Feedback system and method for comprehensive health management of oral cavity of child
By introducing a feedback system and a random forest model into the children's oral health assessment model, combined with incremental pre-training technology, the problem that the model is difficult to select the most predictive features in a multi-character environment is solved, and a more accurate and efficient generation of children's oral health management solutions is achieved.
Patent Information
- Application Number
- CN202510103012.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-22
- Publication Date
- 2025-04-15
AI Technical Summary
Existing children’s oral health assessment models have difficulty selecting the most predictive specific features when processing multiple features, resulting in poor processing of the model on certain features.
By providing a feedback system for all-round oral health management of children, including oral health assessment logic module, evaluation report module, oral health maintenance module, etc., combined with random forest model and incremental pre-training technology, it automatically matches the diagnosis and treatment channels and generates the best health management plan.
The model's processing ability on specific characteristics has been significantly improved, and the generated health management plan is more accurate and conforms to the actual needs of children's oral health, and improves the quality and effectiveness of the health management plan.
Smart Images
Figure CN119833148A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of children's oral health management, and in particular to a feedback system and method for all-round children's oral health management. Background Art
[0002] With the rapid progress of big data analysis technology and artificial intelligence processing technology, children's oral health assessment models can effectively help predict the risk of children suffering from oral diseases such as caries, periodontal disease, and malocclusion in oral health management, so as to take preventive measures in advance, and based on the model evaluation results, personalized oral health management and prevention plans can be formulated for children. However, there are many factors that affect children's oral health. Among the multiple processing features, the children's oral health assessment model faces challenges and it is difficult to select the most predictive feature from multiple processing features. The reason is that most of the current children's oral health assessment models are based on any one of the statistical models, machine learning models, and deep learning models as the data processing base. In order to improve the processing accuracy of the model, a large amount of children's oral health data is often collected to train the model. However, the model often uses a widely used data set for training, lacking in-depth mining and training of specific features, which makes it difficult for the model to select the most predictive specific technical features when processing multiple features. Therefore, the development of a more accurate and efficient children's oral health assessment model has become an important issue that needs to be solved urgently.
[0003] There are some problems in the construction and application of the existing children's oral health assessment model, specifically the following aspects:
[0004] 1. There are various characteristics that affect oral health factors, such as individual characteristics, lifestyle and eating habits characteristics, oral health-related data characteristics, etc. Each characteristic can predict the oral health status of children, but a single database cannot fully cover all characteristics, resulting in poor performance of the model on a certain feature.
[0005] 2. Different features have different degrees of impact on the target's health. A single instruction fine-tuning method is difficult to cover all features that may affect the target's health, and it is impossible to select the most predictive (and most influential) feature from all features. Summary of the invention
[0006] The embodiments of the present application provide a feedback system and method for all-round oral health management of children. The present invention can automatically match corresponding diagnosis and treatment channels for children's oral data information, and generate the best health management plan according to the information diagnosis; and the incremental pre-training injects a large amount of medical knowledge and health management feedback data information, so that the model can perform better when targeting specific features.
[0007] The embodiment of the present application provides a feedback system for comprehensive oral health management of children (S.POM), including an oral health assessment logic module, an oral health assessment learning module, an assessment report module, and an oral health maintenance module;
[0008] The oral health assessment logic module matches the corresponding diagnosis and treatment pathway according to the input data information, further processes the assessment results according to the model, and infers the most predictive feature from multiple features, that is, the feature that has the greatest impact on children's oral health, to provide more comprehensive and accurate information for the issuance of assessment results; the oral health assessment logic module includes the oral hygiene assessment module, the oral function assessment module and the case feature analysis module;
[0009] The assessment report module outputs text information about children’s oral health status;
[0010] The oral health assessment learning module organizes and summarizes the output information of the assessment results and the information output of the expert assessment results to provide data support for the optimization of the oral health assessment logic module;
[0011] The oral health maintenance module executes and maintains according to the optimization or treatment suggestions issued in the evaluation report, and feeds back the maintenance results to the case file records for use as a reference for follow-up comparison and plan adjustment of subsequent long-term maintenance; the oral health maintenance module includes treatment maintenance (S.PODT), professional preventive maintenance (S.PPOC) and family preventive maintenance (S.FPOC).
[0012] The oral hygiene assessment module evaluates the hygiene characteristics of the patient based on clinical examination data (such as tooth cleanliness, caries status, gum health, oral mucosal status, etc.), questionnaire data (such as oral hygiene habits, eating habits, etc.), and oral imaging images (such as X-rays, etc.).
[0013] The oral function assessment module evaluates the functional characteristics of the chewing function (such as chewing efficiency and chewing strength).
[0014] The case characteristic analysis module conducts a comprehensive analysis based on previous case information and current case information to determine the characteristic factors that affect oral health.
[0015] The oral health assessment logic module selects the random forest model as the base model, which performs well when processing various types of features and can provide a ranking of feature importance, which helps to optimize subsequent health management plans.
[0016] The specific process of building the base model is as follows:
[0017] Feature selection: Use correlation analysis to preliminarily screen features; use random forest feature importance scores to further screen key features;
[0018] Model training: input the preprocessed data into the random forest model and adjust the model parameters, such as the number of trees and the depth of the trees, through cross-validation;
[0019] Feature fusion: Random forest combines and filters features through multiple decision trees to automatically complete feature fusion; each tree is trained based on a different feature subset, and the final result is obtained by voting or averaging;
[0020] Recommended health management plans: Based on the importance of the features output by the model, identify the key factors affecting oral health; propose a variety of health management plans for each key factor, such as improving personal hygiene habits, adjusting eating habits, regular oral examinations, etc.; use the model to predict the results, evaluate the effectiveness of each health management plan, and select the best plan.
[0021] An incremental pre-training database is built for the base model. The training task is to let the model predict and generate the most predictive feature in the data, that is, the feature that has the greatest impact on children's oral health, based on the given data information. This method not only enhances the model's ability to process data features, but also enables the continuous learning of medical theoretical knowledge to inject more comprehensive children's oral health knowledge into the model.
[0022] A specific process of database construction is:
[0023] Data collection and integration: Collect data from oral health examinations and questionnaires in hospitals, clinics, and schools, and crawl a large amount of medical theoretical data related to children's oral health from the Internet; clean and unify data from different sources for subsequent analysis;
[0024] Data preprocessing: remove invalid, erroneous or incomplete data records and convert qualitative data into quantitative data;
[0025] Weight distribution of influencing factors: Use linear regression model to analyze data and determine independent variables (influencing factors) and dependent variables (oral health status); use statistical software to fit the linear regression model;
[0026] Data classification and coding: Data is divided into different categories according to the importance of influencing factors, and each category and subcategory is assigned a unique code to facilitate targeted training of the model to improve the accuracy of the model in processing specific features.
[0027] The model can also be further optimized by combining the model's evaluation results and expert evaluation recommendations to facilitate the provision of the best optimization solution later.
[0028] During the implementation and maintenance phase of the health management plan, a feedback reinforcement learning mechanism is introduced to fine-tune the model, further improving the model's health management plan to make it more suitable or more targeted at the characteristics that affect children's oral health. The specific process is as follows:
[0029] Determine mission objectives
[0030] Optimize the generation effect of health management plans so that they can be more suitable or more targeted at the characteristics that affect children's oral health;
[0031] Data Collection and Processing
[0032] Data collection: obtained from patient health management effect feedback and expert feedback data;
[0033] Data processing: cleaning and formatting the collected data for further analysis and model training;
[0034] Feedback signal generation
[0035] Define feedback signals based on quantitative indicators of health management effects, such as treatment success rate, patient satisfaction, relapse rate, etc.;
[0036] Reinforcement learning model training
[0037] Status definition: including information such as the patient’s current oral health status, implemented health management plans, and historical health management results;
[0038] The action space includes all possible health management solutions or adjustments to health management solutions;
[0039] Reward function: The reward function is based on the feedback signal. If the health management effect is improved, a positive reward is given; if the health management effect is poor or leads to negative results, a negative reward is given.
[0040] Policy learning: Use a deep network (DQN) to train the model and learn the strategy of choosing the best action in a given state;
[0041] Real-time optimization
[0042] In the actual health management process, the model conducts online learning based on new patient data and feedback signals and continuously adjusts strategies;
[0043] Evaluation and Monitoring
[0044] Evaluation: Regularly evaluate the effectiveness of the reinforcement learning model in generating health management plans, using statistical tests and clinical indicators;
[0045] Monitoring: Monitor the performance and stability of the model to ensure that it outputs reliable treatment recommendations;
[0046] Feedback loop: Establish a feedback loop to feed new health management results and expert evaluations back to the model for further training and optimization of the model.
[0047] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: the present invention can automatically match the corresponding diagnosis and treatment channels for children's oral data information, and generate the best health management plan according to its information diagnosis; and the incremental pre-training injects a large amount of medical knowledge and treatment feedback data information, so that the model can perform better when targeting specific features; the large model is fine-tuned by combining feedback reinforcement learning and oral health assessment learning, which further improves the model's health management plan to be more in line with or more targeted at the characteristics that affect children's oral health, significantly improving the quality and effectiveness of the health management plan. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 This is the flow chart of the feedback system for comprehensive oral health management of children (S.POM) for this application;
[0049] Figure 2 Construct a flow chart for this application database;
[0050] Figure 3 This is the application flow chart for deploying the base model for this application. DETAILED DESCRIPTION
[0051] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thoroughly understood.
[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0053] Embodiment 1
[0054] See also Figure 1 , a feedback system for children's comprehensive oral health management (S.POM), including oral health assessment logic module, oral health assessment learning module, assessment report module and oral health maintenance module;
[0055] The oral health assessment logic module includes an oral hygiene assessment module, an oral function assessment module, and a case feature analysis module; the oral health assessment logic module matches the corresponding diagnosis and treatment pathway according to the input data information, and performs diagnosis and treatment according to oral hygiene, oral function, or case features;
[0056] The evaluation report module outputs text information about children's oral health status. The report content includes relevant case information, analysis results and recommendations.
[0057] The oral health assessment learning module organizes and summarizes the output information of the assessment results and the information output of the expert assessment results (the doctor gives the corresponding health management plan based on the assessment report), providing data support for the optimization of the oral health assessment logic module;
[0058] The oral health maintenance module includes treatment maintenance (S.PODT), professional preventive maintenance (S.PPOC) and family preventive maintenance (S.FPOC). It is executed and maintained according to the optimization or treatment suggestions issued by the evaluation report, and the maintenance results are fed back to the case file records.
[0059] The oral hygiene assessment module collects information on tooth cleanliness (checking for tartar, plaque and tartar), caries (checking for cavities and whether fillings are intact), gum health (checking for gingivitis, periodontal disease, etc.), tooth alignment and occlusion (checking whether the teeth are aligned and the occlusion is normal), and oral mucosa condition (checking for ulcers, inflammation or other abnormalities).
[0060] The oral function assessment module collects data on chewing function (assessing the efficiency of children in chewing food and checking whether the force of children chewing is appropriate), swallowing function (observing whether the way children swallow food and liquids is normal), pronunciation function (checking whether the children's pronunciation is clear and whether there are pronunciation disorders), oral sensory function (assessing children's ability to perceive the temperature, texture and taste of food), and oral motor function (checking the muscle coordination of children in sucking, chewing, swallowing and other actions).
[0061] The case characteristic analysis module collects oral health history (children's oral hygiene habits, eating habits, previous oral diseases and treatment history), clinical examination data (dental condition: caries, missing teeth, tooth arrangement, etc.; oral mucosal condition: whether there are ulcers, inflammation, etc.; occlusal relationship: whether there are problems such as malocclusion, overbite, deep overjet, etc.), growth and development data (children's age, height, weight and other growth and development indicators, development of facial bones and teeth), and genetic factors (whether there is a genetic tendency for oral diseases in the family).
[0062] Embodiment 2
[0063] See also Figure 2-3 ,For oral health problems, the oral health assessment logic module selects the ,random forest model as the base, which performs well when processing multiple types of ,features, and can provide a ranking of feature importance, which helps in optimizing ,the subsequent health management program.
[0064] The specific process of building the base model is as follows:
[0065] Feature selection: Use correlation analysis to preliminarily screen features (Pearson correlation coefficient can be used for numerical features; chi-square test can be used for categorical features); use random forest feature importance score to further screen key features (for example, set a larger number of trees, such as 100-1000, to ensure model stability; set it to auto or sqrt to increase the randomness of the model; limit the maximum depth of the tree, such as 10-30, to avoid overfitting; after training the random forest model, the feature importance score attribute can be used to obtain the importance score of the feature; set a threshold for the importance score, such as the average importance score, and only retain features with scores above this threshold);
[0066] Model training: Input the preprocessed data into the random forest model, and adjust the model parameters such as the number of trees, the depth of the trees, etc. through cross-validation (define the range of model parameters, which will be used for search in the cross-validation process; use k-fold cross-validation to evaluate the performance of each parameter combination; run GridSearchCV (grid search) to find the best parameter combination. After the cross-validation is completed, the best parameter combination can be obtained and the model can be retrained using the best parameters found);
[0067] Feature fusion: Random forest combines and filters features through multiple decision trees to automatically complete feature fusion; each tree is trained based on a different feature subset, and the final result is obtained by voting or averaging;
[0068] Recommended health management plans: Based on the importance of the features output by the model, identify the key factors affecting oral health; propose a variety of health management plans for each key factor, such as improving personal hygiene habits, adjusting eating habits, regular oral examinations, etc.; use the model to predict the results, evaluate the effectiveness of each health management plan, and select the best plan.
[0069] An incremental pre-training database is built for the base model. The training task is to let the model predict and generate the most predictive feature in the data, that is, the feature that has the greatest impact on children's oral health, based on the given data information. This method not only enhances the model's ability to process data features, but also enables the continuous learning of medical theoretical knowledge to inject more comprehensive children's oral health knowledge into the model.
[0070] A specific process of database construction is:
[0071] Data collection and integration: Collect data from oral health examinations and questionnaires in hospitals, clinics, and schools, and crawl a large amount of medical theoretical data related to children's oral health from the Internet; clean and unify the data from different sources for subsequent analysis (data types include quantitative data such as age, weight, number of caries and qualitative data such as oral hygiene habits, eating habits, etc.);
[0072] Data preprocessing: remove invalid, erroneous or incomplete data records, convert qualitative data into quantitative data, such as converting oral hygiene habits into scores, and standardize the data to compare variables of different magnitudes;
[0073] Weight distribution of influencing factors: Use linear regression model to analyze data, determine independent variables (influencing factors) and dependent variables (oral health status), for example: dependent variable (Y) is oral health score; independent variables (X1, X2, …, Xn) are age, oral hygiene habit score, dietary habit score, socioeconomic status score, etc.; use statistical software (such as Python's scikit-learn library) to fit the linear regression model (the following is the arithmetic expression of linear regression:
[0074] Y = β0 + β1X1 + β2X2 + ... + βnXn + ∈ (where β0 is the intercept term; β1, β2, ..., βn are the coefficients of each variable, representing the weight of their influence on the dependent variable; ∈ is the error term, representing the variation that the model fails to explain); the weight of each factor can be determined through the coefficients of the model. For example, if the model output is as follows:
[0075] Y=2.5+0.8X1-0.5X2+1.2X3
[0076] This means that for every unit increase in age (X1), the oral health score increased by 0.8 points on average; for every unit increase in oral hygiene habit score (X2), the oral health score decreased by 0.5 points on average; for every unit increase in dietary habit score (X3), the oral health score increased by 1.2 points on average);
[0077] Data classification and coding: Data is divided into different categories according to the importance of influencing factors, and each category and subcategory is assigned a unique code to facilitate targeted training of the model to improve the accuracy of the model in processing specific features.
[0078] The constructed database is used to perform incremental pre-training for the base model. The training task allows the model to predict and generate the most predictive feature in the data, that is, the feature that has the greatest impact on children's oral health, based on the given data information. This method not only enhances the model's ability to process data features, but also enables the continuous learning of medical theoretical knowledge to inject more comprehensive children's oral health knowledge into the model.
[0079] Embodiment 3
[0080] See also Figure 3 , combined with the model's evaluation results and expert evaluation suggestions, the model is further optimized to facilitate the provision of the best optimization solution in the later stage. The specific optimization process is as follows:
[0081] Collect and organize data: model evaluation results (collect performance indicators of the model on the test set, such as accuracy), expert evaluation results (organize the health management plans given by doctors based on the model evaluation report, including treatment recommendations, drug prescriptions, etc.);
[0082] Comparative analysis: model performance comparison (comparing the model evaluation results with the previous version of the model to analyze the performance improvement or decline), comparison between the model and expert evaluation results (comparing the model's prediction results with the expert's evaluation results to find the deviation of the model prediction);
[0083] Identify problems and gaps: differences between expert recommendations and the model (identify situations where expert assessments are inconsistent with model predictions and analyze the reasons behind them);
[0084] Formulate optimization strategies: data enhancement (add more relevant features or samples to train the model based on the identified problems), model adjustment (adjust model parameters based on analysis results, such as increasing the number of trees, adjusting the maximum depth of the tree, changing the splitting criteria, etc.), feature selection (perform feature selection or feature transformation to improve model performance);
[0085] Implement optimization plan: Use randomized search (RandomizedSearchCV) to find the optimal parameter adjustment and retrain the model, and use cross-validation to evaluate the stability and generalization ability of the optimized model;
[0086] Feedback loop: The optimized model prediction results are compared with the health management plan recommended by experts.
[0087] Embodiment 4
[0088] See also Figure 3During the implementation and maintenance phase of the health management plan, a feedback reinforcement learning mechanism is introduced to fine-tune the model, further improving the model's health management plan to be more suitable or more targeted at the characteristics that affect children's oral health. The specific process is as follows:
[0089] Determine the mission objectives:
[0090] Optimize the generation of health management plans so that they can be more suitable or more targeted at the characteristics that affect children's oral health.
[0091] Data Collection and Processing:
[0092] Data collection: obtained from patient health management effect feedback and expert feedback data;
[0093] Data processing: Cleaning and formatting the collected data for further analysis and model training.
[0094] Feedback signal generation:
[0095] Define feedback signals based on quantitative indicators of health management effects, such as treatment success rate, recurrence rate, etc.;
[0096]
[0097] Reinforcement learning model training:
[0098] The status definition includes the patient's current oral health status, implemented health management plans, and historical health management results; the action space includes all possible health management plans or health management plan adjustments;
[0099] Reward function, which is based on feedback signals. If the health management effect is improved, a positive reward is given; if the health management effect is poor or leads to negative results, a negative reward is given;
[0100] Policy learning, using a deep network (DQN) to train the model to learn the strategy of choosing the best action in a given state;
[0101] Real-time optimization:
[0102] In the actual treatment process, the model conducts online learning based on new patient data and feedback signals and continuously adjusts its strategy;
[0103] Evaluation and Monitoring:
[0104] Regularly evaluate the effectiveness of the reinforcement learning model in generating health management plans, using statistical tests and clinical indicators;
[0105] Monitor the performance and stability of the model to ensure that it outputs reliable treatment recommendations;
[0106] Establish a feedback loop to feed new treatment results and expert evaluations back to the model for further training and optimization of the model.
[0107] By using the feedback reinforcement learning mechanism, the system can have stronger dynamic adjustment capabilities and significantly improve the quality and effectiveness of health management plans. Specifically, this method enables the model's health management plan to be more suitable or more targeted at the characteristics that affect children's oral health. It is not only based on historical data, but also can generate the latest health management plan based on the latest feedback adjustment.
[0108] Embodiment 5
[0109] The feedback method used for comprehensive oral health management of children is as follows:
[0110] Data input: Children's oral health data, such as children's oral hygiene data or children's oral function data, etc. The oral health assessment logic module matches the corresponding diagnosis and treatment path according to the input data type, and performs diagnosis and treatment according to oral hygiene, oral function or case characteristics;
[0111] Evaluation result output: The evaluation report module outputs text information about children's oral health status. The report content includes relevant case information, analysis results and recommendations;
[0112] Optimization of evaluation results: The oral health evaluation learning module will organize and summarize the output information of the evaluation results and the information output of the expert evaluation results (the doctor will give the corresponding health management plan based on the issuance of the evaluation report) to provide data support for the optimization of the base model;
[0113] Implementation and maintenance of treatment plans: The oral health maintenance module implements and maintains treatment plans through treatment maintenance (S.PODT), professional preventive maintenance (S.PPOC) and family preventive maintenance (S.FPOC) according to the optimization or treatment recommendations issued in the evaluation report, and feeds back the maintenance results to the case file records, and attributes the case file information to the database to provide data support for the optimization of the subsequent model.
[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A feedback system for comprehensive oral health management of children, characterized by: It includes oral health assessment logic module, oral health assessment learning module, assessment report module and oral health maintenance module; The oral health assessment logic module matches the corresponding diagnosis and treatment pathway according to the input data information, and further processes the assessment results according to the model; the oral health assessment logic module includes the oral hygiene assessment module, the oral function assessment module and the case feature analysis module; The assessment report module outputs text information about children’s oral health status; The oral health assessment learning module organizes and summarizes the output information of the assessment results and the information output of the expert assessment results to provide data support for the optimization of the oral health assessment logic module; The oral health maintenance module executes and maintains according to the optimization or treatment suggestions issued in the evaluation report, and feeds back the maintenance results to the case file records for use as a reference for subsequent long-term maintenance tracking and plan adjustment; the oral health maintenance module includes treatment maintenance, professional preventive maintenance and family preventive maintenance.
2. The feedback system for comprehensive oral health management of children according to claim 1, characterized in that: The oral hygiene assessment module collects any one or more of the following information: tooth cleanliness, caries, gum health, tooth alignment and occlusion, oral mucosal condition, daily eating habits, daily oral hygiene, oral condition of family members, and oral care awareness; The oral function assessment module collects any one or more of chewing function, swallowing function, feeding behavior, pronunciation function, oral sensory function, oral motor function, and overall health status; The case characteristic analysis module collects any one or more of oral health history, clinical examination data, growth and development data, genetic factors, and systemic condition data.
3. The feedback system for comprehensive oral health management of children according to claim 1, characterized in that: The oral health assessment logic module selects the random forest model as the base model. The specific process of building the base model is as follows: Feature selection: Use correlation analysis to preliminarily screen features, and use random forest feature importance scores to further screen key features; Model training: input the preprocessed data into the random forest model and adjust the model parameters through cross-validation; Feature fusion: Random forest combines and filters features through multiple decision trees to automatically complete feature fusion; each tree is trained based on a different feature subset, and the final result is obtained by voting or averaging; Recommendation of health management plans: Based on the importance of the features output by the model, identify the key factors affecting oral health, propose multiple health management plans for each key factor, predict the results through the model, evaluate the effect of each health management plan, and select the best plan.
4. The feedback system for comprehensive oral health management of children according to claim 3, characterized in that: An incremental pre-training database is built for the base model. The training task allows the model to predict and generate the most predictive feature in the data based on the given data information, that is, the feature that has a greater impact on children's oral health.
5. The feedback system for comprehensive oral health management of children according to claim 4, characterized in that: A specific process of database construction is: Data collection and integration: Collect data from oral health examinations and questionnaires in hospitals, clinics, and schools, and crawl a large amount of medical theoretical data related to children's oral health from the Internet, clean the data from different sources and unify their formats; Data preprocessing: remove invalid, erroneous or incomplete data records and convert qualitative data into quantitative data; Weight allocation of influencing factors: Use linear regression model to analyze data, determine independent variables and dependent variables, and use statistical software to fit the linear regression model; Data classification and coding: Data is divided into different categories according to the importance of influencing factors, and each category and subcategory is assigned a unique code to facilitate targeted training of the model to improve the accuracy of the model in processing specific features.
6. The feedback system for comprehensive oral health management of children according to claim 1, characterized in that: The model is further optimized based on the evaluation results of the model and the expert evaluation suggestions. The specific optimization process is as follows: Collect and organize data: collect performance indicators of the model on the test set, such as accuracy; organize the health management plans given by doctors based on the model evaluation report; Comparative analysis: Compare the model evaluation results with the previous version of the model to analyze the performance; compare the model's prediction results with the expert's evaluation results to find deviations in the model's predictions; Identify issues and gaps: Identify situations where expert assessments are inconsistent with model predictions; Develop optimization strategies: Based on the identified problems, add more relevant features or samples to train the model, adjust model parameters, and perform feature selection or feature conversion; Implement optimization solutions: Use random search to find the optimal parameter adjustment and retrain the model, and use cross-validation to evaluate the stability and generalization ability of the optimized model; Feedback loop: The optimized model prediction results are compared with the health management plan recommended by experts.
7. The feedback system for comprehensive oral health management of children according to claim 1, characterized in that: The base model introduces a feedback reinforcement learning mechanism to fine-tune the model based on the feedback data during the execution and maintenance of the health management plan. The specific process is as follows: Determine the task objectives: optimize the generation effect of health management programs so that they can be more suitable or more targeted at the characteristics that affect children's oral health; Data collection and processing: Obtain data from patient health management effect feedback and expert feedback; clean and format the collected data; Feedback signal generation: define feedback signals based on quantitative indicators of health management effects; Reinforcement learning model training: The status definition includes the patient’s current oral health status, the implemented oral health management plan, and historical health management effect information; The action space includes all possible health management solutions or adjustments to health management solutions; The reward function is designed based on the feedback signal. If the health management effect is improved, a positive reward is given; if the health management effect is poor or leads to negative results, a negative reward is given. Policy learning: Use deep networks to train models to learn the strategy of choosing the best action in a given state; Real-time optimization: In the actual health management process, the model learns online based on new patient data and feedback signals and continuously adjusts strategies; Evaluation and monitoring: Regularly evaluate the effectiveness of the reinforcement learning model in generating health management solutions, using statistical tests and clinical indicators to measure; monitor the performance and stability of the model to ensure that it outputs reliable treatment recommendations; Feedback loop: Establish a feedback loop to feed new health management results and expert evaluations back to the model for further training and optimization of the model.
8. A method for a feedback system for comprehensive oral health management of children, characterized by: The specific process is as follows: Data input: Children's oral health data, such as children's oral hygiene data or children's oral function data, etc. The oral health assessment logic module matches the corresponding diagnosis and treatment path according to the input data type, and performs diagnosis and treatment according to oral hygiene, oral function or case characteristics; Evaluation result output: The evaluation report module outputs text information about children's oral health status. The report content includes relevant case information, analysis results and recommendations; Optimization of evaluation results: The oral health evaluation learning module will organize and summarize the output information of the evaluation results and the information output of the expert evaluation results to provide data support for the optimization of the base model; Implementation and maintenance of treatment plans: The oral health maintenance module implements and maintains treatment plans through treatment maintenance, professional preventive maintenance, and family preventive maintenance, based on the optimization or treatment recommendations issued in the evaluation report, and feeds back the maintenance results to the case file records, and attributes the case file information to the database to provide data support for the optimization of the subsequent model.
Citation Information
Cited By
Children oral cavity cleaning scoring method and system
CN120199503A