Visual assistant for clinical risk prediction model
Through the visual assistant of clinical risk prediction model, using natural language processing and word vector modeling, the problem of traditional machine learning models not being reliable enough in clinical data processing is solved, personalized risk assessment and automated risk prediction are realized, and the accuracy and efficiency of clinical decision-making are improved.
Patent Information
- Application Number
- CN202510271402.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-08
- Publication Date
- 2025-06-13
AI Technical Summary
When traditional machine learning models process incomplete and unbalanced clinical data, their prediction results are not reliable enough, and the medical decision-making process requires a lot of manual intervention, which is susceptible to subjective biases, resulting in high cost and low efficiency.
Provides a visual assistant for clinical risk prediction models, which helps doctors understand the prediction basis of the model through natural language processing, word vector model, pre-trained risk prediction model and visualization module, and provides personalized risk assessment, automated risk prediction and explanation.
Strengthen doctors’ trust in models, reduce work burden, improve the accuracy of clinical decision-making, and reduce costs and time waste.
Smart Images
Figure CN120148825A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent medical technology, and more specifically, it relates to a visualization assistant for a clinical risk prediction model. Background Art
[0002] The models obtained after training traditional machine learning algorithms on incomplete and unbalanced clinical data sets often cannot achieve satisfactory prediction results. Therefore, there are many models based on machine learning training in current clinical practice.
[0003] However, in the process of model construction, the actual problems of clinical data are not considered, making the prediction functions of many diseases not very reliable. At the same time, many decision-making processes in the medical disease management system require human intervention to complete. This not only wastes a large amount of time and resources, but is also easily affected by personal subjective biases, increasing costs and reducing work efficiency. Summary of the Invention
[0004] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide a visualization assistant for a clinical risk prediction model. Through the visualization tool, doctors can better understand the prediction basis of the model, enhance their trust in the model, be able to provide personalized risk assessments according to the specific conditions of patients, automate risk prediction and interpretation, reduce the work burden of doctors, and doctors can more intuitively understand the results of the risk prediction model, so as to make more accurate clinical decisions.
[0005] To achieve the above purpose, the present invention provides the following technical solutions:
[0006] A clinical risk prediction model visualization assistant, including a processor; a collection module, a preprocessing module, a natural language processing module, a processing module, a comparison module, an optimization module, and a visualization module are set in the processor; the natural language processing module is used to input data into the processing module and output the model to the visualization module; the processing module consists of a word vector model, a pre-trained risk prediction model, and model selection and tuning; the visualization module consists of result presentation and interpretation, as well as visualization and analysis tools; the training sample set is obtained from the database through the collection module and the preprocessing module, and these data are preprocessed. The natural language processing module analyzes the collected text data to better understand the user's status and needs; the preprocessed clinical data is input into a classifier for classification prediction to obtain the predicted classification result; then the comparison module compares the result of the classification prediction with the actual result, and optimizes the classifier according to the comparison result to obtain a disease risk prediction model; then the word vector model is used to convert the text data into a numerical representation, and at the same time, the pre-trained risk prediction model is trained according to the classification prediction result data and the actual result data in the disease risk prediction model to predict the disease risk. Model selection and tuning select the best model and perform parameter tuning through cross-validation and grid search techniques, and then visualize the results generated by the trained model to intuitively observe and understand the output of the model.
[0007] The present invention is further configured such that: the collection module is used to obtain a training sample set in the database, where each sample includes the clinical data and labels of a patient; the database is used for the clinical trial data information of the same type of disease and the pre-clinical trial data information of this sample.
[0008] The present invention is further configured such that: the preprocessing module is used to preprocess the clinical data; the preprocessing includes processing the missing values in the clinical data; the processing of the missing values is to fill them using machine learning algorithms.
[0009] The present invention is further configured such that: the pre-trained risk prediction model is used to show the contribution degree of each data feature to the prediction result, and the SHapley Additive exPlanations (SHAP) value is used to explain the influence of each feature on the individual prediction result. At the same time, for the prediction result of a single patient, local interpretability analysis is provided.
[0010] The present invention is further configured such that: the visualization module is used for risk score distribution, feature contribution graph, decision path visualization, and time series analysis.
[0011] The present invention is further configured such that: the displayed risk score distribution is used for the distribution of risk scores in the patient population to help doctors understand the relative risks of patients; the feature contribution graph is used to display the contributions of various features to the prediction results using bar charts, waterfall charts, etc.; for the tree model, the decision path visualization can display the decision path to help doctors understand the decision-making process of the model; the time series analysis is used to display the changing trend of risks over time.
[0012] The present invention is further configured such that: it further includes a confidentiality module; when processing sensitive clinical data, the confidentiality module ensures data privacy and security.
[0013] The present invention is further configured such that: it further includes an update module; the update module is used to periodically update the data information in the database to maintain its prediction performance.
[0014] The advantages of the present invention are:
[0015] 1. Through the visualization tool of the present invention, doctors can better understand the prediction basis of the model, enhance their trust in the model, be able to provide personalized risk assessments according to the specific conditions of patients, automate risk prediction and interpretation, reduce the workload of doctors, and doctors can more intuitively understand the results of the risk prediction model, thereby making more accurate clinical decisions.
[0016] 2. The present invention shows the distribution of risk scores in the patient population according to the risk score distribution to help doctors understand the relative risks of patients, and shows the contributions of various features to the prediction results by using bar charts, waterfall charts, etc. For the tree model, the decision path can be displayed to help doctors understand the decision-making process of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 It is a framework diagram of a clinical risk prediction model visualization assistant system of the present invention.
[0018] Figure 2 It is a block diagram of the composition of the processing module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0019] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the drawings and in combination with the embodiments.
[0020] It should be pointed out that unless otherwise specified, all technical and scientific terms used in this application have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs.
[0021] Example 1, please refer to Figure 1-2 , the present invention provides the following technical solutions:
[0022] A clinical risk prediction model visualization assistant, specifically, includes a processor; a collection module, a preprocessing module, a natural language processing module, a processing module, a comparison module, an optimization module, and a visualization module are set in the processor.
[0023] The natural language processing module is used to input data into the processing module and output the model to the visualization module; the processing module consists of a word vector model, a pre-trained risk prediction model, and model selection and tuning; the visualization module consists of result presentation and explanation, and visualization and analysis tools.
[0024] The training sample set is obtained from the database through the collection module and the preprocessing module, and these data are preprocessed. The collected text data is analyzed through the natural language processing module to better understand the user's status and needs; the preprocessed clinical data is input into the classifier for classification prediction to obtain the predicted classification result; then the comparison module compares the classification prediction result with the actual result, and optimizes the classifier according to the comparison result to obtain the disease risk prediction model; then the word vector model is used to convert the text data into a numerical representation, and at the same time the pre-trained risk prediction model is trained according to the classification prediction result data and the actual result data in the disease risk prediction model to predict the disease risk. Model selection and tuning select the best model and perform parameter tuning through cross-validation and grid search techniques, and then visualize the results generated by the trained model for intuitive observation and understanding of the model output.
[0025] The working principle of the first embodiment: Through the visualization tool, doctors can better understand the prediction basis of the model, enhance their trust in the model, be able to provide personalized risk assessments according to the specific situation of patients, automate risk prediction and explanation, reduce the doctor's workload, and doctors can more intuitively understand the results of the risk prediction model, thus making more accurate clinical decisions; according to the risk score distribution, display the distribution of risk scores in the patient population to help doctors understand the relative risks of patients. By using bar charts, waterfall charts, etc., display the contribution of each feature to the prediction result. For tree models, the decision path can be displayed to help doctors understand the decision-making process of the model.
[0026] Embodiment Two
[0027] Please refer to Figure 1 , this Embodiment Two makes the following improvements on the basis of Embodiment One. Specifically, the collection module is used to obtain the training sample set in the database, where each sample includes the clinical data and labels of the patient; the database is used for the clinical trial data information of the same type of disease and the pre-clinical trial data information of this sample.
[0028] The preprocessing module is used to preprocess clinical data; the preprocessing includes handling missing values in the clinical data; the handling of missing values is to fill them using machine learning algorithms.
[0029] The working principle of the second embodiment: Considering the possible correlations between medical data indicators, such as body weight and blood lipids, but conventional filling methods only consider factors such as data mean, median, and mode, without considering the associations between indicators. Based on machine learning algorithms to fill data, making better use of the correlations between indicators, so that the filled data is closer to the real data; when using machine learning algorithms to fill data, further consider different data types such as discrete data and continuous data, and further optimize the missing value filling method; in the process of optimizing model parameters, creatively fuse the Jaya algorithm and the dragonfly algorithm. First, use the Jaya algorithm for preliminary global search, and combine the dragonfly algorithm for local search of the optimal solution, which improves the convergence accuracy of the algorithm and the accuracy of model parameter identification.
[0030] Embodiment Three
[0031] Please refer to Figure 1-2 , on the basis of the second embodiment, the third embodiment makes the following improvements. Specifically, the pre-trained risk prediction model is used to show the contribution degree of each data feature to the prediction result. By using SHAP (SHapley Additive exPlanations) values to explain the influence of each feature on the individual prediction result, at the same time, for the prediction result of a single patient, provide local interpretability analysis.
[0032] The visualization module is used for risk score distribution, feature contribution graph, decision path visualization, and time series analysis.
[0033] Showing the risk score distribution is for the distribution of risk scores in the patient population, helping doctors understand the relative risks of patients; the feature contribution graph is used to show the contribution of each feature to the prediction result using bar charts, waterfall charts, etc.; for the tree model, the decision path visualization can show the decision path, helping doctors understand the decision-making process of the model; the time series analysis is used to show the changing trend of risk over time.
[0034] It also includes a confidentiality module; when the confidentiality module processes sensitive clinical data, it ensures data privacy and security.
[0035] It also includes an update module; the update module is used to regularly update the data information in the database to maintain its prediction performance.
[0036] The working principle of the third embodiment:
[0037] SHAP (SHapley Additive exPlanations) provides a unified approach to explain the output of any machine learning model. SHAP connects game theory with local explanations, combines several previous methods, and represents the only possible consistent and locally accurate additive feature attribution method according to expectations.
[0038] The security module includes means for connecting the security module to a terminal, which can be, for example, a mobile station, a portable computer, a telephone, or any corresponding terminal. The security module is a general-purpose module for connecting a telecommunication network and a telecommunication terminal, enabling the execution of required encryption operations, constituting applications that require a high degree of data security; for encrypting electronic data transmissions that act through the security module, decrypting encrypted information, and performing electronic signatures. The encryption device includes a processor for performing encryption, decryption, and electronic signature. In addition, the device includes a memory connected to the processor for storing required keys and parameters.
[0039] This application can help doctors gain a deeper understanding of patients' conditions, etc. Through visualization tools, doctors can better understand the basis for the model's predictions, enhance their trust in the model, be able to provide personalized risk assessments according to the specific conditions of patients, automate risk prediction and explanation, reduce the workload of doctors, and doctors can more intuitively understand the results of the risk prediction model, thereby making more accurate clinical decisions, which can greatly reduce the cost and time of human intervention and improve work efficiency.
[0040] Obviously, the above-described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.
[0041] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0042] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.
[0043] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
[0044] The above is only the preferred implementation manner of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.
Claims
1. A clinical risk prediction model visualization assistant, comprising a processor; characterized in that: The processor is provided with an acquisition module, a preprocessing module, a natural language processing module, a processing module, a comparison module, an optimization module and a visualization module; The natural language processing module is used to input data into the processing module and output the model to the visualization module; the processing module is composed of a word vector model, a pre-trained risk prediction model, and model selection and tuning; the visualization module is composed of result presentation and interpretation as well as visualization and analysis tools; The acquisition module and preprocessing module are used to obtain training sample sets from the database, and these data are preprocessed. The collected text data is analyzed through the natural language processing module to better understand the user's status and needs; the preprocessed clinical data is input into the classifier for classification prediction to obtain the predicted classification results; Then, the classification prediction results are compared with the actual results through the comparison module, and the classifier is optimized according to the comparison results to obtain the disease risk prediction model; The word vector model is then used to convert the text data into numerical representation. At the same time, the pre-trained risk prediction model is trained based on the classified prediction result data and actual result data in the disease risk prediction model to predict disease risk. Model selection and tuning are carried out through cross-validation and grid search technology to select the best model and perform parameter tuning. The results generated by the trained model are then visualized to intuitively observe and understand the output of the model.
2. A clinical risk prediction model visualization assistant according to claim 1, characterized in that: The acquisition module is used to obtain a training sample set in a database, wherein each sample includes the patient's clinical data and labels; the database is used for clinical trial data information of similar diseases and preclinical trial data information of the sample.
3. A clinical risk prediction model visualization assistant according to claim 2, characterized in that: The preprocessing module is used to preprocess the clinical data; the preprocessing includes performing data missing value processing on missing values existing in the clinical data; the data missing value processing is filled by using a machine learning algorithm.
4. A clinical risk prediction model visualization assistant according to claim 3, characterized in that: The pre-trained risk prediction model is used to show the contribution of each data feature to the prediction result. The SHAP (SHapleyAdditive exPlanations) value is used to explain the impact of each feature on the individual prediction result. At the same time, a local interpretability analysis is provided for the prediction result of a single patient.
5. A clinical risk prediction model visualization assistant according to claim 4, characterized in that: The visualization module is used for risk score distribution, feature contribution graphs, decision path visualization, and time series analysis.
6. A clinical risk prediction model visualization assistant according to claim 5, characterized in that: The risk score distribution display is used to display the distribution of risk scores in a patient group, helping doctors understand the relative risk of patients; the feature contribution graph is used to display the contribution of each feature to the prediction result using a bar graph, waterfall graph, etc.; The decision path visualization for the tree model can display the decision path to help doctors understand the decision-making process of the model; the time series analysis is used to display the changing trend of risk over time.
7. A clinical risk prediction model visualization assistant according to claim 6, characterized in that: It also includes a confidentiality module; when the confidentiality module processes sensitive clinical data, it ensures data privacy and security.
8. A clinical risk prediction model visualization assistant according to claim 7, characterized in that: It also includes an updating module; the updating module is used to regularly update the data information in the database to maintain its prediction performance.