A method and system for pushing a children's dental caries prevention and treatment plan

Abstract

The present invention discloses a method and system for pushing a child caries prevention and treatment plan, which is applied to the field of data mining technology, collects data, collects oral examination data for a fixed area, and questionnaire data affecting caries; predicts caries risk and scores caries risk, builds a caries risk prediction model based on the questionnaire data and oral examination data, and calculates caries risk prediction scores; pushes plans, determines the risk level based on the caries risk prediction score, and generates a corresponding personalized prevention and treatment plan. The present invention discloses a method and system for pushing a child caries prevention and treatment plan, which can perform risk prediction in combination with a bound questionnaire after the user submits the examination without the need for large instruments, and the evaluation prediction is simple and efficient, and personalized plans are pushed in a targeted manner.

Description

Technical Field

[0001] The present invention relates to the technical field of data management, and more particularly to a method and system for pushing a children's dental caries prevention and treatment plan. Background Art

[0002] Dental caries is listed by the World Health Organization as the third largest chronic disease after cancer and cardiovascular disease.

[0003] Despite the large-scale oral health education and intervention measures, the prevalence of dental caries among children is still on the rise, and the situation of dental caries among children is not optimistic.

[0004] In the existing dental caries prevention model, the "equal treatment" approach is adopted. However, in reality, the distribution of dental caries in children is not even, but there are people with a higher risk of dental caries, that is, children with high dental caries risk. In addition, in clinical practice, most young children who have undergone dental caries treatment have a high chance of recurring dental caries. The high cost of treatment caused by dental caries and its related complications is undoubtedly a heavy burden on families.

[0005] As caries develops towards early diagnosis of caries lesions and replacement of restorative treatment with prevention, caries risk assessment (CRA) for the population and screening of children at high risk of caries has become an internationally accepted method. However, this method is affected by many factors such as cultural background, ethnicity, economic factors and related oral health policies. The same assessment tool has very poor adaptability in different regions and is very inaccurate.

[0006] At present, there is no dental caries risk assessment tool in China. Existing dental caries risk assessment tools abroad have the problems of overestimating the dental caries risk, low sensitivity in predicting dental caries in young children, lack of accuracy and precision, or need to test microbial indicators, salivary flow rate and salivary buffering capacity, which not only increases the cost of assessment, but also requires operators to have a high level of professionalism. At the same time, these assessment tools do not take low-income people into consideration. How to find a tool that can automatically adapt to different regions, provide accurate prevention and treatment measures for children, make more effective use of health care resources at the group level, and reduce the occurrence of dental caries in children in a targeted manner has become an urgent problem to be solved.

[0007] The development of artificial intelligence and big data has provided a new path to solve the above problems. Based on the use of dental caries risk assessment tools, by training on a large number of data on relevant risk factors of high-risk oral patients in different regions, and through intelligent analysis of big data, more accurate risk assessment and prediction can be made for high-risk dental caries patients in different regions.

[0008] At the same time, how to avoid the "one-size-fits-all" policy of decision-making bodies in the allocation of anti-caries resources is also the key to effective caries control. Therefore, this project will establish a big data analysis visualization terminal for children's oral health management, and use big data science to systematically guide the statistical application of decision-making bodies and optimize resource allocation.

[0009] Therefore, how to provide a system that can achieve multi-terminal intercommunication, more accurately predict the risk level of oral caries in children, and conduct personalized oral health interventions according to different caries risk levels to improve the management and treatment level of oral health is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0010] In view of this, the present invention provides a method and system for pushing a children's dental caries prevention and treatment plan to solve the problems in the background technology.

[0011] In order to achieve the above object, the present invention adopts the following technical solution:

[0012] In one aspect, the present invention discloses a method for delivering a dental caries prevention and treatment plan for children, comprising the following specific steps:

[0013] Collect data, collect oral examination data for fixed areas, and questionnaire data affecting caries;

[0014] A dental caries risk prediction score is constructed based on the questionnaire data and the oral examination data to calculate a dental caries risk prediction score;

[0015] The plan is pushed, the risk level is determined according to the dental caries risk prediction score, and the corresponding personalized prevention and treatment plan is retrieved.

[0016] Preferably, in the above-mentioned method for pushing a child dental caries prevention and treatment plan, the collected data are divided into socioeconomic factors, dietary factors, protective factors, bacterial factors and susceptibility factors; wherein the socioeconomic factors include the mother's education level, ethnicity, per capita monthly income of the family, and migrant / left-behind children; the dietary factors include the frequency of taking sugary snacks or drinks between meals every day; the protective factors include brushing teeth with fluoride toothpaste every day, regularly receiving fluoride application by a dentist, and regularly visiting a hospital or clinic for dental treatment; the bacterial factors include the visible plaque index; the susceptibility factors include the number of carious, missing and filled teeth, special health care needs, active chalky spots, enamel development defects, and intraoral appliances.

[0017] Preferably, in the above-mentioned method for pushing a child dental caries prevention and treatment plan, the dental caries risk prediction model is constructed and the dental caries risk prediction score is calculated in the following specific steps:

[0018] Data preprocessing, including non-null check, integrity check, converting non-numeric features into numeric form using one-hot encoding, and Z-score standardization;

[0019] Scores on the questionnaire and oral examination were calculated;

[0020] Introducing multiple multi-head self-attention sub-layers and multi-layer perceptron sub-layers, the input is a questionnaire sequence or an oral examination sequence, and the output is the feature weights under the corresponding sequence, fully capturing each set of potential features of the user;

[0021] The questionnaire scores and oral examination scores are fused and added together at the fully connected layer to obtain the final caries risk prediction score.

[0022] Preferably, in the above method for pushing a child dental caries prevention and treatment plan, the specific steps of calculating the score of the questionnaire are as follows:

[0023] The set of all patients in the database is defined as The set of question factor types in the user questionnaire is expressed as The set of all problems is Q = {q1,q2,...,q n}, which is equivalent to the question ID, and the set of question option attributes defined is For patients The questionnaire sequence of p is expressed as symbol It represents the answer of patient p to question i. Each answer record is expressed as Each record in the patient's questionnaire sequence contains three parts: question identification, question option attributes, and question factor type; the question identification is represented by question ID, recorded as The question option attributes include multiple aspects. The options include single features and compound options that act as an influencing factor, such as whether the child is a left-behind or under-observation child, whether the child is a binary classification option, and subjective options. is the attribute domain of the question options; the factor type to which the question belongs refers to the factor type corresponding to the question that the patient's answer is included in. is the factor type of the interaction in this question;

[0024] The above expression is used to obtain the score based on the questionnaire Where W i is the weight of the factor selected in question i on the risk of dental caries.

[0025] Preferably, in the above method for pushing a child dental caries prevention and treatment plan, the specific steps of calculating the score of the oral examination are as follows:

[0026] During the input process of the oral examination data on the front-end tooth table page, JS is used to realize real-time calculation. Before submitting the data, the conditions of each tooth position and tooth surface of the patient are converted into feature sequence data; the feature sequence data includes the number of carious, missing and filled tooth surfaces, visible plaque index, the presence or absence of active chalky spots, the presence or absence of enamel hypoplasia, and the presence or absence of intraoral appliances;

[0027] Here it is recorded as Then the score formula of patient p based on oral examination is expressed as Where n represents the number of features, represents the ith eigenvalue of patient p, θ i is the weight of the ith eigenvalue for the risk of dental caries, and the weight is calculated by the multi-head attention mechanism during the encoding and decoding process.

[0028] Preferably, in the above method for pushing a child dental caries prevention and treatment plan, the specific steps of calculating the feature weights under the corresponding sequence are as follows:

[0029] Input feature fe i The i-th layer is a multi-head self-attention sub-layer, defining MultiH (i-1) is the output of the i-1th layer, and h is defined as the total number of attention heads. For each attention head in the multi-head self-attention sub-layer, MultiH (i-1) Transformed into Q∈R by linear projection n×d Query vector, K∈R n×d Key vector, V∈R n×d value vector, n is the sequence length, d is the embedding dimension, and the projection transformation formula is (1):

[0030]

[0031] Directly calculate the attention head, and the scaled dot product attention calculation is as shown in formula (2):

[0032]

[0033] W MultiAtt ∈R d×d and W agg ∈R d×d It is a parameter matrix generated according to the input feature sequence, which is used to input the relationship between different feature combination types in the sequence. MultiAtt For example, W agg and W MultiAtt The generation process is similar to that of parameter matrix If the combined feature type corresponding to the query vector Q is A i , the feature type corresponding to the key vector K is A j ,but See formula (3):

[0034]

[0035] The multi-head self-attention mechanism allows the model to focus on multiple dimensions and aggregate the results of multiple attention heads before output; definition represents the i-th self-attention head of the multi-head self-attention sublayer of the l-th layer, h is the number of attention heads, W is the parameter matrix of the l-th layer, and the aggregation process of multi-head attention is expressed by the following formulas (4) and (5):

[0036]

[0037] Here, QW i Q , KW i K , VW i V It is the result of splitting the parameter matrix W into multiple small matrices and applying them to the query Q, key K, and value V respectively; the results calculated by each attention head are concatenated and then multiplied by the output matrix W O Get the final output.

[0038] Preferably, in the above method for pushing a children's dental caries prevention and treatment plan, in order to introduce nonlinearity, a multilayer perceptron is used to model the behavior semantics and perform feature conversion, and each feature type specifies a different multilayer perceptron sublayer; X is defined as the input of the multilayer perceptron, and the calculation process of the multilayer perceptron layer is shown in the following formula (6):

[0039] X′=ReLU(XW1+b1)W2+b2 (6);

[0040] The multi-head self-attention layer and the connection operation multi-layer perceptron layer are connected through layer normalization and residual connection, see formula (7);

[0041] LayerNorm(X+MultiHead(Q,K,V)); LayerNorm(X′+X)(7);

[0042] Here, LayerNorm represents layer normalization, and the + sign represents residual connection.

[0043] Preferably, in the above-mentioned method for pushing a child dental caries prevention and treatment plan, the questionnaire score and the oral examination score are fused in the fully connected layer for final prediction, and the specific steps are as follows: according to the similarity of each group of option records in the fused questionnaire feature sequence, the attention mechanism is used to aggregate similar items, and the multi-layer perceptron performs feature transformation, which is completed in the fully connected layer and output as a single risk score, as shown in the calculation formula (8) completed in the FC:

[0044]

[0045] The multi-layer perceptron sublayer outputs the feature weight W1 based on the entire questionnaire and the weight W2 based on the replacement of dental caries based on the inspection feature sequence. Then the above formula is used for the final weighted summation to obtain ResultScore p as a predicted caries risk value.

[0046] On the other hand, the present invention discloses a system for pushing a child dental caries prevention and treatment plan, applying the above method, including: a server end, a parent end and a doctor end;

[0047] The server end is signal-connected to the parent end and the doctor end respectively;

[0048] The parent inputs the questionnaire data on dental caries, and the system generates a prevention and treatment plan;

[0049] The server side constructs a caries risk prediction model based on the questionnaire data and the oral examination data, calculates a final caries risk prediction score, and pushes the prediction result;

[0050] The doctor inputs oral examination data, questionnaire data, obtains oral examination scores, questionnaire scores, prediction results, generates personalized prevention and treatment plans, and pushes prevention and treatment plans and oral science knowledge to the parent side to help children prevent dental caries in a personalized and accurate manner.

[0051] Preferably, the above-mentioned system for pushing a child caries prevention and treatment plan also includes a big data analysis visualization terminal; the content displayed by the big data analysis visualization terminal includes: the number of people with caries risk levels in different regions, the caries rate, the number of people with caries in each region, the proportion of children with caries, the caries rate of children, the utilization of children's health services, and caries risk prediction.

[0052] Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses a method and system for pushing children's dental caries prevention and treatment plans, which can predict risks in combination with bound questionnaires after the user submits the examination without the need for large instruments. The evaluation prediction is simple and efficient, and personalized solutions are pushed in a targeted manner. Among them, the design of the questionnaire incorporates more potential influencing factors and designs two different sets of risk assessment questionnaires according to age groups, so as to provide more targeted group data for the model as much as possible, improve the accuracy and robustness of the model prediction; through the multi-head self-attention mechanism, the model can focus on different dimensions, thereby improving the expression ability of the model; in addition, a big data analysis visualization end for children's oral health management is established, and big data science is used to systematically guide the statistical application of decision-making agencies and optimize resource allocation. Incorporate oral health institutions in specific areas into the system to achieve resource sharing, evaluate relevant indicators in different regions, and guide decision-making agencies to re-evaluate the current comprehensive prevention and treatment plans for children's oral diseases, such as fluoride anti-caries projects for children aged 3 to 5 years old, and pit and fissure sealing anti-caries projects for school-age children, and optimize the allocation of relevant medical resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0054] Figure 1 The present invention is a flow chart of the method.

[0055] Figure 2 It is a structural block diagram of the present invention.

[0056] Figure 3 (a)-(b) are diagrams of the big data analysis visualization interface of the present invention. DETAILED DESCRIPTION

[0057] 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.

[0058] The embodiment of the present invention discloses a method for pushing a children's dental caries prevention and treatment plan, such as Figure 1 As shown, the specific steps are as follows:

[0059] Collect data, collect oral examination data for fixed areas, and questionnaire data affecting caries;

[0060] A dental caries risk prediction score is constructed based on the questionnaire data and the oral examination data to calculate a dental caries risk prediction score;

[0061] The plan is pushed, the risk level is determined according to the dental caries risk prediction score, and the corresponding personalized prevention and treatment plan is retrieved.

[0062] In order to further optimize the above technical solution, the collected data are divided into socioeconomic factors, dietary factors, protective factors, bacterial factors and susceptibility factors; wherein the socioeconomic factors include the mother's education level, ethnicity, per capita monthly income of the family, and migrant / left-behind children; the dietary factors include the frequency of taking sugary snacks or drinks between meals every day; the protective factors include brushing teeth with fluoride toothpaste every day, regularly receiving fluoride application by a dentist, and regularly visiting a hospital or clinic for dental care; the bacterial factors include the visible plaque index; the susceptibility factors include the number of carious, missing and filled teeth, special health care needs, active chalky spots, enamel development defects, and intraoral appliances. Specifically, a management system and a small program are established to complete data collection. The dental caries risk prediction data set of this embodiment collects the medical information of 3,897 oral patients in the Guangxi Zhuang Autonomous Region through a self-built website and a small program, including 4,280 sets of oral questionnaire information and 5,016 sets of oral examination test results.

[0063] In order to further optimize the above technical solution, build a dental caries risk prediction model, and calculate the dental caries risk prediction score, the specific steps are as follows:

[0064] Data preprocessing, including non-null check, integrity check, converting non-numeric features into numeric form using one-hot encoding, and Z-score standardization;

[0065] Scores on the questionnaire and oral examination were calculated;

[0066] Introducing multiple multi-head self-attention sub-layers and multi-layer perceptron sub-layers, the input is a questionnaire sequence or an oral examination sequence, and the output is the feature weights under the corresponding sequence, fully capturing each set of potential features of the user;

[0067] The questionnaire scores and oral examination scores are fused and added together at the fully connected layer to obtain the final caries risk prediction score.

[0068] Specifically, in data preprocessing, since the prediction model input only accepts a fixed-length sequence length, the system sets a non-empty check on both the server-side and parent-side questionnaire pages. When the questionnaire is submitted on the front end, it is checked. If there are any unanswered questions, a prompt will be given and the question will be redirected to ensure that there are no missing values ​​before submitting the data to the server for processing.

[0069] In addition, when submitting questionnaire data to the server, the system will also perform data integrity verification in the background to prevent illegal operations such as SQL injection and ensure that the database is not affected by dirty data. This ensures that each patient questionnaire submitted to the database is a complete record, avoiding abnormal sequences with missing values ​​in the data. At the same time, in order to facilitate data processing, the platform improved the traditional dental caries risk assessment form at the beginning of its design, using the t-SNE visualization tool for principal component analysis (PCA) and combining indicators such as the Pearson correlation coefficient for auxiliary feature selection.

[0070] Compared with traditional assessment forms, the system incorporates more potential influencing factors and designs two sets of different risk assessment questionnaires according to age groups, as shown in Tables 1 and 2, in order to provide the model with more targeted group data as much as possible and improve the accuracy and robustness of model predictions.

[0071] Table 1

[0072]

[0073] Table 2

[0074]

[0075]

[0076] Based on the modified questionnaire, some feature values, i.e., the options of some questions, appear in non-numerical form, such as categorical variables. This paper uses one-hot encoding to convert these non-numerical features into numerical form. However, due to the possible dimensional differences between different features, it is easy for some features to have too much impact on the model, while other features have less impact. In order to eliminate the impact of this phenomenon, this paper also performs Z-score standardization on the data so that all features are on the same scale.

[0077] In order to further optimize the above technical solution, the specific steps for calculating the score of the questionnaire are as follows:

[0078] The set of all patients in the database is defined as The set of question factor types in the user questionnaire is expressed as The set of all problems is Q = {q1,q2,...,q n}, which is equivalent to the question ID in this case, and the set of question option attributes defined is For patients The questionnaire sequence of p is expressed as symbol It represents the answer of patient p to question i. Each answer record is expressed as Each record in the patient's questionnaire sequence contains three parts: question identification, question option attributes, and question factor type; the question identification is represented by question ID, recorded as The attributes of question options include multiple aspects. Taking the dental caries risk questionnaire for children over 6 years old as an example, the options include single features and composite options that act as an influencing factor, such as whether the child is a left-behind or under-observation child, whether the child is a binary option, subjective options, etc. It is the attribute domain of the question options; the factor type to which the question belongs refers to the factor type corresponding to the patient's answer to the question. Taking the questionnaire for people over 6 years old as an example, the factor types included include socioeconomic factors, dietary factors, protective factors, bacteriological factors, susceptibility factors, and other factors. is the factor type of the question. (It should be noted that the factor type not included in the questionnaire has a weight of 0) The above expression can be used to derive the score based on the questionnaire. Where W i is the weight of the factor selected in question i on the risk of dental caries.

[0079] In order to further optimize the above technical solution, the specific steps for calculating the score of the oral examination are as follows:

[0080] During the input process of the oral examination data on the front-end tooth table page, JS is used to realize real-time calculation. Before submitting the data, the conditions of each tooth position and tooth surface of the patient are converted into feature sequence data; the feature sequence data includes the number of carious, missing and filled tooth surfaces, visible plaque index, the presence or absence of active chalky spots, the presence or absence of enamel hypoplasia, and the presence or absence of intraoral appliances;

[0081] Here it is recorded as Then the score formula of patient p based on oral examination is expressed as Where n represents the number of features, represents the ith eigenvalue of patient p, θ i is the weight of the ith eigenvalue for the risk of dental caries, and the weight is calculated by the multi-head attention mechanism during the encoding and decoding process.

[0082] Specifically, this step uses a multi-head self-attention mechanism to capture the complex relationship between questions and options. The question ID is fused with each option attribute as a query vector and an independent attention head is used to calculate its weight to extract feature weights. After obtaining the weights of the question and options, these weights are then used as the input of a multi-layer perceptron, which further extracts the nonlinear relationship between features, thereby better understanding the relationship between feature vectors and the risk of caries and obtaining the final feature weights. This module is composed of multiple multi-head self-attention sublayers and multi-layer perceptron sublayers. The input of the module is a questionnaire sequence or an oral examination sequence, and the output is the feature weight under the corresponding sequence. In order to fully capture each set of potential features of the user, an independent self-attention module is constructed for each set of answer records. Each set of self-attention modules is only responsible for processing one type of attribute information, thereby accurately obtaining the feature weights hidden in the answer attribute group.

[0083] Furthermore, the specific steps for calculating the feature weights of the corresponding sequence are as follows:

[0084] Input feature fe i The i-th layer is a multi-head self-attention sub-layer, defining MultiH (i-1) is the output of the i-1th layer, and h is defined as the total number of attention heads. For each attention head in the multi-head self-attention sub-layer, MultiH (i-1) Transformed into Q∈R by linear projection n×d Query vector, K∈R n×d Key vector, V∈R n×d value vector, n is the sequence length, d is the embedding dimension, and the projection transformation formula is (1):

[0085]

[0086] Since the model is based on the prediction model of questionnaires and inspections, the input sequence does not need to distinguish the relationship between the previous and the next sequence, so there is no need to consider the position encoding step, and the attention head calculation is directly performed. The scaled dot product attention calculation is as shown in formula (2):

[0087]

[0088] W MultiAtt ∈R d×d and W agg ∈R d×d It is a parameter matrix generated according to the input feature sequence, which is used to input the relationship between different feature combination types in the sequence. MultiAtt For example, W agg and W MultiAtt The generation process is similar to that of parameter matrix If the combined feature type corresponding to the query vector Q is A i , the feature type corresponding to the key vector K is A j ,but See formula (3):

[0089]

[0090] The multi-head self-attention mechanism allows the model to focus on multiple dimensions and aggregate the results of multiple attention heads before output; definition represents the i-th self-attention head of the multi-head self-attention sublayer of the l-th layer, h is the number of attention heads, W is the parameter matrix of the l-th layer, and the aggregation process of multi-head attention is expressed by the following formulas (4) and (5):

[0091]

[0092] Here, QW i Q , KW i K , VW i V It is the result of splitting the parameter matrix W into multiple small matrices and applying them to the query Q, key K, and value V respectively; the results calculated by each attention head are concatenated and then multiplied by the output matrix W O The final output is obtained. This method enables the model to focus on different dimensions, thereby improving the model's expressiveness.

[0093] In order to further optimize the above technical solution and introduce nonlinearity, a multilayer perceptron is used to model the behavior semantics and perform feature conversion. Each feature type specifies a different multilayer perceptron sublayer. Define X as the input of the multilayer perceptron. The calculation process of the multilayer perceptron layer is shown in the following formula (6):

[0094] X′=ReLU(XW1+b1)W2+b2 (6);

[0095] The multi-head self-attention layer and the connection operation multi-layer perceptron layer are connected through layer normalization and residual connection, see formula (7);

[0096] LayerNorm(X+MultiHead(Q,K,V)); LayerNorm(X′+X)(7);

[0097] Here, LayerNorm represents layer normalization, and the + sign represents residual connection, which ensures the flow of information. This structure can help the model better capture the relationship between features and improve the performance of the model.

[0098] In order to further optimize the above technical solution, this step is to predict the scores of the questionnaire and oral examination of the specified patient. It is necessary to fuse the questionnaire score and the oral examination score in the fully connected layer for the final prediction, rather than predicting the user's caries risk based solely on the questionnaire score. According to the similarity of each group of option records in the fused questionnaire feature sequence, the attention mechanism is used to aggregate items with high similarity, and the multi-layer perceptron performs feature transformation. The caries risk prediction module is completed in the fully connected layer and the output is a single risk score. The calculation formula (8) completed in the FC is shown as follows:

[0099]

[0100] The method for calculating the weight of the oral examination score is the same as above, and it is also necessary to consider the feature type relationship between the query vector Q and the key vector K, see formula (2). The caries risk prediction also has a multi-behavior multi-layer perceptron layer, see formula (6). The multi-layer perceptron layer outputs the feature weight W1 based on the entire questionnaire and the weight W2 based on the replacement of caries based on the inspection feature sequence. Then the above formula is used for the final weighted summation to obtain ResultScore p as a predicted caries risk value.

[0101] Export the trained prediction model to PMML format, and then use the PMML parsing library in Java to load and use the model for prediction. By adding a service for predicting caries risk assessment in the business layer service class based on the springboot backend framework, the caries risk prediction score is divided into two categories: if it is less than 4 points, the system considers the patient to be at low risk, if it is greater than 6 points, it is at high risk, otherwise it is at medium risk. Since the service only needs to be called for prediction when the system submits an oral examination or modifies an oral examination, there is no need to provide a separate interface to the outside world.

[0102] Finally, it is necessary to regularly monitor the accuracy and performance of the prediction results, and optimize and improve the model based on the monitoring results to improve the accuracy and effectiveness of caries risk prediction.

[0103] On the other hand, another embodiment of the present invention discloses a system for pushing a child dental caries prevention and treatment plan, such as Figure 2 As shown, the above method is applied, including: a server side, a parent side and a doctor side;

[0104] The server end is signal-connected to the parent end and the doctor end respectively;

[0105] The parent inputs the questionnaire data on dental caries, and the system generates a prevention and treatment plan;

[0106] The server side constructs a caries risk prediction model based on the questionnaire data and the oral examination data, calculates a final caries risk prediction score, and pushes the prediction result;

[0107] The doctor inputs oral examination data, questionnaire data, obtains oral examination scores, questionnaire scores, prediction results, generates personalized prevention and treatment plans, and pushes prevention and treatment plans and oral science knowledge to the parent side to help children prevent dental caries in a personalized and accurate manner.

[0108] The embodiment of the present invention includes a WeChat applet (parent end), a hospital dental caries grading management system (doctor end), a backend server and a WeChat platform interface service;

[0109] The WeChat applet is connected to the WeChat platform; the WeChat applet obtains the open_id of the WeChat platform and then connects to the backend server; the hospital dental caries grading management system is connected to the backend server.

[0110] In addition, WeChat applet contains view layer (View) and logic layer (App Service):

[0111] The view layer includes: the home page unit, which is used to display information such as risk level;

[0112] Personal center unit: add child information and display user information;

[0113] Questionnaire unit: used to display two sets of questionnaires for different age ranges for submission;

[0114] Prevention and treatment plan unit: After the user submits the questionnaire, the prevention and treatment plan interface will be displayed, including the precautions for children to brush their teeth;

[0115] User authorization unit: used to obtain the authorized avatar and nickname from the user to register or log in to the mini program;

[0116] Dental Encyclopedia Column Unit: used to display columns containing dental encyclopedia articles. When clicked, the article list of the column is displayed;

[0117] Dental Encyclopedia Article Unit: A page used to display information about dental-related articles.

[0118] Moreover, the hospital dental caries grading management system includes a view layer (View) and a logic layer (App Service). The visit management page includes: questionnaire result unit, risk assessment unit, prevention and treatment plan unit, etc.

[0119] Questionnaire result unit: used to view the questionnaire information that the user has filled out in the past. When clicking on Add Assessment, the user can fill out the questionnaire;

[0120] Risk Assessment Unit: The doctor will assess the child's oral condition and fill out an oral examination form for risk assessment;

[0121] Risk assessment submission unit: When the user clicks the submit button, the server interface is called to transmit the data of the oral examination form to the cloud database and perform risk assessment calculations;

[0122] Prevention and treatment plan unit: Doctors can provide precautions and oral articles to users of the mini program to help children with personalized and precise dental prevention.

[0123] The WeChat applet in this embodiment is easy for users to use. They can use it by scanning the code to enter the WeChat applet; the hospital dental caries grading management system is also easy for doctors to operate, which can speed up the doctor's examination, reduce the doctor's examination time, and increase the frequency of doctor's examination.

[0124] Furthermore, doctors formulate and implement personalized prevention and treatment plans based on the dental caries risk level:

[0125] Through system calculation and processing, the system will formulate personalized management plans for children based on the children's age and the level of dental caries risk assessment, and implement more accurate prevention and treatment.

[0126] The personalized comprehensive prevention and treatment plan includes the main aspects of caries repair, diet intervention, toothbrushing intervention, pit and fissure sealing, regular review time, etc. According to the different caries risk levels of children in three different age groups (1-2 years old, 3-5 years old, and over 6 years old), the system invention will automatically formulate and implement different comprehensive caries prevention and treatment intervention plans for children.

[0127] According to the caries risk level, a personalized prevention and treatment plan will be formulated and implemented. The system will automatically select a comprehensive prevention and treatment plan based on the child's age and caries risk assessment level. The comprehensive prevention and treatment plan will include caries repair, dietary intervention, toothbrushing intervention, pit and fissure sealing, regular review time, etc. According to the different caries risks of children in three different age groups (1-2 years old, 3-5 years old, and over 6 years old), the comprehensive caries prevention and treatment intervention plan formulated and implemented is shown in Table 3-5. All intervention measures will be recorded in the computer system.

[0128] Table 3 Prevention and treatment plans for children aged 1 to 2 years with different caries risks

[0129]

[0130]

[0131] Note: Professional advice on dietary intervention for children aged 1 to 2 years old is mainly to correct bad habits such as the frequency of bottle feeding and drinking milk at night.

[0132] Table 4 Prevention and treatment plans for children aged 3 to 5 years with different caries risks

[0133]

[0134]

[0135] Note: Professional recommendations for dietary intervention for children aged 3 to 5 years old are mainly to correct bad habits such as the frequency of eating sweets, eating sweets before bedtime / drinking milk at night, etc.

[0136] Table 5 Prevention and treatment plans for children over 56 years old with different caries risks

[0137]

[0138] Note: Professional advice for dietary intervention for children over 6 years old is mainly to correct the frequency of eating sweets, especially the intake of sweetened drinks.

[0139] This system will record all intervention measures issued by the system, and these personalized comprehensive prevention and treatment plans will be automatically pushed to the parents by the system.

[0140] After obtaining the different dental caries conditions of each child from the system, the system pushes the child's corresponding examination results and personalized treatment plans to the parents, so that the parents can understand the child's oral health-related information in real time and further carry out personalized comprehensive prevention and treatment.

[0141] The doctor's end will also display the child's bad eating and brushing habits data, which can provide guided education to parents and children one by one. It will also push relevant oral health education videos to children of different ages. The visual materials can be pushed by the doctor in the subsequent tracking service according to the child's oral health problems.

[0142] In order to further optimize the above technical solution, a big data analysis visualization terminal is also included; the content displayed by the big data analysis visualization terminal includes: the number of people with caries risk levels in different regions, the caries rate, the number of people with caries in each region, the proportion of children with caries, the caries rate of children, the utilization of children's health services, and caries risk prediction.

[0143] The visualization end of big data analysis connects to decision-making bodies, which helps them to fully grasp the overall situation of children suffering from dental caries in the region and other places, the details of various indicators, as well as risk assessment and risk prediction. Through multi-data fusion and big data analysis and mining, relevant assessment indicators in different regions (such as fluoride coating for dental caries prevention for children aged 3 to 5, pit and fissure sealing for dental caries prevention for school-age children, etc.) are evaluated, so as to provide decision-making bodies with decision-making basis in the comprehensive prevention and treatment plan for children's oral diseases, guide resource allocation, optimize service layout, improve service quality, and achieve precise and intelligent prevention and treatment.

[0144] Specifically, we will establish a visualization terminal for big data analysis of children's oral health management, take oral public health projects as the starting point, strengthen the medical quality control and supervision system of the entire diagnosis and treatment process of oral medical institutions at all levels; ensure the quality of oral diagnosis and treatment in each medical institution, and improve the technical level and service capabilities of primary oral medical institutions. By evaluating relevant indicators in different regions, we will provide data support for the comprehensive prevention and treatment plan for children's oral diseases (such as fluoride coating for caries prevention projects for children aged 3 to 5, pit and fissure sealing for caries prevention projects for school-age children, etc.), and optimize the allocation of relevant medical resources.

[0145] The functions of the big data analysis visualization end include but are not limited to:

[0146] Regional difference analysis: Through data analysis and geographic visualization, the oral health status of children and the distribution of medical resources in different regions are displayed to support regional differentiated decision-making.

[0147] Real-time data update: Use real-time data collection technology to ensure the timeliness and accuracy of data and provide the latest data support for decision-making.

[0148] Intelligent prediction and early warning: Through machine learning and deep learning models, predict future trends and potential risks, and provide early warnings and response measures.

[0149] Interactive visualization: Design interactive dashboards and maps to support user-defined queries and analysis and improve data utilization efficiency.

[0150] Multi-dimensional data fusion: Integrate multi-source data to provide comprehensive and accurate children's oral health information and provide comprehensive data support for decision-making.

[0151] As shown in Figure 3(a)-(b), they are the interface renderings of the big data analysis visualization terminal, where the number of medical institutions in each region: provides the government with an understanding of the distribution of oral medical resources, helps evaluate service coverage and identify areas with service shortages, thereby guiding resource allocation, optimizing service layout, and improving service quality in the government decision-making process.

[0152] Health project implementation: Track the implementation of oral health projects, including project coverage, implementation results and resource utilization. It can provide a comprehensive understanding of oral health projects, help the government evaluate the effectiveness of current policies, identify problems and deficiencies in project implementation, and provide a basis for policy formulation and resource allocation.

[0153] Utilization of children's health services: Evaluate the effectiveness of oral health service policy implementation, guide resource allocation and formulate improvement measures to improve children's health level.

[0154] The proportion of children suffering from dental caries and the dental caries rate in a specific region, taking Guangxi Zhuang Autonomous Region as an example: This helps the government understand the dental caries situation of children of different age groups and genders, and provides a basis for the formulation of targeted health policies.

[0155] Economic factor indicators: Data on indicators such as mother's education level, ethnicity, average monthly household income, and migrant / left-behind children will help the government to fully understand the distribution and trend of dental caries among different socioeconomic groups and assess the impact of socioeconomic factors on oral health.

[0156] The establishment of a big data analysis visualization terminal will adjust the allocation of oral health resources in a targeted manner based on statistical data, tilt it towards the socioeconomically disadvantaged groups with a high incidence of dental caries, ensure that resources are preferentially covered to the areas and populations most in need, and improve social equity.

[0157] Based on data from different socioeconomic groups, the government can formulate precise dental caries prevention and control strategies, such as providing free oral health services to low-income families, strengthening oral health education for migrant / left-behind children, etc., to improve prevention and control effects and resource utilization efficiency.

[0158] Protective factors (protective factors include brushing teeth with fluoride toothpaste every day, receiving regular fluoride application from a dentist, and visiting a hospital or clinic regularly for dental care): Collecting data on protective factors such as brushing teeth with fluoride toothpaste every day, receiving regular fluoride application from a dentist, and visiting a hospital or clinic regularly for dental care will help the government evaluate the effectiveness of the implementation of oral health promotion policies and understand the popularity and changing trends of the public's oral health behaviors. The government can reasonably adjust the allocation of oral health service resources based on statistical data, and focus on oral health promotion and prevention. For example, increase publicity efforts for groups with poor oral health behaviors and provide more oral health education resources. Based on statistical data, the government can carry out more targeted oral health publicity and education activities to popularize correct oral health knowledge and behaviors to the public and promote the improvement of oral health levels.

[0159] Dietary factors (frequency of consuming sugary snacks or drinks between meals every day): The frequency of consuming sugary snacks or drinks is closely related to the occurrence of dental caries. A high-sugar diet can easily lead to oral acidification and promote the formation of dental caries. Therefore, by observing the frequency of consuming sugary snacks or drinks between meals every day, the risk of dental caries for individuals or groups can be predicted more accurately. The government can formulate relevant policies, such as restricting the advertising of sugary snacks or drinks, strengthening food labeling, restricting canteens from entering campuses, and other measures to reduce the intake of high-sugar diets, thereby reducing the incidence of dental caries.

[0160] Caries susceptibility factors: The government can collect and analyze the statistical data of caries susceptibility factors to monitor the prevalence and trend of caries in different populations, so as to timely identify high-incidence areas and high-risk populations of caries. The government can use the statistical data of these indicators to evaluate the effectiveness of caries prevention and control work, understand the progress and achievements of caries prevention and control work, and provide reference for future work.

[0161] Visible plaque index: By collecting data such as the visible plaque index, the government can evaluate the effectiveness of oral health education activities and understand the changes in the public's oral hygiene awareness and behavior. The government can use statistical data to carry out more extensive and in-depth oral health education and publicity activities, raise the public's awareness of the importance of oral hygiene, and guide the public to develop good oral health behavior habits.

[0162] In summary, the big data analysis and visualization end of children's oral health management will provide strong data support and decision-making basis for public health departments and medical institutions, optimize the allocation of children's oral health prevention and treatment resources, and improve children's oral health level.

[0163] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0164] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for promoting a child dental caries prevention and treatment plan, characterized in that: The specific steps include: Collect data, collect oral examination data for fixed areas, and questionnaire data affecting caries; A dental caries risk prediction score is constructed based on the questionnaire data and the oral examination data to calculate a dental caries risk prediction score. The specific steps are as follows: Data preprocessing, non-empty check, integrity check, converting non-numerical features into numerical form using one-hot encoding, and Z-score standardization; Scores on the questionnaire and oral examination were calculated; Introducing multiple multi-head self-attention sub-layers and multi-layer perceptron sub-layers, the input is a questionnaire sequence or an oral examination sequence, and the output is the feature weights under the corresponding sequence, fully capturing each set of potential features of the user; The questionnaire scores and oral examination scores are integrated at the fully connected layer and accumulated to obtain the final caries risk prediction score; The specific steps for calculating the questionnaire score are as follows: The set of all patients in the database is defined as P = {p1, p2, ... p |n| }; The set of question factor types in the user questionnaire is represented by F = {f1,f2,...,f n }; the set of all problems is Q = {q1,q2,...,q n }, which is equivalent to the question ID, and the set of question option attributes defined is A = {a1, a2, ..., a n }; For patient p∈P, p’s questionnaire sequence is expressed as symbol It represents the answer of patient p to question i. Each answer record is expressed as Each record in the patient's questionnaire sequence contains three parts: question identification, question option attributes, and question factor type; the question identification is represented by question ID, denoted as Q i p ; Question option attributes include multiple aspects, options include single features and composite options that together serve as an influencing factor, such as judging whether it is a left-behind or under-observation child category, whether it is a binary classification option, and a subjective option; A i p is the attribute domain of the question options; The factor type to which the question belongs refers to the factor type corresponding to the question in which the patient's answer is counted. i p is the factor type of the interaction in this question; The above expression is used to obtain the score based on the questionnaire Where W i is the weight of the factor selected in question i on the risk of dental caries; The plan is pushed, the risk level is determined according to the dental caries risk prediction score, and a corresponding personalized prevention and treatment plan is generated.

2. A method for pushing a child dental caries prevention and treatment plan according to claim 1, characterized in that: The collected data were divided into socioeconomic factors, dietary factors, protective factors, bacterial factors and susceptibility factors; among them, the socioeconomic factors included the mother's education level, ethnicity, average monthly family income, and migrant / left-behind children; the dietary factors included the frequency of consuming sugary snacks or drinks between meals every day; the protective factors included brushing teeth with fluoride toothpaste every day, regular fluoride application by a dentist, and regular visits to hospitals or clinics for dental care; the bacterial factors included the visible plaque index; the susceptibility factors included the number of carious, missing, and filled teeth, special health care needs, active chalky spots, enamel development defects, and intraoral appliances.

3. A method for pushing a child dental caries prevention and treatment plan according to claim 1, characterized in that: The specific steps for calculating the score of the oral examination are as follows: The oral examination data is input into the front-end tooth table page using the front-end scripting language to achieve real-time calculations, and the patient's tooth positions and tooth surfaces are converted into feature sequence data before submitting the data; The characteristic sequence data include the number of carious, missing and filled tooth surfaces, the visible plaque index, the presence or absence of active chalky plaque, the presence or absence of enamel hypoplasia, and the presence or absence of intraoral appliances; Here it is recorded as Then the score formula of patient p based on oral examination is expressed as Where n represents the number of features, represents the ith eigenvalue of patient p, θ i is the weight of the ith eigenvalue for the risk of dental caries, and the weight is calculated by the multi-head attention mechanism during the encoding and decoding process.

4. A method for pushing a child dental caries prevention and treatment plan according to claim 1, characterized in that: The specific steps for calculating the feature weights under the corresponding sequence are as follows: Input feature fe i The i-th layer is a multi-head self-attention sub-layer, defining MultiH (i-1) is the output of the i-1th layer, and h is defined as the total number of attention heads. For each attention head in the multi-head self-attention sub-layer, MultiH (i-1) Transformed into Q∈R by linear projection n×d Query vector, K∈R n×d Key vector, V∈R n×d value vector, n is the sequence length, d is the embedding dimension, and the projection transformation formula is (1): Directly calculate the attention head, and the scaled dot product attention calculation is as shown in formula (2): W MultiAtt ∈R d×d and W agg ∈R d×d It is a parameter matrix generated according to the input feature sequence, which is used to input the relationship between different feature combination types in the sequence. MultiAtt For example, W agg and W MultiAtt The generation process is similar to that of parameter matrix If the combined feature type corresponding to the query vector Q is A i , the feature type corresponding to the key vector K is A j ,but See formula (3): The multi-head self-attention mechanism allows the model to focus on multiple dimensions and aggregate the results of multiple attention heads before output; definition represents the i-th self-attention head of the multi-head self-attention sublayer of the l-th layer, h is the number of attention heads, W is the parameter matrix of the l-th layer, and the aggregation process of multi-head attention is expressed by the following formulas (4) and (5): Here, QW i Q , KW i K , VW i V It is the result of splitting the parameter matrix W into multiple small matrices and applying them to the query Q, key K, and value V respectively; the results calculated by each attention head are concatenated and then multiplied by the output matrix W O Get the final output.

5. A method for pushing a child dental caries prevention and treatment plan according to claim 4, characterized in that: In order to introduce nonlinearity, a multilayer perceptron is used to model the behavior semantics and perform feature conversion. Each feature type specifies a different multilayer perceptron sublayer. Define X as the input of the multilayer perceptron, and the calculation process of the multilayer perceptron layer is shown in the following formula (6): X′=ReLU(XW1+b1)W2+b2 (6); The multi-head self-attention layer and the connection operation multi-layer perceptron layer are connected through layer normalization and residual connection, see formula (7); LayerNorm(X+MultiHead(Q,K,V)); LayerNorm(X′+X)(7); Here, LayerNorm represents layer normalization, and the + sign represents residual connection.

6. The method for pushing a child dental caries prevention and treatment plan according to claim 1, characterized in that: The questionnaire score and oral examination score are fused in the fully connected layer for final prediction. The specific steps are as follows: Based on the similarity of each set of option records in the fused questionnaire feature sequence, the attention mechanism is used to aggregate similar items, and the multi-layer perceptron performs feature transformation, which is completed in the fully connected layer and output as a single risk score. The calculation formula (8) completed in the FC is shown as follows: The multi-layer perceptron sublayer outputs the feature weight W1 based on the entire questionnaire and the weight W2 for dental caries based on the inspection feature sequence, and then the above formula is used for the final weighted summation to obtain ResultScore p as a predicted caries risk value.

7. A system for pushing a child dental caries prevention and treatment plan, using a method for pushing a child dental caries prevention and treatment plan as described in any one of claims 1 to 6, characterized in that: include: Server side, parent side and doctor side; The server end is signal-connected to the parent end and the doctor end respectively; The parent inputs the questionnaire data on dental caries and checks the prevention and treatment plan generated by the system; The server side constructs a caries risk prediction model based on the questionnaire data and the oral examination data, calculates a final caries risk prediction score, and pushes the prediction result; Calculate the dental caries risk prediction score. The specific steps are as follows: Data preprocessing, including non-null check, integrity check, converting non-numeric features into numeric form using one-hot encoding, and Z-score standardization; Scores on the questionnaire and oral examination were calculated; Introducing multiple multi-head self-attention sub-layers and multi-layer perceptron sub-layers, the input is a questionnaire sequence or an oral examination sequence, and the output is the feature weights under the corresponding sequence, fully capturing each set of potential features of the user; The questionnaire scores and oral examination scores are integrated at the fully connected layer and accumulated to obtain the final caries risk prediction score; The specific steps for calculating the questionnaire score are as follows: The set of all patients in the database is defined as P = {p1, p2, ... p |n| }; The set of question factor types in the user questionnaire is represented by F = {f1,f2,...,f n }; the set of all problems is Q = {q1,q2,...,q n }, which is equivalent to the question ID, and the set of question option attributes defined is A = {a1, a2, ..., a n }; For patient p∈P, p’s questionnaire sequence is expressed as symbol It represents the answer of patient p to question i. Each answer record is expressed as Each record in the patient's questionnaire sequence contains three parts: question identification, question option attributes, and question factor type; the question identification is represented by question ID, denoted as Q i p ; Question option attributes include multiple aspects, options include single features and composite options that together serve as an influencing factor, such as judging whether it is a left-behind or under-observation child category, whether it is a binary classification option, and a subjective option; A i p is the attribute domain of the question options; The factor type to which the question belongs refers to the factor type corresponding to the question in which the patient's answer is counted. i p is the factor type of the interaction in this question; The above expression is used to obtain the score based on the questionnaire Where W i is the weight of the factor selected in question i on the risk of dental caries; The doctor inputs oral examination data, questionnaire data, obtains oral examination scores, questionnaire scores, prediction results, generates personalized prevention and treatment plans, and pushes prevention and treatment plans and oral science knowledge to the parent side to help children prevent dental caries in a personalized and accurate manner.

8. A system for pushing a child dental caries prevention and treatment plan according to claim 7, characterized in that: It also includes a big data analysis visualization terminal; the content displayed by the big data analysis visualization terminal includes: the number of people with caries risk levels in different regions, the caries rate, the number of people with caries in each region, the proportion of children with caries, the caries rate of children, the utilization of children's health services, and caries risk prediction.

Images