Handheld portable oral cavity digital model observation and analysis system

Abstract

The invention relates to the technical field of medical treatment, and discloses a handheld portable oral cavity digital model observation and analysis system, which is characterized in that after data obtained by handheld three-dimensional optical scanning equipment is subjected to digital model observation and analysis processing, an intraoral digital three-dimensional model established by three-dimensional digital information images of upper and lower dentitions is rotated, magnified and shrunk at any angle, and a digital three-dimensional model is obtained; according to the observation angle and direction, the position, depth and angle of the undercut are automatically displayed, observation lines corresponding to the angle and direction are displayed, sectioning and measurement are carried out on the intraoral digital three-dimensional model, and two-dimensional intelligent planning design is carried out. According to the invention, the whole process is dynamic and visible, the observation position is variable, random adjustment, measurement and repetition can be realized, handwriting recording is not needed, the generated digital information is convenient for mass storage, analysis design and remote communication, and the true color three-dimensional occlusion conditions of upper and lower dentitions and the influence of the true color three-dimensional occlusion conditions on the observation angle can be displayed at the same time.

Description

TECHNICAL FIELD

[0001] The present application relates to the medical technology field, and in particular to a hand-held portable oral digital model observation and analysis system. BACKGROUND

[0002] Removable partial denture (removable denture) must be model observed before design, tooth preparation and production, that is, the soft and hard tissue morphology of teeth and gums in the mouth is analyzed. These morphologies need to be analyzed by using impression material to obtain the morphology of the patient's oral tissue, and then pouring into a dental plaster model, and then installing the dental plaster model on a mechanical observation platform (model observation instrument) for model observation and analysis, that is, analyzing the denture wearing angle. At the same time, the tooth surface is marked with a contour point, and finally the removable partial denture is designed according to the analysis and marking results on the model.

[0003] These methods must have been used for nearly a hundred years and must rely on impression materials placed in the patient's mouth, plaster models, and mechanical observation platforms to complete.

[0004] The prior art is as follows:

[0005] 1. The operator needs to use impression material to obtain the soft and hard tissue morphology of the patient's mouth, as shown in Figure 1 A, Figure 1 A is a schematic diagram of using impression material to obtain the soft and hard tissue morphology of the patient's mouth. Figure 1 A is a schematic diagram of using impression material to obtain the soft and hard tissue morphology of the patient's mouth.

[0006] 2. The operator or technician needs to pour the prepared impression into a dental model using plaster, as shown in Figure 1 B, Figure 1 B is a schematic diagram of pouring a dental model using plaster.

[0007] 3. The poured plaster model needs to be installed on a model observation instrument for tracing, or installed on a large desktop scanner for optical scanning, as shown in Figure 2 A, Figure 2 A is a schematic diagram of installing a plaster model on a mechanical model observation instrument and designing a fixed observation angle. Figure 2 A is a schematic diagram of installing a plaster model on a mechanical model observation instrument and designing a fixed observation angle.

[0008] 4. The operator needs to set an observation angle, and after setting, the model angle is locked on the observation instrument and cannot be shifted or loosened, as shown in Figure 2 A.

[0009] 5. Manual model marking is required to draw model high points and record the undercut angle value, as shown in Figure 2 B.

[0010] 6. Existing model observation technology such as Figure 3 As shown in the figure, the gypsum model is installed on a large desktop scanning instrument for optical scanning, and the undercut data is marked in the software: the gypsum model is placed on the desktop optical scanning device, and a fixed observation angle is designed, as shown in Figure 3 A; the digital model is directly marked in the system, and the model high point is drawn and the undercut angle value is recorded, as shown in Figure 3 B.

[0011] The existing technology has the following technical problems:

[0012] 1. High clinical time consumption, about 50 minutes are needed throughout the process.

[0013] 2. Impression material, gypsum material, and tray material are needed.

[0014] 3. It cannot display the undercut depth and angle of all teeth. The measured site angle and depth value needs to be recorded by hand.

[0015] 4. It cannot obtain the tooth arrangement relationship when occlusion and the undercut information when occlusion.

[0016] 5. The impression material has an odor, the patient's impression time is long, the body feeling is poor, and there is obvious nausea and foreign body sensation.

[0017] 6. Whether it is mechanical or model observation, or large desktop optical scanning observation, the gypsum model needs to be obtained first. The gypsum model is large in volume, heavy in weight, needs to be sterilized and stored, and is not easy to store in large quantities. If it is lost or damaged, it cannot be copied, and it cannot be quickly communicated and transmitted remotely.

[0018] 7. The mechanical observation instrument is relatively fixed in observation angle and lead position, and is not easy to change.

[0019] 8. When the mechanical observation is performed, the measured object is small, and the human error is large. SUMMARY

[0020] The purpose of the present application is to provide a handheld portable digital model observation analysis system for oral cavity, which solves the technical problems existing in the prior art.

[0021] To achieve the above purpose, a handheld portable digital model observation analysis system for oral cavity, the system comprises a data capture unit, a data processing unit, an undercut model generation unit, an inference model construction unit, an information display unit and a data storage unit;

[0022] The data capture unit is used to capture the data in the patient's mouth, and clearly display the colored upper and lower tooth digital three-dimensional model in the mouth on the display, and the digital three-dimensional model in the mouth can be rotated at any angle, enlarged or reduced.

[0023] The data processing unit is used for patient profiling and generating a health report, and adjusting the observation angle of the intraoral digital three-dimensional model according to the oral condition, and in the adjusted intraoral digital three-dimensional model, the precise full dentition digital information of the guide line, the undercut depth, and the angle displayed on the digital dentition, and in the full dentition digital information, the undercut depth is distinguished by color, and the deeper the color, the greater the undercut angle;

[0024] The undercut model generation unit is used for undercut detection and generation of an intraoral digital undercut model, and specifically used for rotating, enlarging or reducing the intraoral digital three-dimensional model, automatically displaying the position, depth, and angle of the undercut according to the observation angle and direction, and displaying the observation line corresponding to the observation angle and direction, the greater the rotation angle, the greater the undercut filling amount, the smaller the rotation angle, the smaller the undercut filling amount, and when performing three-dimensional observation of the intraoral digital three-dimensional model, automatically filling the undercut, and displaying the position, slope, depth of the undercut, and the position of the observation line;

[0025] The inference model construction unit is used for constructing a two-dimensional planning scheme inference model and performing model optimization evaluation, and specifically used for constructing a two-dimensional bracket planning scheme inference model based on deep learning according to the intraoral digital undercut model, completing scheme selection, clasp design, and connector design;

[0026] The information display unit is used for design scheme display and tooth preparation information display, wherein the design scheme display is a two-dimensional bracket scheme completed by visual interface, and the tooth preparation information display is a detailed tooth preparation guidance scheme containing tooth preparation amount, preparation form, and seating path direction, which is intelligently generated based on the clasp type of the stem clasp and the round clasp and the corresponding tooth position in the design scheme, and provides precise technical guidance for clinical operation;

[0027] The data storage unit is used for automatically saving data into classified data, facilitating subsequent design and modification;

[0028] If the common seating path of the selected intraoral digital three-dimensional model is not conducive to denture wearing, that is, the guide line position and the undercut depth do not meet the requirements, the observation angle is adjusted, the common seating path is reselected, and the best angle is selected until the best angle is selected.

[0029] Further, the data processing unit has the following specific working process:

[0030] (1) Establish a new patient intraoral data storage file, or use a default storage file, and after capturing the patient intraoral data, the scanning data in the corresponding folder will be automatically obtained according to the storage file position and path and automatically named;

[0031] (2) According to the patient information, data classification and patient file are made, and patient information and repair information are inputted;

[0032] (3) After the patient file is filled and confirmed, the undercut model option and the intraoral digital three-dimensional model scanning data are selected, and the imported intraoral digital three-dimensional model scanning data is displayed in a small window.

[0033] Further, when the model observation is performed by the undercut model generation unit, the undercut is automatically filled, the position, slope, depth and position of the observation line of the undercut are displayed, the depth of the undercut is distinguished by color, at the same time, the undercut is filled with "virtual red wax", the slope of the undercut is displayed on the side of the intraoral digital three-dimensional model, the color of the intraoral digital three-dimensional model is in sharp contrast with the color of the undercut, and the intersection of the two colors is the position of the observation line.

[0034] Further, the undercut model generation unit is also used for performing the intraoral digital three-dimensional model observation, and according to the observation result, the intraoral digital three-dimensional model is sequentially subjected to the common seating path determination, undercut depth and slope display and observation line display processing;

[0035] Among them, the common seating path determination is to extract a directional vector data (dir[x, y, z]) as the common seating path direction of the intraoral digital three-dimensional model;

[0036] The undercut model generation unit performs undercut detection and generates an undercut model on the intraoral digital three-dimensional model through the undercut depth and slope display and the observation line display processing, and the specific work flow is as follows:

[0037] (1) Set the direction and space transformation: rotate the system three-dimensional space to the observation direction dir corresponding to the x axis, project all the vertices of the intraoral digital three-dimensional model to this space, and express the x value of each point as "depth";

[0038] (2) Construct a two-dimensional depth map: construct a pixel grid in the rotated yz plane, rasterize all the facets on the intraoral digital three-dimensional model, fill in the depth value of the minimum x facet, and store the structure as C, C contains a list of pixels m X n, and the structure of the list node is {face: the facet with the minimum depth value of the current pixel, x: the depth value, m: the index of pixel m, n: the index of pixel n};

[0039] (3) Judge whether the undercut is visible: judge whether each vertex on the intraoral digital three-dimensional model is blocked by a certain face with an x value less than the existing value on C, mark the points with an x value less than the existing value on C to judge whether they are in the undercut area;

[0040] (4) Calculate the undercut depth for each point in the undercut region, find the nearest boundary in the 2D depth map, calculate the distance between the point and the boundary in the plane perpendicular to the undercut direction as the undercut depth;

[0041] (5) Connectivity filtering: remove isolated undercut regions recursively to enhance robustness;

[0042] (6) Filter the surface based on the detection results, only keep the surface that does not belong to the undercut region for subsequent filling;

[0043] (7) Extend the point cloud to generate a supplementary depth map, construct a new point set cloud_points by interpolating or projecting the surrounding points to generate x coordinates for the blank area with high precision;

[0044] (8) Generate connecting surfaces, group cloud_points by 2D position to form small regions, and construct triangular surfaces that pass the condition check, which includes: avoid creating surfaces with an area of 0, avoid surface intersection, avoid self-intersection with existing boundaries, if a conflict occurs, discard the construction of this surface;

[0045] (9) Generate the bottom surface, detect the points in all cloud_points that are at the maximum depth to form the bottom surface floor_faces;

[0046] (10) Generate the undercut model, merge the non-undercut surface, connecting surface, and bottom surface region to generate the undercut model, and perform local smoothing on the connecting surface and bottom surface region;

[0047] (11) Assign depth information to all vertices on the intraoral digital 3D model for display, calculate the vertices with depth information of 0, and connect them in sequence to generate the observation line of the intraoral digital 3D model.

[0048] Further, the inference model construction unit is based on a deep learning two-dimensional stent planning scheme inference model construction, and the specific working process is as follows:

[0049] (1) Train the inference model: based on the Transformer deep learning model, construct an end-to-end removable partial denture two-dimensional planning scheme inference model, the model training input is a patient information feature matrix, and the model training output is a removable partial denture two-dimensional planning scheme inference model, wherein the model training adopts a supervised learning method, generates the patient information feature matrix from the intraoral digital three-dimensional undercut model, intraoral situation and patient information, evaluates the loss through cross-validation and planning scheme, and continuously optimizes the model parameters to improve the accuracy of the design;

[0050] (2) The reasoning model is applied to the intelligent two-dimensional planning scheme. Based on the input intraoral digital three-dimensional model and intraoral condition, the AI ​​automatically plans, generates an initial scheme and displays it visually, and finally reviews, fine-tunes and confirms it.

[0051] Furthermore, the specific steps for training the inference model using the inference model building unit are as follows:

[0052] (1) Data preparation;

[0053] (2) Propagation forward;

[0054] (3) Loss calculation;

[0055] (4) Backpropagation and optimization;

[0056] (5) Validation set evaluation;

[0057] (6) Model saving.

[0058] further,

[0059] (1) Data preparation, specifically the processing of training data, including determining the oral scan sequence: the 2D feature sequence Xoral∈Rn×128 after 3D point cloud segmentation, where n is the number of teeth and / or regions; determining clinical parameters: the vectorized Kennedy classification, and patient data for loose teeth Xclin∈R64; determining label data: determining the standard scheme Y∈Rn×5 containing clasp type and position parameters.

[0060] (2) Forward propagation, specifically self-attention encoding, including the oral sequence passing through a Transformer layer with relative position encoding to capture the relationship between teeth; determining cross attention: clinical parameters interact with oral features as key / value pairs to generate conditional features; output prediction: linear layers are mapped to design parameters Ypred;

[0061] (3) Loss calculation, specifically multi-objective optimization, including: multi-objective loss function L=λ1L cls +λ2L reg ; Ring type classification loss: cross-entropy loss L cls For N samples and C categories, the formula is:

[0062]

[0063] y_{i,c}: The true label of the i-th sample. The value is 1 if the category is c, and 0 otherwise.

[0064] p_{i,c}: The probability that the model predicts the i-th sample belongs to class c;

[0065] Coordinate / size regression loss: Smooth L1 loss Lreg, formula:

[0066]

[0067] x: predicted value; y: true value;

[0068] j: regression target dimension, center coordinates x, y, width w, height h;

[0069] N: sample size;

[0070] (4) Back propagation and optimization, specifically to determine the optimizer: AdamW, learning rate 1e-4, weight decay 0.01; Gradient clipping is performed: limit the gradient norm ≤2.0, prevent explosion;

[0071] (5) Validation set evaluation, specifically geometric accuracy evaluation, functional compliance evaluation, and model stability evaluation;

[0072] (6) Model saving, specifically obtaining a two-dimensional bracket planning scheme inference model that meets the evaluation conditions, and saving the inference model.

[0073] Further, the inference model construction unit is also used to determine whether the placed clasp conforms to the design concept according to the selection of the abutment undercut condition, the design concept being to avoid the situation that the clasp retention force is not enough or the clasp is difficult to dislodge due to incorrect undercut, and the software automatically prompts to replace the clasp that does not conform to the design concept and recommends a clasp that conforms to the design concept. Finally, it is determined whether to perform hand-drawing adjustment according to the above judgment result.

[0074] Further, the inference model construction unit performs information labeling on the clasp that conforms to the design concept according to the patient's oral situation.

[0075] The application also provides a handheld portable oral cavity analysis device, comprising:

[0076] A handheld intraoral scanner is used to acquire real-time three-dimensional topographic data in the patient's oral cavity, and the three-dimensional topographic data includes high-precision three-dimensional point cloud information of teeth, gums and occlusal surfaces.

[0077] The handheld portable oral digital model observation and analysis system;

[0078] Computer hardware;

[0079] The software of the handheld portable oral digital model observation and analysis system is installed on the computer for operation;

[0080] The software of the handheld portable oral digital model observation and analysis system is provided with a three-dimensional digital modeling engine and intelligent analysis algorithm, data obtained by the handheld intraoral scanner is subjected to digital modeling to generate an intraoral digital three-dimensional model, and the intraoral digital three-dimensional model of the upper and lower dentition is rendered in real time, interactive operations such as arbitrary angle free rotation, zooming, sectioning and measuring of the intraoral digital three-dimensional model are supported, automatic detection of undercut is performed, and the position, depth and angle of the undercut are automatically displayed, and an observation line corresponding to the observation angle and direction is generated.

[0081] The method has the following advantages:

[0082] The optical scanning device is used to quickly and non-contactly obtain three-dimensional digital information of soft and hard tissues in the mouth of a patient, professional doctor operation is not needed, the whole process needs 2-3 minutes, the patient feels comfortable, and there is no foreign body sensation.

[0083] Since the three-dimensional true color digital model of the mouth of the patient is obtained, no alginate impression, no gypsum tooth model and other consumables are needed, and the gypsum model does not need to be installed on a mechanical model observation instrument or a desktop three-dimensional optical scanner.

[0084] In the software system, the position and shape of the undercut of all soft and hard tissues in the mouth, the depth and angle of the undercut and other specific values can be intuitively displayed. A guide line is also displayed, and the position of the guide line can dynamically change with different observation angles. All the depths of the undercut are distinguished by colors, and the depth of the undercut changes when the observation angle changes. The whole process is dynamically visible. The observation position can be changed and adjusted at will, and the measurement can be repeated without handwriting recording.

[0085] The generated digital shape can be three-dimensionally printed at any time. The digital information is convenient for mass storage, analysis, design and remote communication. The invention can simultaneously display the true color three-dimensional occlusion of the upper and lower dentition and the influence of the occlusion on the observation angle.

[0086] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the content of the specification can be implemented, and in order to make the above and other purposes, characteristics and advantages of the present application more obvious and easy to understand, the following preferred embodiments are described in detail below, and the accompanying drawings are described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0087] Figure 1 A diagram for taking a model with impression material and then pouring a gypsum model is shown in the following figure: Figure 1 A diagram for taking a model with impression material and then pouring a gypsum model is shown in the following figure: Figure 1 B is a diagram for pouring a tooth model with gypsum;

[0088] Figure 2 A diagram for installing a gypsum model on a model observation instrument to manually draw a line is shown in the following figure:Figure 2 A is a schematic diagram of installing the gypsum model on a mechanical model observation instrument and designing a fixed observation angle, Figure 2 B is a schematic diagram of manually marking the model, drawing the model high point, and recording the undercut angle value;

[0089] Figure 3 For the existing model observation technology, the gypsum model is installed on a large desktop scanning instrument for optical scanning, and the undercut data is marked and measured in the software as shown in the schematic diagram: Figure 3 A is a schematic diagram of placing the gypsum model on a desktop optical scanning device and designing a fixed observation angle, Figure 3 B is a schematic diagram of directly marking the digital model in the system, drawing the model high point, and recording the undercut angle value;

[0090] Figure 4 A schematic diagram of a handheld portable oral digital model observation and analysis system;

[0091] Figure 5 A schematic diagram of scanning all the teeth and gingival tissue in the patient's oral cavity to capture digital images in the patient's mouth;

[0092] Figure 6 A schematic diagram of presenting precise digital information such as guide lines and undercut depth and angle on the digital dentition;

[0093] Figure 7 A two-dimensional depth map in the step of generating an undercut model;

[0094] Figure 8 A training and inference flowchart;

[0095] Figure 9 A data flow schematic diagram;

[0096] Figure 10 A schematic diagram of selecting a common abutment for the denture;

[0097] Figure 11 A schematic diagram of generating a two-dimensional denture design according to the undercut depth and the missing tooth condition. DETAILED DESCRIPTION

[0098] The technical solutions of the present application will be described in detail below in conjunction with specific embodiments, but those skilled in the art should understand that the embodiments described below are only used to illustrate the present application and should not be regarded as limiting the scope of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0099] The prior art is generally a mechanical or optical digital model observation technology outside the oral cavity. The present application is an optical digital rapid intraoral observation technology inside the oral cavity. The present application is a complete clinical practical technology. The technology is upgraded and innovated by means of the existing intraoral scanning equipment, so as to expand it into a new tool.

[0100] The prior art needs a skilled doctor to prepare the soft and hard tissue morphology in the patient's mouth with a smelly impression material. The patient feels bad, has obvious nausea and foreign body sensation, and the whole process takes 3-5 minutes. The present application uses an optical scanning device to quickly and non-contactly obtain the three-dimensional digital information of the soft and hard tissue in the patient's mouth. No professional doctor is needed, and the whole process takes 2-3 minutes. The patient feels comfortable and has no foreign body sensation.

[0101] The prior art needs a doctor or technician to pour the prepared impression into a dental model with gypsum. The whole process takes about 30-40 minutes. The maxillary or mandibular dental model also needs to be installed on a mechanical model observation instrument or a desktop optical scanning device to obtain a three-dimensional digital model. The present application obtains a three-dimensional digital model in the patient's mouth, so it can only display the three-dimensional true color digital morphology of the full dentition in the mouth in the software. No gypsum dental model needs to be poured, and no mechanical model observation instrument needs to be installed.

[0102] The prior art needs a skilled doctor or technician to draw lines and make marks on the maxillary or mandibular dental model with a pencil or a tracing needle on the model observation instrument to determine the most convex point of the tooth or soft tissue shape. Then a measuring instrument is used to mark the depth and angle of the undercut at the measured position, and the data is recorded with paper and pen. Once the observation angle position and guide line are set, it is not easy to change. If it needs to be changed, it needs to be re-drawn on the model. The present application obtains the three-dimensional true color digital morphology of the upper and lower dentition in the mouth, so the present application provides a model observation analysis system. The system is improved and innovated based on the existing system, integrates the original data collection and CAD design software, realizes the basic functions of data storage, management, query, display, download, sharing, etc., and provides users with online design, remote assistance design, collaborative design, etc. The model observation analysis system of the present application can directly display the specific values of the position and shape of the undercut, the depth and angle of the undercut of all soft and hard tissues in the mouth. This function integrates the model observation analysis function in the oral data collection software based on the existing three-dimensional modeling technology, which makes up for the defect that the existing data collection software cannot directly reflect the guide line position and undercut shape.

[0103] The guide wire is displayed simultaneously, and the position of the guide wire can also dynamically change with the observation angle, all the undercut depths are distinguished by colors, when the observation angle changes, the undercut depths also change, and the whole process is dynamically visible. The observation position can be changed and adjusted arbitrarily, and the measurement can be repeated without handwriting records. The existing oral three-dimensional data acquisition software cannot directly perform model observation and guide wire drawing, and needs to transmit data to CAD design software for analysis and processing, which makes it difficult for doctors and technicians to communicate. The system innovatively integrates data acquisition and model observation functions, thereby simplifying the series of processes and facilitating communication between the doctor end and the technician end.

[0104] The prior art is to observe and analyze by using a plaster model, and both mechanical model observation instruments and desktop optical scanners need to take the plaster model. Since the plaster model is large in size, heavy in weight, needs to be sterilized and stored, and is not easy to store in large quantities, it cannot be replicated after being lost or damaged, and it is also difficult to quickly perform remote information exchange and transmission of the plaster model. Therefore, the present application does not need a plaster model, uses a full digital technology to record and save a true color three-dimensional digital image inside an oral cavity, and the generated digital form can be three-dimensionally printed at any time. The digital information is convenient for large storage, analysis and design, and remote communication.

[0105] The prior art needs about 50 minutes from taking the model, pouring the model, and mapping the model, and the present application needs 5-8 minutes by using an oral scanning device.

[0106] The prior art cannot simultaneously observe the undercut conditions of the upper and lower tooth rows and the influence of the occlusion condition on the observation angle. The present application can simultaneously display the true color three-dimensional occlusion condition of the upper and lower tooth rows and the influence of the occlusion condition on the observation angle.

[0107] Embodiment 1

[0108] The present application can quickly obtain the three-dimensional digital form of the soft and hard tissues in the oral cavity of a patient before surgery, and the digital information of the path, the guide wire, the undercut depth and angle, thereby providing a design basis for denture design and production. Therefore, the present application provides a handheld portable oral digital model observation and analysis system, as shown in Figure 4 The system includes a data capture unit, a data processing unit, an undercut model generation unit, an inference model construction unit, an information display unit and a data storage unit.

[0109] The data capture unit is used to capture the data in the mouth of a patient.

[0110] Specifically, first, a handheld three-dimensional optical scanning device (hereinafter referred to as an "intraoral scanner (IOS)") is used to scan all tooth rows and gingival tissues inside the oral cavity of a patient.

[0111] The data capture unit captures digital images in the patient's mouth, and clearly displays a color three-dimensional digital information image model of the upper and lower teeth (hereinafter referred to as "intraoral digital three-dimensional model") on the display, which can be rotated, enlarged or reduced at any angle, as shown in Figure 5

[0112] The data processing unit is used for patient filing and health report generation, and adjusts the model observation angle according to the oral condition. In the adjusted model, precise full-arch digital information such as guide lines, undercut depth, and angle is displayed on the digital dentition. At the same time, the undercut depth is distinguished by color in the full-arch digital information, and the deeper the color, the greater the undercut angle.

[0113] The specific working process of the data processing unit is as follows:

[0114] (1) Establish a new patient intraoral data storage file, or use the default storage file. After the patient's intraoral data is captured, the corresponding folder scanning data will be automatically obtained according to the storage file location and path and automatically named.

[0115] (2) According to the patient information, data classification and patient filing are carried out, and patient information and repair information are input. There are two ways to import model scanning data: 1) Open the new capture data in data management, add a new storage path, and select file data for classification; 2) Open the classified data in data management, create new data, upload local data, and input patient information.

[0116] (3) After confirming the patient filing, select the model scanning data, and the small window displays the imported model scanning data.

[0117] The undercut model generation unit is used for undercut detection and generation of intraoral digital undercut model.

[0118] The intraoral digital three-dimensional model can be rotated, enlarged or reduced at any angle.

[0119] The undercut model generation unit automatically displays the position, depth, and angle of the undercut according to the observation angle and direction, and displays the corresponding observation line. The greater the rotation angle, the greater the undercut filling amount, and the smaller the rotation angle, the smaller the undercut filling amount. The system defaults the undercut angle to 3°.

[0120] ​The reverse concave model generating unit automatically fills in the reverse concave when observing the model, and displays the position, slope, depth and position of the observation line of the reverse concave. The depth of the reverse concave is distinguished by color, the deeper the depth of the reverse concave, the more the color deviates from red, the shallower the depth of the reverse concave, the more the color deviates from yellow, the reverse concave is filled in using "virtual red wax", the slope of the reverse concave is observed from the side of the model, the model color and the reverse concave color have a clear contrast, and the intersection of the two colors is the position of the observation line, as shown in Figure 6 At the same time, the intraoral digital three-dimensional model can be cut and measured, helping doctors to more intuitively understand the specific situation of the abutment reverse concave.

[0121] The reverse concave model generating unit is also used for observing the intraoral digital three-dimensional model, and sequentially performing "determining common seating path", "reverse concave depth and slope display" and "observation line display" processing on the three-dimensional model according to the observation result.

[0122] Among them, the determination of the common seating path is to extract a direction vector data (dir[x, y, z]) as the direction of the common seating path of the three-dimensional model;

[0123] The reverse concave model generating unit performs reverse concave detection and generates a reverse concave model on the three-dimensional model through the "reverse concave depth and slope display" and "observation line display" processing, as shown in Figure 7 .

[0124] The specific working process of the reverse concave model generating unit is as follows:

[0125] (1) The reverse concave model generating unit sets the direction and space transformation, rotates the system three-dimensional space to the observation direction dir corresponding to the x axis. Project all the vertices of the three-dimensional model to this space, and represent the x value of each point as "depth";

[0126] (2) The reverse concave model generating unit constructs a two-dimensional depth map, constructs a pixel grid in the rotated yz plane, rasterizes all the facets on the intraoral digital three-dimensional model, fills in the depth value of the minimum x facet, and stores the structure as C. C contains a list of pixel m X n, and the structure of the list node is {face: the facet with the minimum depth of the current pixel, x: depth value, m: index of pixel m, n: index of pixel n}

[0127] (3) The reverse concave model generating unit judges the visibility (whether it is a reverse concave), judges whether each vertex on the three-dimensional model is blocked by a certain face (i.e. whether the x value of each point is greater than the existing value on C) and marks whether the point is in the reverse concave area;

[0128] (4) The undercut model generation unit calculates the undercut depth of each point in the undercut region. The distance between the point and the boundary in the vertical plane perpendicular to the undercut direction is calculated as the undercut depth by searching for the nearest boundary in the two-dimensional depth map.

[0129] (5) The undercut model generation unit filters connectivity by recursively removing isolated undercut regions to enhance robustness.

[0130] (6) The undercut model generation unit filters the surface based on the detection results, and only keeps the surface that does not belong to the undercut region for subsequent filling.

[0131] (7) The undercut model generation unit extends the point cloud to construct a supplementary depth map with high precision, and generates x coordinates by surrounding interpolation or projection for the blank area to generate a new point set (cloud_points).

[0132] (8) The undercut model generation unit generates connecting surfaces, and constructs triangular surfaces that pass the condition check by grouping cloud_points into small regions according to two-dimensional positions. The condition check includes: avoiding creating a surface with an area of 0, avoiding surface intersection, and avoiding self-intersection with existing boundaries. If a conflict occurs, the surface is abandoned.

[0133] (9) The undercut model generation unit generates a floor surface by detecting points in all cloud_points that are at the maximum depth to form a floor surface (floor_faces).

[0134] (10) The undercut model generation unit generates an undercut model by merging non-undercut surface patches, connecting surfaces, and floor surface regions to generate an undercut model, and performs local smoothing on the connecting surface and the floor surface region.

[0135] (11) The undercut model generation unit assigns depth information to all vertices on the three-dimensional model for display. Meanwhile, it calculates the vertices with depth information of 0 and connects them in sequence to generate observation lines.

[0136] The inference model construction unit is used to construct a two-dimensional planning scheme inference model and perform model optimization evaluation.

[0137] The inference model construction unit constructs a two-dimensional stent planning scheme inference model based on the results of three-dimensional observation of the intraoral digital model, completes scheme selection, ring design, and connector design.

[0138] The inference model construction unit constructs a two-dimensional stent planning scheme inference model based on deep learning, and the specific working process is as follows:

[0139] The inference model construction unit trains the inference model.

[0140] The inference model construction unit is specifically used for designing an end-to-end removable partial denture two-dimensional planning scheme inference model based on a Transformer deep learning model. The model training inputs a patient information feature matrix, and outputs a removable partial denture two-dimensional planning scheme inference model. The model training adopts a supervised learning method, generates a patient information feature matrix from an intraoral digital undercut model, an intraoral situation and patient information, evaluates a loss through cross-validation and a planning scheme, and continuously optimizes model parameters to improve the design accuracy.

[0141] As shown in Figure 8 , the specific steps of training the inference model by the inference model construction unit are as follows:

[0142] (1) Data preparation;

[0143] Training data processing:

[0144] Determine the oral scan sequence: the 2D feature sequence Xoral of the 3D point cloud after slicing, Xoral ∈ Rn×128 (n is the number of teeth / regions).

[0145] Determine the clinical parameters: the vectorized patient data Xclin, Xclin ∈ R64 (such as Kennedy classification, tooth looseness, etc.).

[0146] Determine the label data: the standard scheme Y designed by an expert, Y ∈ Rn×5 (including clasp type, position and other parameters).

[0147] The data flow is as shown in Figure 9 :

[0148] (2) Forward propagation;

[0149] Self-attention encoding: the oral sequence passes through a Transformer layer with relative position encoding to capture the relationship between teeth.

[0150] Cross-attention: the clinical parameters are used as Key / Value to interact with the oral features to generate conditional features.

[0151] Output prediction: linear layer mapping to design parameters Y pred .

[0152] (3) Loss calculation, specifically multi-objective optimization, including: multi-objective loss function L = λ1L cls + λ2L reg ; clasp type classification loss: cross-entropy loss L cls , for N samples and C categories, formula:

[0153]

[0154] yi,c: the true label of the ith sample (value 1 if the class is c, otherwise 0).

[0155] pi,c: the probability that the model predicts the ith sample belongs to class c;

[0156] Coordinate / size regression loss: Smooth L1 loss L reg , formula:

[0157]

[0158] x: predicted value; y: true value;

[0159] j: regression target dimension (center coordinates x, y, width w, height h, etc.);

[0160] N: sample size.

[0161] (4) Backpropagation and optimization;

[0162] Optimizer: AdamW (learning rate 1e-4, weight decay 0.01).

[0163] Gradient clipping: limit the gradient norm ≤ 2.0 to prevent explosion.

[0164] (5) Validation set evaluation;

[0165] Geometric accuracy evaluation:

[0166] Key point positioning error (KPE), formula:

[0167] Clinical significance: accuracy of clasp / retainer position, pass threshold: <1.0mm

[0168] The meanings and values of each parameter in the above key point positioning error (KPE) calculation formula are shown in Table 1.

[0169] Table 1 Parameter definitions in the key point positioning error (KPE) calculation formula

[0170]

[0171] Size relative error (SRE), formula:

[0172] Clinical significance: reasonableness of denture component size, pass threshold: <10%

[0173] The meanings and values of each parameter in the above size relative error (SRE) calculation formula are shown in Table 2.

[0174] Table 2 Parameter definitions in the size relative error (SRE) calculation formula

[0175]

[0176]

[0177] Functional compliance evaluation: the evaluation method and description are shown in Table 3.

[0178] Table 3 Functional compliance evaluation method and description

[0179]

[0180] Model stability evaluation: the evaluation method and description are shown in Table 4.

[0181] Table 4 Model stability evaluation method and description

[0182]

[0183] (6) Obtain a two-dimensional support planning scheme reasoning model that meets the evaluation conditions, and save the model.

[0184] The reasoning model construction unit applies the reasoning model to the intelligent two-dimensional planning scheme. According to the inputted intraoral digital three-dimensional model and intraoral conditions, etc., the AI automatically plans, generates an initial scheme, and visualizes the scheme through software. Finally, the scheme is audited, fine-tuned, and confirmed.

[0185] The reasoning model construction unit determines whether the clasp placed according to the selection of the abutment undercut meets the design concept, which is to avoid the situation that the clasp retention force is not enough due to incorrect small undercut or the clasp is difficult to be seated and dislocated due to incorrect large undercut. The software will automatically prompt to replace the inappropriate clasp and recommend a suitable clasp. Finally, it is determined whether to perform hand-drawing adjustment according to the situation, and information is marked at the corresponding position according to the patient's intraoral situation, such as Figure 10

[0186] The information display unit is used for information display process, which is divided into two parts: design scheme display and tooth preparation information display. The design scheme display part directly presents the completed two-dimensional support scheme through a visual interface, including support structure, connector layout, and clasp position, etc. The tooth preparation information display part intelligently generates detailed tooth preparation guidance scheme including tooth preparation amount, preparation form, seating path direction, etc. based on the clasp type (such as bar clasp, round clasp, etc.) and corresponding tooth position in the design scheme, providing precise technical guidance for clinical operation.

[0187] The data storage unit is used for automatically saving data into classified data, facilitating subsequent design and modification.

[0188] ​The data storage unit checks whether the tooth preparation in the patient's mouth is in place according to the result of three-dimensional observation, and ensures the rationality of the repair scheme. Finally, the scanning data module file is uploaded to the Internet.

[0189] If the common seating path of the selected digital dentition model is not conducive to the insertion of the denture, that is, the guide wire position and the depth of the undercut do not meet the requirements, the observation angle is adjusted, the common seating path is reselected, and the adjustment is performed until an optimal angle is selected.

[0190] After the observation angle is determined, the operator can design the denture support form according to the guide wire and the undercut angle size marked on the dentition. This information can guide the operator to prepare the abutment in the patient's mouth. The obtained guide wire position and undercut depth data can also generate a two-dimensional denture design graph according to the undercut depth and the missing tooth condition, as shown in Figure 11 The operator can mark information on the software or can hand-draw adjustment. The generated three-dimensional design graph can be remotely communicated, analyzed, stored, and used for various types of digital additive or subtractive manufacturing.

[0191] Embodiment 2

[0192] The application provides a handheld portable oral cavity analysis device, which is an efficient and intelligent oral cavity digital diagnosis and treatment tool. The system realizes the handheld portable oral cavity analysis device, and comprises:

[0193] The handheld intraoral scanner is used for acquiring three-dimensional topographic data in the patient's oral cavity in real time. The three-dimensional topographic data includes high-precision three-dimensional point cloud information of teeth, gums and occlusal surfaces.

[0194] The handheld portable oral cavity digital model observation and analysis system is an analysis system composed of multiple units in the embodiment 1 of the application.

[0195] The computer hardware specifically comprises:

[0196] CPU: Intel Core i7 or higher, to ensure smooth modeling calculation;

[0197] GPU: NVIDIA RTX 2060 and above, to guarantee 3D rendering performance;

[0198] Memory: 32 GB, to improve multitasking processing capacity;

[0199] Hard disk: 1 TB NVMe SSD, to accelerate data reading and writing;

[0200] Display: 2K, 2560x1440 and above resolution, to clearly present details.

[0201] The handheld portable oral digital model observation and analysis system is internally provided with a three-dimensional digital modeling engine and an intelligent analysis algorithm, data obtained by the handheld intraoral scanner is subjected to digital modeling, and three-dimensional digital information images of upper and lower tooth rows are rendered in real time, interactive operations of arbitrary angle free rotation, scaling, sectioning and measurement of the intraoral digital three-dimensional model are supported, automatic detection of undercut is performed, and the position, depth and angle of the undercut are automatically displayed to generate corresponding angle and direction observation lines.

[0202] Although the present application has been described in detail with general description and specific embodiments above, some modifications or improvements can be made on the basis of the present application, which is obvious to those skilled in the art. Therefore, these modifications or improvements made on the basis of not deviating from the spirit of the present application are within the scope of the present application.

Claims

1. A handheld portable oral digital model observation and analysis system, comprising a data capture unit, a data processing unit, an undercut model generation unit, an inference model construction unit, an information display unit, and a data storage unit; in, The data capture unit is used to capture intraoral data of the patient and clearly display a color digital three-dimensional model of the upper and lower teeth in the mouth on the monitor. The intraoral digital three-dimensional model can be rotated and zoomed in or out at any angle. The data processing unit is used to create patient files and generate health reports, and adjust the observation angle of the intraoral digital three-dimensional model according to the oral cavity condition. In the adjusted intraoral digital three-dimensional model, the digital dentition displays accurate digital information of the entire dentition, including guide wires, undercut depth, and angle. The digital information of the entire dentition is distinguished by color, with darker colors indicating larger undercut angles. The indentation model generation unit is used to detect indentations and generate intraoral digital indentation models. Specifically, it is used to rotate and zoom out the intraoral digital three-dimensional model, and automatically display the position, depth, and angle of the indentation according to the observation angle and direction, and display the observation line corresponding to the observation angle and direction. The larger the rotation angle, the greater the amount of indentation to be filled; the smaller the rotation angle, the smaller the amount of indentation to be filled. When performing three-dimensional observation of the intraoral digital three-dimensional model, the indentation is automatically filled, and the position, slope, depth, and position of the observation line of the indentation are displayed. The reasoning model building unit is used to build a two-dimensional planning scheme reasoning model and perform model optimization and evaluation. Specifically, it is used to build a two-dimensional stent planning scheme reasoning model based on deep learning according to the intraoral digital inverted concave model, and complete the scheme selection, clasp design and connector design. The information display unit is used for design scheme display and tooth preparation information display. The design scheme display is a visual presentation of the completed two-dimensional framework scheme through a visual interface. The tooth preparation information display is based on the clasp type of bar clasp and round clasp determined in the design scheme and their corresponding tooth positions. It intelligently generates a detailed tooth preparation guidance scheme including the amount of tooth preparation, preparation shape and placement path direction, providing precise technical guidance for clinical operation. The data storage unit is used to automatically save data to the classified data, which facilitates subsequent design and modification; If the common path of insertion of the selected intraoral digital 3D model is not conducive to denture insertion, that is, the position of the guide wire and the depth of the undercut do not meet the requirements, then adjust the observation angle and reselect the common path of insertion until an optimal angle is selected.

2. The handheld portable oral digital model observation and analysis system according to claim 1, characterized in that, The specific workflow of the data processing unit is as follows: (1) Create a new patient intraoral data storage file, or use the default storage file. After the patient intraoral data is captured, the scan data in the corresponding folder will be automatically obtained and named according to the storage file location and path. (2) Based on patient information, classify data and create patient files, and input patient information and repair information; (3) After the patient's file is filled out and confirmed, select the inverted concave model option and the intraoral digital three-dimensional model scan data. The imported intraoral digital three-dimensional model scan data will be displayed in a small window.

3. The handheld portable oral digital model observation and analysis system according to claim 1, characterized in that, When the inverted concave model generation unit performs model observation, it automatically fills in the inverted concave areas and displays the location, slope, depth, and position of the observation line. The depth of the inverted concave areas is distinguished by color. At the same time, "virtual red wax" is used to fill in the inverted concave areas. The slope of the inverted concave areas is displayed in the side view of the digital 3D model inside the mouth. The color of the digital 3D model inside the mouth has a sharp contrast with the color of the inverted concave areas. The place where the two colors meet is the position of the observation line.

4. The handheld portable oral digital model observation and analysis system according to claim 3, characterized in that, The inverted concave model generation unit is also used to observe the intraoral digital three-dimensional model, and according to the observation results, to sequentially determine the common placement path, display the inverted concave depth and slope, and display the observation line of the intraoral digital three-dimensional model. The determination of the common placement path involves extracting a direction vector data (dir[x,y,z]) as the direction of the common placement path of the intraoral digital 3D model; The indentation model generation unit performs indentation detection and generates an indentation model from the intraoral digital 3D model through indentation depth and slope display and observation line display processing. The specific workflow is as follows: (1) Set direction and spatial transformation: Rotate the system's three-dimensional space to the x-axis corresponding to the observation direction dir, project all vertices of the digital three-dimensional model inside the mouth onto this system's three-dimensional space, and represent the "depth" by the x-value of each point; (2) Construct a two-dimensional depth map: Construct a pixel grid in the rotated yz plane, rasterize all the faces on the digital three-dimensional model inside the mouth, fill in the depth value of the smallest x face, and store the structure as C, which contains a list of pixels m x n. The structure of the list node is {face: the face with the smallest current pixel depth, x: depth value, m: index of pixel m, n: index of pixel n}; (3) Determine the visibility of the under-concave area: Determine whether each vertex on the digital 3D model inside the mouth is occluded by a face whose x value is less than the existing value on C, and mark the points whose x value is less than the existing value on C to determine whether they are in the under-concave area. (4) Calculate the indentation depth for each point in the indentation area. By finding the nearest boundary in the two-dimensional depth map, calculate the distance between the point and the boundary on the plane perpendicular to the indentation direction, and use it as the indentation depth. (5) Connectivity filtering: Recursively remove isolated inverted regions to enhance robustness; (6) Based on the detection results, only surfaces that do not belong to the concave area are retained for subsequent filling; (7) Extend the generated point cloud to construct a supplementary depth map with high precision, perform surrounding interpolation or projection on the blank area to generate x coordinates, and generate a new point set cloud_points; (8) Generate connecting surfaces: The cloud_points are arranged into small regions according to their two-dimensional positions, and triangular surfaces that pass the condition check are constructed. The condition check includes: avoiding the creation of surfaces with an area of ​​0, avoiding surface intersections, avoiding self-intersection with existing boundaries. If a conflict occurs, the construction of this surface is abandoned. (9) Generate floor surfaces: Detect all points in the cloud_points that are at the maximum depth and form floor surfaces; (10) Generate an inverted concave model: merge the non-inverted concave surfaces, connecting surfaces, and bottom surfaces to generate an inverted concave model, and perform local smoothing on the connecting surfaces and bottom surfaces; (11) Assign depth information to all vertices on the intraoral digital three-dimensional model, display them, calculate the vertices with a depth information of 0, and connect the vertices with a depth information of 0 in sequence to generate the intraoral digital three-dimensional model observation line.

5. The handheld portable oral digital model observation and analysis system according to claim 1, characterized in that, The reasoning model building unit is based on the construction of a two-dimensional scaffolding planning scheme reasoning model using deep learning. The specific working process is as follows: (1) Training the inference model: Based on the Transformer deep learning model, an end-to-end two-dimensional planning scheme inference model for removable partial dentures is constructed. The model training input is the patient information feature matrix, and the model training output is the two-dimensional planning scheme inference model for removable partial dentures. The model training adopts a supervised learning method, which generates the patient information feature matrix from the intraoral digital three-dimensional undercut model, intraoral condition and patient information. The loss is evaluated by cross-validation and planning scheme, and the model parameters are continuously optimized to improve the accuracy of the design. (2) The reasoning model is applied to the intelligent two-dimensional planning scheme. Based on the input intraoral digital three-dimensional model and intraoral condition, the AI ​​automatically plans, generates an initial scheme and displays it visually, and finally reviews, fine-tunes and confirms it.

6. The handheld portable oral digital model observation and analysis system according to claim 5, characterized in that, The specific steps for training the inference model using the inference model construction unit are as follows: (1) Data preparation; (2) Propagation forward; (3) Loss calculation; (4) Backpropagation and optimization; (5) Validation set evaluation; (6) Model saving.

7. The handheld portable oral digital model observation and analysis system according to claim 6, characterized in that, (1) Data preparation, specifically the processing of training data, including determining the oral scan sequence: the 2D feature sequence Xoral∈Rn×128 after 3D point cloud segmentation, where n is the number of teeth and / or regions; determining clinical parameters: the vectorized Kennedy classification, and patient data for loose teeth Xclin∈R64; determining label data: determining the standard scheme Y∈Rn×5 containing clasp type and position parameters. (2) Forward propagation, specifically self-attention encoding, including the oral sequence passing through a Transformer layer with relative position encoding to capture the relationship between teeth; determining cross attention: clinical parameters interact with oral features as Key / Value to generate conditional features; Output prediction: Linear layer mapped to design parameter Y pred ; (3) Loss calculation, specifically multi-objective optimization, including: multi-objective loss function L=λ1L cls +λ2L reg ; Ring type classification loss: cross-entropy loss L cls For N samples and C categories, the formula is: y_{i,c}: The true label of the i-th sample. The value is 1 if the category is c, and 0 otherwise. p_{i,c}: The probability that the model predicts the i-th sample belongs to class c; Coordinate / Size Regression Loss: Smooth L1 Loss L reg ,formula: x: predicted value; y: actual value; j: Target dimension for regression, center coordinates x, y, width w, height h; N: Sample size; (4) Backpropagation and optimization, specifically: determining the optimizer: AdamW, learning rate 1e-4, weight decay 0.01; performing gradient clipping: limiting the gradient norm to ≤2.0 to prevent explosion; (5) Validation set evaluation, specifically geometric accuracy evaluation, functional compliance evaluation, and model stability evaluation; (6) Model saving, specifically, obtaining a two-dimensional support planning scheme reasoning model that meets the evaluation conditions and saving the reasoning model.

8. The handheld portable oral digital model observation and analysis system according to claim 5, characterized in that, The reasoning model construction unit is also used to determine whether the placed clasp conforms to the design concept based on the selection of the undercut of the abutment tooth. The design concept is to avoid insufficient clasp retention or clasp dislocation due to incorrect undercut. For clasps that do not conform to the design concept, the software automatically prompts for replacement and recommends clasps that conform to the design concept. Finally, based on the above judgment results, it is selected whether to make hand-drawn adjustments.

9. The handheld portable oral digital model observation and analysis system according to claim 8, characterized in that, The reasoning model construction unit marks information at the clasp location that conforms to the design concept, based on the patient's intraoral condition.

10. A handheld portable oral cavity analysis device, comprising: A handheld intraoral scanner is used to acquire real-time three-dimensional topographic data of the patient's oral cavity, including high-precision three-dimensional point cloud information of teeth, gums and occlusal surfaces. The handheld portable oral digital model observation and analysis system according to any one of claims 1-9; Computer hardware; The software of the handheld portable oral digital model observation and analysis system is installed and runs on the computer. The software of the handheld portable oral digital model observation and analysis system has a built-in three-dimensional digital modeling engine and intelligent analysis algorithm. It uses the data obtained from the handheld intraoral scanner to digitally model and generate an intraoral digital three-dimensional model. It also renders the intraoral digital three-dimensional models of the upper and lower teeth in real time. The system supports interactive operations such as free rotation, scaling, sectioning, and measurement of the intraoral digital three-dimensional model at any angle. It can also automatically detect undercuts, automatically display the position, depth, and angle of the undercuts, and generate observation lines with corresponding observation angles and directions.

Images