Deep learning-based periodontal atrophy prediction method and system

Abstract

The invention discloses a periodontal atrophy prediction method and system based on deep learning, and relates to the technical field of data processing, and the system comprises a periodontal recognition end, a grading detection end and a prediction correction end. The periodontal recognition end is used for acquiring periodontal data of a patient through image recognition and judging whether recognition artifacts appear or not in real time in combination with multi-angle and multi-position centralized recognition; the grading detection end is used for setting periodontal atrophy detection schemes of different patients according to periodontal bones and predicting whether periodontal atrophy prediction can be executed or not in time; and the prediction correction end is used for automatically generating a periodontal atrophy early warning solution according to the predicted periodontal atrophy prediction result. According to the periodontal atrophy prediction method and system based on deep learning, multi-angle and multi-position concentrated recognition is combined, whether recognition artifacts appear or not is judged, overlapping artifacts existing in periodontal data are avoided, the situation that the detection angle is abnormal when a periodontal detection scheme is executed is avoided, and the precision, safety and reliability of periodontal atrophy prediction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a periodontal atrophy prediction method and system based on deep learning. BACKGROUND

[0002] In simple terms, periodontal atrophy prediction based on deep learning is an advanced artificial intelligence technology that automatically analyzes oral medical data (such as X-rays) to discover early signs that are difficult to detect with the naked eye and predict the risk of future periodontal atrophy. Periodontal atrophy prediction based on deep learning essentially brings periodontal disease diagnosis and management into a new era of intelligence, precision, and foresight. It is no longer just a description of the current state of the bone around the teeth, but a powerful "early warning system" that uses AI to gain insight into the future. It is one of the key directions for the future development of oral medicine.

[0003] Currently, there are some deficiencies in periodontal atrophy prediction: 1. Due to the different shapes, gaps, and other data of patients' own teeth, the periodontal atrophy prediction model is not fixed when predicting periodontal atrophy, and overlapping artifacts may occur when identifying periodontal atrophy, resulting in large errors in periodontal atrophy prediction; 2. When obtaining periodontal data for different patients based on imaging features, due to the different tooth capture positions, tooth arrangement shapes, and tooth adjacent gaps of different patients, it is not possible to perform hierarchical detection of teeth in different regions during periodontal detection, which may result in angle abnormalities during periodontal detection, affecting the accuracy of tooth data recognition during periodontal atrophy prediction; 3. Existing periodontal prediction models are generally applicable to a wide range of people and are difficult to cover all rare cases or atypical atrophy patterns, resulting in periodontal atrophy prediction based on imaging features (such as trabecular bone structure, alveolar crest height, etc.) being unable to determine whether the periodontal atrophy model is abnormal in a timely manner based on the predicted periodontal atrophy prediction results, and being unable to dynamically generate a periodontal atrophy warning solution, resulting in the inability to timely warn and solve periodontal atrophy.

[0004] Therefore, the present application proposes a periodontal atrophy prediction method and system based on deep learning to solve the above problems. SUMMARY

[0005] The main purpose of the present application is to provide a periodontal atrophy prediction method and system based on deep learning to solve the problems raised in the background.

[0006] To achieve the above purpose, the technical scheme adopted by the present application is as follows: a periodontal atrophy prediction method and system based on deep learning, comprising a periodontal recognition end, a hierarchical detection end, and a prediction correction end, wherein the periodontal recognition end, the hierarchical detection end, and the prediction correction end are provided with a prediction warning module. The periodontal recognition terminal is used to set up a periodontal atrophy prediction model and obtain the patient's periodontal data through image recognition. Combined with multi-angle and multi-position centralized recognition, it can determine in real time whether recognition artifacts occur. The graded detection terminal is used to set different periodontal atrophy detection plans for different patients based on periodontal bone, and to perform graded detection of periodontal atrophy detection areas to predict in a timely manner whether periodontal atrophy prediction is feasible. The prediction and correction end is used to automatically generate a periodontal atrophy early warning solution based on the predicted periodontal atrophy results, or to correct the periodontal atrophy detection angle based on the detection angle deviation results, and to track the correction results in real time to determine whether the secondary prediction periodontal atrophy detection solution is feasible. The prediction alarm module is used to report to the system and issue an abnormal warning reminder for periodontal atrophy prediction when it determines that there are overlapping artifacts in image recognition, that periodontal atrophy prediction is not feasible, or that the detection angle is deviated.

[0007] The periodontal recognition terminal includes a data acquisition module, a periodontal recognition module, and an artifact detection module; The data acquisition module includes a patient data acquisition unit and a prediction model setting unit; The patient data acquisition unit is used to collect basic patient data through a data acquisition device. The basic data includes age, gender, height, physical indicators, medical history, hygiene habits, dental diagnosis history, tooth shape, tooth arrangement, number of teeth, lifestyle habits, and favorite foods, and is recorded in real time through a data recorder. The prediction model setting unit is used to set the periodontal atrophy prediction model. The periodontal atrophy prediction model is set according to the basic data of different patients. The periodontal atrophy prediction model is implemented by one of the linear model, logistic regression model and machine learning model. Different periodontal atrophy prediction models are corresponding to patients of different ages. The periodontal atrophy prediction model sets the standard periodontal data corresponding to the basic data. The standard periodontal data includes the standard distance from the cementoenamel junction to the gingival margin, the gap between adjacent teeth, the amount of dental plaque accumulation, the height of the alveolar crest, the height of the alveolar bone, and the root morphology.

[0008] The periodontal recognition module includes a multi-dimensional recognition unit and a multi-angle, multi-position unit; The multidimensional recognition unit is used to achieve multidimensional recognition of different teeth through imaging equipment to obtain periodontal data. The imaging equipment includes an intraoral scanner, cone-beam CT, two-dimensional X-ray imaging instrument and three-dimensional imaging instrument. The periodontal data includes the distance from the cementoenamel junction to the gingival margin, the gap between adjacent teeth, the amount of dental plaque accumulation, the height of the alveolar crest, the height of the alveolar bone, and the morphology of the tooth root. The multi-angle, multi-position unit is used to adjust the angle and position of the imaging device in real time during the multi-dimensional recognition process, so as to realize the recognition of periodontal data from multiple angles and positions.

[0009] The artifact detection module includes an artifact detection unit, a data judgment unit, and a data recording unit; The artifact recognition unit is used to calculate the weighted sum of the deviations of the periodontal data features of the current time, current angle, and current region from the standard periodontal data features; The data judgment unit is used to set a deviation threshold T. If the value is greater than T, it is determined that an artifact has occurred during periodontal data recognition, and the system will issue a voice alarm to remind the user. If not, it means that no artifact has occurred during periodontal data recognition. The data recording unit is used to record the identification results of periodontal data features in real time through a data recorder.

[0010] The graded detection terminal includes a detection scheme module, a graded detection module, and a judgment and early warning module; The detection scheme module includes a periodontal bone unit and a detection scheme unit; The periodontal bone unit is used to obtain periodontal bone parameters of different patients through image recognition equipment. The periodontal bone parameters include the distance from the cementoenamel junction to the bottom of the periodontal pocket, the distance from the gingival margin to the bottom of the periodontal pocket, the distance from the cementoenamel junction to the gingival margin, bone dehiscence, bone fenestration, the vertical distance from the cementoenamel junction to the alveolar ridge crest, bone resorption morphology, bone density, trabecular bone structure, alveolar bone thickness, and root bifurcation morphology. Standard periodontal bone parameters corresponding to different patients are set. The detection scheme unit is used to set corresponding periodontal atrophy detection schemes according to the periodontal bone parameters of different patients. The periodontal atrophy detection scheme includes setting the detection angle, detection position and detection area of ​​the periodontium of different patients, and judging whether the periodontal bone parameters of different patients are abnormal by combining the standard deviation of the periodontal bone parameter measurement values ​​of different patients from the average value of the standard periodontal bone parameters of healthy people of the same age.

[0011] The graded detection module includes a detection area unit, a light following unit, and a graded detection unit. The detection area unit is used to divide the patient's detection area into multiple equal small areas, and to arrange the small areas in ascending order using Arabic numerals (this numbering is used to better identify the areas during tiered detection and to prevent confusion during tiered detection). The light following unit is used to install corresponding lights in each small area. The lighting angle of the lights can be adjusted in real time according to the detection angle of the patient's periodontium, and the lighting angle of the lights can be calculated in real time to see if it deviates from the detection angle of the periodontium. The graded detection unit is used to perform real-time deviation detection of periodontal detection angles at different locations in different small areas.

[0012] The judgment and early warning module includes a detection and judgment unit and a feasible prediction unit; The detection and judgment unit is used to determine that the periodontal bone detection angle is normal when the deviation value between the periodontal bone detection angle and other areas is less than or equal to 0; otherwise, it determines that the periodontal bone detection angle is abnormal and reports to the system to issue a voice alarm. The feasible prediction unit is used to receive the periodontal bone detection angle judgment results in real time through the data receiver, record the periodontal bone detection angle judgment results for 3 cycles, and calculate the deviation of the periodontal bone detection angle from the average value for 3 cycles. If the deviation value of the periodontal bone detection angle is greater than the average value for at least 2 cycles, it means that the periodontal atrophy detection plan cannot continue to be executed, and it also means that the patient's periodontal atrophy has occurred. A periodontal atrophy warning reminder can be issued. Otherwise, it means that the periodontal atrophy detection plan can continue to be executed, and the periodontal atrophy prediction can continue.

[0013] The prediction and correction module includes a shrinkage assessment module, an early warning resolution module, a judgment and correction module, and an adjustment and detection module. The atrophy assessment module includes a periodontal trend change unit and a periodontal risk assessment unit; The periodontal trend change unit is used to track the trend of periodontal bone parameters of the same tooth position and the same periodontal bone parameter over time by combining the patient's basic data, as shown in the following formula: Y_ij=β0+β1*Time_ij+u0i+u1i*Time_ij+ε_ij; Wherein, Yi_ij represents the periodontal bone parameter observation results of the i-th patient at time j, Time_ij represents the time variable of the i-th patient at time j, β0 represents the initial value of the same periodontal bone parameter for the same tooth position (representing the average periodontal probing depth of all patients before treatment or at the start of the study, and the overall average value of Yi_ij when Time_ij=0 (i.e., at baseline), and β1 represents the overall average effect of Time_ij on Yi_ij; The periodontal risk assessment unit is used to indicate that when β1 is greater than 0, the patient's periodontal bone parameters have deteriorated and there is a risk of periodontal atrophy; when β1 is less than 0, the patient's periodontal bone parameters have improved and there is a risk of periodontal atrophy; when β1 is equal to 0, the patient's periodontal bone parameters have not changed over time and there is no risk of periodontal atrophy. The early warning resolution module includes a report generation unit and a precautions unit; The report generation unit is used to automatically generate periodontal atrophy prediction and detection reports from the periodontal risk assessment results of different patients using NLG, and record them in a table. The precautions unit is used to automatically generate precautions for different patients based on the periodontal risk assessment results. The precautions include the types of food supplements, sleep duration, teeth cleaning, whether gingival transplantation is recommended, improving brushing methods, using desensitizing toothpaste, using fluoride mouthwash, quitting smoking, and maintaining a balanced diet.

[0014] The judgment and correction module includes an angle correction unit, which is used to correct the periodontal atrophy detection angle based on the detection angle deviation result. The correction angle is determined based on the calculated angle deviation value. The adjustment detection module includes a self-tracking correction unit and a result prediction feasibility unit; The self-tracking correction unit is used to track the deviation of the periodontal bone detection angle in real time through the data tracker, and to track the correction results in real time, and to predict the feasibility of the periodontal atrophy detection scheme in a secondary manner. The result prediction feasibility unit is used to continue predicting periodontal atrophy based on the grading test results when the secondary prediction periodontal atrophy detection plan is feasible, and automatically generate periodontal atrophy warning solutions and precautions. When the secondary prediction periodontal atrophy detection plan is not feasible, the periodontal atrophy detection plan is adjusted according to the tooth shape, and the grading test process is repeated.

[0015] A deep learning-based method for predicting periodontal atrophy includes the following steps: Step 1: Configure the IP address information of the remote control region server for periodontal atrophy prediction; Step 2: Enter the periodontal recognition terminal, collect basic data of different patients in real time through the data acquisition device, set up a periodontal atrophy prediction model, and obtain the patient's periodontal data through image recognition. Combined with multi-angle and multi-position centralized recognition, it can judge in real time whether recognition artifacts occur. Step 3: Enter the grading detection terminal, receive the periodontal bones of different patients, set the periodontal atrophy detection plan for different patients according to the periodontal bones, perform grading detection on the periodontal atrophy detection area, and judge whether the detection angle deviates in real time, and predict whether the periodontal atrophy prediction is feasible in a timely manner. Step 4: Enter the prediction and correction end. If the periodontal atrophy prediction is executable, the periodontal atrophy situation is predicted based on the graded detection results, and a periodontal atrophy warning solution is automatically generated. If the periodontal atrophy prediction is not executable, the periodontal atrophy detection angle is corrected based on the detection angle deviation result, and the correction result is tracked in real time to determine whether the periodontal atrophy detection solution is feasible. Step 5: If the secondary prediction of periodontal atrophy detection plan is feasible, continue to predict the periodontal atrophy situation based on the grading detection results and automatically generate a periodontal atrophy warning solution. If the secondary prediction of periodontal atrophy detection plan is not feasible, adjust the periodontal atrophy detection plan according to the tooth shape and conduct a secondary grading detection.

[0016] The present invention has the following beneficial effects: 1. In this invention, by setting up a periodontal recognition terminal, during the periodontal atrophy prediction operation based on deep learning, basic data of different patients are collected in real time through a data acquisition device. Based on the basic data, a periodontal atrophy prediction model matching different patients is set. Different patients correspond to different periodontal atrophy prediction models. The periodontal atrophy prediction model is set in a timely manner according to the basic data of different patients, which increases the applicability of periodontal atrophy prediction. During image recognition, multi-angle and multi-position centralized recognition is combined to determine whether recognition artifacts occur. This increases the accuracy of periodontal data recognition when acquiring patients' periodontal data, avoids overlapping artifacts in periodontal data, and corrects and re-identifies in a timely manner when overlapping artifacts occur. This reduces the error in periodontal atrophy prediction when performing periodontal atrophy prediction.

[0017] 2. In this invention, by setting up a graded detection end, during the periodontal atrophy prediction operation based on deep learning, different periodontal atrophy detection schemes are set according to the periodontal bones of different patients. The detection angles of the periodontal bones at different positions of the patient are judged in real time to determine whether they deviate. The periodontal atrophy detection scheme in the periodontal atrophy prediction process is executable. This enables graded detection of teeth in different areas when obtaining periodontal data of different patients based on imaging features, avoiding abnormal detection angles during the execution of periodontal detection schemes, and improving the accuracy of tooth data recognition during periodontal atrophy prediction.

[0018] 3. In this invention, by setting up a prediction and correction end, during the periodontal atrophy prediction operation based on deep learning, the periodontal atrophy status of the patient is predicted according to the risk assessment results. A periodontal atrophy warning solution and precautions are automatically generated. When the predicted periodontal atrophy is not feasible, the periodontal atrophy detection angle is corrected according to the deviation from the detection angle result, and the correction result is tracked in real time to determine the feasibility of a second periodontal atrophy detection plan. If the second periodontal atrophy detection plan is not feasible, the periodontal atrophy detection plan is adjusted according to the tooth shape, and the graded detection process is repeated. By predicting feasibility, a periodontal atrophy warning solution can be dynamically generated, enabling timely generation of precautions and warning solutions after changes in periodontal atrophy occur, thus improving the accuracy and reliability of periodontal atrophy prediction. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a deep learning-based method for predicting periodontal atrophy according to the present invention. Figure 2 This is a schematic diagram of the overall architecture of a deep learning-based periodontal atrophy prediction system according to the present invention. Figure 3 This is a schematic diagram of the periodontal recognition end of a deep learning-based periodontal atrophy prediction system according to the present invention. Figure 4 This is a schematic diagram of the architecture of the graded detection end of a deep learning-based periodontal atrophy prediction system of the present invention. Figure 5 This is a schematic diagram of the prediction and correction end of a deep learning-based periodontal atrophy prediction system according to the present invention. Detailed Implementation

[0020] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0021] Example 1, please refer to Figures 1-3 As shown: A method and system for predicting periodontal atrophy based on deep learning, including a periodontal identification end, a grading detection end, and a prediction and correction end, wherein the periodontal identification end, the grading detection end, and the prediction and correction end are all equipped with a prediction alarm module; The periodontal recognition terminal is used to acquire patients’ periodontal data through image recognition, set up a periodontal atrophy prediction model, and combine multi-angle and multi-position centralized recognition to determine in real time whether recognition artifacts occur. The grading detection terminal is used to set different periodontal atrophy detection plans for different patients based on periodontal bone, and to perform grading detection of periodontal atrophy detection areas to predict in a timely manner whether periodontal atrophy prediction is feasible. The prediction and correction end is used to automatically generate a periodontal atrophy warning solution based on the predicted periodontal atrophy results, or to correct the periodontal atrophy detection angle based on the deviation of the detection angle, and to track the correction results in real time to determine whether the secondary prediction periodontal atrophy detection solution is feasible. The prediction alarm module is used to report to the system and issue an abnormal warning reminder when the image recognition shows overlapping artifacts, the periodontal atrophy prediction is not feasible, or the detection angle is deviated.

[0022] The periodontal recognition terminal includes a data acquisition module, a periodontal recognition module, and an artifact detection module; The data acquisition module includes a patient data acquisition unit and a prediction model setting unit; The patient data acquisition unit is used to collect basic patient data through a data acquisition device. The basic data includes age, gender, height, body indicators, medical history, hygiene habits, dental diagnosis history, tooth shape, tooth arrangement, number of teeth, lifestyle habits, and favorite foods. This data is recorded in real time by a data recorder. (The basic data here is used to collect the corresponding body indicators of different patients. The basic data of different patients is combined to match the corresponding periodontal atrophy prediction model. At the same time, the periodontal atrophy prediction model can be dynamically adjusted according to the patient's body indicators and tooth shape. The factors considered in the dynamic adjustment include age, gender, height, body indicators, medical history, hygiene habits, dental diagnosis history, tooth shape, tooth arrangement, number of teeth, lifestyle habits, and favorite foods.) The prediction model setting unit is used to set the periodontal atrophy prediction model. The periodontal atrophy prediction model is set according to the basic data of different patients. The periodontal atrophy prediction model is implemented through one of the linear model, logistic regression model and machine learning model. Different periodontal atrophy prediction models are corresponding to patients of different ages. The periodontal atrophy prediction model sets the standard periodontal data corresponding to the basic data. The standard periodontal data includes the standard distance from the cementoenamel junction to the gingival margin, the gap between adjacent teeth, the amount of dental plaque accumulation, the height of the alveolar crest, the height of the alveolar bone, and the root morphology.

[0023] The periodontal recognition module includes multi-dimensional recognition units and multi-angle, multi-position units; The multidimensional recognition unit is used to identify different teeth in multiple dimensions through imaging equipment to obtain periodontal data. The imaging equipment includes an intraoral scanner, cone-beam CT, two-dimensional X-ray imaging instrument and three-dimensional imaging instrument. The periodontal data includes the distance from the cementoenamel junction to the gingival margin, the gap between adjacent teeth, the amount of dental plaque accumulation, the height of the alveolar crest, the height of the alveolar bone, and the morphology of the tooth root. The multi-angle, multi-position unit is used to adjust the angle and position of the imaging device in real time during the multi-dimensional recognition process, so as to realize the recognition of periodontal data from multiple angles and positions.

[0024] The artifact detection module includes an artifact detection unit, a data judgment unit, and a data recording unit; The artifact recognition unit is used to calculate the weighted sum of the deviations of the periodontal data features from the standard periodontal data features at the current time, current angle, and current region, as shown in the following formula: E_i=Σ[w_j*|F_j_i-F_j_healthy|^p]; in, This represents the weighted sum of all periodontal data features in the current time, angle, and region that deviate from the standard periodontal data features. F_j_i represents the periodontal data value of the j-th feature measured at the i-th position / angle (including periodontal texture, periodontal margin sharpness, alveolar crest height, and gingival margin clarity). F_j_healthy represents the standard periodontal data value of the j-th feature in healthy periodontal tissue. F_j_i-F_j_healthy represents the absolute deviation between the periodontal data features and the standard periodontal data features of the healthy standard. p represents the exponent, usually 1 (sum of absolute deviations) or 2 (sum of squared deviations, i.e., Euclidean distance). w_j represents the periodontal data weight of the j-th feature (this is used to perform the first periodontal atrophy prediction based on the basic data of different patients. If the absolute deviation between the periodontal data features and the standard periodontal data features of the healthy standard is not equal to 1, it means that the corresponding patient has periodontal atrophy. Otherwise, it means that the corresponding patient has not experienced periodontal atrophy. This periodontal atrophy prediction is the first step in the periodontal atrophy prediction process). The data judgment unit is used to set the deviation threshold T. If If the value is greater than T, it is determined that an artifact has occurred during periodontal data recognition. The system will issue a voice alarm and re-recognize the periodontal data, or pulse elimination will be used to calculate the overlap artifact correction value. The calculation formula is as follows:

[0025] in, To correctly correct the entropy value of the corresponding overlapping artifact image, i.e., the overlapping artifact correction value, This represents the entropy value of the overlapping artifact image corresponding to each inverse simulation correction. If not, it indicates that no artifacts were found in the periodontal data identification (this is used to determine whether the patient's periodontal data collected during periodontal atrophy prediction is abnormal, further improving the accuracy of periodontal atrophy prediction). The data recording unit is used to record the identification results of periodontal data features in real time through a data recorder.

[0026] Based on the basic data of different patients, a periodontal atrophy prediction model is set up to increase the applicability of periodontal atrophy prediction. The model is combined with image recognition to obtain the patient's periodontal data, including the shape of the patient's own teeth, gaps, and other data. During image recognition, multi-angle and multi-position centralized recognition is combined to determine whether recognition artifacts occur. This increases the accuracy of periodontal data recognition when acquiring the patient's periodontal data and avoids the existence of overlapping artifacts in the periodontal data, which would affect the accuracy of the periodontal atrophy prediction.

[0027] Example 2, please refer to Figure 4 As shown: Based on Embodiment 1, the graded detection terminal includes a detection scheme module, a graded detection module, and a judgment and early warning module; The detection protocol module includes a periodontal bone unit and a detection protocol unit; The periodontal skeleton unit is used to obtain periodontal skeleton parameters for different patients through image recognition equipment. The periodontal skeleton parameters include the distance from the cementoenamel junction to the bottom of the periodontal pocket, the distance from the gingival margin to the bottom of the periodontal pocket, the distance from the cementoenamel junction to the gingival margin, bone fissures, bone fenestrations, the vertical distance from the cementoenamel junction to the alveolar ridge crest, bone resorption morphology, bone density, trabecular bone structure, alveolar bone thickness, and root bifurcation morphology. Standard periodontal skeleton parameters corresponding to different patients are set. The detection protocol unit is used to set corresponding periodontal atrophy detection protocols based on the periodontal bone parameters of different patients. The periodontal atrophy detection protocol includes setting the detection angle, detection position, and detection area for different patients' periodontal atrophy (this is a periodontal atrophy detection protocol set according to the periodontal bone parameters of different patients; for example, if a patient has large gaps between teeth (due to the patient's own tooth growth), dental malformation (due to the patient's own tooth growth), or tooth loss, the periodontal atrophy detection protocol will exclude the patient's large gaps and compare other abnormalities in dental bone parameters; in cases of dental malformation, the detection will be graded according to the protruding position of the tooth; in cases of tooth loss, the periodontal area other than the lost tooth will be detected). It also combines the standard deviation of the periodontal bone parameter measurements of different patients from the average value of standard periodontal bone parameters of healthy individuals of the same age to determine whether the periodontal bone parameters of different patients are abnormal, as detailed below: Step 1: Calculate the standard deviation Z of the periodontal and skeletal parameter measurements of different patients from the average periodontal and skeletal parameters of healthy individuals of the same age. The formula is as follows: Z=(X_patient-μ_age) / σ_age; Where X_patient represents the measured periodontal and skeletal parameter values ​​of different patients, μ_age represents the average value of the standard periodontal and skeletal parameters of healthy people of the same age, and σ_age represents the standard deviation of the standard periodontal and skeletal parameters of healthy people of the same age. Step 2: Set the standard deviation threshold. The standard deviation threshold is determined based on the standard periodontal bone parameters corresponding to different ages. If Z is greater than the standard deviation threshold ±2σ, it indicates that the patient's periodontal bone parameters are abnormal. The system will issue a voice alarm and make the first periodontal atrophy warning. If not, it indicates that the patient's periodontal bone parameters are normal. (Here, the standard deviation of the periodontal bone parameters of different patients and the average of the standard periodontal bone parameters of healthy people of the same age is used to determine whether the patient's periodontal bone parameters are abnormal and whether a periodontal atrophy warning has occurred. The standard deviation of periodontal bone parameters for different ages includes the average alveolar crest height μ_age=2.1mm (standard deviation σ_age=0.5mm), etc. The standard deviation represents the "normal reference range" established through "big data of healthy people".)

[0028] The graded detection module includes a detection area unit, a light following unit, and a graded detection unit; The detection area unit is used to divide the patient's detection area into multiple equal small areas, and to number the small areas with Arabic numerals in ascending order. The light-following unit is used to install corresponding lights in each small area. The illumination angle of the lights can be adjusted in real time according to the detection angle of the patient's periodontium. It calculates in real time whether the illumination angle of the lights deviates from the detection angle of the periodontium. The specific method is as follows: Establish a coordinate system by setting the rotation center point of the lighting angle of the lamp as the origin, and derive the adjustment angle vector when adjusting the lighting angle of the lamp using the calculation formula: ; ; in, Adjust the angle between the distance to the lighting fixture and the positive X-axis. Adjust the angle between the distance to the lighting fixture and the positive Y-axis, where A is the height of the lighting fixture. This is the adjustment angle vector when the lighting fixture is adjusted in the X-axis direction. This is the adjustment angle vector when the lighting fixture is adjusted in the Y-axis direction; Calculate the adjustment angle vector of the lighting lamp and The difference between the periodontal measurement angle and the standard deviation is calculated, and the standard deviation is set as O and M. -O≥1 or If -M≥1, it is determined that the adjustment angle of the lighting is abnormal, and the lighting angle is adjusted in real time according to the angle vector. If not, it is determined that the adjustment angle of the lighting is normal (this is to ensure the accuracy of the imaging equipment in recognizing periodontal bone parameters in real time when performing periodontal atrophy prediction, and reduce the error in the recognition process during periodontal atrophy prediction). The graded detection unit is used to perform real-time deviation detection of periodontal examination angles at different locations in different small areas, as detailed below: Step 1: Calculate the coefficient of the linear term in the regression equation, using the following formula: ; in, Let be the coefficient of the first-order term, representing the quadratic relationship between the detection angle and the plaque presence angle during periodontal bone testing, as detailed below: This represents the baseline value for the deviation of the detection angle during the first periodontal bone examination. This represents the baseline value indicating the deviation of the first periodontal bone detection angle from the specified point. This represents the baseline value indicating the deviation of the second periodontal bone measurement angle. This represents the baseline value indicating the deviation of the second periodontal bone detection angle from the specified point. This represents the baseline value of the deviation of the periodontal bone detection angle at the current moment. Here, it represents the deviation value of the periodontal bone detection angle at different times and different locations. Step 2: Calculate the coefficient of the quadratic term in the regression equation, using the following formula: ; in, It represents the coefficient of the quadratic term and the binary quadratic relationship between the detection angle and the plaque presence angle data during periodontal bone testing. ; ; in, Represents the coefficient of the constant term. This represents the average deviation value of the periodontal bone detection angle at the current moment. The periodontal bone detection angle here indicates the deviation angle of the periodontal bone detection angle from the periodontal bone detection angle of other areas. Indicates the first Deviation value of the periodontal bone detection angle. This represents the average value of the deviation angle data deformation for position detection of the periodontal bone. This represents the average deviation angle data of all periodontal bone detections in the angle dataset. Based on the coefficients of the first term, the second term, and the constant term, a regression equation is established to obtain the deviation value of the periodontal bone detection angle between the periodontal bone detection angle and other regions. (Here, the periodontal detection area is divided during the periodontal atrophy prediction process, and the direction of dental plaque obstructing the periodontal detection angle is captured in real time when capturing the periodontal detection angle, so as to realize the real-time detection of the periodontal detection angle deviation, ensure the safety and reliability of periodontal detection, reduce the detection obstacles in periodontal detection, and improve the accuracy of periodontal atrophy prediction.)

[0029] By judging in real time whether the detection angle of the periodontal bone in different positions of the patient deviates, the feasibility of the periodontal atrophy detection plan in the periodontal atrophy prediction process can be predicted. When obtaining periodontal data of different patients based on imaging features, the multi-factor features of different patients' tooth capture position, tooth arrangement shape and different gaps between adjacent teeth can be comprehensively considered. During periodontal detection, teeth in different areas can be graded and detected, avoiding abnormal detection angles when the periodontal detection plan is executed.

[0030] Example 3, please refer to Figure 5 As shown: Based on Embodiment 1, the early warning module includes a detection and judgment unit and a feasibility prediction unit; The detection and judgment unit is used to determine that the periodontal bone detection angle is normal if the deviation between the periodontal bone detection angle and other areas is less than or equal to 0; otherwise, it determines that the periodontal bone detection angle is abnormal and reports to the system to issue a voice alarm. The feasibility prediction unit receives the periodontal bone detection angle judgment results in real time through the data receiver, records the periodontal bone detection angle judgment results for 3 cycles, and calculates the deviation of the periodontal bone detection angle from the average value for 3 cycles. If the deviation of the periodontal bone detection angle is greater than the average value for at least 2 cycles, it means that the periodontal atrophy detection plan cannot continue to be executed, and it also means that the patient's periodontal atrophy has occurred, so a periodontal atrophy warning reminder can be issued. Otherwise, it means that the periodontal atrophy detection plan can continue to be executed, and periodontal atrophy prediction can continue (here, at least 2 cycles greater than the average value means that the amount of dental plaque is large and has seriously affected the execution of the periodontal atrophy detection plan, so the corresponding patient's periodontal atrophy can be immediately judged, and a periodontal atrophy warning reminder can be issued).

[0031] The prediction and correction module includes a shrinkage assessment module, an early warning and resolution module, a judgment and correction module, and an adjustment and detection module. The atrophy assessment module includes a periodontal trend change unit and a periodontal risk assessment unit; The periodontal trend change unit is used to track the trend of periodontal bone parameters of the same tooth position and the same periodontal bone parameter over time by combining the patient's basic data. The formula is as follows: Y_ij=β0+β1*Time_ij+u0i+u1i*Time_ij+ε_ij; Wherein, Yi_ij represents the periodontal bone parameter observation result of the i-th patient at time j (here, it represents the periodontal bone parameter observation value of the same tooth position and the same periodontal bone parameter of the patient over a period of time), Time_ij represents the time variable of the i-th patient at time j, β0 represents the initial value of the same tooth position and the same periodontal bone parameter (representing the average periodontal probing depth of all patients before treatment or at the beginning of the study, and the overall average value of Yi_ij when Time_ij=0), and β1 represents the overall average effect of Time_ij on Yi_ij; The periodontal risk assessment unit is used to indicate that when β1 is greater than 0, it means that the patient's periodontal bone parameters have deteriorated and there is a risk of periodontal atrophy; when β1 is less than 0, it means that the patient's periodontal bone parameters have improved and there is a risk of periodontal atrophy; when β1 is equal to 0, it means that the patient's periodontal bone parameters have not changed over time and there is no risk of periodontal atrophy (the period of time here includes 7 days, 14 days, 21 days, 30 days, etc.). The early warning resolution module includes a report generation unit and a precautions unit; The report generation unit is used to automatically generate periodontal atrophy prediction reports from the periodontal risk assessment results of different patients through NLG, and record them in a table (existing technology has been disclosed). The "Precautions" section is used to automatically generate corresponding precautions for different patients based on the periodontal risk assessment results. The precautions include the types of food supplements, sleep duration, teeth cleaning, whether gingival transplantation is recommended, improving brushing techniques, using desensitizing toothpaste, using fluoride mouthwash, quitting smoking, and maintaining a balanced diet.

[0032] The judgment and correction module includes an angle correction unit, which is used to correct the periodontal atrophy detection angle based on the detection angle deviation result. The correction angle is determined based on the calculated angle deviation value. The adjustment detection module includes a self-tracking correction unit and a result prediction feasibility unit; The self-tracking correction unit is used to track the deviation of the periodontal bone detection angle in real time through the data tracker, and to track and correct the results in real time, and to predict the feasibility of the periodontal atrophy detection plan in a secondary manner. The result prediction feasibility unit is used to continue predicting periodontal atrophy based on the grading test results when the secondary prediction periodontal atrophy detection plan is feasible, and automatically generate periodontal atrophy warning solutions and precautions. When the secondary prediction periodontal atrophy detection plan is not feasible, the periodontal atrophy detection plan is adjusted according to the tooth shape, and the grading test process and steps are repeated.

[0033] Based on the risk assessment results, the system predicts the patient's periodontal atrophy and automatically generates periodontal atrophy warning solutions and precautions. When the predicted periodontal atrophy is not feasible, the detection angle is corrected according to the deviation of the detection angle. By predicting feasibility, the system can dynamically generate periodontal atrophy warning solutions, so that precautions and warning solutions can be generated in a timely manner after periodontal atrophy changes, thereby improving the accuracy and reliability of periodontal atrophy prediction.

[0034] In this invention, see reference Figures 1-5As shown, a deep learning-based method and system for predicting periodontal atrophy is presented. During operation, the system first configures the IP address information of the remote control area server for periodontal atrophy prediction. Then, it enters the periodontal recognition terminal and collects basic data from different patients in real time using a data acquisition device. A periodontal atrophy prediction model is set, and periodontal data is obtained through image recognition. Combined with multi-angle, multi-location centralized recognition, the system continuously assesses for recognition artifacts. By collecting basic data from different patients in real time and setting periodontal atrophy prediction models tailored to each patient, different periodontal atrophy prediction models are established for different patients. This allows existing periodontal prediction models to be applicable beyond a broad population, covering all rare cases or atypical atrophy patterns, and providing timely... Based on the basic data of different patients, a periodontal atrophy prediction model is set up to increase the applicability of periodontal atrophy prediction. This model combines image recognition to acquire patients' periodontal data, including the shape, gaps, and other data related to the patient's own tooth growth. During image recognition, multi-angle and multi-position centralized identification and judgment are used to determine whether recognition artifacts occur, thereby increasing the accuracy of periodontal data identification and avoiding overlapping artifacts that could affect the accuracy of periodontal atrophy prediction. Furthermore, when overlapping artifacts appear, timely correction and re-identification are performed to reduce the prediction error of periodontal atrophy. The model then enters the grading detection terminal, receiving periodontal bone data from different patients, and setting different patient-specific criteria based on the periodontal bone data. The system implements a periodontal atrophy detection protocol, grading the detection area and assessing the angle of deviation in real time to predict the feasibility of the periodontal atrophy prediction. Upon entering the prediction and correction phase, if the prediction is feasible, it predicts the extent of periodontal atrophy based on the grading results and automatically generates a periodontal atrophy warning solution. If the prediction is not feasible, it corrects the angle of deviation based on the deviation and tracks the correction results in real time, performing a secondary prediction to assess the feasibility of the periodontal atrophy detection protocol. A data acquisition device collects basic data from different patients in real time, and based on this data, a periodontal atrophy prediction model is set to match each patient. Different patients correspond to different periodontal atrophy prediction models, thus improving the existing periodontal prediction model. The model is no longer limited to a broad population; it can cover all rare cases or atypical atrophy patterns. It can set up periodontal atrophy prediction models in a timely manner based on the basic data of different patients, increasing the applicability of periodontal atrophy prediction. It also combines image recognition to obtain patients' periodontal data, including the shape, gaps, and other data of the patient's own teeth. During image recognition, it combines multi-angle and multi-position centralized recognition and judgment to determine whether recognition artifacts occur. This increases the accuracy of periodontal data recognition when acquiring patients' periodontal data, avoids the existence of overlapping artifacts in periodontal data, and avoids affecting the accuracy of periodontal atrophy prediction. At the same time, when overlapping artifacts occur, they are corrected and re-identified in a timely manner, reducing the error of periodontal atrophy prediction.If the secondary periodontal atrophy prediction detection plan is feasible, then the periodontal atrophy situation is predicted again based on the grading detection results, and a periodontal atrophy warning solution is automatically generated. If the secondary periodontal atrophy prediction detection plan is not feasible, then the periodontal atrophy detection plan is adjusted according to tooth shape, and a secondary grading detection is performed. During the periodontal atrophy prediction operation based on deep learning, when the periodontal atrophy prediction is executable, the periodontal trend changes of different patients are analyzed in real time for risk assessment. Based on the risk assessment results, the patient's periodontal atrophy situation is predicted, and a periodontal atrophy warning solution and precautions are automatically generated. When the periodontal atrophy prediction is not executable, then the detection angle is used... The system corrects deviations in the periodontal atrophy detection angle and tracks the correction results in real time. It then predicts the feasibility of a second periodontal atrophy detection plan. If the second plan is feasible, it continues to predict periodontal atrophy based on the graded detection results and automatically generates early warning solutions and precautions. If the second plan is not feasible, it adjusts the plan according to tooth shape and repeats the graded detection process. By dynamically generating early warning solutions based on predictive feasibility, it ensures timely generation of warnings and precautions after changes in periodontal atrophy occur, improving the accuracy and reliability of periodontal atrophy prediction.

[0035] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A deep learning-based periodontal atrophy prediction system, characterized in that, The system includes a periodontal identification terminal, a grading detection terminal, and a prediction and correction terminal, and the periodontal identification terminal, the grading detection terminal, and the prediction and correction terminal are all equipped with a prediction alarm module; The periodontal recognition terminal is used to acquire the patient's periodontal data through image recognition, set a periodontal atrophy prediction model, and combine multi-angle and multi-position centralized recognition to determine in real time whether recognition artifacts occur. The graded detection terminal is used to set different periodontal atrophy detection plans for different patients based on periodontal bone, and to perform graded detection of periodontal atrophy detection areas to predict in a timely manner whether periodontal atrophy prediction is feasible. The prediction and correction end is used to automatically generate a periodontal atrophy early warning solution based on the predicted periodontal atrophy results, or to correct the periodontal atrophy detection angle based on the detection angle deviation results, and to track the correction results in real time to determine whether the secondary prediction periodontal atrophy detection solution is feasible. The prediction alarm module is used to report to the system to issue an abnormal periodontal atrophy prediction warning when it determines that overlapping artifacts occur in image recognition, periodontal atrophy prediction is not feasible, or the detection angle is deviated.

2. The system according to claim 1, characterized in that: The periodontal recognition terminal includes a data acquisition module, a periodontal recognition module, and an artifact detection module; The data acquisition module includes a patient data acquisition unit and a prediction model setting unit; The patient data acquisition unit is used to collect basic patient data through a data acquisition device. The basic data includes age, gender, height, physical indicators, medical history, hygiene habits, dental diagnosis history, tooth shape, tooth arrangement, number of teeth, lifestyle habits, and favorite foods, and is recorded in real time through a data recorder. The prediction model setting unit is used to set the periodontal atrophy prediction model. The periodontal atrophy prediction model is set according to the basic data of different patients. The periodontal atrophy prediction model is implemented by one of the linear model, logistic regression model and machine learning model. Different periodontal atrophy prediction models are corresponding to patients of different ages. The periodontal atrophy prediction model sets the standard periodontal data corresponding to the basic data. The standard periodontal data includes the standard distance from the cementoenamel junction to the gingival margin, the gap between adjacent teeth, the amount of dental plaque accumulation, the height of the alveolar crest, the height of the alveolar bone, and the root morphology.

3. The system according to claim 2, characterized in that: The periodontal recognition module includes a multi-dimensional recognition unit and a multi-angle, multi-position unit; The multidimensional recognition unit is used to achieve multidimensional recognition of different teeth through imaging equipment to obtain periodontal data. The imaging equipment includes an intraoral scanner, cone-beam CT, two-dimensional X-ray imaging instrument and three-dimensional imaging instrument. The periodontal data includes the distance from the cementoenamel junction to the gingival margin, the gap between adjacent teeth, the amount of dental plaque accumulation, the height of the alveolar crest, the height of the alveolar bone, and the morphology of the tooth root. The multi-angle, multi-position unit is used to adjust the angle and position of the imaging device in real time during the multi-dimensional recognition process, so as to realize multi-angle, multi-position recognition of periodontal data.

4. The system according to claim 3, characterized in that: The artifact detection module includes an artifact detection unit, a data judgment unit, and a data recording unit; The artifact recognition unit is used to calculate the weighted sum of the deviations of the periodontal data features from the standard periodontal data features at the current time, current angle, and current region, as shown in the following formula: E_i=Σ[w_j*|F_j_i-F_j_healthy|^p]; in, This represents the weighted sum of all periodontal data features in the current time, current angle, and current region that deviate from the standard periodontal data features. F_j_i represents the periodontal data value of the j-th feature measured at the i-th position / angle (including periodontal texture, periodontal margin sharpness, alveolar crest height, and gingival margin clarity). F_j_healthy represents the standard periodontal data value of the j-th feature in healthy periodontal tissue. F_j_i-F_j_healthy represents the absolute deviation between the periodontal data features and the standard periodontal data features of the healthy standard. p represents the exponent, usually taking 1 (sum of absolute deviations) or 2 (sum of squared deviations, i.e., Euclidean distance). w_j represents the periodontal data weight of the j-th feature. The data judgment unit is used to set a deviation threshold T, if If the value is greater than T, it indicates that an artifact has occurred during periodontal data recognition. The system will issue a voice alarm and re-recognize the periodontal data. If not, it means that no artifact has occurred during periodontal data recognition. The data recording unit is used to record the identification results of periodontal data features in real time through a data recorder.

5. The system according to claim 1, characterized in that: The graded detection terminal includes a detection scheme module, a graded detection module, and a judgment and early warning module; The detection scheme module includes a periodontal bone unit and a detection scheme unit; The periodontal bone unit is used to obtain periodontal bone parameters of different patients through image recognition equipment. The periodontal bone parameters include the distance from the cementoenamel junction to the bottom of the periodontal pocket, the distance from the gingival margin to the bottom of the periodontal pocket, the distance from the cementoenamel junction to the gingival margin, bone dehiscence, bone fenestration, the vertical distance from the cementoenamel junction to the alveolar ridge crest, bone resorption morphology, bone density, trabecular bone structure, alveolar bone thickness, and root bifurcation morphology. Standard periodontal bone parameters corresponding to different patients are set. The detection scheme unit is used to set corresponding periodontal atrophy detection schemes based on the periodontal bone parameters of different patients. The periodontal atrophy detection scheme includes setting the detection angle, detection position, and detection area for the periodontium of different patients, and judging whether the periodontal bone parameters of different patients are abnormal by combining the standard deviation of the measured values ​​of periodontal bone parameters of different patients from the average value of standard periodontal bone parameters of healthy people of the same age. The details are as follows: Step 1: Calculate the standard deviation Z of the periodontal and skeletal parameter measurements of different patients from the average periodontal and skeletal parameters of healthy individuals of the same age. The formula is as follows: Z=(X_patient-μ_age) / σ_age; Where X_patient represents the measured periodontal and skeletal parameter values ​​of different patients, μ_age represents the average value of the standard periodontal and skeletal parameters of healthy people of the same age, and σ_age represents the standard deviation of the standard periodontal and skeletal parameters of healthy people of the same age. Step 2: Set the standard deviation threshold. The standard deviation threshold is determined based on the standard periodontal bone parameters corresponding to different ages. If Z is greater than the standard deviation threshold ±2σ, it indicates that the patient's corresponding periodontal bone parameters are abnormal. The system will issue a voice alarm and make the first periodontal atrophy warning. If not, it indicates that the patient's corresponding periodontal bone parameters are normal.

6. The system according to claim 5, characterized in that: The graded detection module includes a detection area unit, a light following unit, and a graded detection unit. The detection area unit is used to divide the patient's detection area into multiple equal small areas, and to number the small areas using Arabic numerals, in ascending order. The light-following unit is used to install corresponding lights in each small area. The illumination angle of the lights can be adjusted in real time according to the detection angle of the patient's periodontium. The unit calculates in real time whether the illumination angle of the lights deviates from the detection angle of the periodontium. The specific method is as follows: Establish a coordinate system by setting the rotation center point of the lighting angle of the lamp as the origin, and derive the adjustment angle vector when adjusting the lighting angle of the lamp using the calculation formula: ； ； in, Adjust the angle between the distance to the lighting fixture and the positive X-axis. Adjust the angle between the distance to the lighting fixture and the positive Y-axis, where A is the height of the lighting fixture. This is the adjustment angle vector when the lighting fixture is adjusted in the X-axis direction. This is the adjustment angle vector when the lighting fixture is adjusted in the Y-axis direction; Calculate the adjustment angle vector of the lighting lamp and The difference between the periodontal measurement angle and the standard deviation is calculated, and the standard deviation is set as o and M. -o≥1 or If -M≥1, it is determined that the adjustment angle of the lighting is abnormal, and the lighting angle is adjusted in real time according to the angle vector; otherwise, it is determined that the adjustment angle of the lighting is normal. The graded detection unit is used to perform real-time deviation detection operations on periodontal detection angles at different locations in different small areas, as detailed below: Step 1: Calculate the coefficient of the linear term in the regression equation, using the following formula: ； in, Let be the coefficient of the first-order term, representing the quadratic relationship between the detection angle and the plaque presence angle during periodontal bone testing, as detailed below: This represents the baseline value for the deviation of the detection angle during the first periodontal bone examination. This represents the baseline value indicating the deviation of the first periodontal bone detection angle from the specified point. This represents the baseline value indicating the deviation of the second periodontal bone measurement angle. This represents the baseline value indicating the deviation of the second periodontal bone detection angle from the specified point. This represents the baseline value of the deviation of the periodontal bone detection angle at the current moment. Here, it represents the deviation value of the periodontal bone detection angle at different times and different locations. Step 2: Calculate the coefficient of the quadratic term in the regression equation, using the following formula: ； in, It represents the coefficient of the quadratic term and the binary quadratic relationship between the detection angle and the plaque presence angle data during periodontal bone testing. ； ； in, Represents the coefficient of the constant term. This represents the average deviation value of the periodontal bone detection angle at the current moment. The periodontal bone detection angle here indicates the deviation angle of the periodontal bone detection angle from the periodontal bone detection angle of other areas. Indicates the first Deviation value of the periodontal bone detection angle. This represents the average value of the deviation angle data deformation for position detection of the periodontal bone. The value represents the average deviation angle data of all periodontal bone detections in the angle dataset. Based on the coefficients of the first term, the second term, and the constant term, a regression equation is established to obtain the deviation value of the periodontal bone detection angle between the periodontal bone detection angle and other regions.

7. The system according to claim 6, characterized in that: The judgment and early warning module includes a detection and judgment unit and a feasible prediction unit; The detection and judgment unit is used to determine that the periodontal bone detection angle is normal when the deviation value between the periodontal bone detection angle and other areas is less than or equal to 0; otherwise, it determines that the periodontal bone detection angle is abnormal and reports to the system to issue a voice alarm. The feasible prediction unit is used to receive the periodontal bone detection angle judgment results in real time through the data receiver, record the periodontal bone detection angle judgment results for 3 cycles, and calculate the deviation of the periodontal bone detection angle from the average value for 3 cycles. If the deviation value of the periodontal bone detection angle is greater than the average value for at least 2 cycles, it means that the periodontal atrophy detection plan cannot continue to be executed, and it also means that the patient's periodontal atrophy has occurred. A periodontal atrophy warning reminder can be issued. Otherwise, it means that the periodontal atrophy detection plan can continue to be executed, and the periodontal atrophy prediction can continue.

8. The system according to claim 1, characterized in that: The prediction and correction module includes a shrinkage assessment module, an early warning resolution module, a judgment and correction module, and an adjustment and detection module. The atrophy assessment module includes a periodontal trend change unit and a periodontal risk assessment unit; The periodontal trend change unit is used to track the trend of periodontal bone parameters of the same tooth position and the same periodontal bone parameter over time by combining the patient's basic data, as shown in the following formula: Y_ij=β0+β1*Time_ij+u0i+u1i*Time_ij+ε_ij; Wherein, Yi_ij represents the periodontal bone parameter observation results of the i-th patient at time j, Time_ij represents the time variable of the i-th patient at time j, β0 represents the initial value of the same periodontal bone parameter for the same tooth position (representing the average periodontal probing depth of all patients before treatment or at the start of the study, and the overall average value of Yi_ij when Time_ij=0 (i.e., at baseline), and β1 represents the overall average effect of Time_ij on Yi_ij; The periodontal risk assessment unit is used to indicate that when β1 is greater than 0, the patient's periodontal bone parameters have deteriorated and there is a risk of periodontal atrophy; when β1 is less than 0, the patient's periodontal bone parameters have improved and there is a risk of periodontal atrophy; when β1 is equal to 0, the patient's periodontal bone parameters have not changed over time and there is no risk of periodontal atrophy. The early warning resolution module includes a report generation unit and a precautions unit; The report generation unit is used to automatically generate periodontal atrophy prediction and detection reports from the periodontal risk assessment results of different patients using NLG, and record them in a table. The precautions unit is used to automatically generate precautions for different patients based on the periodontal risk assessment results. The precautions include the types of food supplements, sleep duration, teeth cleaning, whether gingival transplantation is recommended, improving brushing methods, using desensitizing toothpaste, using fluoride mouthwash, quitting smoking, and maintaining a balanced diet.

9. The system according to claim 8, characterized in that: The judgment and correction module includes an angle correction unit, which is used to correct the periodontal atrophy detection angle based on the detection angle deviation result. The correction angle is determined based on the calculated angle deviation value. The adjustment detection module includes a self-tracking correction unit and a result prediction feasibility unit; The self-tracking correction unit is used to track the deviation of the periodontal bone detection angle in real time through the data tracker, and to track the correction results in real time, and to predict the feasibility of the periodontal atrophy detection scheme in a secondary manner. The result prediction feasibility unit is used to continue predicting periodontal atrophy based on the grading test results when the secondary prediction periodontal atrophy detection plan is feasible, and automatically generate periodontal atrophy warning solutions and precautions. When the secondary prediction periodontal atrophy detection plan is not feasible, the periodontal atrophy detection plan is adjusted according to the tooth shape, and the grading test process is repeated.

10. A deep learning-based method for predicting periodontal atrophy, comprising the deep learning-based periodontal atrophy prediction system according to any one of claims 1-9, characterized in that, Includes the following steps: Step 1: Configure the IP address information of the remote control region server for periodontal atrophy prediction; Step 2: Enter the periodontal recognition terminal, collect basic data of different patients in real time through the data acquisition device, set up a periodontal atrophy prediction model, and obtain the patient's periodontal data through image recognition. Combined with multi-angle and multi-position centralized recognition, it can judge in real time whether recognition artifacts occur. Step 3: Enter the grading detection terminal, receive the periodontal bones of different patients, set the periodontal atrophy detection plan for different patients according to the periodontal bones, perform grading detection on the periodontal atrophy detection area, and judge whether the detection angle deviates in real time, and predict whether the periodontal atrophy prediction is feasible in a timely manner. Step 4: Enter the prediction and correction end. If the periodontal atrophy prediction is executable, the periodontal atrophy situation is predicted based on the graded detection results, and a periodontal atrophy warning solution is automatically generated. If the periodontal atrophy prediction is not executable, the periodontal atrophy detection angle is corrected based on the detection angle deviation result, and the correction result is tracked in real time to determine whether the periodontal atrophy detection solution is feasible. Step 5: If the secondary prediction of periodontal atrophy detection plan is feasible, continue to predict the periodontal atrophy situation based on the grading detection results and automatically generate a periodontal atrophy warning solution. If the secondary prediction of periodontal atrophy detection plan is not feasible, adjust the periodontal atrophy detection plan according to the tooth shape and conduct a secondary grading detection.

Images