Gout onset risk prediction system and method

Abstract

The invention relates to the technical field of data processing, in particular to a gout onset risk prediction system and method, and the system comprises a collection module which is used for collecting follow-up visit data of a patient, and a feature extraction module which is used for extracting text features of unstructured data, extracting normalized numerical features of structured data, and extracting time sequence features; the feature splicing module is used for splicing the text features, the normalized numerical value features and the time sequence features to obtain a comprehensive feature vector; the model training module is used for inputting the comprehensive feature vector into a neural network structure for training to obtain a gout risk prediction model; and the risk prediction module is used for predicting the follow-up data to be predicted based on the gout risk prediction model to obtain a gout risk prediction result. Therefore, the problems that in the background technology, a gout risk prediction model lacks high-dimensional indexes, structured and unstructured data are difficult to integrate, personalized prediction is lacked and compatibility to time series data is poor are solved, and prediction precision is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a gout risk prediction system and method. Background Art

[0002] Gout is a common metabolic disease with an increasing incidence. It is primarily caused by a sharp rise in uric acid levels in the body, which leads to the deposition of urate crystals in the subcutaneous tissue, causing localized swelling and excruciating pain. Although control medications such as nonsteroidal anti-inflammatory drugs, glucocorticoids, and colchicine can relieve acute gout, long-term medication use can cause organ damage. Therefore, gout management systems are becoming increasingly popular, guiding lifestyle and dietary habits to regulate uric acid levels. The core of these systems is to predict gout risk by analyzing blood and urine metabolites and personal sociological information (such as age and gender).

[0003] Among related technologies, most existing gout risk prediction models are based on traditional machine learning methods, such as logistic regression and support vector machines. They perform poorly when processing high-dimensional, nonlinear data and have difficulty effectively integrating structured and unstructured data. At the same time, existing models typically require a large amount of manual feature engineering, which is time-consuming and labor-intensive, and has difficulty capturing complex feature interactions. Furthermore, gout is a long-term chronic disease with strong spatiotemporal characteristics. Current deep neural networks have difficulty learning to capture the spatiotemporal information of time series data, which reduces the model's generalization ability and urgently needs to be addressed. Summary of the Invention

[0004] The present application provides a gout risk prediction system and method to solve the problems in the background technology of gout risk prediction models, such as lack of high-dimensional indicators, difficulty in integrating structured and unstructured data, lack of personalized predictions, and poor compatibility with time series data. By pre-training a large model to extract text features and combining numerical features for normalization and feature splicing, the accuracy of gout risk prediction is significantly improved.

[0005] The first embodiment of the present application provides a gout risk prediction system, comprising:

[0006] An acquisition module, configured to acquire follow-up data of the patient, wherein the follow-up data includes structured data and unstructured data;

[0007] A feature extraction module, configured to extract text features of the unstructured data, extract normalized numerical features of the structured data, and extract temporal features from the structured data and the unstructured data;

[0008] A feature splicing module, configured to splice the text features, the normalized numerical features, and the time series features to obtain a comprehensive feature vector;

[0009] A model training module, configured to input the comprehensive feature vector into a preset neural network structure for training to obtain a gout risk prediction model;

[0010] The risk prediction module is used to predict the follow-up data to be predicted based on the gout risk prediction model to obtain a gout risk prediction result.

[0011] According to one embodiment of the present application, the feature extraction module includes:

[0012] A pre-trained model, wherein the pre-trained model is used to extract text features of the unstructured data;

[0013] A normalization unit, configured to perform normalization processing on the structured data to obtain the normalized numerical features;

[0014] A deep learning model is used to extract time series features from the structured data and the unstructured data.

[0015] According to one embodiment of the present application, the gout risk prediction system further includes:

[0016] The data cleaning module is used to clean the follow-up data to obtain cleaned follow-up data.

[0017] According to one embodiment of the present application, the gout risk prediction system further includes:

[0018] The interpolation module is used to perform interpolation processing on the cleaned follow-up data to obtain the interpolated follow-up data.

[0019] According to one embodiment of the present application, the gout risk prediction system further includes:

[0020] The prediction result output module is used to output the gout risk prediction result.

[0021] According to one embodiment of the present application, the gout risk prediction system further includes:

[0022] A gout risk scoring module is used to generate a gout risk score based on the gout risk prediction result.

[0023] According to the gout risk prediction system of the embodiment of the present application, text features are extracted from the patient's unstructured data, normalized numerical features are extracted from the patient's structured data, and time series features are extracted. The above-mentioned extracted features are integrated through the feature splicing module to form a comprehensive feature vector, and the comprehensive feature vector is input into a preset neural network structure for training, thereby obtaining a gout risk prediction model, and the gout risk prediction model is used to analyze new follow-up data to obtain a gout risk prediction result. Thus, the problems of the gout risk prediction model in the background technology, such as the lack of high-dimensional indicators, the difficulty in integrating structured and unstructured data, the lack of personalized predictions, and the poor compatibility with time series data, are solved, and the prediction accuracy is improved.

[0024] A second embodiment of the present application provides a method for predicting the risk of gout, using the above-mentioned gout risk prediction system, wherein the method comprises the following steps:

[0025] Collecting follow-up data of the patient using a collection module, wherein the follow-up data includes structured data and unstructured data;

[0026] Using a feature extraction module to extract text features of the unstructured data, extract normalized numerical features of the structured data, and extract temporal features from the structured data and the unstructured data;

[0027] Using a feature splicing module to splice the text features, the normalized numerical features, and the time series features to obtain a comprehensive feature vector;

[0028] Using a model training module, the comprehensive feature vector is input into a preset neural network structure for training to obtain a gout risk prediction model;

[0029] The risk prediction module is used to predict the follow-up data to be predicted based on the gout risk prediction model to obtain a gout risk prediction result.

[0030] According to one embodiment of the present application, before using a feature extraction module to extract text features of the unstructured data, extract normalized numerical features of the structured data, and extract temporal features from the structured data and the unstructured data, the method further includes:

[0031] The follow-up data is cleaned using a data cleaning module to obtain cleaned follow-up data.

[0032] According to one embodiment of the present application, after obtaining the follow-up data after cleaning, the method further includes:

[0033] An interpolation module is used to perform interpolation processing on the cleaned follow-up data to obtain interpolated follow-up data.

[0034] According to one embodiment of the present application, the method for predicting the risk of gout further includes:

[0035] A gout risk score is generated based on the gout risk prediction result, and the gout risk score is output.

[0036] According to the gout risk prediction method of the embodiment of the present application, text features are extracted from the patient's unstructured data, normalized numerical features are extracted from the patient's structured data, and time series features are extracted. The above-mentioned extracted features are integrated through a feature splicing module to form a comprehensive feature vector, and the comprehensive feature vector is input into a preset neural network structure for training, thereby obtaining a gout risk prediction model, and the gout risk prediction model is used to analyze new follow-up data to obtain a gout risk prediction result. Thus, the problems of the gout risk prediction model in the background technology, such as the lack of high-dimensional indicators, the difficulty in integrating structured and unstructured data, the lack of personalized predictions, and the poor compatibility with time series data, are solved, and the prediction accuracy is improved.

[0037] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0039] Figure 1 Schematic diagram of a system for predicting the risk of gout according to an embodiment of the present application;

[0040] Figure 2 This is a flowchart of the operation of a gout risk prediction model according to one embodiment of the present application;

[0041] Figure 3 A technical roadmap for predicting gout recurrence risk according to one embodiment of the present application is provided;

[0042] Figure 4 This is a flowchart of a method for predicting the risk of gout according to an embodiment of the present application. DETAILED DESCRIPTION

[0043] The following describes in detail embodiments of the present application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0044] Those skilled in the art will understand that in related technologies, prediction models are usually built based on traditional machine learning models, which are simple and easy to implement. However, the constructed prediction models have a single structure and are very sensitive to data dimensions and structure, and have the following significant problems:

[0045] (1) Lack of high-dimensional indicators: Traditional methods usually only consider a few indicators and ignore a large amount of unstructured data (such as patients' follow-up records, medical history descriptions, etc.), resulting in low prediction accuracy.

[0046] (2) Lack of personalization: Traditional methods ignore individual difference information for personalized prediction and are unable to meet the needs of precision medicine.

[0047] (3) Limited applicability of model data types: Traditional methods mainly train structured data. Without preprocessing, unstructured data cannot be used for training.

[0048] Therefore, to address the problem of the lack of high-dimensional indicators in the above-mentioned existing technologies, this application introduces advanced deep learning architectures (such as Transformer, etc.) to efficiently process high-dimensional, multi-source data (including structured data, unstructured data, and time series data). For example, it automatically extracts key features from unstructured text such as patient follow-up records and medical history descriptions, and combines them with traditional structured indicators to significantly improve the model's prediction accuracy and generalization ability.

[0049] Furthermore, in response to the lack of personalization in existing technologies, this application adopts a personalized modeling method to construct an individual gout risk prediction model by integrating multi-dimensional data such as the patient's genetic information, lifestyle habits, medical history, etc. This personalized method can more accurately reflect the patient's unique risk factors and provide reliable support for precision medicine.

[0050] Furthermore, to address the limited applicability of existing model data types and poor compatibility with time series data, this application has designed a multimodal data fusion mechanism that can simultaneously process structured data (such as laboratory test results), unstructured data (such as text), and time series features. By using pre-trained models and adaptive feature extraction techniques, training can be performed directly on multi-source data without complex preprocessing, significantly expanding the model's applicable scenarios.

[0051] The following describes a gout risk prediction system and method according to an embodiment of the present application with reference to the accompanying drawings.

[0052] Specifically, Figure 1 A block diagram of a gout risk prediction system provided in an embodiment of the present application.

[0053] like Figure 1As shown, the gout risk prediction system 10 includes: an acquisition module 100, a feature extraction module 200, a feature splicing module 300, a model training module 400 and a risk prediction module 500.

[0054] Among them, the acquisition module 100 is used to collect the follow-up data of the patient, wherein the follow-up data includes structured data and unstructured data; the feature extraction module 200 is used to extract text features of unstructured data, extract normalized numerical features of structured data, and extract time series features in structured data and unstructured data; the feature splicing module 300 is used to splice text features, normalized numerical features and time series features to obtain a comprehensive feature vector; the model training module 400 is used to input the comprehensive feature vector into a preset neural network structure for training to obtain a gout risk prediction model; the risk prediction module 500 is used to predict the follow-up data to be predicted based on the gout risk prediction model to obtain a gout risk prediction result.

[0055] Specifically, the patient's follow-up data, including structured data and unstructured data, is collected through the acquisition module 100. The structured data may be the patient's laboratory test results, such as blood uric acid levels, renal function indicators, etc.; the structured data may be text data such as the patient's medical history description, follow-up records, etc.

[0056] Furthermore, after collecting the follow-up data of the patient through the acquisition module 100 , the embodiment of the present application performs time alignment on the follow-up data of the patient to unify the timestamp formats of each data source.

[0057] Optionally, in some embodiments, the gout risk prediction system 10 further includes: a data cleaning module, configured to clean the follow-up data to obtain cleaned follow-up data.

[0058] Specifically, the embodiment of the present application can use a data cleaning module to clean the collected follow-up data, remove noise and outliers, and thus ensure the quality of the data.

[0059] Optionally, in some embodiments, the gout risk prediction system 10 further includes: an interpolation module, configured to interpolate the cleaned follow-up data to obtain interpolated follow-up data.

[0060] For example, the embodiment of the present application collects 39,968 gout patients and 67,096 follow-up data, including laboratory test results and medical history descriptions, and uses a data cleaning module to systematically clean and preprocess the collected data. First, through data integrity screening, sample data with missing key information are deleted; secondly, based on feature missing rate analysis, feature variables with too high a proportion of missing values are eliminated. After the above processing, 67,080 valid follow-up data are finally obtained, and 52 feature variables are retained, including 25 text features and 27 numerical features. For numerical data, a reasonable numerical interval is determined based on clinical experience and data distribution, and outliers are eliminated accordingly; then the interpolation module can be used to interpolate the remaining missing values using the K-Nearest Neighbor (KNN) algorithm to ensure the integrity and reliability of the data.

[0061] Optionally, in some embodiments, the feature extraction module 200 includes: a pre-trained model, which is used to extract text features of unstructured data; a normalization unit, which is used to normalize structured data to obtain normalized numerical features; and a deep learning model, which is used to extract time series features from structured data and unstructured data.

[0062] Specifically, the pre-trained large model in the feature extraction module 200 is used to extract text features of unstructured data, normalize structured data (such as laboratory test results), eliminate dimensional differences between different indicators to make them comparable, and extract time series features from structured data and unstructured data.

[0063] Optionally, the pre-trained large model of the embodiment of the present application can be a BERT (Bidirectional Encoder Representations from Transformers, bidirectional transformer model) model. BERT is a very powerful natural language processing (NLP) model specifically designed to understand and process human language. The model understands the relationship between words in a sentence in a "bidirectional" way (using the Transformer architecture and self-attention mechanism) to more accurately capture the meaning of the language. BERT is first pre-trained through large-scale text to learn language rules, and then fine-tuned to adapt to specific tasks. BERT is widely used in scenarios such as text classification, question-answering systems, machine translation, and information extraction, which significantly improves the ability of machines to understand human language.

[0064] It should be noted that the BERT model is a pre-trained large model mentioned above, which is only an example. Other pre-trained models (such as GPT (Generative Pre-trained Transformer), XLNet (Extended Language Model), etc.) can also be used as a replacement, and there is no specific limitation here.

[0065] Therefore, the embodiment of the present application can perform text feature extraction on unstructured text data (such as medical history descriptions) by using a pre-trained BERT model, thereby capturing the contextual information in the text and extracting rich semantic features.

[0066] Furthermore, the embodiments of the present application can also extract time series features in structured data and unstructured data through a preset neural network architecture, such as an RNN (Recurrent Neural Network) architecture, which is not specifically limited here.

[0067] For example, the pre-set neural network architecture in the embodiments of this application is an artificial neural network architecture specifically designed for time-series heterogeneous data. By combining recurrent neural networks (RNNs) and attention mechanisms, it enables efficient processing of heterogeneous data. This architecture can preserve the temporal and heterogeneous nature of data while demonstrating excellent performance in complex data environments, providing a new solution for modeling and analyzing time-series heterogeneous data.

[0068] Specifically, the embodiment of the present application can mine key information from the patient's dynamic medical data that changes over time. These data can include laboratory test results such as blood uric acid levels, renal function indicators, inflammatory factors, and other laboratory tests detected by multiple follow-up visits, as well as clinical events such as medication records and symptom changes. The sliding window statistical method can calculate the mean, variance, change slope and other statistics within a specific time period to reveal the short-term fluctuation characteristics of the indicator; and deep learning models such as recurrent neural networks (such as RNN / LSTM (Long Short-Term Memory, long short-term memory network) / GRU (Gated Recurrent Unit, gated recurrent unit)) can capture more complex long-term dependencies, such as identifying specific patterns of abnormal fluctuations in indicators, predicting future disease development trends, etc. In addition, the time attention mechanism can automatically focus on clinical key time points (such as before and after acute attacks) and enhance the weight of important time period features. These time series features can not only reflect the dynamic evolution of the disease (such as the cumulative effect of gout crystal deposition caused by the continuous increase in blood uric acid), but also capture the treatment response pattern (such as the improvement trend of indicators after the use of uric acid-lowering drugs), providing a more comprehensive and accurate prediction basis for clinical decision-making. Compared with using only static data from a single test, the time series feature extraction in the embodiment of the present application can better grasp the inherent laws and evolution trajectory of disease development, thereby significantly improving the accuracy and clinical practicality of the gout risk prediction model.

[0069] Therefore, according to the deep learning-based time series data feature extraction process of the embodiment of the present application, the dynamic features and long-term dependencies in the time series data can be automatically captured. Through the innovative neural network architecture design, the time series characteristics of the time series data are protected, while the accuracy and efficiency of feature extraction are significantly improved, providing strong technical support for time series data analysis and prediction.

[0070] Furthermore, the feature splicing module 300 is used to splice the extracted text features, normalized numerical features and time series features to form a comprehensive feature vector, thereby effectively integrating structured and unstructured data, making full use of various types of information such as patients' follow-up records and medical history descriptions, and improving the performance of the prediction model.

[0071] Therefore, the feature splicing module 300 of the present embodiment uses efficient fusion technology for heterogeneous data types such as text, tables, and time series, and achieves information integration across data types by designing a unified data mapping framework. This method can preserve the original characteristics of the data while improving the compatibility and consistency of heterogeneous data, providing a solid foundation for data-driven applications in complex scenarios.

[0072] Furthermore, the comprehensive feature vector is input into the fully connected layer for training through the model training module 400. The fully connected layer can capture the complex interactive relationship between features and generate a high-precision gout risk prediction model.

[0073] It's important to note that fully connected layers are a key structure in neural networks, typically located in the last few layers of the network. They integrate all information extracted by previous layers. Each neuron in a fully connected layer is connected to all neurons in the previous layer. Input data is adjusted through weights and biases, and activation functions are combined to introduce nonlinearity, ultimately helping the model complete tasks such as classification and regression. A fully connected layer acts like an "information integrator," synthesizing complex features to provide the model with ultimate decision-making or prediction capabilities.

[0074] In addition, the fully connected layer of the embodiment of the present application can also be replaced by other types of neural network structures (such as convolutional neural networks, recurrent neural networks, etc.) to achieve similar prediction effects, which is not specifically limited here.

[0075] Furthermore, the trained model is used to predict the follow-up data of new patients to obtain gout risk prediction results. The gout risk prediction results can be standardized reports containing risk levels, key influencing factors, and clinical recommendations, which facilitate doctors to formulate personalized treatment plans based on the prediction results.

[0076] Therefore, the embodiments of the present application provide personalized risk prediction results based on individual differences among patients, which helps doctors develop personalized treatment plans.

[0077] In order to help those skilled in the art understand more clearly and intuitively the performance of the gout risk prediction model of the embodiment of the present application in gout-related prediction tasks, the following is an explanation with reference to specific examples.

[0078] For example, Figure 2As shown, the gout risk prediction model of the embodiment of the present application can use a 12-layer BERT (output 768 dimensions) to extract features, superimpose a 1D convolution layer (kernel_size = 3, padding = 1) for local feature enhancement, fuse the temporal feature information extracted by the RNN convolutional network, and fuse multi-dimensional information through a fully connected layer (768+n→512, ReLU activation, Dropout = 0.1). Finally, the prediction result is output by the binary classification layer (512→2). The AdamW optimizer (lr = 2e-5), cross entropy loss function and dynamic learning rate scheduling strategy (ReduceLROnPlateau) are used to train 40 rounds with a batch size of 16. On the test set, the model showed excellent performance: the accuracy reached 96.40%, the F1 score that comprehensively measures the precision and recall rate was 95.69%, and the ROC-AUC value that characterizes the overall discrimination ability of the classifier was as high as 98.79%. The results show that the model has high precision and strong generalization ability in gout-related prediction tasks, especially in class differentiation of complex samples (ROC-AUC is close to 0.99), verifying the effectiveness of the "BERT+1D convolution+RNN+feature enhancement" architecture design, which can meet the needs of high-reliability health management tools in actual scenarios.

[0079] Optionally, in some embodiments, the gout risk prediction system further includes: a prediction result output module for outputting the gout risk prediction result.

[0080] Specifically, the prediction result output module of the embodiment of the present application can output the gout risk prediction results generated by the gout risk prediction model in a user-friendly form, ensuring that doctors, patients or other medical systems can clearly understand and utilize the prediction information.

[0081] Optionally, in some embodiments, the gout risk prediction system further includes: a gout risk scoring module, configured to generate a gout risk score based on the gout risk prediction result.

[0082] Specifically, the embodiments of the present application are used to convert the original predicted value of the gout risk prediction model into a clinically significant gout risk score, and output the gout risk score to facilitate risk stratification and cross-population comparison.

[0083] In order to facilitate those skilled in the art to more clearly and intuitively understand the technical circuit of the gout risk prediction system of the embodiment of the present application, the following is combined with Figure 3 Provide detailed explanation.

[0084] Specifically, if Figure 3 As shown, the gout recurrence risk prediction technology includes:

[0085] First, the system collects raw data from the National Rheumatology Immunology Platform and screens eligible case data based on inclusion and exclusion criteria. During the data preprocessing phase, the collected data undergoes cleaning and baseline statistical analysis to ensure reliable data quality. Subsequently, the processed data is proportionally divided into a training set (80%) and a test set (20%) to prepare for subsequent model training and validation.

[0086] Secondly, advanced natural language processing techniques are used to extract features from text data (such as medical history records and follow-up reports). Numerical data (such as laboratory test indicators) is normalized to eliminate dimensional differences. After feature extraction, the system fuses features from different sources to form a comprehensive feature vector.

[0087] Furthermore, the classifier model is trained based on the fused features, and the model performance is verified and evaluated through the test set to select the optimal model.

[0088] Finally, the optimal model selected by the system is validated on an independent clinical dataset to ensure that the model has reliable generalization ability and clinical practicality.

[0089] According to the gout risk prediction system of the embodiment of the present application, text features are extracted from the patient's unstructured data, normalized numerical features are extracted from the patient's structured data, and time series features are extracted. The above-mentioned extracted features are integrated through the feature splicing module to form a comprehensive feature vector, and the comprehensive feature vector is input into a preset neural network structure for training, thereby obtaining a gout risk prediction model, and the gout risk prediction model is used to analyze new follow-up data to obtain a gout risk prediction result. Thus, the problems of the gout risk prediction model in the background technology, such as the lack of high-dimensional indicators, the difficulty in integrating structured and unstructured data, the lack of personalized predictions, and the poor compatibility with time series data, are solved, and the prediction accuracy is improved.

[0090] Next, a method for predicting the risk of gout according to an embodiment of the present application will be described with reference to the accompanying drawings.

[0091] like Figure 4 As shown, the gout risk prediction method includes the following steps:

[0092] In step S401 , a collection module is used to collect follow-up data of the patient, wherein the follow-up data includes structured data and unstructured data.

[0093] In step S402, a feature extraction module is used to extract text features of unstructured data, extract normalized numerical features of structured data, and extract temporal features from structured data and unstructured data.

[0094] In step S403, a feature concatenation module is used to concatenate text features, normalized numerical features, and time series features to obtain a comprehensive feature vector.

[0095] In step S404, the comprehensive feature vector is input into a preset neural network structure for training using a model training module to obtain a gout risk prediction model.

[0096] In step S405, the risk prediction module is used to predict the follow-up data to be predicted based on the gout risk prediction model to obtain a gout risk prediction result.

[0097] Furthermore, in some embodiments, before using the feature extraction module to extract text features of unstructured data, extracting normalized numerical features of structured data, and extracting time series features from structured data and unstructured data, it also includes: using a data cleaning module to clean the follow-up data to obtain cleaned follow-up data.

[0098] Furthermore, in some embodiments, after obtaining the cleaned follow-up data, the method further includes: performing interpolation processing on the cleaned follow-up data using an interpolation module to obtain the interpolated follow-up data.

[0099] Furthermore, in some embodiments, the gout risk prediction method further includes: generating a gout risk score based on the gout risk prediction result, and outputting the gout risk score.

[0100] According to the gout risk prediction method of the embodiment of the present application, text features are extracted from the patient's unstructured data, normalized numerical features are extracted from the patient's structured data, and time series features are extracted. The above-mentioned extracted features are integrated through a feature splicing module to form a comprehensive feature vector, and the comprehensive feature vector is input into a preset neural network structure for training, thereby obtaining a gout risk prediction model, and the gout risk prediction model is used to analyze new follow-up data to obtain a gout risk prediction result. Thus, the problems of the gout risk prediction model in the background technology, such as the lack of high-dimensional indicators, the difficulty in integrating structured and unstructured data, the lack of personalized predictions, and the poor compatibility with time series data, are solved, and the prediction accuracy is improved.

[0101] It should be noted that the aforementioned explanation of the gout risk prediction system embodiment is also applicable to the gout risk prediction method of this embodiment and will not be repeated here.

[0102] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0103] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. Throughout the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0104] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A gout risk prediction system, characterized in that: include: An acquisition module, configured to acquire follow-up data of the patient, wherein the follow-up data includes structured data and unstructured data; A feature extraction module, configured to extract text features of the unstructured data, extract normalized numerical features of the structured data, and extract temporal features from the structured data and the unstructured data; A feature splicing module, configured to splice the text features, the normalized numerical features, and the time series features to obtain a comprehensive feature vector; A model training module, configured to input the comprehensive feature vector into a preset neural network structure for training to obtain a gout risk prediction model; The risk prediction module is used to predict the follow-up data to be predicted based on the gout risk prediction model to obtain a gout risk prediction result.

2. The system according to claim 1, wherein: The feature extraction module includes: A pre-trained model, wherein the pre-trained model is used to extract text features of the unstructured data; A normalization unit, configured to perform normalization processing on the structured data to obtain the normalized numerical features; A deep learning model is used to extract time series features from the structured data and the unstructured data.

3. The system according to claim 1, wherein: Also includes: The data cleaning module is used to clean the follow-up data to obtain cleaned follow-up data.

4. The system according to claim 3, characterized in that Also includes: The interpolation module is used to perform interpolation processing on the cleaned follow-up data to obtain the interpolated follow-up data.

5. The system according to claim 1, wherein: Also includes: The prediction result output module is used to output the gout risk prediction result.

6. The system according to claim 1, wherein: Also includes: A gout risk scoring module is used to generate a gout risk score based on the gout risk prediction result.

7. A method for predicting the risk of gout, characterized in that: The gout risk prediction system according to any one of claims 1 to 6 is used, wherein the method comprises the following steps: Collecting follow-up data of the patient using a collection module, wherein the follow-up data includes structured data and unstructured data; Using a feature extraction module to extract text features of the unstructured data, extract normalized numerical features of the structured data, and extract temporal features from the structured data and the unstructured data; Using a feature splicing module to splice the text features, the normalized numerical features, and the time series features to obtain a comprehensive feature vector; Using a model training module, the comprehensive feature vector is input into a preset neural network structure for training to obtain a gout risk prediction model; The risk prediction module is used to predict the follow-up data to be predicted based on the gout risk prediction model to obtain a gout risk prediction result.

8. The method according to claim 7, characterized in that Before extracting the text features of the unstructured data, extracting the normalized numerical features of the structured data, and extracting the time series features in the structured data and the unstructured data using the feature extraction module, the method further includes: The follow-up data is cleaned using a data cleaning module to obtain cleaned follow-up data.

9. The method according to claim 8, characterized in that After obtaining the post-cleaning follow-up data, it also includes: An interpolation module is used to perform interpolation processing on the cleaned follow-up data to obtain interpolated follow-up data.

10. The method according to claim 7, characterized in that Also includes: A gout risk score is generated based on the gout risk prediction result, and the gout risk score is output.

Images