Real-time oral health monitoring system based on intelligent imaging technology

Abstract

The invention relates to the technical field of oral health monitoring, and discloses a real-time oral health monitoring system based on an intelligent imaging technology. The system comprises an image capture module, a feature analysis module, a health parameter calculation module, an anomaly detection module and a feedback control module. The image capture module captures dynamic image data in the oral cavity in real time and transmits the data to the feature analysis module; the feature analysis module preprocesses the image data, identifies key oral cavity structural features and generates a structured feature data set; the health parameter calculation module quantifies related health parameters based on the data set and integrates the related health parameters into an oral health parameter set; the anomaly detection module identifies the oral cavity area with the parameter value deviating from the normal range and outputs an anomaly detection result set; and the feedback control module generates optimization suggestions according to the optimization suggestions and drives the user interface display module to output a visual report. According to the system, real-time monitoring, automatic analysis and timely feedback of oral health are realized, and the monitoring efficiency and accuracy are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of oral health monitoring, in particular to a real-time oral health monitoring system based on intelligent imaging technology. BACKGROUND

[0002] In the modern field of oral medicine, oral health monitoring is an important means to maintain oral function and overall health. Traditional oral health examination mainly relies on the naked eye observation and manual recording of doctors. This method is not only limited by the experience and subjective judgment of doctors, but also difficult to capture subtle changes inside the mouth. For example, early distribution of dental plaque, early signs of gum swelling and small cracks on the surface of teeth, etc. These potential health problems are often overlooked under naked eye observation, leading to delayed disease discovery and affecting treatment effectiveness.

[0003] In the traditional monitoring process, the clarity and stability of image acquisition are insufficient, making it difficult to provide high-quality image data for accurate analysis. Even if relevant images are obtained, subsequent feature recognition and parameter calculation still rely on manual operation, which is not only inefficient, but also greatly affected by human factors. Different operators may draw different conclusions, lacking objectivity and consistency.

[0004] Traditional methods cannot achieve real-time monitoring and feedback. Patients usually need to go to the hospital for examination regularly, and the health status in the mouth may change during the interval, which cannot be discovered and intervened in time. This lag makes some oral diseases develop to a more serious stage when discovered, increasing the difficulty and cost of treatment and causing more pain to patients.

[0005] With the increasing emphasis on oral health, there is an urgent need for efficient, accurate and real-time oral health monitoring methods. Although there are some oral image analysis systems in existing technologies, there are still deficiencies in real-time performance, automation level and feature recognition accuracy, which cannot meet the diversified needs in actual applications. Therefore, it is of great practical significance to develop a monitoring system that can capture oral images in real time, automatically analyze features, calculate health parameters and provide timely feedback. SUMMARY

[0006] The present application relates to the technical field of oral health monitoring, in particular to a real-time oral health monitoring system based on intelligent imaging technology.

[0007] To achieve the above-mentioned purpose, the present application provides a real-time oral health monitoring system based on intelligent imaging technology, which comprises:

[0008] The image capture module is used to capture a sequence of intraoral images in real time, obtain dynamic image data of the tooth surface, gum margin and soft tissue area through a high-resolution camera, and transmit the image data to the feature analysis module;

[0009] The feature analysis module performs image preprocessing operations including noise reduction and contrast enhancement based on the received image data, identifies key oral structure features including plaque distribution area, gum inflammation area and tooth crack position, and generates a structured feature data set;

[0010] The health parameter calculation module performs parameterized modeling operations based on the structured feature data set, quantifies plaque coverage, gum inflammation degree and tooth wear degree, and integrates them into an oral health parameter set;

[0011] The anomaly detection module performs pattern matching operations based on the oral health parameter set, identifies oral areas with parameter values deviating from the normal range, and outputs an anomaly detection result set;

[0012] The feedback control module performs real-time adjustment operations based on the anomaly detection result set, generates oral health optimization suggestions and drives the user interface display module to output a visual report.

[0013] Preferably, the image capture module includes a light source adjustment sub-module and an image acquisition sub-module. The light source adjustment sub-module automatically adjusts the light intensity according to the oral environment lighting conditions to ensure the uniformity of the image data;

[0014] The image acquisition sub-module captures image data through a multi-angle camera array, synchronizes timestamp information, filters motion blur frames, and generates a standardized image sequence;

[0015] The feature analysis module receives the standardized image sequence, applies an edge detection algorithm to locate the tooth profile, uses region segmentation technology to separate the gum tissue, and outputs a feature marker map containing coordinate information.

[0016] Preferably, the health parameter calculation module includes a parameter modeling sub-module and a data set integration sub-module. The parameter modeling sub-module analyzes the structured feature data set through a decision tree algorithm and calculates the plaque area proportion value;

[0017] The data set integration sub-module fuses gum inflammation area size data and tooth crack depth data, and generates a comprehensive health score through weighted averaging;

[0018] The health parameter calculation module compares the comprehensive health score with the historical health parameter set, updates the oral health parameter set and transmits it to the anomaly detection module.

[0019] Preferably, the anomaly detection module comprises a threshold comparison submodule and a pattern recognition submodule, the threshold comparison submodule compares each parameter in the oral health parameter set based on a preset normal range value, and marks abnormal items exceeding the threshold value;

[0020] The pattern recognition submodule applies a support vector machine algorithm to analyze the correlation of abnormal items, clusters similar abnormal patterns, and generates an anomaly detection result set;

[0021] The anomaly detection module outputs the anomaly detection result set to the feedback control module and synchronously stores it to the database module.

[0022] Preferably, the feedback control module comprises a suggestion generation submodule and an interface driving submodule, the suggestion generation submodule matches a pre-defined optimization rule library based on the anomaly detection result set, and generates personalized oral care suggestions;

[0023] The interface driving submodule converts the personalized oral care suggestions into interactive instructions to control the user interface display module to render the three-dimensional oral model in real time;

[0024] The feedback control module receives feedback data from the user interface display module and dynamically adjusts the suggestion content.

[0025] Preferably, the system further comprises an optimization module, which performs parameter optimization operations based on the oral health parameter set to adjust the acquisition frequency and resolution of the image capture module;

[0026] The optimization module comprises a parameter adjustment submodule and a performance evaluation submodule, the parameter adjustment submodule applies a random forest algorithm to predict the best acquisition parameter combination;

[0027] The performance evaluation submodule verifies the effectiveness of the parameter combination, updates the acquisition parameters, and outputs them to the image capture module.

[0028] Preferably, the system further comprises a threshold setting module, which performs dynamic threshold calculation operations based on a structured feature data set to define health parameter boundary values;

[0029] The threshold setting module comprises a data sampling submodule and a boundary calculation submodule, the data sampling submodule extracts the mean value of historical feature data;

[0030] The boundary calculation submodule applies a K-means clustering algorithm to group data and set a dynamic threshold range;

[0031] The threshold setting module transmits the dynamic threshold range to the anomaly detection module for comparison operations.

[0032] Preferably, the system further comprises a prediction module, which performs trend prediction operations based on the oral health parameter set to generate future health state values;

[0033] The prediction module comprises a time series analysis submodule and a prediction output submodule, the time series analysis submodule processes time series data through a long short-term memory network model;

[0034] The prediction output submodule integrates the prediction result and real-time health parameters, and outputs a risk warning signal to the feedback control module.

[0035] Preferably, the system further comprises a data integration module, which performs a multi-source data fusion operation based on image data, a structured feature data set and an oral health parameter set;

[0036] The data integration module comprises a data cleaning submodule and a fusion output submodule, the data cleaning submodule removes redundant image frames;

[0037] The fusion output submodule applies a principal component analysis algorithm to reduce the dimension of feature data, and generates a unified data stream;

[0038] The data integration module distributes the unified data stream to the health parameter calculation module and the prediction module.

[0039] Preferably, the system further comprises a user interface display module, which performs a visual rendering operation based on oral health optimization suggestions;

[0040] The user interface display module comprises a model construction submodule and a report generation submodule, the model construction submodule dynamically reconstructs a three-dimensional oral structure;

[0041] The report generation submodule compiles a health data summary and outputs an interactive report;

[0042] The user interface display module receives a driving signal from the feedback control module and updates the display content.

[0043] Compared with the prior art, the present application has the following advantages:

[0044] The image capture module captures a sequence of images inside the oral cavity in real time, and a high-resolution camera is used to obtain dynamic image data of the tooth surface, gum edge and soft tissue area, providing high-quality raw materials for subsequent analysis. Compared with the traditional method, this real-time and high-resolution image acquisition can more comprehensively present the details of the inside of the oral cavity, including those subtle and easily overlooked parts, making the understanding of the oral health condition more in-depth and meticulous.

[0045] The image preprocessing operations performed by the feature analysis module, such as noise reduction and contrast enhancement, can effectively improve the quality of the image data and reduce the influence of interference factors on subsequent analysis. On this basis, key oral structure features such as plaque distribution area, gum inflammation area, and tooth crack position are identified, and a structured feature data set is generated, realizing the automatic and accurate identification of oral features. This automated process avoids the subjectivity and errors in manual identification, making the feature recognition results more objective and consistent, providing a reliable basis for subsequent health assessment.

[0046] The health parameter calculation module performs parameterized modeling based on the structured feature data set, quantifies plaque coverage, gum inflammation degree, and tooth wear degree, integrates them into an oral health parameter set, and converts abstract oral features into specific and quantifiable parameters. These parameters can more intuitively reflect the severity and changes of oral health status, allowing users to clearly understand the health status of various aspects of the oral cavity, facilitating horizontal and vertical comparison and grasping the development trend of oral health.

[0047] The anomaly detection module identifies oral areas with parameter values deviating from the normal range based on the oral health parameter set, outputs an anomaly detection result set, and can timely detect abnormal conditions in the oral cavity. This anomaly detection does not rely on manual judgment and realizes rapid positioning of abnormal areas through pattern matching operations, ensuring that abnormal conditions can be identified in a timely manner and avoiding missed detection due to human negligence.

[0048] The feedback control module generates oral health optimization suggestions based on the anomaly detection result set and drives the user interface display module to output a visual report, realizing real-time feedback. Users can understand their oral health problems in a timely manner based on these suggestions and visual reports and take appropriate intervention measures. This real-time feedback mechanism breaks the time and space limitations of traditional monitoring, allowing users to monitor and manage their oral health status in their daily lives, which helps to prevent and control oral diseases in the early stages and reduces the risk of disease progression. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 A working principle diagram of the real-time oral health monitoring system based on intelligent imaging technology described in the present application;

[0050] Figure 2 A flowchart of the health parameter calculation module;

[0051] Figure 3 A flowchart of the anomaly detection module;

[0052] Figure 4 A flowchart of the optimization module;

[0053] Figure 5Flowchart of the prediction module. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0055] Referring to Figure 1 The present application provides a real-time oral health monitoring system based on intelligent imaging technology, which comprises:

[0056] Real-time evaluation and feedback of oral health status are achieved through the cooperation of multiple modules. The image capture module uses a high-resolution camera to dynamically capture internal oral images, covering the tooth surface, gum edge, and soft tissue area. The feature analysis module performs noise reduction and contrast enhancement processing on the original image, identifies plaque distribution, gum inflammation, and tooth crack features through edge detection and region segmentation technology, and generates a structured data set containing coordinate information. The health parameter calculation module quantifies plaque coverage, gum inflammation, and tooth wear, forming a standardized health parameter set. The anomaly detection module compares the health parameters with the preset threshold, and identifies abnormal areas through pattern matching. The feedback control module generates care recommendations based on the abnormal results and drives the user interface to output a three-dimensional visual report. During system operation, the data of each module is transmitted through a standardized interface, forming a closed-loop processing flow.

[0057] Embodiment 1: Referring to Figure 2 The image capture module includes a light source adjustment submodule and an image acquisition submodule, which work together to obtain high-quality oral image data. The light source adjustment submodule uses a multi-spectral LED array that can automatically adjust the light intensity according to the light reflection characteristics of different areas inside the mouth. The submodule has an ambient light sensor that monitors the light distribution inside the mouth in real time and dynamically adjusts the brightness and color temperature of the LEDs through a closed-loop feedback control algorithm to ensure the uniformity of the image data. In specific implementation, the light source adjustment submodule can identify the light differences of the tooth surface, gum edge, and soft tissue area, reduce the light intensity for the highly reflective enamel area, and appropriately increase the illumination for the gum tissue that strongly absorbs light, thereby avoiding overexposure or underexposure of the image.

[0058] The image acquisition sub-module adopts a multi-camera cooperative working mode, including a wide-angle camera and multiple macro cameras. The wide-angle camera is responsible for capturing the overall structure of the oral cavity, while the macro cameras are used for high-precision shooting of specific teeth or gum areas. All cameras are equipped with optical image stabilization and synchronized by hardware signals to ensure time consistency of multi-angle images. During the acquisition process, the image acquisition sub-module adds precise timestamps and spatial position information to each frame of image, facilitating subsequent feature matching and three-dimensional reconstruction. The motion blur detection algorithm analyzes the pixel displacement between adjacent frames. If a frame with blur exceeding the set threshold is detected, a re-acquisition mechanism is automatically triggered to ensure that the final output image sequence has high clarity and stability.

[0059] After receiving the standardized image sequence from the image capture module, the feature analysis module first performs preprocessing operations. The preprocessing stage includes noise suppression and contrast optimization. An adaptive filtering algorithm is used to remove random noise in the image, while a histogram equalization technique is used to enhance the contrast between different tissues, making the boundaries of teeth, gums, and soft tissues more clear and distinguishable. An edge detection algorithm is used to accurately locate the outline of the teeth. This algorithm can adapt to different tooth shapes and arrangements, accurately identifying the boundaries between the tooth crown, neck, and root. For gum tissue segmentation, the feature analysis module uses a region growing-based segmentation technique, combined with color space conversion and texture analysis, to effectively separate the gums from teeth, oral mucosa, and other structures.

[0060] After completing the basic feature extraction, the feature analysis module further identifies key oral health indicators. The plaque detection algorithm analyzes the color and texture features of the tooth surface to distinguish between normal enamel and plaque deposition areas, and calculates the coverage area and distribution density of plaque. The gum health assessment is based on color, swelling, and bleeding point detection of the gum tissue to quantify the degree of inflammation. The tooth crack detection uses a multi-scale analysis strategy, combining local enhanced images and high-frequency component extraction techniques to identify the location and direction of micro-cracks. All detection results are output in a structured data format, including coordinate information of feature points, area statistics, and health status classification labels.

[0061] The health parameter calculation module receives the structured feature data set generated by the feature analysis module and performs further parameterized modeling. The plaque coverage rate calculation uses an area ratio algorithm to quantify the degree of plaque accumulation by calculating the percentage of plaque area relative to the total tooth crown surface area. The assessment of gum inflammation degree is based on color analysis and morphological measurement. The RGB values of the gum redness area are converted to the HSV color space, and the saturation component is extracted as an inflammation indicator, which is then graded based on clinical standards. The quantification of tooth wear degree is achieved by analyzing the crack depth, enamel thickness, and changes in the morphology of the occlusal surface to calculate the wear index.

[0062] The data set integration submodule is responsible for synthesizing the above-mentioned health parameters to generate a unified oral health score. This submodule uses a weighted fusion algorithm to assign weights based on the impact of different parameters on overall oral health, and finally outputs a standardized comprehensive health index. The health parameter calculation module also has a historical data comparison function, which can compare and analyze the current detection results with the user's past oral health records, identify abnormal trends, and update the oral health parameter set for subsequent modules.

[0063] The implementation process of the entire embodiment 1 emphasizes the data connection and processing accuracy between modules. The image capture module ensures high-quality acquisition of raw images, the feature analysis module realizes accurate feature extraction and classification, and the health parameter calculation module completes data quantization and integration. Each submodule communicates through a standardized data interface to ensure smooth and reliable system operation. The image processing algorithm uses a real-time optimization strategy to achieve efficient operation under limited computing resources, meeting the needs of daily oral health monitoring. Users can intuitively understand their oral health status through the structured report output by the system and take appropriate care measures based on the system's recommendations.

[0064] Embodiment 2: Refer to Figure 3 , the abnormal detection module and the feedback control module work together to focus on abnormal identification of oral health parameters and generation of corresponding visual feedback. The abnormal detection module consists of a threshold comparison submodule and a pattern recognition submodule. Its core function is to analyze the oral health parameter set output by the health parameter calculation module in real time, identify abnormal indicators that deviate from the normal range, and pass the detection results to the feedback control module to generate targeted optimization suggestions. The threshold comparison submodule has a built-in clinical standard database that stores reference threshold ranges for gum sulcus probing depth, plaque index, and gingival bleeding index, among other oral health parameters. This submodule uses a sliding window scanning algorithm to check each value in the health parameter set. When the probing depth exceeds 3mm or the plaque index is greater than 20%, it is automatically marked as an abnormal item and its spatial location information is recorded. For continuously changing parameters such as the percentage of gum swelling area, the system sets a dynamic detection window to track the parameter's trend rather than a single measurement value, avoiding false positives caused by transient fluctuations.

[0065] The pattern recognition submodule adopts machine learning methods to analyze the relevance between the labeled abnormal items. This submodule trains a support vector machine classifier with a radial basis function (RBF) kernel function, mapping multi-dimensional health parameters to a high-dimensional feature space for non-linear classification. In the feature space, similar abnormal patterns are clustered into tight groups, while discrete abnormal points are identified as independent events. The density peak algorithm is used to automatically determine the number of cluster centers during clustering, avoiding the subjectivity of predefining the number of categories. For the identified abnormal pattern groups, the system analyzes their spatial distribution characteristics, such as determining whether plaque is clustered on adjacent tooth surfaces or whether gum inflammation is regionally distributed, to distinguish between local problems and systemic health risks. The abnormal detection results are output in the form of a structured data set, including abnormal type codes, severity scores, and spatial distribution coordinates.

[0066] The feedback control module generates specific oral care recommendations based on the abnormal detection result set. The recommendation generation submodule has a built-in rule engine containing hundreds of diagnosis-recommendation mapping rules defined by oral medicine experts. When interproximal plaque aggregation is detected, the system triggers flossing recommendations and marks the specific interproximal space positions that need to be cleaned; for extensive gum redness, it recommends an antibacterial mouthwash solution with usage frequency and duration guidance. The rule engine uses a priority mechanism to handle multiple abnormal situations, such as when plaque and gum bleeding coexist, it prioritizes recommending a gentle cleaning solution to avoid exacerbating bleeding. The recommendation content is converted into user-readable text through natural language generation technology, while retaining machine-parsable structured tags for interface rendering.

[0067] The interface driving submodule is responsible for converting the generated recommendations into visual interactive content. This submodule calls the WebGL graphics library to construct a STL format three-dimensional oral model based on the tooth three-dimensional coordinate data provided by the feature analysis module. The model rendering uses the Phong lighting model to enhance the stereoscopic effect, and the abnormal areas are displayed with high-light blocks superimposed, supporting view rotation and zoom operations. For microscopic features such as tooth cracks, the system provides a local magnification view, enhancing surface detail expression through normal mapping technology. The health parameter change trend is presented in a dynamic line chart, with the horizontal axis representing the detection time sequence and the vertical axis representing the parameter standardized value, assisting users in understanding the evolution process of health status. All visual elements are encapsulated as responsive components, which can automatically adjust the layout according to the size of the display device.

[0068] In terms of abnormal score calculation, the system uses the following formula to quantify the overall severity of abnormalities:

[0069]

[0070] where S represents the overall abnormality score, n is the number of detected abnormal parameters, w i is the clinical importance weight of the i-th parameter, and vi is the measured value, t i is the reference threshold, r i is the normal range radius. This score is used to drive the priority ranking of feedback control module's recommendations, high score anomalies will trigger more prominent interface warnings and more frequent review reminders. User interaction data such as viewing duration, zooming operations, etc. will be recorded and used to optimize the presentation of interface elements, for example, automatically improving rendering accuracy for frequently viewed areas.

[0071] During system operation, the anomaly detection module and the feedback control module form a real-time response loop. Newly collected image data triggers the anomaly detection process immediately after feature extraction and parameter calculation, and the detection results drive interface updates within milliseconds. Historical anomaly records are stored in a circular buffer for comparison of current state trends. When persistent anomalies or deteriorating trends are detected, the system automatically adjusts the recommendation content, such as upgrading daily care recommendations to professional appointment reminders. All user operations and system decisions are logged, including timestamps, operation types, and impact parameters, providing a data foundation for subsequent system optimization.

[0072] The visual report generation uses template engine technology to automatically arrange detection results, anomaly analysis, and care recommendations into multi-page PDF documents. The document structure includes a summary dashboard, detailed parameter table, and personalized recommendations, supporting mixed layout of text, charts, and three-dimensional screenshots. The report style follows medical document specifications, with key data displayed in bold and abnormal values marked with red warning signs. Electronic reports are sent to user terminals via secure transmission protocols, with cloud copies retained for historical queries. For content requiring professional interpretation, the system provides auxiliary annotation functions, displaying popularized explanations when users click on professional terms.

[0073] Example 3: Refer to Figure 4 , focusing on the collaborative working mechanism of the optimization module and the threshold setting module, emphasizing dynamic adjustment of system parameters and adaptive updating of health assessment standards. The optimization module automatically adjusts image acquisition parameters to improve data quality by continuously monitoring system operation status and detection accuracy; the threshold setting module dynamically calculates reasonable boundary values for various oral health indicators based on the distribution characteristics of historical health data, enabling anomaly detection standards to adapt to individual differences and long-term trends of users.

[0074] The optimization module includes a parameter adjustment submodule and a performance evaluation submodule, forming a closed-loop optimization system. The parameter adjustment submodule collects multi-dimensional state data generated during system operation, including image signal-to-noise ratio under different lighting conditions, feature recognition accuracy under different resolutions, processing delay under different acquisition frequencies, etc. These data are organized into a feature matrix and input into a regression prediction model based on random forests. The model analyzes the pros and cons of hundreds of historical parameter combinations and establishes a nonlinear mapping relationship between acquisition parameters and detection accuracy. For anterior tooth detection scenarios, the model may recommend using high resolution (3840x2160) with medium intensity ring light (1200 lux); while for posterior molar area, it may suggest switching to standard resolution (1920x1080) with diffuse strong light (1500 lux) to overcome the problem of insufficient lighting caused by narrow space in the area. The parameter adjustment strategy not only considers the static optimal solution, but also analyzes the system stability after parameter adjustment to avoid performance fluctuations caused by frequent switching.

[0075] The performance evaluation submodule is responsible for verifying the actual effect of the new parameter combination. This submodule designs a cross-validation process, collecting image data of the same oral area before and after parameter change, and quantifying the improvement by comparing the output differences of the feature analysis module. Evaluation indicators include edge detection continuity, region segmentation boundary accuracy, and feature label position deviation, etc. The evaluation process adopts a double-blind mechanism to avoid subjective factors. When the new parameter combination performs stably in continuous multiple tests, the performance evaluation submodule generates parameter update instructions, which are delivered to the image capture module through the system configuration interface. Detailed logs of this optimization operation are also recorded, including parameter comparison before and after adjustment, performance improvement amplitude, and system resource occupation changes, etc., accumulating experience data for subsequent optimization.

[0076] The threshold setting module is composed of a data collection submodule and a boundary calculation submodule, focusing on the individual adaptation of health assessment standards. The data collection submodule regularly extracts historical detection results from the health parameter calculation module to build a user-specific oral health baseline database. The extracted data are processed through cleaning and normalization to eliminate measurement deviations caused by device status or environmental factors. For each type of health parameter, the submodule calculates its statistical characteristic quantities, including mean, variance, skewness, and kurtosis, etc., forming a metadata set describing the data distribution pattern. These metadata are organized into a time series to analyze the variation of parameters with season, lifestyle, or physiological state.

[0077] The boundary calculation sub-module applies an improved K-means clustering algorithm to group historical health data. The algorithm first determines the optimal number of clusters automatically through the silhouette coefficient method, avoiding the subjectivity of manually setting the number of groups. For continuous variables such as the gingival bleeding index, the clustering process uses kernel density estimation techniques to handle non-uniformly distributed data points, making the grouping boundaries more consistent with the actual distribution characteristics. After grouping, the system establishes a three-layer threshold system for each parameter: the first layer is the ideal range, corresponding to the typical value interval of the healthy population; the second layer is the warning range, indicating the critical state that requires enhanced care; the third layer is the abnormal range, suggesting clear indications for professional intervention. The threshold calculation uses a dynamic buffer mechanism, adding a moderate margin to the statistical boundary to prevent false alarms caused by physiological fluctuations. The threshold updating strategy uses gradual adjustment, with the change amplitude controlled within a reasonable range to avoid user discomfort caused by sudden changes in standards.

[0078] In terms of optimizing the objective function, the system uses the following formula to balance the multiple constraints of parameter adjustment:

[0079]

[0080] where O represents the optimization target value, Q represents the image quality score, T is the time consumed for a single detection, T max is the maximum allowed time consumption, R is the system resource occupancy rate, R max is the upper limit of resources, and a, b, and g are the adjustment coefficients for quality, efficiency, and resource consumption, respectively. This objective function guides the system to reasonably balance processing speed and resource consumption while ensuring detection accuracy, achieving multi-objective collaborative optimization. The coefficients in the formula can be dynamically configured according to device performance and user preferences, for example, the efficiency coefficient weight can be appropriately increased on mobile terminals, while quality optimization can be emphasized when deploying workstations.

[0081] At the system implementation level, the optimization module and the threshold setting module adopt a loosely coupled architecture design. The optimization module receives performance indicators from various processing links through a message queue and triggers the parameter evaluation process using an event-driven mode. The threshold setting module establishes a dual mechanism of timing tasks and event triggers, ensuring regular updates of standards and enabling rapid responses to sudden changes in health status. The two modules share a monitoring data board that displays key information such as current parameter configuration, historical optimization trajectory, and threshold distribution map, assisting administrators in understanding the system's adaptive process. The data persistence layer uses a time series database to store massive running data, supporting fast retrieval and trend analysis.

[0082] In terms of user interaction, the system provides parameter adjustment explanations with controllable transparency. The basic view briefly hints at the optimization focus of the current detection mode; the expert view presents detailed parameter adjustment records and threshold change history, including the technical basis and expected impact of each change. For important threshold adjustments, the system generates easy-to-understand health status migration reports to help users understand the clinical significance behind standard changes. All adaptive adjustment operations are accompanied by complete audit logs, recording adjustment time, decision basis, execution results, and other key information to meet the traceability requirements of medical devices.

[0083] The implementation process of Embodiment 3 embodies the multi-level design of the system's self-optimization capability. At the data acquisition level, intelligent parameter adjustment is used to achieve precise matching of hardware performance and detection needs; at the analysis and evaluation level, dynamic health standards are established based on personalized historical data; at the system architecture level, modular design is used to balance the contradiction between automation and controllability. This adaptive mechanism enables the system to continuously adapt to changes in the user's oral state, changes in the use environment, and evolution of detection needs, maintaining high precision and reliability over a long period of use. Optimization decisions not only consider technical indicators but also incorporate soft factors such as human-computer interaction experience and clinical applicability, forming a multi-dimensional intelligent optimization system. The system learns the user's unique oral characteristics and health patterns through continuous learning, gradually evolving from a general-purpose device to a personalized health partner, and achieving closed-loop management of preventive oral care.

[0084] Embodiment 4: Referring to Figure 5 , the prediction module and the data integration module work together to achieve trend prediction and unified management of oral health status through time series analysis and multi-source data fusion. In a typical scenario in actual application, the system processes 7 days of plaque detection data of a user and predicts possible changes in health risks in the next 3 days, while integrating scattered detection information into structured data streams.

[0085] The time series analysis submodule of the prediction module uses a long short-term memory network architecture to process time series data of oral health parameters. The network input layer receives hourly sampling values of key indicators such as plaque coverage and gingival index for the past 168 hours (7 days), and after nonlinear transformation by three layers of hidden units, outputs prediction values for the next 72 hours (3 days). During network training, sliding window technology is used to expand sample size, with 96-hour historical data input at each time step to predict 24-hour changes. Through hundreds of iterations, the network learns the variation patterns of oral health parameters. In actual prediction, the system adjusts prediction weights in combination with user brushing frequency, dietary records, and other auxiliary information, such as appropriately increasing plaque growth prediction values for subsequent periods after high-sugar diet days.

[0086] The prediction output submodule spatiotemporally aligns the real-time detection data with the prediction results to generate a risk heat map. The heat map displays the health status change trend of different oral regions in future time periods in a matrix form, with color gradients representing risk levels. When the crack propagation speed of a certain tooth exceeds the safety threshold, the system automatically triggers a three-level early warning mechanism: the primary warning displays a prompt message on the user interface; the intermediate warning sends a mobile phone notification to suggest an appointment for examination; and the advanced warning directly contacts the preset dental clinic to generate an urgent appointment. The confidence level of the prediction results is estimated by the Monte Carlo Dropout method, and low-confidence predictions automatically trigger a supplementary detection process.

[0087] The data integration module processes information streams from multiple sources, including raw images from the image capture module, structured data sets from the feature analysis module, and quantitative indicators from the health parameter calculation module. The data cleaning submodule uses a hybrid method based on rules and machine learning to filter low-quality data. For image data, the system detects and removes invalid frames caused by lens stains or motion blur; for parameter data, it identifies and corrects outliers caused by sensor abnormalities. The cleaned data is arranged in chronological order, with missing values filled due to device offline, forming a complete time series.

[0088] The fusion output submodule uses a feature-level fusion strategy to map data from different sources to a unified feature space. Table 1 shows a comparison example before and after multi-source data fusion.

[0089] Table 1: Data fusion processing example.

[0090] Data type Original feature dimension Fused feature Processing mode Image data 256×256×3 Texture feature Wavelet transform Plaque detection 18-dimensional vector Coverage index Principal component analysis Gingival parameter 6 indices Inflammation score Weighted aggregation Occlusion record Time series Pressure distribution Fourier transform

[0091] The output of the data integration module uses a hierarchical data structure: the base layer stores raw measurement values; the intermediate layer contains cleaned and standardized data; and the application layer is a customized data set for specific functions (such as prediction and visualization). This structure maintains data integrity while meeting the specific needs of different modules. The system performs full data reorganization every 24 hours to ensure that long-term stored data uses the optimal compression format, while establishing a comprehensive index to speed up queries.

[0092] In terms of system architecture implementation, the prediction module uses a microservice design, with each prediction task running as an independent process to avoid long-time computation blocking real-time detection processes. The data integration module is built on a stream processing framework, supporting both real-time data access and batch processing modes. The two modules exchange high-frequency access data through shared memory and transfer processing results through a message queue, ensuring data freshness and avoiding resource competition. The system resource manager dynamically monitors the computational load of each module, automatically allocating more computing resources during intensive prediction task periods and starting background data optimization tasks during idle periods.

[0093] In terms of user interaction, the system provides a multi-perspective display of the prediction results. The timeline view shows the convergence of parameter change history and prediction trend horizontally; the oral map view uses a three-dimensional heat map to mark the risk areas; the calendar view highlights the warning dates that require special attention. All views support linked operations, such as clicking on a warning point on the timeline automatically locates to the corresponding position on the oral map. The system also generates a prediction report that explains the causes of risk and preventive measures in simple language, such as "the second molar on the left lower jaw is predicted to accelerate plaque growth, it is recommended to strengthen the use of dental floss in this area."

[0094] The data management strategy embodies intelligent storage optimization. Recent data with high frequency access is saved in high-speed storage devices with dual backup; historical data uses columnar compression storage, significantly reducing storage space occupation; temporary data such as prediction intermediate results are automatically cleaned up after processing. The data access interface realizes fine permission control, and clinical doctors can view complete prediction details, while ordinary users see simple health recommendations. All data operation records detailed metadata, including collection time, processing flow, access record, etc., to meet the compliance requirements of medical data management.

[0095] Example 5: The user interface display module focuses on transforming complex oral health data into intuitive and visual interactive content. This module uses three-dimensional reconstruction technology and dynamic rendering engine to present detection results, analysis data and health recommendations in a user-friendly way, achieving seamless integration of professional medical information and daily care guidance.

[0096] The model construction submodule uses a surface reconstruction algorithm based on point cloud data to process the oral structure coordinate information from the feature analysis module. The system first performs spatial sorting and topological connection on the discrete feature points to generate a preliminary triangular mesh. Subsequently, the mesh density is increased by applying the surface subdivision technique, which maintains the accuracy of the original anatomical structure while improving the visual smoothness. For high-reflectivity areas such as tooth surfaces, the algorithm automatically adjusts the vertex normal direction to simulate the real lighting reflection characteristics; while soft tissues such as gums use subsurface scattering shading technology to reproduce their translucent texture. The reconstruction process is performed in real time, and the three-dimensional model can be updated within 30 seconds after each new detection is completed, reflecting the latest changes in oral status.

[0097] The report generation submodule compiles various health data into structured visual elements. Detection values are displayed in ring progress bars to show the current value in comparison to the ideal range; trend data is converted into annotated timeline charts; abnormal areas are highlighted with pulsing halo effects on the 3D model. The report layout uses responsive design, automatically adjusting the element arrangement according to the display device size. On the desktop interface, the 3D model, parameter charts, and suggestion text are displayed side by side; on mobile devices, it switches to a tabbed mode, allowing users to view different content panels by swiping left or right. All visual elements are linked to detailed data source information, and users can access the corresponding original detection records by clicking on any chart or model area.

[0098] The interactive function design emphasizes natural operation logic. The 3D model supports multi-touch zooming and rotating, with the zooming range limited to a clinically reasonable magnification interval (0.5x-8x) to avoid misinterpretation due to excessive magnification. The rotating operation is set with damping effect to make the rotation more consistent with physical inertia, improving the smoothness of operation. For specific diagnostic needs, the system provides preset viewing angle quick buttons, such as "occlusal view" to one-click position to observe the wear of the occlusal surface, and "gingival line view" to automatically adjust to the best angle to check the edge state of the gum. All user operations are accompanied by delicate tactile feedback, such as a slight vibration prompt when switching tabs, enhancing the certainty of interaction.

[0099] The visual presentation follows the design specifications for medical visualization. The color scheme uses the clinically common identification system, with pink color used to mark plaque deposition areas, amber color used to indicate gum inflammation areas, and natural color used for healthy tissues. Important warning information uses a contrast ratio that meets the WCAG2.1 standard, ensuring that users with color vision impairments can also clearly identify it. The text content uses sans-serif font families, with key data highlighted in bold, and paragraph spacing set to 1.5 times the font size to improve readability. Animation effects follow the functional principle, such as using fade-in and fade-out transitions for data refresh, avoiding distracting and dazzling effects that distract attention.

[0100] The data update mechanism decouples the front-end display from the back-end processing. When the feedback control module generates new health recommendations, the interface driving submodule receives a JSON format instruction package containing semantic tags. The instruction package defines in detail the interface elements that need to be updated, the content changes, and the transition animation parameters. Interface components subscribe to their relevant data channels, and only trigger re-rendering when there is substantive content update, avoiding unnecessary performance consumption. For complex elements such as 3D models, the system uses a difference comparison algorithm to only update the grid parts that have actually changed, significantly improving rendering efficiency.

[0101] Accessibility features cater to a wide range of user needs. A voice navigation system reads out interface content and operation instructions, with adjustable speech rate. A high-contrast mode simplifies interface elements to black and white, removing all visual decorations. A keyboard operation system supports sequential access to all interactive elements via the Tab key, with focus outlines displayed. For specialized terms and medical concepts, an instant dictionary function is embedded in the system, with long-pressing any specialized term bringing up a plain-English explanation. All accessibility settings are saved in user profiles, automatically synchronized when logging in across devices.

[0102] The history view function adopts a timeline navigation design. Users can slide along the horizontal timeline to select any detection date, with the system instantly loading the corresponding three-dimensional model snapshot and detection report. Important health events such as tooth decay treatment, dental cleaning, etc. are marked as points on the timeline, with clicking allowing viewing of event details. The comparison mode allows selecting two time points of detection results for side-by-side display, with the difference highlighted in color contours. History data download supports multiple formats, including the medical standard DICOM format, PDF reports for easy sharing, and CSV data sets for further analysis.

[0103] System status feedback remains transparent and timely. When performing data processing or model updates, a progress indicator is displayed in a fixed area of the interface, distinguishing between the type and estimated completion time of background tasks. When detecting insufficient hardware performance that may affect the experience, an optimization suggestion dialog box is automatically popped up, providing options such as reducing rendering precision. Network connection status is monitored in real time, automatically switching to local cache mode when offline, and synchronizing all operation records after restoring connection. System maintenance notifications are pushed 24 hours in advance, explaining the maintenance period and the scope of possible affected functions.

[0104] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one from another entity or action, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0105] While the embodiments of the application have been illustrated and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and alterations can be made therein without departing from the spirit and scope of the application, which is defined by the following claims and their equivalents.

Claims

1. A real-time oral health monitoring system based on intelligent imaging technology, characterized in that, include: The image capture module is used to capture real-time image sequences inside the oral cavity. It acquires dynamic image data of the tooth surface, gingival margin and soft tissue area through a high-resolution camera and transmits these image data to the feature analysis module. Based on the received image data, the feature analysis module performs image preprocessing operations, including noise reduction and contrast enhancement, and identifies key oral structural features, including plaque distribution areas, gingival redness and swelling areas, and tooth crack locations, generating a structured feature dataset. Based on the structured feature dataset, the health parameter calculation module performs parametric modeling operations to quantify dental plaque coverage, gingival inflammation, and tooth wear, and integrates them into an oral health parameter set. Based on the oral health parameter set, the anomaly detection module performs pattern matching to identify oral regions where parameter values ​​deviate from the normal range and outputs anomaly detection result set. Based on the abnormal detection result set, the feedback control module performs real-time adjustment operations, generates oral health optimization suggestions, and drives the user interface display module to output a visual report.

2. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, The image capture module includes a light source adjustment submodule and an image acquisition submodule. The light source adjustment submodule automatically adjusts the supplementary light intensity according to the oral cavity lighting conditions to ensure the uniformity of image data. The image acquisition submodule captures image data through a multi-angle camera array, synchronizes timestamp information, filters motion-blurred frames, and generates a standardized image sequence. The feature analysis module receives a standardized image sequence, applies an edge detection algorithm to locate the tooth contour, uses region segmentation technology to separate the gingival tissue, and outputs a feature map containing coordinate information.

3. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, The health parameter calculation module includes a parameter modeling submodule and a dataset integration submodule. The parameter modeling submodule parses the structured feature dataset using a decision tree algorithm and calculates the percentage of dental plaque area. The dataset integration submodule merges data on the size of swollen gingival areas and the depth of tooth cracks, and generates a comprehensive health score by weighted averaging. The health parameter calculation module compares the comprehensive health score with the historical health parameter set, updates the oral health parameter set, and transmits it to the anomaly detection module.

4. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, The anomaly detection module includes a threshold comparison submodule and a pattern recognition submodule. The threshold comparison submodule compares each parameter in the oral health parameter set with a preset normal range value and marks abnormal items that exceed the threshold. The pattern recognition submodule uses the support vector machine algorithm to analyze the correlation of outliers, cluster similar outlier patterns, and generate an outlier detection result set. The anomaly detection module outputs the anomaly detection result set to the feedback control module and stores it synchronously in the database module.

5. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, The feedback control module includes a suggestion generation submodule and an interface-driven submodule. The suggestion generation submodule generates personalized oral care suggestions based on the abnormal detection result set and a predefined optimization rule base. The interface-driven submodule converts personalized oral care suggestions into interactive commands, controlling the user interface display module to render a 3D oral model in real time. The feedback control module receives feedback data from the user interface display module and dynamically adjusts the suggested content.

6. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, It also includes an optimization module, which performs parameter optimization operations based on the oral health parameter set, adjusting the acquisition frequency and resolution of the image capture module; The optimization module includes a parameter tuning submodule and a performance evaluation submodule. The parameter tuning submodule uses the random forest algorithm to predict the optimal combination of acquisition parameters. The performance evaluation submodule verifies the effectiveness of the parameter combination, updates the acquisition parameters, and outputs them to the image capture module.

7. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, It also includes a threshold setting module, which performs dynamic threshold calculation based on a structured feature dataset and defines boundary values ​​for health parameters. The threshold setting module includes a data sampling submodule and a boundary calculation submodule. The data sampling submodule extracts the mean of historical feature data. The boundary calculation submodule applies the K-means clustering algorithm to group data and sets a dynamic threshold range; The threshold setting module passes the dynamic threshold range to the anomaly detection module for comparison.

8. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, It also includes a prediction module, which performs trend prediction operations based on a set of oral health parameters to generate future health status values; The prediction module includes a time series analysis submodule and a prediction output submodule. The time series analysis submodule processes time series data through a long short-term memory network model. The prediction output submodule integrates the prediction results with real-time health parameters and outputs a risk warning signal to the feedback control module.

9. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, It also includes a data integration module, which performs multi-source data fusion operations based on image data, structured feature datasets, and oral health parameter sets; The data integration module includes a data cleaning submodule and a fusion output submodule. The data cleaning submodule removes redundant image frames. The fusion output submodule applies principal component analysis algorithm to reduce the dimensionality of feature data and generate a unified data stream. The data integration module distributes a unified data stream to the health parameter calculation module and the prediction module.

10. The real-time oral health monitoring system based on intelligent imaging technology according to claim 1, characterized in that, It also includes a user interface display module, which performs visualization rendering operations based on oral health optimization suggestions; The user interface display module includes a model building submodule and a report generation submodule. The model building submodule dynamically reconstructs the three-dimensional oral cavity structure. The report generation submodule compiles a health data summary and outputs an interactive report. The user interface display module receives drive signals from the feedback control module and updates the displayed content.

Images