Early tumor screening and predicting system based on information analysis

Abstract

The invention discloses an early tumor screening prediction system based on information analysis, and belongs to the technical field of medical data screening. The invention discloses an early tumor screening prediction system based on information analysis. The system comprises a multi-source data acquisition module, a dynamic model updating module, a reinforcement learning prediction model, a personalized health management scheme generator and a user-friendly display interface. The problems that in the prior art, due to the fact that the model cannot be updated in time, the timeliness and prediction accuracy of the model are low, model optimization cannot be conducted according to user feedback, and a personalized health management scheme cannot be generated according to the tumor risk probability and risk level of the user are solved. According to the method, latest research results are automatically captured from a plurality of databases at regular intervals, a user feedback mechanism is combined, dynamic and real-time updating of the model is realized, and a personalized health management scheme generator is arranged, so that each user can obtain health management suggestions aiming at own conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data screening, and in particular to an early screening and prediction system for tumors based on information analysis. Background Art

[0002] Early cancer screening is a key step to improve cancer cure and survival rates. Currently, many early cancer screening prediction systems based on information analysis have some significant technical defects in practical applications, which are mainly reflected in the following aspects:

[0003] 1. Data lag: Existing prediction models often rely on fixed data sets for training, while the field of medical research is progressing very rapidly. New research results, clinical trial data, and epidemiological survey results continue to emerge, but these new data are usually not incorporated into existing models in a timely manner, resulting in reduced timeliness and accuracy of the models.

[0004] Second, the model cannot be optimized based on user feedback: The existing system usually transmits information in a one-way direction, that is, from the system to the user, and lacks an effective user feedback mechanism. The user's actual health changes and the implementation of health management recommendations cannot be fed back to the system, making it impossible to adjust and optimize the model based on the user's real experience. Due to the lack of user feedback, the system cannot form a closed-loop continuous improvement mechanism, making it difficult for the model's performance to improve over time and as user needs change.

[0005] 3. Unable to generate personalized health management plans based on the user's tumor risk probability and risk level: Existing systems usually provide general health management suggestions, fail to fully consider individual differences and specific risk levels, lack multi-dimensional comprehensive evaluations, and fail to comprehensively consider the user's multi-dimensional health data, thus failing to generate comprehensive and personalized health management plans. Summary of the invention

[0006] The purpose of the present invention is to provide an early screening and prediction system for tumors based on information analysis. By automatically capturing the latest research results from multiple databases on a regular basis and combining it with a user feedback mechanism, the model can be dynamically and real-time updated. A personalized health management program generator is set up so that each user can obtain health management suggestions tailored to their own situation, thereby solving the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] An early screening and prediction system for tumors based on information analysis, comprising:

[0009] Multi-source data acquisition module, used to collect users' personal information from multiple channels, and clean and standardize the data to ensure that data from different sources can be effectively integrated and utilized;

[0010] Dynamic model update module, which regularly introduces the latest medical research results and technological advances to automatically update the prediction model;

[0011] Enhanced learning prediction model, based on reinforcement learning algorithm, continuously optimizes the model according to user feedback to improve prediction accuracy;

[0012] A personalized health management program generator generates specific prevention recommendations based on the user's risk assessment results and provides a link to an online consultation service, allowing users to upload new health data or provide feedback on their actual conditions;

[0013] User-friendly display interface, simple and intuitive user interface, supports instant viewing of the latest assessment report and tracking of historical records. It is an application suitable for smartphones and tablets.

[0014] Preferably, the multi-source data acquisition module includes:

[0015] Data collection unit, which collects users' personal information from multiple channels, including but not limited to electronic medical record systems, wearable monitoring devices, genetic testing reports, and lifestyle questionnaires;

[0016] Data cleaning unit, used to remove noise data and process missing values ​​to ensure data quality and consistency;

[0017] Standardization: unify the data formats from different sources to facilitate subsequent analysis and modeling;

[0018] The privacy protection mechanism encrypts and anonymizes data to ensure the personal privacy of users.

[0019] Preferably, the dynamic model updating module includes:

[0020] Data subscription service, which is used to subscribe to multiple medical research databases and journals, automatically identify and download the latest research results by filtering keywords;

[0021] The model update algorithm uses automated scripts to extract new research results, extract new features that affect predictions, evaluate new and old features, and integrate highly correlated features into existing prediction models; select the corresponding model update strategy based on the amount of data, update frequency, and changes in model structure.

[0022] Verification mechanism: used for internal verification after the model is updated. The model is trained using the training set data, the model's performance indicators are calculated on the verification set, and the model parameter combination with the best performance on the verification set is selected. The selected best model is finally evaluated and the performance of the new model is compared with the old model for final verification.

[0023] Preferably, the reinforcement learning prediction model includes:

[0024] Reinforcement learning framework, using Deep Q-Network reinforcement learning algorithm, continuously adjusting model parameters through reward mechanism;

[0025] User feedback loop, which allows users to submit changes in their actual health status, and adjust model parameters accordingly, forming a closed-loop optimization process;

[0026] Model training platform, used to provide a high-performance platform to support large-scale data training and model optimization.

[0027] Preferably, the personalized health management program generator includes:

[0028] The risk assessment engine assesses the user's tumor risk based on multi-dimensional data;

[0029] Generation of prevention recommendations: generating targeted and personalized health management and prevention recommendations based on assessment results and risk levels, including but not limited to lifestyle recommendations, work and rest adjustments, diet adjustments, medical examination recommendations, living environment improvement recommendations, exercise plans, and mental health support;

[0030] The online consultation interface is used to provide online service links to help users directly contact medical experts in related fields and obtain professional opinions.

[0031] Preferably, the user-friendly display interface includes:

[0032] Visual reports, which provide visual reports in various forms such as charts and graphs, so that users can easily understand their health status and potential risks;

[0033] Mobile applications, apps for smartphones and tablets, allow users to access and manage their health information anytime, anywhere;

[0034] The real-time update function supports instant viewing of the latest assessment report and tracking of changing trends in historical records, helping users to understand changes in their own health in a timely manner while tracking changes in user health behaviors.

[0035] Preferably, the reward mechanism adjusts parameters according to user feedback results, and the feedback results include but are not limited to changes in the user's actual health status and user satisfaction.

[0036] Preferably, the model updating strategy includes incremental learning and complete retraining.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. On the one hand, the present invention ensures that the model is always based on the most cutting-edge scientific research by regularly capturing the latest research results from multiple medical research databases and journals. On the other hand, it collects actual feedback from users, adjusts and optimizes the health management plan based on the feedback, and forms a closed-loop mechanism for continuous improvement. Through the combination of new research results and user feedback, the dynamic update and flexible optimization of the prediction model are achieved to ensure the accuracy and practicality of the model.

[0039] 2. Based on the user's personal health data, the present invention uses a preset health status model to calculate the probability of disease occurrence and risk level, and generates corresponding personalized and diversified health management plans and suggestions to fully cover the user's health needs.

[0040] 3. The present invention improves the accuracy of disease risk prediction by adopting advanced algorithms such as DQN, selects appropriate model update strategies according to the amount of data and changes in model structure, and ensures the adaptability of the model to new data.

[0041] 4. The present invention displays the generated personalized health management plan to the user in a visual form. The user can view his or her own tumor risk assessment and health management plan through the smart terminal and website, which is convenient for the user to understand and execute and improves the interactivity with the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the overall structure of the system of the present invention;

[0043] Figure 2 The present invention is a flow chart for users. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] In order to solve the problem that the existing technology cannot update the model in time, resulting in low timeliness and prediction accuracy of the model, cannot optimize the model according to user feedback, and cannot generate personalized health management plans according to the user's tumor risk probability and risk level, please refer to Figure 1-2 , this embodiment provides the following technical solutions:

[0046] An early screening and prediction system for tumors based on information analysis, comprising:

[0047] Multi-source data acquisition module, used to collect users' personal information from multiple channels, and clean and standardize the data to ensure that data from different sources can be effectively integrated and utilized;

[0048] Dynamic model update module, which regularly introduces the latest medical research results and technological advances to automatically update the prediction model;

[0049] Enhanced learning prediction model, based on reinforcement learning algorithm, continuously optimizes the model according to user feedback to improve prediction accuracy;

[0050] A personalized health management program generator generates specific prevention recommendations based on the user's risk assessment results and provides a link to an online consultation service, allowing users to upload new health data or provide feedback on their actual conditions;

[0051] User-friendly display interface, simple and intuitive user interface, supports instant viewing of the latest assessment report and tracking of historical records. It is an application suitable for smartphones and tablets.

[0052] The multi-source data acquisition module includes:

[0053] Data collection unit, which collects users' personal information from multiple channels, including but not limited to electronic medical record systems, wearable monitoring devices, genetic testing reports, and lifestyle questionnaires;

[0054] Data cleaning unit, used to remove noise data and process missing values ​​to ensure data quality and consistency;

[0055] Standardization: unify the data formats from different sources to facilitate subsequent analysis and modeling;

[0056] The privacy protection mechanism encrypts and anonymizes data to ensure the personal privacy of users.

[0057] The dynamic model update module includes data subscription service, model update algorithm and verification mechanism.

[0058] Data subscription service is used to subscribe to multiple medical research databases and journals, such as PubMed, ClinicalTrials.gov, Nature and Science. The subscription frequency can be set before downloading, and filtering can be performed by setting keywords, including but not limited to: tumors, early cancer screening and risk assessment, etc. The latest research results are automatically identified and downloaded to supplement the model, thereby improving the timeliness and cutting-edge nature of the model.

[0059] Model update algorithm, using automated scripts such as web crawlers to integrate new research results into existing prediction models to ensure that the models are always up to date; using recursive feature elimination and LASSO regression to extract new features that affect predictions, and evaluate the importance of new and old features to select the most relevant features for model training and updating. There are two model update strategies: one is to gradually update model parameters without completely retraining the model. This strategy is suitable for situations where the amount of data is large and frequent updates are required; the other is to retrain the entire model using the merged data set. This strategy is suitable for situations where the amount of data is small or the model structure has undergone major changes.

[0060] Verification mechanism: used to perform internal verification after the model is updated to ensure the accuracy and stability of the new model and to have higher prediction accuracy.

[0061] The specific process of the verification mechanism is as follows: divide the data set into a training set (80%), a validation set (10%), and a test set (10%); use the training set data to train the DQN model, and use different hyperparameter combinations for multiple training; calculate the performance indicators of the model on the validation set, with an average accuracy of 91% and a recall of 84% on the validation set; select the model parameter combination that performs best on the validation set; perform a final evaluation on the selected best model, compare the performance of the new model with the old model, and ensure that the performance of the new model is not lower than that of the old model; use 5-fold cross-validation to ensure the stability and generalization ability of the model; use the new model to collect user feedback and adjust the model accordingly.

[0062] The reinforcement learning prediction model includes a reinforcement learning framework, a user feedback loop, and a model training platform.

[0063] The reinforcement learning framework uses the Deep Q-Network reinforcement learning algorithm to process high-dimensional input data such as multi-source health data, and continuously adjusts the model parameters through the reward mechanism. The specific implementation is as follows:

[0064] Step 1: Initialization

[0065] Define the state space, which includes the user's age, gender, family medical history, living habits, physical examination data, etc.

[0066] Define the action space, where the actions include the generated risk assessment results and the corresponding health management recommendations;

[0067] Define a reward function based on the actual health status change and satisfaction reported by the user. If the user's health status improves or the user is satisfied with the suggestion, a positive reward is given, otherwise a negative reward is given.

[0068] Step 2: Build a DQN model

[0069] Neural network structure:

[0070] Input layer: receiving user status data

[0071] Hidden layer: multi-layer fully connected layer, using ReLU activation function

[0072] Output layer: Outputs the Q value of each possible action

[0073] Experience Replay Buffer:

[0074] Store past (state, action, reward, next state) quadruplets for training models.

[0075] Step 3: Training process

[0076] Initialize the DQN model: Randomly initialize model parameters

[0077] Sampling data: Randomly extract a batch of data from the experience replay buffer for training

[0078] Calculate the target Q value: Use the Bellman equation to calculate the target Q value

[0079] Q target (s,a)=r+γmax a' Q(s',a')

[0080] Where r is the immediate reward, γ is the discount factor, s' is the next state, and a' is the optimal action in the next state;

[0081] Update model parameters: Use gradient descent to minimize the mean square error (MSE) between the predicted Q value and the target Q value.

[0082] Step 4: Explore strategies

[0083] In the early stages of training, random actions are selected with a higher probability (exploration). As training progresses, the probability of random actions is gradually reduced and the probability of selecting the optimal action is increased (exploitation).

[0084] Step 5: User Feedback Loop

[0085] Users upload new health data or provide feedback on their actual conditions through the system;

[0086] Add the new (state, action, reward, next state) quadruple to the experience replay buffer;

[0087] Regularly retrain the DQN model using the latest experience replay data and adjust the model parameters.

[0088] Step 6: Model deployment

[0089] Deploy the trained DQN model to the online system to provide users with real-time risk assessment;

[0090] The model is continuously updated and optimized based on continuous feedback from users.

[0091] The user feedback loop is used to allow users to submit changes in their actual health status, and adjust model parameters accordingly to form a closed-loop optimization process. Through user feedback, problems in the actual application of the model, such as false positives and missed negatives, can be discovered in a timely manner. At the same time, there is the possibility of revealing new features or variables. Introducing these new features into the model can further improve the model's predictive ability.

[0092] Model training platform, used to provide a high-performance platform to support large-scale data training and model optimization.

[0093] The personalized health management program generator includes a risk assessment engine, prevention advice generation, and an online consultation interface.

[0094] The risk assessment engine assesses the user's tumor risk based on multi-dimensional data. Specifically, the user's health status data is input into the model, the user's disease risk is calculated, and logistic regression is used to predict the risk of disease.

[0095] Generation of prevention recommendations: Generate personalized health management and prevention recommendations based on the assessment results, including but not limited to lifestyle recommendations, work and rest adjustments, diet adjustments, medical examination recommendations, living environment improvement recommendations, exercise plans, and mental health support.

[0096] Specifically, according to the user's disease risk assessment results, the corresponding prevention instruction plan is called from the pre-established database. The above database stores preventive measures and suggestions for various diseases. At the same time, the risks are divided into three levels, and different degrees of suggestions and solutions are generated according to different levels. The level division is as follows:

[0097] Low risk: Provides basic lifestyle advice and encourages users to continue to maintain good living habits, such as regular physical examinations, maintaining a healthy diet and exercise habits.

[0098] Moderate risk: In addition to lifestyle recommendations, more specific interventions are provided, such as regular specific check-ups, more professional dietary guidance, and reduced exposure to harmful environments.

[0099] High risk: Provide detailed intervention measures, which may include further medical examinations such as tumor marker testing, early diagnosis and treatment recommendations, more frequent follow-up and tracking, etc.

[0100] The online consultation interface is used to provide online service links to help users directly contact medical experts in related fields and obtain professional opinions.

[0101] The user-friendly display interface includes:

[0102] Visual reports, which provide visual reports in various forms such as charts and graphs, so that users can easily understand their health status and potential risks;

[0103] Mobile applications are applications for smartphones and tablets that allow users to access and manage their health information anytime and anywhere, set health reminders, or record health logs, etc.

[0104] The real-time update function supports instant viewing of the latest assessment report and tracking of changing trends in historical records, helping users to promptly understand changes in their own health. At the same time, it tracks changes in user health behaviors, detects whether users follow recommendations, whether their health status has changed, etc., so as to continuously adjust and optimize health management plans.

[0105] Working principle: Users input personal health status parameters through smart terminals or websites, including basic information, family medical history, life factors and other evaluation items. The system inputs the user's health data into the preset health status model, calculates the user's disease risk probability, and calls the corresponding prevention indication plan from the database according to the risk level to generate corresponding health management suggestions. During this period, users can submit actual feedback, and the system adjusts and optimizes the health management plan based on the feedback to form a closed-loop improvement mechanism.

[0106] During daily operation, the system regularly captures the latest research results from medical research databases and journals, and performs data cleaning, screening and classification. New research results are integrated into existing data sets, new features are extracted, and models are updated and trained to ensure that the system always provides accurate health predictions and management recommendations based on the latest scientific research results.

[0107] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0108] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. An early screening and prediction system for tumors based on information analysis, characterized in that: include: Multi-source data acquisition module, used to collect users' personal information from multiple channels, and clean and standardize the data to ensure that data from different sources can be effectively integrated and utilized; Dynamic model update module, which regularly introduces the latest medical research results and technological advances to automatically update the prediction model; Enhanced learning prediction model, based on reinforcement learning algorithm, continuously optimizes the model according to user feedback to improve prediction accuracy; A personalized health management program generator generates specific prevention recommendations based on the user's risk assessment results and provides a link to an online consultation service, allowing users to upload new health data or provide feedback on their actual conditions; User-friendly display interface, simple and intuitive user interface, supports instant viewing of the latest assessment report and tracking of historical records. It is an application suitable for smartphones and tablets.

2. The early screening and prediction system for tumors based on information analysis according to claim 1, characterized in that: The multi-source data acquisition module includes: Data collection unit, which collects users' personal information from multiple channels, including but not limited to electronic medical record systems, wearable monitoring devices, genetic testing reports, and lifestyle questionnaires; Data cleaning unit, used to remove noise data and process missing values ​​to ensure data quality and consistency; Standardization: unify the data formats from different sources to facilitate subsequent analysis and modeling; The privacy protection mechanism encrypts and anonymizes data to ensure the personal privacy of users.

3. The early screening and prediction system for tumors based on information analysis according to claim 2, characterized in that: The dynamic model updating module comprises: Data subscription service, which is used to subscribe to multiple medical research databases and journals, automatically identify and download the latest research results by filtering keywords; The model update algorithm uses automated scripts to extract new research results, extract new features that affect predictions, evaluate old and new features, and integrate highly correlated features into existing prediction models; select the corresponding model update strategy based on the amount of data, update frequency, and changes in model structure. Verification mechanism: used for internal verification after the model is updated. The model is trained using the training set data, the model's performance indicators are calculated on the verification set, and the model parameter combination with the best performance on the verification set is selected. The selected best model is finally evaluated and the performance of the new model is compared with the old model for final verification.

4. The early screening and prediction system for tumors based on information analysis according to claim 3 is characterized by: The reinforcement learning prediction model includes: Reinforcement learning framework, using Deep Q-Network reinforcement learning algorithm, continuously adjusting model parameters through reward mechanism; User feedback loop, where users submit changes in their actual health status, and the model parameters are adjusted accordingly, forming a closed-loop optimization process; Model training platform, used to provide a high-performance platform to support large-scale data training and model optimization.

5. The early screening and prediction system for tumors based on information analysis according to claim 4 is characterized by: The personalized health management program generator includes: The risk assessment engine assesses the user's tumor risk based on multi-dimensional data; Generation of prevention recommendations: generating targeted and personalized health management and prevention recommendations based on assessment results and risk levels, including but not limited to lifestyle recommendations, work and rest adjustments, diet adjustments, medical examination recommendations, living environment improvement recommendations, exercise plans, and mental health support; The online consultation interface is used to provide online service links to help users directly contact medical experts in related fields and obtain professional opinions.

6. The early screening and prediction system for tumors based on information analysis according to claim 5, characterized in that: The user-friendly display interface includes: Visual reports, which provide visual reports in various forms such as charts and graphs, so that users can easily understand their health status and potential risks; Mobile applications, apps for smartphones and tablets, allow users to access and manage their health information anytime, anywhere; The real-time update function supports instant viewing of the latest assessment report and tracking of changing trends in historical records, helping users to understand changes in their own health in a timely manner while tracking changes in user health behaviors.

7. The early screening and prediction system for tumors based on information analysis according to claim 6, characterized in that: The reward mechanism adjusts parameters according to user feedback results, and the feedback results include but are not limited to changes in the user's actual health status and user satisfaction.

8. The early screening and prediction system for tumors based on information analysis according to claim 7, characterized in that: The model updating strategies include incremental learning and complete retraining.

Images