evolution relationship network is constructed based on the evolution relationships and cognitive distance weights. Construct a learning model, obtain the concepts already mastered by interns, their learning behaviors, and the target learning concepts as inputs to the learning model, and output the learning path; A semantic graph database was constructed based on semantic vectors from the Department of Gastroenterology, evolutionary relationship networks, and learning model data; Personalized learning plans are generated based on semantic graph databases and learning paths.

[0004] As a further aspect of the present invention, a triple semantic analysis is performed on the gastroenterology medical concept to obtain anatomical space vectors, pathological space vectors, and clinical space vectors, which are then combined to form a gastroenterology semantic vector, including: The digestive system is hierarchically divided from upstream to downstream, and an anatomical space vector is constructed based on the hierarchical anatomical location of the digestive system. The correlation logic of common pathologies in gastroenterology is classified, the pathological mechanism classification results are obtained, and a pathological space vector is constructed based on the pathological mechanism classification results. Obtain typical clinical manifestation patterns of gastrointestinal diseases, and construct clinical space vectors based on the clinical manifestation patterns; Semantic vectors for gastroenterology are obtained based on anatomical spatial vectors, pathological spatial vectors, and clinical spatial vectors.

[0005] As a further aspect of the present invention, obtaining the conceptual evolution relationship between gastroenterological medical concepts includes: A prerequisite relationship is defined as a relationship in which one concept is a fundamental premise for understanding another concept. A refinement relationship is defined as one concept being a specific manifestation of another concept. The relationship between the two concepts that need to be distinguished and compared is defined as a contrastive relationship; The relationship in which one concept is applied to another concept is defined as an application relationship. Based on the aforementioned premise relationships, refinement relationships, comparison relationships, and application relationships, we can obtain the conceptual evolution relationships between gastroenterological medical concepts.

[0006] As a further aspect of the present invention, the cognitive distance weights of concept evolution relationships are quantified, and a concept evolution relationship network is constructed based on the evolution relationships and cognitive distance weights, including: Obtain the co-occurrence frequency of concepts in teaching materials and literature; Obtain the related expressions between concepts in clinical guideline texts and quantify the strength of the association based on semantic similarity algorithms; Based on co-occurrence frequency and association strength, and combined with the AHP (Analytic Hierarchy Process) method, the cognitive distance weights of premise relations, refinement relations, contrast relations and application relations are obtained respectively; Using each concept as a node, and with premise relationships, refinement relationships, comparison relationships, and application relationships as edges, and labeling the corresponding cognitive distance weights, a concept evolution relationship network is constructed.

[0007] As a further aspect of the present invention, a learning model is constructed, which takes the concepts already mastered by the intern, their learning behaviors, and the target learning concepts as input to the learning model, and outputs a learning path, including: The system acquires the concepts and learning behaviors that interns have mastered, including their learning speed, error types, and learning preferences, and simultaneously acquires the interns' target learning concepts. The learning model is constructed using the minimization of total cognitive load, the maximization of learning benefits, and the constraint of learning coherence as optimization functions; The learning model inputs the concepts, learning behaviors, and target learning concepts already mastered by interns into the learning model, and outputs the learning path by combining the concept evolution relationship network.

[0008] As a further aspect of the present invention, a learning model is constructed using the minimization of total cognitive load, the maximization of learning gains, and learning coherence constraints as optimization functions, including: Conceptual complexity is obtained based on the level of abstraction of each semantic space in the semantic vector of the gastroenterology department; learning distance is obtained based on the shortest path distance in the conceptual evolution relationship network; and cognitive load is obtained based on conceptual complexity and learning distance. Concept importance is determined by the frequency of occurrence of each concept in clinical practice; concept relevance is determined by the target learning concept; and learning benefits are determined by concept importance and concept relevance. A similarity threshold is set to obtain the semantic similarity of adjacent concept nodes in the concept evolution relationship network. Constraints are imposed on adjacent concept nodes that are less than the similarity threshold, and the learning order is also constrained. The learning order is expressed as learning the premise concept first and then learning the target concept.

[0009] As a further aspect of the present invention, the concepts, learning behaviors, and target learning concepts already mastered by the intern are input into the learning model, and a learning path is output by combining the concept evolution relationship network, including: During the learning process, real-time testing nodes are added to test interns and obtain quantitative values ​​of the test results. Set a first threshold and a second threshold. If the quantitative value of the test result is less than the first threshold, it means that the current concept has not been mastered. Then, return to the premise relation concept of the current concept to relearn. If the quantitative value of the test result is greater than or equal to the first threshold and less than the second threshold, it indicates that the current concept has been partially mastered. In this case, the learning of the detailed relationship, comparative relationship and application relationship of the current concept should be strengthened. If the quantified value of the test result is greater than or equal to the second threshold, it indicates that the current concept has been mastered, and then the learning of the next concept is advanced based on the optimization function.

[0010] As a further aspect of the present invention, a semantic graph database is constructed based on gastroenterology semantic vectors, evolutionary relationship networks, and learning model data, including: The teaching attributes are based on the concept complexity, concept importance, and prerequisite concepts in the learning model data; Evolutionary relation edges are constructed based on target concept identifiers, relation types, and corresponding cognitive distance weights; The concept node structure is composed of concept identifiers, gastroenterology semantic vectors, teaching attributes, and evolutionary relationship edges; Knowledge status is acquired based on the target learning concepts and corresponding mastery levels in the learning model data; The learning model structure is composed of knowledge state and learning mode; A semantic graph database is constructed based on the concept node structure and the learning model structure.

[0011] As a further aspect of the present invention, a personalized learning scheme is generated based on a semantic graph database and a learning path, including: Based on the learning path, the learning materials in the learning path are adjusted according to the teaching attributes and visual resources of the gastroenterology semantic vector association in the semantic graph database, and combined with the learning patterns in the semantic graph database, to generate an initial personalized learning plan. Based on the real-time test nodes of the learning path, the system obtains the current learning concept and the subsequent related concepts of the learning path, and simultaneously identifies the intern's knowledge gaps in the learning path. It then recommends comparative concepts with the current learning concept for enhanced comparative learning, forming a personalized learning plan.

[0012] Furthermore, embodiments of the present invention also provide an analysis system based on gastroenterology data, including: The parsing module performs triple semantic parsing on gastroenterology medical concepts to obtain anatomical space vectors, pathological space vectors, and clinical space vectors, and then forms a gastroenterology semantic vector. The acquisition module is used to acquire the conceptual evolution relationship between gastroenterological medical concepts, and to acquire the concepts mastered by interns, their learning behaviors, and the concepts they are aiming to learn. The calculation module is used to quantify the cognitive distance weights of concept evolution relationships; The construction module is used to build a learning model, construct a semantic graph database based on gastroenterology semantic vectors, evolutionary relationship networks and learning model data, and construct a concept evolutionary relationship network based on evolutionary relationships and cognitive distance weights; The processing module inputs the concepts, learning behaviors, and target learning concepts already mastered by the interns into the learning model, outputs a learning path, and generates a personalized learning plan based on the semantic graph database and the learning path.

[0013] Compared with existing technologies, this invention has the following beneficial effects: It clarifies four types of logical relationships in gastroenterology knowledge—premise, refinement, comparison, and application—through a conceptual evolution relationship network, thereby ensuring the logical coherence of the learning path. Simultaneously, it achieves a deep connection between theoretical learning and clinical practice by using clinically important concept data marked in a semantic graph database and combining it with associated clinical visualization resources, helping interns establish a connection between theory and practice. Furthermore, by combining the semantic graph database with interns' knowledge status data and multi-objective optimization planning of the learning path, it can identify different interns' knowledge blind spots, learning speeds, and content preferences, generating personalized learning plans and resolving the contradiction between standardized teaching and personalized needs in traditional methods. Through this technical solution with multi-dimensional semantic analysis and personalized path optimization capabilities, it improves the accuracy and adaptability of learning plans in the training of gastroenterology interns. Attached Figure Description

[0014] Figure 1 This is a flowchart of the steps of the analysis method based on gastroenterology data of the present invention; Figure 2 This is a flowchart of step S4 in the analysis method based on gastroenterology data of the present invention, which involves constructing a semantic graph database. Figure 3 This is a schematic diagram of the analysis system based on gastroenterology data of the present invention. Detailed Implementation

[0015] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating the steps of an analysis method based on gastroenterology data provided in one embodiment of the present invention. Figure 2 This is a flowchart of step S4 in the analysis method based on gastroenterology data of the present invention, which involves constructing a semantic graph database. The analysis method based on gastroenterology data will be described in detail below.

[0016] Step S1: Perform triple semantic parsing on the gastroenterology medical concept to obtain the anatomical space vector, pathological space vector, and clinical space vector, and form the gastroenterology semantic vector.

[0017] It should be noted that the concept of gastroenterology refers to the core elements of medical knowledge related to the digestive system, covering anatomical structures such as the stomach and duodenum, pathological mechanisms such as mucosal damage-repair imbalance, disease entities such as peptic ulcers, clinical manifestations such as periodic upper abdominal pain, and diagnostic and treatment methods such as proton pump inhibitor therapy, etc. It is a basic module for interns to build a gastroenterology knowledge system.

[0018] In this embodiment, step S1 includes: Step S1-1: Divide the digestive system into anatomical locations from upstream to downstream, and construct an anatomical space vector based on the anatomical location hierarchy of the digestive system.

[0019] Specifically, anatomical locations are hierarchically classified according to the upstream and downstream order of the digestive system's physiological structure to obtain a complete anatomical link from the oral cavity to the small intestine; the anatomical location level corresponding to each gastroenterological medical concept is labeled, and then the anatomical location information is transformed into a numerical anatomical space vector through vector encoding, where the dimension of the vector corresponds to the anatomical level, and the value represents the strength of the association between the concept and the anatomical location.

[0020] In some possible embodiments, the anatomical hierarchy of the digestive system is divided into dimensions such as oral cavity, esophagus, stomach, duodenum, and small intestine. Taking the concept of peptic ulcer as an example, its main anatomical location is first determined to be the stomach and duodenum. Peptic ulcer is strongly associated with the stomach and duodenum dimension, and the association strength is set to 0.9. It is weakly associated with other anatomical dimensions, and the association strength is set to 0.1. The anatomical space vector is obtained as [0.1,0.1,0.9,0.9,0.1,...].

[0021] Steps S1-2 involve classifying the correlation logic of common pathologies in gastroenterology, obtaining the pathological mechanism classification results, and constructing a pathological space vector based on the pathological mechanism classification results.

[0022] Specifically, we organize the common pathological mechanisms of digestive diseases and classify them into core pathological types such as inflammatory, functional, neoplastic, and metabolic diseases according to their correlation logic, thus establishing a pathological mechanism classification system. We analyze the pathological mechanism category to which each medical concept belongs and transform the pathological type information into a pathological space vector through vector encoding. In this vector, the dimension corresponds to the pathological classification, and the value represents the degree of matching between the concept and the pathological type.

[0023] In some possible embodiments, the pathological mechanism is classified into dimensions such as inflammatory, functional, and neoplastic. Taking the concept of peptic ulcer as an example, its pathological mechanism is the imbalance between mucosal damage and repair, which belongs to the inflammatory pathological type. The value of peptic ulcer in the inflammatory dimension is set to 0.85, and the values ​​of other pathological dimensions are set to 0.05-0.15, and the pathological space vector is obtained as [0.85,0.1,0.05,...].

[0024] Steps S1-3: Obtain typical clinical manifestation patterns of gastroenterological diseases, and construct clinical space vectors based on the clinical manifestation patterns.

[0025] Specifically, the typical clinical manifestations of various gastroenterological diseases are obtained and classified according to symptom type, such as pain, bleeding, and functional impairment, to establish a clinical manifestation model system; the typical clinical manifestations corresponding to each medical concept are extracted, and the symptom information is transformed into a clinical space vector through vector encoding, with the vector dimension corresponding to the symptom category and the value representing the frequency or feature intensity of the symptom.

[0026] In some possible embodiments, clinical symptoms are divided into dimensions such as pain, bleeding, and functional impairment. Taking the concept of peptic ulcer as an example, its typical clinical manifestations include periodic upper abdominal pain, acid reflux, and belching. Among them, the value of periodic upper abdominal pain in the pain dimension is set to 0.9, the value of acid reflux in the functional impairment dimension is set to 0.7, and the value of other symptom dimensions is set to 0.1, so that the clinical space vector is [0.9,0.1,0.7,...].

[0027] Steps S1-4: Obtain the gastroenterology semantic vector based on the anatomical spatial vector, pathological spatial vector, and clinical spatial vector.

[0028] Specifically, the anatomical space vector, pathological space vector, and clinical space vector obtained in steps S1-1 to S1-3 are spliced ​​and fused to obtain the gastroenterology semantic vector. During the fusion process, the dimensional independence of each sub-vector is maintained to ensure that the gastroenterology semantic vector can simultaneously possess the anatomical location, pathological essence, and clinical characteristics of the concept, providing a semantic basis for subsequent conceptual relationship analysis.

[0029] In some possible embodiments, continuing with the example of peptic ulcer, its anatomical space vector [0.1,0.1,0.9,0.9,0.1,...], pathological space vector [0.85,0.1,0.05,...], and clinical space vector [0.9,0.1,0.7,...] are concatenated in sequence to obtain the gastroenterology semantic vector [0.1,0.2,0.9,0.9,0.1,...,0.85,0.1,0.05,...,0.9,0.1,0.7,...].

[0030] Step S2: Obtain the conceptual evolutionary relationships between gastroenterological medical concepts. The evolutionary relationships include premise relationships, refinement relationships, comparison relationships, and application relationships. Quantify the cognitive distance weights of the conceptual evolutionary relationships and construct a conceptual evolutionary relationship network based on the evolutionary relationships and cognitive distance weights.

[0031] In this embodiment, step S2 includes: Step S2-1: Define a premise relationship as a fundamental premise for understanding one concept.

[0032] For example, to understand concept B, peptic ulcer, one must first grasp concept A, the gastric acid secretion mechanism, because abnormal gastric acid secretion is the core mechanism of peptic ulcer pathogenesis. Without understanding the gastric acid secretion mechanism, it is impossible to deeply understand the formation principle of ulcer. Therefore, the gastric acid secretion mechanism and peptic ulcer constitute a prerequisite relationship.

[0033] Step S2-2 defines a relationship as a specific manifestation of one concept.

[0034] For example, concept B, peptic ulcer, includes two main types: gastric ulcer and duodenal ulcer. Concept C, duodenal ulcer, is a specific manifestation of peptic ulcer in the duodenum. The two differ in terms of location of onset and pain patterns. Therefore, duodenal ulcer and peptic ulcer constitute a detailed relationship.

[0035] Step S2-3: Define the relationship between the two concepts that need to be distinguished and compared as a contrastive relationship.

[0036] For example, both concept B, peptic ulcer, and concept D, gastric cancer, may present with symptoms such as upper abdominal pain and black stools. In clinical diagnosis, they need to be differentiated through methods such as pathological biopsy and imaging examinations. During the learning process, it is necessary to compare the pain patterns, accompanying symptoms, and prognostic differences between the two. Therefore, peptic ulcer and gastric cancer constitute a comparative relationship.

[0037] Step S2-4 defines the relationship of applying one concept to another as an application relationship.

[0038] For example, concept F, proton pump inhibitors, reduce gastric acid secretion by inhibiting H+-K+-ATPase in gastric parietal cells. This can be directly applied to the treatment of concept B, peptic ulcers, to guide clinical drug selection. Therefore, the two constitute an application relationship.

[0039] Steps S2-5: Based on the aforementioned premise relationship, refinement relationship, comparison relationship, and application relationship, obtain the conceptual evolution relationship between gastroenterology medical concepts.

[0040] For example, when parsing digestive text, concept pairs such as gastric acid secretion mechanism-peptic ulcer, peptic ulcer-duodenal ulcer, peptic ulcer-gastric cancer, and proton pump inhibitor-peptic ulcer are extracted and labeled as premise relationship, refinement relationship, comparison relationship, and application relationship, respectively, forming a set of concept evolution relationships related to peptic ulcer.

[0041] Steps S2-6: Obtain the co-occurrence frequency of concepts in teaching materials and literature.

[0042] Specifically, we collected teaching materials such as classic textbooks and core journal articles in gastroenterology, and used text mining technology to count the number of times each pair of concepts co-occurred in the materials. We then weighted the co-occurrence based on text length and co-occurrence position. The higher the co-occurrence frequency and the closer the position, the closer the relationship between the concepts.

[0043] In some possible embodiments, it is assumed that the co-occurrence of gastric acid secretion mechanism and peptic ulcer in classic gastroenterology textbooks and core journal articles is statistically analyzed. It is found that the two co-occurred 35 times in multiple articles, of which 20 times were co-occurrences in the same paragraph. The number of co-occurrences per thousand words is calculated to be 2.8, and the weighted score of co-occurrence frequency is assumed to be 0.8.

[0044] Steps S2-7: Obtain the related expressions between concepts in the clinical guideline text and quantify the strength of the association based on the semantic similarity algorithm.

[0045] Specifically, we extract the related expressions between concepts in clinical guideline texts, such as leading to, recommended for, and key identification points. We then use algorithms such as cosine similarity and BERT semantic similarity to obtain the semantic similarity of concept pairs, quantifying the strength of the association between concepts. The higher the obtained similarity value, the stronger the association.

[0046] It should be noted that the BERT semantic similarity algorithm is a pre-trained language model based on the Transformer architecture, which can transform text into low-dimensional vectors through bidirectional contextual semantic understanding. First, the conceptual text in the clinical guidelines is segmented, such as gastric acid secretion mechanism and peptic ulcer. Then, the model's multi-layer attention mechanism captures the semantic association features of the concepts in the context, and finally outputs a fixed-dimensional vector. This vector can reflect the semantic connotation of the concept and provide a high-quality semantic foundation for subsequent similarity calculation.

[0047] In some possible embodiments, it is assumed that the clinical guidelines extract statements that the pathogenesis of peptic ulcers is closely related to excessive gastric acid secretion. In this embodiment, the BERT semantic similarity algorithm is used to obtain the BERT semantic similarity between gastric acid secretion mechanism and peptic ulcer, which is set to 0.82, which is the association strength score between the two.

[0048] Steps S2-8: Based on co-occurrence frequency and association strength, and combined with the AHP (Analytic Hierarchy Process) method, obtain the cognitive distance weights of premise relations, refinement relations, contrast relations and application relations respectively.

[0049] Specifically, the weights of co-occurrence frequency of teaching materials and correlation strength of clinical guidelines can be determined by using the Analytic Hierarchy Process (AHP). The scores of the two dimensions are weighted and summed to obtain the cognitive distance weight of each concept evolution relationship. The cognitive distance weight ranges from 0 to 1, and the smaller the value, the stronger the correlation.

[0050] It should be noted that the AHP (Analytic Hierarchy Process) is a multi-criteria decision-making method that combines qualitative and quantitative analysis. By constructing a hierarchical structure, such as the target layer determining the cognitive distance weight; the criterion layer including co-occurrence frequency and association strength; and the scheme layer outputting the evolutionary relationship of each concept, a judgment matrix can be established by pairwise comparison, and the eigenvectors of the matrix can be calculated to determine the weight of each criterion layer, ensuring the scientific and objective nature of the weight allocation and avoiding the bias of subjective experience judgment.

[0051] Understandably, the co-occurrence frequency of teaching materials directly reflects the inherent logical connections between concepts and is therefore more important in learning path planning, hence it is given a higher weight, such as 0.6. The correlation strength of clinical guidelines reflects the application of concepts in clinical practice, and its weight is lower than that of co-occurrence frequency, such as 0.4.

[0052] In some possible embodiments, for the premise relationship of gastric acid secretion mechanism-peptic ulcer, the co-occurrence frequency score is 0.8 and the association strength score is 0.82. The median value is obtained by weighted summation using the AHP (Analytic Hierarchy Process) method: 0.8×0.6+0.82×0.4=0.808. Finally, the cognitive distance weight of this premise relationship is 1-0.808=0.192. The cognitive distance weights of the refinement relationship, comparison relationship, and application relationship are obtained using the same method in this step.

[0053] It should be noted that the co-occurrence frequency score and the association strength score are both within the range of 0-1, so they can be directly weighted without preprocessing.

[0054] Furthermore, the co-occurrence frequency score and the association strength score are both positive vectorized values, that is, the larger the value, the stronger the association between concepts. The cognitive distance weight is the opposite; the closer the cognitive distance, that is, the closer the quantized value is to 0, the stronger the association. Therefore, the cognitive distance weight needs to be represented by 1 minus the median value of the weighted sum.

[0055] Steps S2-9: Using each concept as a node, with premise relations, refinement relations, comparison relations and application relations as edges, and labeling the corresponding cognitive distance weights, construct a concept evolution relationship network.

[0056] Specifically, the concept of gastroenterology is used as a network node, and the four evolutionary relationships obtained in steps S2-5 are used as edges between nodes. Cognitive distance weights are labeled on each edge, and a concept evolution relationship network is constructed using a graph structure, such as an undirected graph or a directed graph, to achieve visualization and topological analysis of concept relationships.

[0057] In some possible embodiments, this step can use a directed graph structure method to construct a concept evolution relationship network, wherein the gastric acid secretion mechanism, peptic ulcer, duodenal ulcer, gastric cancer, and proton pump inhibitor are used as nodes, and the corresponding nodes are connected by premise relations, refinement relations, contrast relations, and application relations, respectively. Assuming that their cognitive distance weights are 0.192, 0.21, 0.18, and 0.15, respectively, a local concept evolution relationship network with peptic ulcer as the core is constructed, and then the concept evolution relationship network is composed of each local concept evolution relationship network.

[0058] It should be noted that the cognitive distance weight is a quantitative value of the association strength of a single edge, rather than the proportional allocation of edges under the same node. Therefore, the sum of the obtained cognitive distance weights does not need to be 1.

[0059] Step S3: Construct a learning model, obtain the concepts already mastered by the interns, their learning behaviors, and the target learning concepts as inputs to the learning model, and output the learning path.

[0060] In this embodiment, step S3 includes: Step S3-1: Obtain the concepts and learning behaviors that the interns have mastered. The learning behavior patterns include the interns' learning speed, error types, and learning preferences. Simultaneously, obtain the interns' target learning concepts.

[0061] Specifically, the hospital's learning management system can be used to collect the historical learning discipline of interns, obtain the set of concepts that interns have mastered, and simultaneously analyze learning behavior data to obtain learning speed, error types, and learning preferences. The learning management system can then be used to input the target learning concepts for interns.

[0062] In some possible embodiments, suppose that intern A has mastered concepts such as the mechanism of gastric acid secretion and the protective function of the gastric mucosa through previous tests, with a learning speed of 2 basic concepts per hour, and the error types are mostly in differentiating the symptoms of peptic ulcers from gastric cancer. The learning preference is clinical case videos with graphic explanations, and the target learning concept is the diagnosis and treatment process of peptic ulcers.

[0063] Step S3-2: Construct a learning model using the minimization of total cognitive load, the maximization of learning benefits, and the constraint of learning coherence as optimization functions.

[0064] It should be noted that the learning model framework in this embodiment adopts a three-layer architecture: input layer, processing layer, and output layer. The input layer receives three types of data: the concepts already mastered by the interns, their learning behaviors (including learning speed, error types, and learning preferences), and the target learning concept. This data is then standardized and transformed into feature vectors that the model can compute. The processing layer embeds a concept evolution relationship network topology and uses minimizing total cognitive load, maximizing learning benefits, and learning coherence constraints as multi-objective optimization functions. A heuristic search algorithm, such as the A* algorithm, is used to select the optimal concept sequence in the network. The output layer outputs the initial learning path and reserves a real-time testing interface for dynamic path adjustment. By quantifying the total cognitive load and learning benefits, combined with learning coherence constraints, multi-objective optimization of the learning path is achieved.

[0065] In this embodiment, step S3-2 includes: Step S3-2-1: Obtain the concept complexity based on the level of abstraction of each semantic space in the gastroenterology semantic vector, obtain the learning distance based on the shortest path distance in the concept evolution relationship network, and obtain the cognitive load based on the concept complexity and learning distance.

[0066] Specifically, concept complexity is obtained by weighting and summing the scores based on the degree of abstraction of the anatomical, pathological, and clinical spaces in the gastroenterology semantic vector and their assigned weights; learning distance is the shortest path length between the current concept and the target concept in the concept evolution relationship network; its cognitive load = concept complexity × learning distance, and the larger the obtained value, the higher the learning difficulty.

[0067] It should be noted that the anatomical spatial score reflects the clarity of the association between the concept and the specific anatomical location; the higher the value, the more precise the association. The pathological spatial score reflects the clarity of the pathological mechanism of the concept; the higher the value, the clearer the mechanism. The clinical spatial score reflects the typicality of the clinical manifestation of the concept; the higher the value, the more recognizable the symptoms. After analysis, the anatomical spatial score, pathological spatial score, and clinical spatial score were found to be 0.6, 0.8, and 0.9, respectively.

[0068] Furthermore, pathological mechanisms are the core link between anatomical foundations and clinical applications, and have the greatest impact on learning difficulty, so they are given the highest weight, such as 0.4; anatomical and clinical spaces are the foundation and application carrier of medical knowledge, respectively, and are equally important, so they can each be given a weight of 0.3.

[0069] In some possible embodiments, assuming that in the conceptual complexity calculation of the peptic ulcer diagnosis and treatment process, the anatomical space score is 0.6, the pathological space score is 0.8, and the clinical space score is 0.9, and then the conceptual complexity is obtained by weighted summation according to their corresponding weights as 0.6×0.3+0.8×0.4+0.9×0.3=0.79; the shortest path distance between it and the already mastered concept of gastric acid secretion mechanism is 0.192, which is the edge weight of the premise relation; then the cognitive load is 0.79×0.164≈0.13.

[0070] Step S3-2-2: Obtain concept importance based on the frequency of occurrence of each concept in clinical practice; obtain concept relevance based on the target learning concept; and obtain learning benefits based on concept importance and concept relevance.

[0071] Specifically, the importance of a concept is determined by the frequency of its occurrence in clinical cases and treatment guidelines. For example, the annual occurrence frequency of peptic ulcer is 500 times per thousand cases, with a score of 0.9. Concept relevance is determined by calculating the semantic similarity between the concept and the target learning concept. For example, the similarity between the diagnosis and treatment process of duodenal ulcer and peptic ulcer is 0.85. Learning benefit = concept importance × concept relevance. The higher the value obtained, the higher the learning value.

[0072] It should be noted that the importance of a concept is determined by statistically analyzing its frequency of occurrence in clinical cases and treatment guidelines. This involves collecting recent clinical case databases of gastroenterology and authoritative treatment guidelines, counting the number of times the target concept appears in patient complaints, diagnoses, and treatment plans, and then normalizing the result by combining this with the total number of cases or the length of the guideline text. For example, the annual frequency is calculated as: annual frequency of concept / total number of cases × 1000. The normalized result is then mapped to a score range of 0-1. The higher the frequency, the closer the score is to 1. For instance, if the annual frequency of peptic ulcer is 500 times per 1000 cases, the score would be 0.9.

[0073] Furthermore, the same BERT semantic similarity algorithm as in steps S2-7 can be used to obtain a quantitative value of the semantic similarity between the current concept and the target learning concept. The larger the value, the closer the semantic association. For example, the similarity between the diagnosis and treatment process of duodenal ulcer and peptic ulcer is 0.85.

[0074] In some possible implementations, assuming that the clinical frequency score of duodenal ulcer is 0.8 and the correlation score with the target learning concept of peptic ulcer diagnosis and treatment process is 0.85, then the learning gain = 0.8 × 0.85 = 0.68; the frequency score of gastric cancer is 0.7 and the correlation score is 0.7, then the learning gain = 0.7 × 0.7 = 0.49. Therefore, the learning gain for duodenal ulcer is higher.

[0075] Step S3-2-3: Set a similarity threshold, obtain the semantic similarity of adjacent concept nodes in the concept evolution relationship network, constrain adjacent concept nodes with similarity less than the threshold, and constrain the learning order at the same time. The learning order means learning the premise concept first and then learning the target concept.

[0076] Specifically, a semantic similarity threshold is set, and the similarity of the digestive semantic vectors of adjacent nodes in the concept evolution relationship network is calculated. Edges with similarity less than the threshold are removed. At the same time, the learning order is defined through the premise relationship, which forces the premise concept to appear first and the target concept to appear later in the path to avoid the learning logic gap.

[0077] It should be noted that by statistically analyzing the semantic similarity distribution among gastroenterology concepts, we found that 0.6 is the critical value for distinguishing between topic-related and unrelated concepts; concepts with a similarity higher than 0.6 have a clear association in the anatomical, pathological, or clinical dimensions, such as peptic ulcer and duodenal ulcer, while concepts with a similarity lower than 0.6 have a loose association, such as peptic ulcer and acute pancreatitis.

[0078] In some possible implementations, assuming a similarity threshold of 0.6 is set, the semantic similarity between peptic ulcer and duodenal ulcer is calculated to be 0.8. If the similarity is greater than or equal to the similarity threshold of 0.6, the edge is retained. If the similarity with acute pancreatitis is 0.4, which is less than the similarity threshold of 0.6, the edge is removed. In the learning order constraint, the path must first arrange the prerequisite concept of gastric acid secretion mechanism and then arrange the target concept of peptic ulcer. The order cannot be reversed.

[0079] Step S3-3: Input the concepts, learning behaviors, and target learning concepts that the interns have mastered into the learning model, and output the learning path by combining the concept evolution relationship network.

[0080] In this embodiment, step S3-3 includes: Step S3-3-1: During the learning process, add real-time testing nodes to test interns and obtain quantitative values ​​of the test results.

[0081] Understandably, after every 2-3 concepts are learned in the learning path, a real-time test node is set up. The test content may include concept comprehension questions, such as pathological mechanism selection questions, and clinical case analysis questions, such as simulated diagnosis and treatment cases and pathological mechanism judgment questions. The test results are quantitatively scored on a percentage basis to obtain the quantitative value of the test result. For example, if intern A scores 75 points on the test, this score is the quantitative value of the test result.

[0082] Step S3-3-2: Set a first threshold and a second threshold. If the quantified value of the test result is less than the first threshold, it means that the current concept has not been mastered. Then, return to the premise relation concept of the current concept to relearn.

[0083] Specifically, a first threshold is set, such as 60 points, and a second threshold is set, such as 80 points. When the quantitative value of the test result is less than the first threshold, it means that the intern has not mastered the core knowledge of the current concept. It is necessary to go back to the premise relationship concept of the current concept in the concept evolution relationship network, relearn the basic content, and then proceed.

[0084] It should be noted that, referring to the general grading standards in the field of medical education, 60 points is the basic passing score for knowledge mastery; a score below this indicates that the student has not reached the introductory level of understanding. 80 points is the threshold for mastering knowledge proficiently, corresponding to the ability to independently apply concepts to solve basic clinical problems in medical learning. A score between 60 and 80 points indicates basic cognition but with knowledge gaps. Considering the learning stage characteristics of interns, this threshold range can accurately distinguish between three levels: not mastered, partially mastered, and mastered. This avoids learning progress stagnation due to an excessively high threshold, and also prevents a lack of solid knowledge mastery due to an excessively low threshold, thus meeting the ability assessment needs of interns transitioning from theoretical learning to clinical practice.

[0085] In some possible embodiments, if intern A scores 55 points on a test after learning about peptic ulcers, which is less than the first threshold of 60 points, it indicates that the concept has not been mastered. In this case, the intern will relearn the concept of its prerequisite relationship, gastric acid secretion mechanism. By reviewing the basic content such as the cellular mechanism and regulatory factors of gastric acid secretion, the intern will lay the foundation for learning about peptic ulcers again.

[0086] Step S3-3-3: If the quantitative value of the test result is greater than or equal to the first threshold and less than the second threshold, it indicates that the current concept has been partially mastered. Then, the learning of the detailed relationship, comparative relationship and application relationship of the current concept should be strengthened.

[0087] Specifically, when the quantitative value of the test result is greater than or equal to the first threshold and less than the second threshold, it indicates that the intern has a basic understanding of the current concept but has knowledge gaps. It is necessary to design reinforcement exercises targeting the refinement, comparison, and application of the current concept, and deepen understanding through multi-dimensional associative learning.

[0088] In some possible embodiments, assuming that intern A scores 75 points on the test, the quantitative value of the test result is greater than or equal to the first threshold and less than the second threshold, indicating that he has partially mastered peptic ulcers. He should strengthen his learning of the detailed relationship concept of duodenal ulcer, the difference in the location of the ulcer, the key points of gastric cancer identification, and the application relationship concept of proton pump inhibitor treatment. He should fill in the knowledge gaps through case comparison, medication simulation and other methods.

[0089] Step S3-3-4: If the quantified value of the test result is greater than or equal to the second threshold, it indicates that the current concept has been mastered, and then the learning of the next concept is advanced based on the optimization function.

[0090] Specifically, when the test result is greater than or equal to the second threshold, it indicates that the intern has mastered the core knowledge and application points of the current concept. The learning model then selects the next concept from the concept evolution relationship network based on the optimization function, and continues to advance the learning path.

[0091] In some possible implementations, assuming that intern A scores 85 points on the test, and the test result is greater than or equal to the second threshold, it indicates that peptic ulcer has been mastered. The learning model filters the next concept as a complication of peptic ulcer based on the optimization function, and advances to the learning of that concept, maintaining the coherence of the path.

[0092] Step S4: Construct a semantic graph database based on gastroenterology semantic vectors, evolutionary relationship networks, and learning model data.

[0093] In this embodiment, step S4 includes: Step S4-1: The teaching attributes are based on the concept complexity, concept importance, and prerequisite concepts in the learning model data.

[0094] Specifically, the conceptual complexity, significance, and prerequisite concepts of each concept are extracted from the learning model data. These three types of information are then integrated into the pedagogical attributes of the concept to guide the design of the depth and sequence of learning materials.

[0095] In some possible embodiments, the teaching properties of peptic ulcers that can be obtained include concept complexity (0.79), concept importance (0.9), prerequisite concepts such as gastric acid secretion mechanism and gastric mucosal protection. These properties will be used to determine the depth of explanation of the learning materials for this concept and the prerequisites for learning.

[0096] Step S4-2: Based on the target concept identifier, relationship type, and corresponding cognitive distance weight, construct the evolutionary relationship edge.

[0097] Specifically, a unique identifier, such as an ID number, is assigned to the target concept associated with each concept. Combined with the relationship types defined in step S2, including premise relationship, refinement relationship, comparison relationship, and application relationship, as well as the cognitive distance weights obtained in steps S2-8, structured evolutionary relationship edge data is constructed to clarify the association attributes between concepts.

[0098] In some possible embodiments, the evolutionary relationship edge data of peptic ulcers includes target concept identifier duodenal ulcer, relationship type refinement relationship, and cognitive distance weight 0.21; target concept identifier gastric cancer, relationship type comparison relationship, and cognitive distance weight 0.18, etc.

[0099] Step S4-3: Based on concept identifiers, gastroenterology semantic vectors, teaching attributes, and evolutionary relationship edges, a concept node structure is formed.

[0100] Specifically, a unique concept identifier is assigned to each gastroenterology medical concept. The concept identifier is then integrated with the gastroenterology semantic vector, teaching attributes, and evolutionary relationships to form a complete concept node structure, thereby achieving a multi-dimensional structured description of the concept.

[0101] In some possible embodiments, the conceptual node structure of peptic ulcer is as follows: conceptual identifier C001, gastroenterology semantic vector [0.1,0.2,0.9,...,0.85,0.1,...,0.9,0.1,...], teaching attributes {complexity 0.79, importance 0.9, premise concept [C002]}, evolutionary relation edges [C003-refinement relation-0.21,C004-comparison relation-0.18,...], where C002 represents the gastric acid secretion mechanism, C003 represents duodenal ulcer, and C004 represents gastric cancer.

[0102] Step S4-4: Obtain the knowledge status based on the target learning concepts and corresponding mastery levels in the learning model data.

[0103] Specifically, the learning model data is used to extract the interns' target learning concepts and their mastery of each concept, which are then integrated into knowledge status data to record the interns' learning progress and knowledge level.

[0104] In some possible embodiments, it is assumed that the knowledge status data of intern A is the target learning concept [C001, C003] and the level of mastery [C001 partially mastered, C003 not mastered], which intuitively reflects his current learning status.

[0105] Steps S4-5: Based on the knowledge state and learning mode, a learning model structure is formed.

[0106] Specifically, the learning model includes the intern's learning speed, error type, and learning preferences. By integrating knowledge status with the learning model, a learning model structure is formed, thereby obtaining the intern's personalized learning characteristics and providing a basis for personalized adjustments to the learning plan.

[0107] In some possible embodiments, it is assumed that the learning model structure of intern A is a knowledge state [C001 partially mastered, C003 not mastered], a learning mode {learning speed: 2 concepts / hour, error type: symptom identification, learning preference: case video combined with text and images}, and its personalized learning attributes are recorded.

[0108] Steps S4-6: Construct a semantic graph database based on the concept node structure and the learning model structure.

[0109] For example, in a semantic graph database, node C001 is connected to C003 through the edge refinement relation -0.21 and to C004 through the edge comparison relation -0.18. At the same time, it is associated with the learning model structure of intern B: {knowledge state [C001: partially mastered], learning mode [preference for case videos]}, realizing the integrated storage of conceptual knowledge and learner data.

[0110] Step S5: Generate a personalized learning plan based on the semantic graph database and learning path.

[0111] In this embodiment, step S5 includes: Step S5-1: Based on the learning path, the learning materials in the learning path are adjusted according to the teaching attributes and visual resources related to the semantic vectors of the gastroenterology department in the semantic graph database, and the learning patterns in the semantic graph database, to generate an initial personalized learning plan.

[0112] Specifically, following the conceptual sequence of the learning path, the teaching attributes of each concept and the visualization resources associated with the semantic vectors of the gastroenterology department are retrieved from the semantic graph database, such as anatomical vectors corresponding to anatomical diagrams and clinical vectors corresponding to case videos. Combined with the learning mode of each intern, an initial personalized learning plan is formed, consisting of concept explanation, resource display, and basic exercises.

[0113] In some possible embodiments, the learning path is gastric acid secretion mechanism - peptic ulcer - duodenal ulcer. The teaching attributes of peptic ulcer and typical case videos of peptic ulcer associated with clinical vectors are called up. The learning preference case videos are combined with text and images to generate an initial learning scheme module that includes text and image explanation of pathological mechanism - case video analysis - basic identification exercises.

[0114] Step S5-2: Based on the real-time test nodes of the learning path, obtain the current learning concept and the subsequent related concepts of the learning path, and simultaneously obtain the intern's knowledge blind spots in the learning path. Recommend comparative concepts with the current learning concept for enhanced comparative learning to form a personalized learning plan.

[0115] Specifically, at the real-time test node of the learning path, the current learning concept and subsequent related concepts are located from the semantic graph database based on the test results. Knowledge gaps are identified by analyzing the types of test errors, such as symptom identification errors. The comparative concepts of the current concept are queried, and reinforcement comparative learning content is designed and integrated into the initial plan to form a complete personalized learning plan.

[0116] In some possible embodiments, intern A identifies a knowledge gap by making a mistake in the differential diagnosis of upper abdominal pain during a peptic ulcer test. The intern then retrieves the comparative relationship between the current concept of peptic ulcer and the concept of gastric cancer from the database, recommends reinforcement content such as a comparison table of pain patterns between peptic ulcer and gastric cancer, and simulated differential diagnosis cases of patients with upper abdominal pain. After integrating the reinforcement content into the initial plan, a personalized learning plan is formed that includes basic learning, reinforcement comparison, and error correction exercises.

[0117] It should be noted that by storing standardized knowledge data such as the teaching attributes and evolutionary relationships of concepts in a semantic graph database, and simultaneously linking them with personalized data such as the interns' knowledge status and learning patterns, all connected through learning paths, the learning content is ensured to remain within the core framework of the gastroenterology knowledge system. At the same time, the material format can be adjusted according to the interns' knowledge gaps and learning preferences, thus resolving the contradiction between standardized content and personalized needs in traditional solutions.

[0118] Furthermore, the progressive logic of the learning path inherits the hierarchical association of the concept evolution relationship network, enabling the initial learning plan to have knowledge coherence. Finally, through step S5 and the feedback mechanism of the real-time test node, knowledge blind spots can be quickly located with the help of the semantic graph database, and comparative relationship concepts can be retrieved to carry out reinforcement learning. This not only avoids knowledge gaps but also enables precise error correction, helping interns to improve their knowledge system.

[0119] Figure 3 The diagram illustrates a gastroenterology data analysis system based on some embodiments of this application, which can realize the ideas of this application.

[0120] Specifically, the analysis system based on gastroenterology data includes: The parsing module performs triple semantic parsing on gastroenterology medical concepts to obtain anatomical space vectors, pathological space vectors, and clinical space vectors, and then forms a gastroenterology semantic vector. The acquisition module is used to acquire the conceptual evolution relationship between gastroenterological medical concepts, and to acquire the concepts mastered by interns, their learning behaviors, and the concepts they are aiming to learn. The calculation module is used to quantify the cognitive distance weights of concept evolution relationships; The construction module is used to build a learning model, construct a semantic graph database based on gastroenterology semantic vectors, evolutionary relationship networks and learning model data, and construct a concept evolutionary relationship network based on evolutionary relationships and cognitive distance weights; The processing module inputs the concepts, learning behaviors, and target learning concepts already mastered by the interns into the learning model, outputs a learning path, and generates a personalized learning plan based on the semantic graph database and the learning path.

[0121] The specific usage and function of this embodiment are explained below: First, a triple semantic analysis of gastroenterology medical concepts is performed to obtain anatomical, pathological, and clinical spatial vectors, which are then combined to form a gastroenterology semantic vector. Next, the conceptual evolution relationships between gastroenterology medical concepts are identified, and the cognitive distance weights of these relationships are quantified. A conceptual evolution relationship network is constructed using these relationships and cognitive distance weights. This network clarifies four types of logical connections within gastroenterology knowledge: premises, refinements, comparisons, and applications, ensuring the logical coherence of the learning path. Finally, a learning model is constructed, using the intern's existing concepts, learning behaviors, and target learning concepts as inputs, and outputting... The learning path is then developed, and a semantic graph database is constructed based on the semantic vectors, evolutionary relationship networks, and learning model data of the gastroenterology department. Finally, personalized learning plans are generated through the semantic graph database and the learning path. By combining the semantic graph database with the interns' knowledge status data and multi-objective optimization planning of the learning path, the knowledge blind spots, learning speed, and content preferences of different interns can be obtained, and personalized learning plans can be generated, resolving the contradiction between the uniform teaching of traditional plans and personalized needs. Through this technical solution with multi-dimensional semantic analysis and personalized path optimization capabilities, the accuracy and adaptability of learning plans in the training of gastroenterology interns can be improved.

[0122] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via limited means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0123] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0124] It should be understood that, in the embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0125] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An analysis method based on gastroenterology data, characterized in that, It includes the following steps: A triple semantic analysis of gastroenterology medical concepts is performed to obtain anatomical space vectors, pathological space vectors, and clinical space vectors, which are then combined to form a gastroenterology semantic vector. The concept evolution relationships between gastroenterological medical concepts are obtained, including premise relationships, refinement relationships, comparison relationships, and application relationships. The cognitive distance weights of the concept evolution relationships are quantified, and a concept evolution relationship network is constructed based on the evolution relationships and cognitive distance weights. Construct a learning model, obtain the concepts already mastered by interns, their learning behaviors, and the target learning concepts as inputs to the learning model, and output the learning path; A semantic graph database was constructed based on semantic vectors from the Department of Gastroenterology, evolutionary relationship networks, and learning model data; Personalized learning plans are generated based on semantic graph databases and learning paths.

2. The analysis method based on gastroenterology data according to claim 1, characterized in that, A triple semantic analysis of gastroenterological medical concepts is performed to obtain anatomical spatial vectors, pathological spatial vectors, and clinical spatial vectors, which are then combined to form a gastroenterological semantic vector, including: The digestive system is hierarchically divided from upstream to downstream, and an anatomical space vector is constructed based on the hierarchical anatomical location of the digestive system. The correlation logic of common pathologies in gastroenterology is classified, the pathological mechanism classification results are obtained, and a pathological space vector is constructed based on the pathological mechanism classification results. Obtain typical clinical manifestation patterns of gastrointestinal diseases, and construct clinical space vectors based on the clinical manifestation patterns; Semantic vectors for gastroenterology are obtained based on anatomical spatial vectors, pathological spatial vectors, and clinical spatial vectors.

3. The analysis method based on gastroenterology data according to claim 1, characterized in that, To obtain the conceptual evolutionary relationships between gastroenterological medical concepts, including: A prerequisite relationship is defined as a relationship in which one concept is a fundamental premise for understanding another concept. A refinement relationship is defined as one concept being a specific manifestation of another concept. The relationship between the two concepts that need to be distinguished and compared is defined as a contrastive relationship; The relationship in which one concept is applied to another concept is defined as an application relationship. Based on the aforementioned premise relationships, refinement relationships, comparison relationships, and application relationships, we can obtain the conceptual evolution relationships between gastroenterological medical concepts.

4. The analysis method based on gastroenterology data according to claim 1, characterized in that, The cognitive distance weights of quantified concept evolution relationships are used to construct a concept evolution relationship network based on evolution relationships and cognitive distance weights, including: Obtain the co-occurrence frequency of concepts in teaching materials and literature; Obtain the related expressions between concepts in clinical guideline texts and quantify the strength of the association based on semantic similarity algorithms; Based on co-occurrence frequency and association strength, and combined with the AHP (Analytic Hierarchy Process) method, the cognitive distance weights of premise relations, refinement relations, contrast relations and application relations are obtained respectively; Using each concept as a node, and with premise relationships, refinement relationships, comparison relationships, and application relationships as edges, and labeling the corresponding cognitive distance weights, a concept evolution relationship network is constructed.

5. The analysis method based on gastroenterology data according to claim 1, characterized in that, Construct a learning model, taking the intern's existing concepts, learning behaviors, and target learning concepts as input, and outputting a learning path, including: The system acquires the concepts and learning behaviors that interns have mastered, including their learning speed, error types, and learning preferences, and simultaneously acquires the interns' target learning concepts. The learning model is constructed using the minimization of total cognitive load, the maximization of learning benefits, and the constraint of learning coherence as optimization functions; The learning model inputs the concepts, learning behaviors, and target learning concepts already mastered by interns into the learning model, and outputs the learning path by combining the concept evolution relationship network.

6. The analysis method based on gastroenterology data according to claim 5, characterized in that, A learning model is constructed using the minimization of total cognitive load, the maximization of learning gains, and the constraint of learning coherence as optimization functions, including: Conceptual complexity is obtained based on the level of abstraction of each semantic space in the semantic vector of the gastroenterology department; learning distance is obtained based on the shortest path distance in the conceptual evolution relationship network; and cognitive load is obtained based on conceptual complexity and learning distance. Concept importance is determined by the frequency of occurrence of each concept in clinical practice; concept relevance is determined by the target learning concept; and learning benefits are determined by concept importance and concept relevance. A similarity threshold is set to obtain the semantic similarity of adjacent concept nodes in the concept evolution relationship network. Constraints are imposed on adjacent concept nodes that are less than the similarity threshold, and the learning order is also constrained. The learning order is expressed as learning the premise concept first and then learning the target concept.

7. The analysis method based on gastroenterology data according to claim 5, characterized in that, The learning model inputs the concepts, learning behaviors, and target learning concepts already mastered by the interns, and outputs the learning path by combining the concept evolution relationship network, including: During the learning process, real-time testing nodes are added to test interns and obtain quantitative values ​​of the test results. Set a first threshold and a second threshold. If the quantitative value of the test result is less than the first threshold, it means that the current concept has not been mastered. Then, return to the premise relation concept of the current concept to relearn. If the quantitative value of the test result is greater than or equal to the first threshold and less than the second threshold, it indicates that the current concept has been partially mastered. In this case, the learning of the detailed relationship, comparative relationship and application relationship of the current concept should be strengthened. If the quantified value of the test result is greater than or equal to the second threshold, it indicates that the current concept has been mastered, and then the learning of the next concept is advanced based on the optimization function.

8. The analysis method based on gastroenterology data according to claim 1, characterized in that, A semantic graph database is constructed based on gastroenterology semantic vectors, evolutionary relationship networks, and learning model data, including: The teaching attributes are based on the concept complexity, concept importance, and prerequisite concepts in the learning model data; Evolutionary relation edges are constructed based on target concept identifiers, relation types, and corresponding cognitive distance weights; The concept node structure is composed of concept identifiers, gastroenterology semantic vectors, teaching attributes, and evolutionary relationship edges; Knowledge status is acquired based on the target learning concepts and corresponding mastery levels in the learning model data; The learning model structure is composed of knowledge state and learning mode; A semantic graph database is constructed based on the concept node structure and the learning model structure.

9. The analysis method based on gastroenterology data according to claim 1, characterized in that, Personalized learning solutions are generated based on semantic graph databases and learning paths, including: Based on the learning path, the learning materials in the learning path are adjusted according to the teaching attributes and visual resources of the gastroenterology semantic vector association in the semantic graph database, and combined with the learning patterns in the semantic graph database, to generate an initial personalized learning plan. Based on the real-time test nodes of the learning path, the system obtains the current learning concept and the subsequent related concepts of the learning path, and simultaneously identifies the intern's knowledge gaps in the learning path. It then recommends comparative concepts with the current learning concept for enhanced comparative learning, forming a personalized learning plan.

10. An analysis system based on gastroenterology data, characterized in that, include: The parsing module performs triple semantic parsing on gastroenterology medical concepts to obtain anatomical space vectors, pathological space vectors, and clinical space vectors, and then forms a gastroenterology semantic vector. The acquisition module is used to acquire the conceptual evolution relationship between gastroenterological medical concepts, and to acquire the concepts mastered by interns, their learning behaviors, and the concepts they are aiming to learn. The calculation module is used to quantify the cognitive distance weights of concept evolution relationships; The construction module is used to build a learning model, construct a semantic graph database based on gastroenterology semantic vectors, evolutionary relationship networks and learning model data, and construct a concept evolutionary relationship network based on evolutionary relationships and cognitive distance weights; The processing module inputs the concepts, learning behaviors, and target learning concepts already mastered by the interns into the learning model, outputs a learning path, and generates a personalized learning plan based on the semantic graph database and the learning path.

Images