Gout patient dynamic follow-up visit method and device based on multi-modal fusion and medium

Abstract

The invention discloses a gout patient dynamic follow-up visit method and device based on multi-modal fusion and a medium, and relates to the technical field of medical information processing, and the method comprises the steps: obtaining gout patient end data, and carrying out the adaptive configuration of a follow-up visit plan for the gout patient end data, so as to obtain a gout patient follow-up visit initial plan; determining real-time interaction data of the gout patient based on the initial follow-up plan of the gout patient, and performing special disease scene semantic analysis on the real-time interaction data of the gout patient to obtain structured symptom information; according to the structured symptom information, determining a current follow-up session record through multi-modal information fusion; performing dynamic feature construction of gout symptoms on the current follow-up session record to obtain a symptom change state of the gout patient; and performing doctor-seeing suggestion decision on the symptom change state of the gout patient to determine a structured follow-up visit result. Through the method, the technical problem that dynamic follow-up visit cannot be performed for gout cases in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of medical information processing technology, and in particular to a method, device and medium for dynamic follow-up of gout patients based on multimodal fusion. Background Technology

[0002] In the field of medical information processing technology, case follow-up is a core component of chronic disease management, especially for specific diseases such as hyperuricemia and gout, which require long-term monitoring and are prone to recurrence. Systematic follow-up is crucial for controlling the disease and preventing complications. Traditional case follow-up methods generally rely on manual telephone calls or outpatient follow-ups, which are time-consuming, labor-intensive, and difficult to scale up or personalize.

[0003] Several intelligent follow-up systems based on deep learning have emerged in the existing technology. Chinese patent application CN118230878A discloses a method and storage medium for an intelligent case follow-up system based on deep learning, which achieves partial automation of follow-up through a follow-up customization module, a follow-up server, and a result analysis module, effectively improving efficiency. However, this method is a general-purpose design, lacking deep adaptation to specific diseases such as hyperuricemia and gout. Its natural language processing model cannot accurately understand the key terms and descriptions of gout cases. Secondly, this method is relatively weak in data analysis, typically only able to perform simple information classification and storage, lacking an intelligent decision-making mechanism that integrates real-time symptoms, historical data, and clinical knowledge. It cannot quantitatively assess the urgency of a patient's condition, nor can it provide clinically significant grading recommendations. Therefore, there is an urgent need for an intelligent follow-up analysis method specifically for gout patients. Summary of the Invention

[0004] This application provides a method, device, and medium for dynamic follow-up of gout patients based on multimodal fusion, which solves the technical problem that the prior art cannot perform dynamic follow-up of gout cases.

[0005] In a first aspect, embodiments of this application provide a method for dynamic follow-up of gout patients based on multimodal fusion. The method includes: acquiring gout patient-side data and adaptively configuring a follow-up plan based on the gout patient-side data to obtain an initial follow-up plan; determining real-time interaction data of the gout patient based on the initial follow-up plan, and performing disease-specific contextual semantic analysis on the real-time interaction data to obtain structured symptom information; determining the current follow-up session record based on the structured symptom information through multimodal information fusion; constructing dynamic features of gout symptoms on the current follow-up session record to obtain the symptom change status of the gout patient; and making medical advice decisions based on the symptom change status of the gout patient to determine the structured follow-up result.

[0006] In one implementation of this application, adaptive configuration of follow-up plans for gout patient data is performed to obtain an initial follow-up plan for gout patients. Specifically, this includes: standardizing the gout patient data to obtain standard patient data; the standardization process includes RobustScaler standardization and Min-Max standardization; determining a patient feature matrix based on the standard patient data and gout-specific features; performing unsupervised learning clustering on the patient feature matrix to obtain labeled gout patient feature clusters; determining gout patient follow-up tasks based on follow-up rule mapping according to the gout patient feature clusters; and configuring an initial task queue for the gout patient follow-up tasks to obtain the initial follow-up plan for gout patients.

[0007] In one implementation of this application, specific contextual semantic analysis is performed on real-time interactive data of gout patients to obtain structured symptom information. Specifically, this includes: determining text classification results based on real-time interactive data of gout patients through interactive intent recognition; performing named entity recognition on the text classification results to obtain a list of entity fragments; constructing a gout-specific dictionary based on the list of entity fragments; and determining fuzzy interaction mappings based on the gout-specific dictionary through fuzzy matching using an edit distance algorithm; and obtaining structured symptom information related to the severity of symptoms through dependency parsing based on the fuzzy interaction mappings.

[0008] In one implementation of this application, the current follow-up session record is determined based on structured symptom information through multimodal information fusion. Specifically, this includes: extracting text features from the structured symptom information to obtain a gout symptom text feature vector; extracting image features from the structured symptom information to obtain a gout symptom visual feature vector; normalizing the numerical features of a preset VAS score to determine scalar features; obtaining a gout symptom fusion feature through weighted fusion using an attention mechanism based on the gout symptom text feature vector, the gout symptom visual feature vector, and the scalar features; and determining the current follow-up session record based on the gout symptom fusion feature and by calculating the severity of gout symptoms.

[0009] In one implementation of this application, dynamic features of gout symptoms are constructed from the current follow-up session record to obtain the symptom change status of the gout patient. Specifically, this includes: acquiring historical disease data corresponding to the gout patient, and determining the patient's medication status based on the historical disease data through sentiment analysis of the patient's medication description; calculating the patient's dynamic treatment characteristics based on the patient's medication status through medication records; assessing the severity of gout symptoms in the current follow-up session record to determine the absolute difference between the current patient's symptoms and the historical average symptom severity; and updating the absolute difference to the historical disease data to obtain the symptom change status of the gout patient.

[0010] In one implementation of this application, a medical advice decision is made based on the changing symptom status of a gout patient to determine a structured follow-up result. Specifically, this includes: determining the urgency score of the current gout patient's symptoms based on the changing symptom status using gradient boosting decision tree analysis; determining a medical advice threshold based on the urgency score to obtain a medical advice decision; determining a safe medical advice decision based on the medical advice decision through a medical rule channel; and generating a language explanation for the safe medical advice decision to determine the structured follow-up result.

[0011] In one implementation of this application, a safe medical treatment recommendation decision is determined based on the medical treatment recommendation decision and through a medical treatment rule channel. Specifically, this includes: obtaining the urgency parameter of the medical treatment rule and configuring a response signal for the urgency parameter to obtain the medical treatment rule channel; inputting the medical treatment recommendation decision into the medical treatment rule channel and determining the safe medical treatment recommendation decision through the urgency signal.

[0012] In one implementation of this application, after making medical advice decisions based on the changes in symptoms of gout patients to determine the structured follow-up results, the method further includes: monitoring key indicators of the structured follow-up results to obtain dynamic parameters of the key indicators; and visualizing the dynamic parameters of the key indicators to obtain a follow-up result evaluation report.

[0013] Secondly, embodiments of this application also provide a dynamic follow-up device for gout patients based on multimodal fusion, characterized in that the device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: acquire gout patient terminal data, and adaptively configure a follow-up plan based on the gout patient terminal data to obtain an initial follow-up plan for gout patients; determine real-time interaction data of gout patients based on the initial follow-up plan for gout patients, and perform disease-specific contextual semantic analysis on the real-time interaction data of gout patients to obtain structured symptom information; determine the current follow-up session record based on the structured symptom information through multimodal information fusion; construct dynamic features of gout symptoms on the current follow-up session record to obtain the symptom change status of gout patients; and make medical advice decisions based on the symptom change status of gout patients to determine the structured follow-up results.

[0014] Thirdly, embodiments of this application also provide a non-volatile computer storage medium for dynamic follow-up of gout patients based on multimodal fusion, storing computer-executable instructions. The computer-executable instructions are characterized by: acquiring gout patient-side data and adaptively configuring a follow-up plan based on the gout patient-side data to obtain an initial follow-up plan for the gout patient; determining real-time interaction data of the gout patient based on the initial follow-up plan, and performing disease-specific contextual semantic analysis on the real-time interaction data of the gout patient to obtain structured symptom information; determining the current follow-up session record based on the structured symptom information through multimodal information fusion; constructing dynamic features of gout symptoms on the current follow-up session record to obtain the symptom change status of the gout patient; and making medical advice decisions based on the symptom change status of the gout patient to determine the structured follow-up result.

[0015] This application provides a method, device, and medium for dynamic follow-up of gout patients based on multimodal fusion. By combining domain-adaptive semantic context analysis of gout, multimodal information fusion, and gout symptom severity assessment, it solves the technical problem that existing technologies cannot perform dynamic follow-up of gout cases. It realizes intelligent follow-up of gout patients based on semantic analysis, and can provide personalized follow-up and medical advice according to the patient's follow-up status, thereby improving the follow-up efficiency of gout patients and improving the health status of patients with high uric acid and gout. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart of a dynamic follow-up method for gout patients based on multimodal fusion provided in this application embodiment; Figure 2 This is a schematic diagram of the internal structure of a dynamic follow-up device for gout patients based on multimodal fusion, provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] This application provides a method, device, and medium for dynamic follow-up of gout patients based on multimodal fusion. By combining domain-adaptive semantic context analysis of gout, multimodal information fusion, and gout symptom severity assessment, it solves the technical problem that existing technologies cannot perform dynamic follow-up of gout cases. It realizes intelligent follow-up of gout patients based on semantic analysis, and can provide personalized follow-up and medical advice according to the patient's follow-up status, thereby improving the follow-up efficiency of gout patients and improving the health status of patients with high uric acid and gout.

[0019] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This document provides a flowchart of a dynamic follow-up method for gout patients based on multimodal fusion, as illustrated in an embodiment of this application. Figure 1 As shown in the figure, the present application provides a method for dynamic follow-up of gout patients based on multimodal fusion, which specifically includes the following steps: Step 101: Obtain gout patient data and adaptively configure the follow-up plan based on the gout patient data to obtain the initial follow-up plan for gout patients.

[0021] For example, due to the common presence of outliers in medical data and the blurred boundaries of patient status in medical scenarios, this application achieves personalized follow-up for gout patients by adaptively configuring follow-up plans for gout patient data, thereby improving the accuracy of customized follow-up plans.

[0022] Specifically, adaptive configuration of follow-up plans for gout patients is performed to obtain an initial follow-up plan. This includes: standardizing the gout patient data to obtain standard patient data; the standardization process includes RobustScaler standardization and Min-Max standardization; based on the standard patient data, a patient feature matrix is ​​determined by constructing gout-specific features; unsupervised learning clustering is performed on the patient feature matrix to obtain labeled gout patient feature clusters; follow-up tasks are determined based on the gout patient feature clusters through follow-up rule mapping; and an initial task queue is configured for the gout patient follow-up tasks to obtain the initial follow-up plan.

[0023] In one embodiment, firstly, multi-source heterogeneous data is obtained from a gout patient information database to obtain the basic information of patient A, whose blood uric acid test sequences are 520 μmol / L, 480 μmol / L, and 510 μmol / L, and whose medical history is gout for 3 years with tophi. Parameters such as age and BMI are standardized using Min-Max and linearly scaled to the [0,1] interval.

[0024] Data preprocessing employed RobustScaler standardization to normalize serum uric acid levels. A BMI of 26.1 was calculated and marked as overweight. Based on creatinine levels, the eGFR was calculated using the CKD-EPI formula to be 85 mL / min / 1.73 m². Using the processed data, specific features for gout were constructed, resulting in a patient feature matrix containing basic information and derived specific disease characteristics.

[0025] Then, an unsupervised learning clustering algorithm was used to analyze the patient feature matrix. Key indicators selected for clustering included serum uric acid trends, historical highest serum uric acid levels, and history of tophi. Patients were probabilistically stratified using a Gaussian mixture model, classifying patient A into the high-frequency attack cluster. The cluster characteristics of the high-frequency attack cluster included: consistently high serum uric acid levels above 480 μmol / L, BMI greater than 25, and recent history of acute attacks, thus obtaining labeled gout patient feature clusters.

[0026] It is important to note that Gaussian mixture models were chosen instead of K-Means for patient stratification because Gaussian mixture models can provide probabilistic clustering results, which are more in line with the characteristics of blurred patient state boundaries in medical scenarios.

[0027] Finally, based on the characteristic clusters of gout patients, and through follow-up rule mapping, the initial follow-up plan for gout patients was obtained by selecting the follow-up parameters corresponding to each cluster ID. The follow-up plan is to have a follow-up frequency of 7 days, use the GT-03 script template (focusing on asking about joint symptoms and medication side effects), and set the next follow-up time to 14:00 on [Date].

[0028] Furthermore, the above plan is transformed into a structured task record {"patient_id":"P01xxx","next_followup":"20xx-xx-xx14:00:00","talk_track_id":"GT-03"} and stored in the Redis task queue.

[0029] Among them, patient_id is a unique identifier for each patient, a unique ID assigned to each patient, used to accurately locate a specific patient in all data and processes; next_followup is the next follow-up time, specifying the specific date and time for the next automatic follow-up of the patient; talk_track_id is the ID of a dialogue template, representing a dialogue script template for a specific scenario in the knowledge base.

[0030] Step 102: Based on the initial follow-up plan for gout patients, determine the real-time interaction data of gout patients, and perform disease-specific contextual semantic analysis on the real-time interaction data of gout patients to obtain structured symptom information.

[0031] For example, since the existing technology lacks interactive analysis for gout, in order to improve the recognition accuracy of gout patient follow-up interactive data, this application realizes intelligent analysis of gout patient follow-up interaction by performing disease-specific contextual semantic analysis on real-time interactive data of gout patients.

[0032] Specifically, semantic analysis of real-time interactive data from gout patients is performed to obtain structured symptom information. This includes: determining text classification results based on real-time interactive data from gout patients through interaction intent recognition; performing named entity recognition on the text classification results to obtain a list of entity fragments; constructing a gout-specific dictionary based on the entity fragment list; and determining fuzzy interaction mappings based on the gout-specific dictionary through fuzzy matching using an edit distance algorithm; and obtaining structured symptom information related to the severity of symptoms through dependency parsing based on the fuzzy interaction mappings.

[0033] In one embodiment, when the planned follow-up time arrives, the gout transition interaction process is initiated, the follow-up task is retrieved from the task queue through the batch outbound call unit, and the professional inquiry in the GT-03 script template is broadcast through the speech synthesis engine.

[0034] Following the output vector of the [CLS] token in the pre-trained model, a Dropout layer and a fully connected layer are applied, and the Softmax function is used to output the probability (i.e., confidence) of each intent category. Each character or word corresponding to the patient interaction information is converted into a context-sensitive embedding vector to capture long-distance bidirectional dependencies in the text.

[0035] For example, after a patient replies with a voice message, "My right big toe suddenly swelled up last night, and it hurt so much that I couldn't sleep all night. It felt like being cut with a knife. I drank some beer and ate some seafood yesterday," the disease-optimized voice recognition service will convert the voice into text information.

[0036] Next, the disease-specific contextual semantic understanding module performs deep analysis of the text. Intent recognition uses a fine-tuned RoBERTa model, determining the dialogue intent as reporting symptoms (with a confidence level of 0.96). Entity recognition employs a BERT-BiLSTM-CRF architecture for sequence labeling. This three-layer architecture is chosen because BERT generates high-quality context-sensitive word vectors, BiLSTM effectively captures long-distance dependencies, and the CRF layer learns the constraint rules between labels.

[0037] Finally, the entity was identified as: {"B-BodyPart":"Right side","I-BodyPart":"Big toe","B-Symptom":"Swelling","B-Symptom":"Pain","B-Degree":"Cut","B-Trigger":"Beer","B-Trigger":"Seafood"}.

[0038] The entity standardization module maps "big toe" to the standard term MTP1 joint and "cutting" to the severity level "Severe" by querying the disease-specific knowledge graph. Based on intent analysis and real-time interactive data from gout patients, structured symptom information is obtained.

[0039] Step 103: Based on the structured symptom information, determine the current follow-up session record through multimodal information fusion.

[0040] For example, the use of attention mechanisms in the prior art can meet the learning needs of symptom information, but it cannot learn the importance weights of different modal information dynamically. Since patients may provide not only voice data but also image data of lesions during the follow-up process, this application improves the robustness and accuracy of subsequent assessment of the severity of the condition through multimodal information fusion.

[0041] Specifically, based on structured symptom information, the current follow-up session record is determined through multimodal information fusion, including: extracting text features from the structured symptom information to obtain a gout symptom text feature vector; extracting image features from the structured symptom information to obtain a gout symptom visual feature vector; normalizing the numerical features of the preset VAS score to determine scalar features; obtaining a gout symptom fusion feature through weighted fusion using an attention mechanism based on the gout symptom text feature vector, gout symptom visual feature vector, and scalar features; and determining the current follow-up session record based on the gout symptom fusion feature and by calculating the severity of gout symptoms.

[0042] In one embodiment, gout patients upload images of their affected areas via their own mobile app. The image feature extraction module uses the EfficientNet-B2 model to extract feature vectors, identify case features such as redness and swelling, and then performs a weighted fusion of text, images, and VAS scores based on an attention mechanism.

[0043] First, key information from the structured symptom information (JSON), such as symptom type, severity level, and onset speed, is converted into feature vectors. Simultaneously, a multi-dimensional sentence-level semantic vector is extracted from the [CLS] token to capture the overall sentiment and urgency of the text.

[0044] For numerical features, the VAS score is normalized as a scalar feature, and the text feature vector, image feature vector, and VAS value are projected onto a unified feature space through different fully connected layers.

[0045] The projected vectors are concatenated and input into an attention network, where an attention weight is calculated for each modality. The current follow-up session record is determined by calculating the weighted severity of gout symptoms. After uploading images of the affected area, gout patients use the EfficientNet-B2 model to extract 1408-dimensional image feature vectors and identify redness and swelling features. The feature fusion module uses an attention mechanism to weight and fuse text feature vectors, image feature vectors, and the patient's self-reported VAS score (8 points), calculating attention weights of 0.6, 0.3, and 0.1, respectively. Finally, a comprehensive severity index of 0.87 is output through a multilayer perceptron.

[0046] Furthermore, when the images uploaded by gout patients are of poor quality, the image weight can be reduced through automatic learning, and the current follow-up session records will rely more on descriptions obtained from patient follow-ups.

[0047] Step 104: Construct dynamic features of gout symptoms from the current follow-up session records to obtain the symptom change status of gout patients.

[0048] Specifically, dynamic features of gout symptoms are constructed from the current follow-up conversation records to obtain the symptom change status of gout patients. This includes: obtaining historical disease data corresponding to gout patients, and determining the patient's medication status based on the historical disease data through sentiment analysis of the patient's medication description; calculating the patient's dynamic treatment characteristics based on the patient's medication status through medication records; assessing the severity of gout symptoms in the current follow-up conversation records to determine the absolute difference between the current patient's symptoms and the historical average symptom severity; and updating the historical disease data with the absolute difference to obtain the symptom change status of gout patients.

[0049] In one embodiment, firstly, current follow-up session records and historical condition data are retrieved from the gout patient follow-up session database. Patient B's current session information is obtained, with a medication description of "recently frequently forgetting to take uric acid-lowering medication, and feeling worsening joint pain." The historical symptom severity sequence is 2, 3, and 5 points (higher scores indicate greater severity). Historical medication records show that Patient B has been treated with allopurinol for one year. For parameters such as symptom scores, Z-score standardization can be used to eliminate the influence of dimensions.

[0050] Data preprocessing employed a BERT-based classifier to perform sentiment analysis on patient medication descriptions, outputting a sentiment score of -0.8 (range [-1, 1], where values ​​less than zero indicate a negative state and values ​​greater than zero indicate a positive state), which was then marked as a negative medication state. Dynamic treatment characteristics were calculated based on medication records. The medication adherence rate was calculated to be 65% by using the ratio of the number of medications taken in the past 30 days to the planned number of doses, and recent dose adjustments were also recorded.

[0051] Simultaneously, the severity of gout symptoms was assessed based on the current follow-up session records. Key symptom terms were extracted using natural language processing technology, and the severity of the current symptoms was quantified into a score of 5 based on a pre-defined rule base. The historical average score was calculated based on the historical symptom severity sequence, which is 3 points. Therefore, the absolute difference between the current symptom severity and the historical average score is 2 points.

[0052] The absolute difference is updated to the historical disease data, and the average symptom score of the sliding window is recalculated to obtain the symptom change status of gout patients as symptom worsening.

[0053] Then, the symptom changes of gout patients are comprehensively assessed in conjunction with dynamic treatment characteristics. Assessment characteristics can include key indicators such as medication status, absolute symptom difference, and medication adherence rate. A logistic regression model is used to probabilistically classify symptom changes, categorizing patients into high-risk symptom progression categories. These high-risk categories are characterized by: negative medication use, an absolute symptom difference greater than 1 point, and a medication adherence rate below 70%, thus obtaining labeled symptom changes for gout patients.

[0054] It is important to note that logistic regression is chosen over simple threshold judgment for symptom state classification because logistic regression can provide probability output and feature weights, which is more in line with the multi-factor influence of symptom changes in medical scenarios.

[0055] Finally, based on the changes in gout patients' symptoms, a follow-up adjustment recommendation was obtained by mapping early warning rules and selecting the corresponding early warning parameters for each status category. The adjustment recommendation was to increase the follow-up frequency to once every 3 days, and the language template used emphasized medication adherence and symptom monitoring, triggering immediate doctor intervention notifications.

[0056] Step 105: Make medical advice decisions based on changes in the symptoms of gout patients to determine the structured follow-up results.

[0057] For example, due to the special requirements of medical safety, the rule engine acts as a safety net to ensure that the highest level of alarm is triggered immediately in the event of a critical situation. This application improves the safety and clinical reliability of gout patients by making medical advice decisions based on the changes in the symptoms of gout patients and implementing a dual protection mechanism.

[0058] Specifically, the process involves making medical advice decisions based on changes in gout patients' symptoms to determine structured follow-up outcomes. This includes: determining the urgency score of the current gout patient's symptoms using gradient boosting decision tree analysis based on the changes in the patient's symptoms; determining a medical advice threshold based on the urgency score to obtain a medical advice decision; determining a safe medical advice decision based on the medical advice decision through a medical advice rule channel; and generating a language explanation for the safe medical advice decision to determine structured follow-up outcomes.

[0059] Furthermore, based on the medical treatment recommendation decision, a safe medical treatment recommendation decision is determined through the medical treatment rule channel. Specifically, this includes: obtaining the urgency parameter of the medical treatment rule and configuring the response signal for the urgency parameter to obtain the medical treatment rule channel; inputting the medical treatment recommendation decision into the medical treatment rule channel and determining the safe medical treatment recommendation decision through the urgency signal.

[0060] Furthermore, after making medical advice decisions based on the changes in symptoms of gout patients to determine the structured follow-up results, the method also includes: monitoring key indicators of the structured follow-up results to obtain dynamic parameters of key indicators; and visualizing the dynamic parameters of key indicators to obtain a follow-up result evaluation report.

[0061] In one embodiment, a medical advice decision is made based on the changing symptom status of a gout patient. The structured feature vector is immediately input into a pre-trained LightGBM gradient boosting decision tree model. By evaluating the current situation, an urgency score between 0 and 1 is output to represent the probability that the patient needs to seek medical attention immediately.

[0062] The initial recommendation for medical treatment decision-making is: emergency score < 0.3, and the model generates the initial recommendation: green (home management). If the emergency score is 0.3 or less and the emergency score is less than 0.7, the model generates the initial suggestion: Yellow (follow-up visit recommended). An emergency score ≥ 0.7 triggers the model's initial recommendation: Red (Seek immediate medical attention). The generated dynamic feature vector is {"severity_index": 0.87,} "has_fever":false, "ua_trend":-0.15, "attack_frequency":2}.

[0063] Among them, severity_index represents the severity of the patient's condition, has_fever indicates whether the patient has fever symptoms, ua_trend represents the trend of changes in blood uric acid, which can be calculated by linear regression to characterize the direction and rate of change in the patient's recent blood uric acid level, and attack_frequency represents the frequency of acute attacks in the patient within one month.

[0064] The LightGBM model inference outputs an urgency score of 0.82, while the hard safety rule engine scans the patient's complaint text and finds no keywords such as "chest pain". Since the urgency score of 0.82 is greater than the threshold of 0.7 and the rule engine does not trigger a rejection, the system generates a final medical advice in red (seek immediate medical attention).

[0065] The decision-making basis generation module automatically creates an explanation: "The system determines that you need to seek medical attention immediately, primarily based on: high severity of the condition (0.87 / 1.0), severe pain (VAS score 8), and a history of a high-purine diet prior to the attack. This recommendation and basis are pushed to both the doctor and patient through the follow-up feedback module. After the doctor confirms the recommendation, the system automatically adjusts the follow-up plan, shortening the next follow-up interval to 3 days and marking the patient as a high-risk patient for intensive monitoring."

[0066] Finally, the system initiated the model optimization process. The semantic understanding confidence score of this interaction was 0.96, higher than the threshold of 0.9, and was directly stored in the database as a training sample. The system performed weekly incremental training tasks, fine-tuning the RoBERTa model with a learning rate of 5e-6. After A / B testing, the new model improved its intent recognition accuracy from 92.1% to 93.8%, and its entity recognition F1 score from 89.5% to 90.7%. Simultaneously, the knowledge graph was expanded based on the newly discovered symptom "describing a needle-like pain," and added to the degree vocabulary after review by doctors.

[0067] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a dynamic follow-up device for gout patients based on multimodal fusion, the structure of which is as follows: Figure 2 As shown.

[0068] Figure 2 This is a schematic diagram of the internal structure of a dynamic follow-up device for gout patients based on multimodal fusion, provided as an embodiment of this application. Figure 2 As shown, the device includes: At least one processor 201; And a memory 202 that is communicatively connected to at least one processor; The memory 202 stores instructions executable by at least one processor, which are executed by at least one processor 201 to enable at least one processor 201 to: Data from gout patients is acquired, and adaptive configuration of follow-up plans is performed on this data to obtain an initial follow-up plan. Based on this plan, real-time interaction data is determined, and contextual semantic analysis is conducted to obtain structured symptom information. The current follow-up session record is then determined through multimodal information fusion based on the structured symptom information. Dynamic features of gout symptoms are constructed from the current session record to reveal the changing symptom status of gout patients. Finally, medical advice decisions are made based on the changing symptom status to determine the structured follow-up outcome.

[0069] Some embodiments of this application provide corresponding to Figure 1 A non-volatile computer storage medium for dynamic follow-up of gout patients based on multimodal fusion, storing computer-executable instructions, wherein the computer-executable instructions are configured as follows: Data from gout patients is acquired, and adaptive configuration of follow-up plans is performed on this data to obtain an initial follow-up plan. Based on this plan, real-time interaction data is determined, and contextual semantic analysis is conducted to obtain structured symptom information. The current follow-up session record is then determined through multimodal information fusion based on the structured symptom information. Dynamic features of gout symptoms are constructed from the current session record to reveal the changing symptom status of gout patients. Finally, medical advice decisions are made based on the changing symptom status to determine the structured follow-up outcome.

[0070] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0071] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0072] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0073] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0074] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0075] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0076] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0077] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0078] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0079] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0080] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for dynamic follow-up of gout patients based on multimodal fusion, characterized in that, The method includes: Acquire gout patient data and adaptively configure the follow-up plan based on the gout patient data to obtain an initial follow-up plan for gout patients; Based on the initial follow-up plan for gout patients, real-time interactive data of gout patients is determined, and disease-specific contextual semantic analysis is performed on the real-time interactive data of gout patients to obtain structured symptom information. Based on the structured symptom information, the current follow-up session record is determined through multimodal information fusion; Dynamic features of gout symptoms are constructed from the current follow-up session records to obtain the symptom change status of gout patients; Medical advice decisions are made based on the changes in symptoms of the gout patients to determine the structured follow-up results.

2. The method for dynamic follow-up of gout patients based on multimodal fusion according to claim 1, characterized in that, Adaptive configuration of the follow-up plan for the gout patient data is performed to obtain an initial follow-up plan for gout patients, specifically including: The gout patient data is subjected to data standardization processing to obtain standard patient data; wherein, the data standardization processing includes RobustScaler standardization and Min-Max standardization; Based on the aforementioned standard patient data, a patient feature matrix was determined by constructing gout-specific disease features. Unsupervised learning clustering is performed on the patient feature matrix to obtain labeled gout patient feature clusters; Based on the gout patient characteristic clusters, follow-up tasks for gout patients are determined through follow-up rule mapping. An initial task queue is configured for the gout patient follow-up task to obtain the initial follow-up plan for the gout patients.

3. The method for dynamic follow-up of gout patients based on multimodal fusion according to claim 1, characterized in that, Specific contextual semantic analysis was performed on the real-time interactive data of gout patients to obtain structured symptom information, specifically including: Based on the real-time interaction data of the gout patients, the text classification result is determined through interaction intent recognition; Named entity recognition is performed on the text classification results to obtain a list of entity fragments; Based on the list of entity fragments, a gout-specific dictionary is constructed, and based on the gout-specific dictionary, a fuzzy interaction mapping is determined through fuzzy matching using an edit distance algorithm. Based on the fuzzy interaction mapping, structured symptom information related to the severity of the disease is obtained through dependency parsing.

4. The method for dynamic follow-up of gout patients based on multimodal fusion according to claim 1, characterized in that, Based on the structured symptom information, the current follow-up session record is determined through multimodal information fusion, specifically including: Text feature extraction is performed on the structured symptom information to obtain a text feature vector of gout symptoms; Image features are extracted from the structured symptom information to obtain visual feature vectors of gout symptoms; The pre-defined VAS scores are normalized numerically to determine scalar features; Based on the textual feature vector of gout symptoms, the visual feature vector of gout symptoms, and the scalar features, a weighted fusion of gout symptoms is obtained through an attention mechanism. Based on the fusion characteristics of gout symptoms, the current follow-up session record is determined by calculating the severity of gout symptoms.

5. The method for dynamic follow-up of gout patients based on multimodal fusion according to claim 1, characterized in that, Dynamic features of gout symptoms are constructed from the current follow-up session records to obtain the symptom change status of gout patients, specifically including: Obtain historical medical data for gout patients, and determine the patient's medication status based on the historical medical data and through sentiment analysis of the patient's medication description. Based on the patient's medication status, dynamic treatment characteristics of the patient are calculated through medication records; The severity of gout symptoms is assessed on the current follow-up session record to determine the absolute difference between the current patient's symptoms and the historical average symptom severity. The absolute difference is then updated to the historical disease data to obtain the symptom change status of the gout patient.

6. The method for dynamic follow-up of gout patients based on multimodal fusion according to claim 1, characterized in that, Medical advice decisions are made based on the changes in symptoms of the gout patients to determine structured follow-up outcomes, specifically including: Based on the changes in the symptoms of the gout patient, the urgency score of the current gout patient's symptoms is determined through gradient boosting decision tree analysis. The urgency score is used to determine a threshold for medical advice to obtain a medical advice decision. Based on the aforementioned medical treatment recommendations, a safe medical treatment recommendation decision is determined through the medical treatment rules channel; The recommended safe medical treatment decisions are interpreted using language generation to determine the structured follow-up results.

7. The method for dynamic follow-up of gout patients based on multimodal fusion according to claim 1, characterized in that, Based on the aforementioned medical treatment recommendations, a safe medical treatment recommendation decision is determined through the medical treatment rules channel, specifically including: Obtain the urgency parameter of the medical treatment rule, and configure the response signal for the urgency parameter of the medical treatment rule to obtain the medical treatment rule channel; The medical treatment recommendation decision is input into the medical treatment rule channel, and a safe medical treatment recommendation decision is determined by judging the urgency signal.

8. The method for dynamic follow-up of gout patients based on multimodal fusion according to claim 1, characterized in that, After making medical advice decisions based on the changes in symptoms of the gout patients to determine structured follow-up results, the method further includes: The structured follow-up results are monitored for key indicators to obtain dynamic parameters of the key indicators; The dynamic parameters of the key indicators are visualized to obtain a follow-up result evaluation report.

9. A dynamic follow-up device for gout patients based on multimodal fusion, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to: Acquire gout patient data and adaptively configure the follow-up plan based on the gout patient data to obtain an initial follow-up plan for gout patients; Based on the initial follow-up plan for gout patients, real-time interactive data of gout patients is determined, and disease-specific contextual semantic analysis is performed on the real-time interactive data of gout patients to obtain structured symptom information. Based on the structured symptom information, the current follow-up session record is determined through multimodal information fusion; Dynamic features of gout symptoms are constructed from the current follow-up session records to obtain the symptom change status of gout patients; Medical advice decisions are made based on the changes in symptoms of the gout patients to determine the structured follow-up results.

10. A non-volatile computer storage medium for dynamic follow-up of gout patients based on multimodal fusion, storing computer-executable instructions, characterized in that, The computer-executable instructions are set as follows: Acquire gout patient data and adaptively configure the follow-up plan based on the gout patient data to obtain an initial follow-up plan for gout patients; Based on the initial follow-up plan for gout patients, real-time interactive data of gout patients is determined, and disease-specific contextual semantic analysis is performed on the real-time interactive data of gout patients to obtain structured symptom information. Based on the structured symptom information, the current follow-up session record is determined through multimodal information fusion; Dynamic features of gout symptoms are constructed from the current follow-up session records to obtain the symptom change status of gout patients; Medical advice decisions are made based on the changes in symptoms of the gout patients to determine the structured follow-up results.

Images