Method for characterizing intraoral organs

Abstract

The invention relates to a method for characterizing an intraoral organ to be characterized, comprising the following steps: 1) modeling the intraoral organ to be characterized in the form of a digital three-dimensional model or "model to be characterized", the model to be characterized comprising a mesh of points defining a surface; 2) setting the model to be characterized in a standardized configuration relative to a digital three-dimensional model of a reference intraoral organ, called "initial reference model", the initial reference model comprising a mesh of points called "initial reference points"; then 3) determining for each initial reference point a final reference point by means of a morphing algorithm, then determining a set of values determining the position of the final reference points and / or determining a final basic surface from the final reference points, the morphing algorithm determining the position of the final reference points so that the final reference model composed of the mesh of final reference points matches as closely as possible the model to be characterized; 4) generating a feature vector grouping in an ordered manner the values determined in step 3) for all the initial reference points.

Description

Technical Field

[0001] This invention relates to a method for characterizing intraoral organs.

[0002] The present invention also relates to methods for generating a database of dental models, methods for correcting dental models based on the database, methods for generating models of intraoral organs, methods for detecting shape abnormalities of intraoral organs, methods for identifying individuals, methods for evaluating attributes based on three-dimensional representations of intraoral organs, and methods for computerized compression of historical model databases. These methods advantageously implement the characterization methods according to the invention. Background Technology

[0003] Recent orthodontic treatments utilize 3D models of the dental arch, which in turn consist of 3D tooth models. These tooth models can be moved, especially for modeling future dental conditions, such as the alignment of teeth at the end of orthodontic treatment.

[0004] However, scanning an individual's dental arch at the initial time (usually at the start of orthodontic treatment) does not allow for the definition of a complete dental model, especially since some parts of the teeth being scanned may be hidden by other teeth or orthodontic appliances.

[0005] Therefore, when moving dental models, they may show "white areas" in regions that were not initially observable. Consequently, they especially do not allow for subsequent detection of tooth or gum deformities in these areas.

[0006] This issue is particularly critical when dental models are used to manufacture orthodontic appliances (e.g., in cases of reversing ineffective treatment) or to produce competing orthodontic appliances after treatment.

[0007] PCT / EP2019 / 052121 describes a method for allowing the restriction of white areas in a model, particularly a method for restricting white areas in a reference model described in WO2016066651. However, this method requires acquiring an image. Such images are not always available or are of satisfactory quality.

[0008] Therefore, it is necessary to limit the white area of ​​the dental model.

[0009] More generally, there is a need for methods to quickly search for missing information related to a dental model and / or to correct any errors in the dental model.

[0010] The purpose of this invention is to at least partially address these needs. Summary of the Invention

[0011] This invention relates to a method for characterizing intraoral organs, preferably teeth, to be characterized, the method comprising the following steps:

[0012] 1) Model the intraoral organ to be characterized in the form of a digital 3D model or a “model to be characterized”, wherein the model to be characterized includes a mesh of points defining the surface;

[0013] 2) The model to be characterized is set up in a normalized configuration relative to a digital 3D model (preferably a digital 3D model of an intraoral organ) called the “initial reference model,” the initial reference model comprising a grid of points called “initial reference points”; then

[0014] 3) A final reference point is determined for each initial reference point using a deformation algorithm, and then a set of values ​​is determined, which determines the position of the final reference point and / or the final basic surface is determined based on the final reference point. The final basic surface is preferably defined by interpolation based on the final reference point, using each free initial reference point and a number of pairs of associated (i.e., determined by the deformation algorithm based on the initial reference point) final reference points.

[0015] The deformation algorithm determines the position of the final reference point, so that the final reference model composed of the mesh of the final reference point matches the model to be represented as closely as possible;

[0016] 4) Generate feature vectors that group the values ​​determined for all the initial reference points in step 3) in an ordered manner.

[0017] In step 3), the final reference point is determined to correspond to a deformation of the initial reference model in order to obtain a final reference model that is as close as possible to the model to be represented, i.e., minimizing the shape difference between the two models. The deformation algorithm specifies the rules guiding the deformation and the criteria that allow the determination of the differences. These criteria allow the determination of the shape closeness between the deformed initial reference model and the model to be represented, thereby allowing the determination of a deformed initial reference model that matches the model to be represented "as closely as possible" according to these criteria specific to the deformation algorithm. Thus, the deformed initial reference model is the final reference model. The final reference model can vary depending on the deformation algorithm implemented for its acquisition.

[0018] In one implementation, the deformation algorithm deforms the mesh of the initial reference model by specifying constraints on the deformation, such as by setting the elasticity of the mesh. This deformation is thus referred to as "constrained deformation." Preferably, the closest possible match to the model to be represented is considered achieved when the cumulative distance between the final reference point and the surface of the model to be represented is minimized.

[0019] In a preferred embodiment, the deformation algorithm includes a function for projecting initial reference points onto the surface of the model to be characterized. Preferably, the deformation algorithm calculates points on the surface of the model to be characterized that are as close as possible to each initial reference point. The points on the surface of the model to be characterized determined in this way constitute the final reference points, wherein the final reference model matches the model to be characterized as closely as possible.

[0020] The deformation algorithm can specify that the final reference point is a point on the surface of the model to be represented, preferably a mesh point of the model to be represented.

[0021] Preferably, in step 3), the deformation algorithm is either a constrained deformation algorithm for the mesh of the initial reference model or an algorithm for projecting initial reference points onto the surface of the model to be represented. Unlike optimization methods (“best fit”), these algorithms do not provide an “optimal” set of final points for the final reference model for all initial point searches, i.e., the set of final points that are closest in absolute value to the model to be represented. The constraints specified by the constrained deformation algorithm effectively limit the possibilities of deformation, and the projection algorithm does not iteratively search for multiple final reference points, but rather searches for final reference points continuously for each initial reference point.

[0022] Constraint and projection deformation algorithms advantageously allow for very fast processing, much faster than optimization methods. These algorithms are particularly suitable when eigenvectors serve as “signatures” of the initial reference model, where the signature is not intended to include the information needed to construct the model it specifies, but rather simply to specify the model in a particular way.

[0023] The set of values ​​determined in step 3) allows for the direct or indirect definition of the final reference point and / or the location of the final basic surface in space.

[0024] These values ​​can be, in particular, the values ​​of the following parameters:

[0025] - Preferably, the parameters of the movement vector are: the starting point of the movement vector is the initial reference point, and the ending point is the corresponding final reference point; or

[0026] - Preferably, the parameters of the function that determines the final reference point position based on the position of the initial reference point, especially the parameters of the function that ensure the initial reference point is projected onto the surface of the model to be characterized; or

[0027] - Parameters that help define the location of the initial reference point (e.g., its x-coordinate).

[0028] Preferably, in step 3), for each initial reference point, a set of values ​​is determined, which allows for the construction of a final base surface approximating an associated final reference point based on the initial reference point, preferably a final base surface passing through the associated final reference point. The values ​​in such a set are preferably parameterized values ​​of an interpolation function that determines the final base surface by interpolation based on several pairs of pairs consisting of each free initial reference point and the associated final reference point, preferably based on the interpolation of all said pairs.

[0029] Preferably, the interpolation function allows for the determination of a final underlying surface around each final reference point. The interpolation function is preferably a radial basis function. The set of values ​​determined in step 3) preferably includes parameterized values ​​of the radial basis function.

[0030] Therefore, by combining all the final basic surfaces together, a “final” surface can be provided that matches the model to be characterized as closely as possible.

[0031] Preferably, in step 4), the feature vector includes a parameterization of a determined interpolation function, preferably composed of a parameterization of a determined interpolation function, such that the interpolation function parameterized by one of the parameterizations generates a final basic surface for the initial reference point.

[0032] The larger the number of initial reference points, the closer the shape of the final reference model will be to the shape of the model to be represented. The smaller the number of initial reference points, the smaller the feature vector, and therefore the easier it is to manipulate and store.

[0033] For example, generating a final reference model is not essential, so that it can be viewed or manipulated by a computer.

[0034] As will be seen in further detail throughout the remainder of the description, the representation method or “vectorization method” according to the invention allows for the generation of an indexed historical model database.

[0035] Therefore, the present invention relates to a method for generating an index history model database, the index history model database being referred to as an "index history database," the method comprising, for each intraoral organ in a set of intraoral organs comprising more than 500 intraoral organs:

[0036] -Characterization is performed based on the characterization method according to the invention to determine a model or "historical model" of the intraoral organs, and a historical feature vector corresponding to the historical model; then

[0037] - Form a record including the historical feature vector and the historical model, this record is called the "historical record"; then

[0038] - Add the aforementioned historical records to the index history library.

[0039] Preferably, the oral organs are teeth.

[0040] In one implementation, all teeth in the set of teeth have the same tooth number or the same type, and the initial reference model represents a reference tooth with the number or the same type, respectively.

[0041] In one embodiment, all teeth in the set of teeth have the same tooth number, and preferably, the reference tooth has the number.

[0042] Preferably, the teeth in the set of teeth have different tooth numbers or different types, and the initial reference model used to determine the historical feature vector is the same, regardless of the teeth considered in the set of teeth. The common initial reference model for the set of teeth is not necessarily a model of a tooth. The initial reference model can have any shape, such as a sphere.

[0043] Therefore, each historical model is indexed using historical feature vectors, which at least partially transform the deformation required to transform the initial reference model into a final reference model that is substantially identical to the historical model after the historical model is set up with respect to the initial reference model in a standardized configuration. Thus, the historical feature vector of the historical model is an index that can be used to quickly identify historical models in the indexed history library.

[0044] This method can be implemented in particular for multiple sets of said intraoral organs, wherein said intraoral organs are teeth, each set containing only teeth with the same number or the same type, and reference teeth used in characterizing the teeth in said sets having said number or said type respectively, wherein the number of initial reference points preferably depends on said number or said type.

[0045] The historical database allows for the calibration of the "analysis" model.

[0046] This invention particularly relates to a method for correcting a digital three-dimensional model, referred to as an "analysis model," which models intraoral organs (preferably teeth) referred to as "intraoral analysis organs," the method comprising the following steps:

[0047] - Generate an index history database based on the method according to the present invention;

[0048] - The intraoral analytical organ is characterized based on the characterization method according to the present invention in order to generate the analytical model and the “analysis” feature vector corresponding to the analytical model;

[0049] - Search the index history database for historical feature vectors that best match the analytical feature vectors, and use the historical model of the historical data to correct the analytical model, wherein the correction may involve replacing the analytical model with the historical model.

[0050] In one implementation, all historical models, initial reference models, and analysis models in the index history database are used to model teeth with the same number or type.

[0051] Preferably, the historical model in the index history database models teeth that may have different numbers or may be of different types, and the initial reference model used to determine the historical feature vector is always the same and independent of the historical model. In one embodiment, for the search, the following steps are performed:

[0052] e1) Define a filter associated with one or more parameters of the historical feature vector;

[0053] e2) Filter the index history database using the filter to retain a subset of the index history database;

[0054] e3) Modify the filter by increasing the number of parameters involved in the filter and / or by making the filter conditions more stringent.

[0055] e4) Filter the subset using the modified filter to define a new subset, repeating the loop of steps e3) and e4) until the subset obtained in step e3) includes fewer than 5 historical records.

[0056] e5) Correct the analysis model using one of the historical records obtained from step e4).

[0057] Preferably, the historical model found after the search is used to fill the white areas (i.e., areas without points) of the analysis model and / or remove any errors in the analysis model and / or replace a portion of the analysis model with the surface of the historical model, and / or replace the analysis model with the historical model.

[0058] The characterization method according to the present invention also allows for several other applications according to the present invention.

[0059] For some of these applications, a history library must be created, or only a historical feature vector database must be created.

[0060] Therefore, the present invention relates to a method for generating a historical feature vector database, the method comprising, for each intraoral organ (referred to as a "historical intraoral organ") in a set of intraoral organs comprising more than 500 intraoral organs, characterizing the historical intraoral organ based on a characterization method according to the present invention, and then the following step 5'):

[0061] 5') Add the vector or "historical representation vector" determined by the representation to the historical feature vector database.

[0062] Based on this historical feature vector database and the initial reference model, for each historical feature vector, it is advantageous to obtain the following model, which essentially represents the model obtained during step 1) of the generation of the historical intraoral organ representation of the historical feature vector.

[0063] Preferably, the intraoral organ is a tooth. In one embodiment, all teeth in the set of teeth have the same tooth number or the same type, and an initial reference model is used to model reference teeth with the same number or the same type, respectively. Preferably, the teeth in the set of teeth may have different numbers or different types, and it may or may not be indicated that the initial reference model of the teeth is always the same, regardless of the teeth in the set under consideration.

[0064] Therefore, the present invention relates to a method for generating a model of intraoral organs, the method comprising: generating a feature vector based on a representation method according to the present invention, and then deforming an initial reference model by moving an initial reference point according to the feature vector based on the feature vector and the initial reference model.

[0065] Preferably, the initial reference model is a model representing a reference intraoral organ. To ensure that the model generated by this method has a shape close to that of the model to be represented that generates the feature vectors, the initial reference model preferably has a shape as close as possible to the model to be represented. Specifically, if the intraoral organ is a tooth, the initial reference model and the model to be represented preferably represent teeth with the same tooth number or the same type.

[0066] Preferably, the method includes:

[0067] According to the present invention, a historical feature vector database is generated, and then, based on the historical feature vectors included in the historical feature vector database and the initial reference model used for the generation, the initial reference model is deformed by moving the initial reference point according to the historical feature vectors.

[0068] Historical eigenvectors essentially provide values ​​that allow defining final reference points and / or final surfaces in space based on initial reference points. These values, in turn, allow for the realization of a final reference model with a shape closely resembling that of the historical model. Applying these values ​​to the initial reference points produces a final reference model that is substantially identical to the historical model corresponding to the historical eigenvectors. Therefore, a model very close to the historical model can be reconstructed, thus avoiding the need to store these models.

[0069] Furthermore, if the intraoral organ is not one of the organs used to generate the historical feature vector database, a search can be conducted to find the historical feature vector that best corresponds to that intraoral organ. Preferably, a representation method is applied to the intraoral organ, and the historical feature vector that most closely approximates the obtained feature vector of that intraoral organ is sought. Reconstructing the historical model associated with this historical feature vector allows for the acquisition of a model that substantially represents the intraoral organ.

[0070] Based on a historical feature vector database, abnormalities in intraoral organs can also be detected. Specifically, this invention relates to a method for detecting shape abnormalities in an intraoral organ to be tested, the method comprising the following steps:

[0071] c) Characterize the intraoral organ to be tested based on the characterization method according to the present invention to obtain the “test” feature vector;

[0072] d) For at least one parameter of the feature vector to be tested, compare the value of the parameter with a predetermined range of “acceptable” values;

[0073] e) If the value to be measured is not within the range, a notification indicating the presence of a shape abnormality is generated, the notification preferably indicating the area of ​​the intraoral organ to be measured that is affected by the shape abnormality.

[0074] An anomaly is considered to exist when the initial reference point moves excessively during the deformation of the initial reference model in order to obtain the final reference model. This separation is typically provided by the values ​​of three parameters in the eigenvector to be measured (e.g., they provide the movement along the three axes of an orthogonal coordinate system). Therefore, three ranges corresponding to these three parameters determine the set of acceptable movements of the initial reference point, i.e., the tolerance of the reference intraoral organ at the location of the initial reference point under consideration. This notification is an information item indicating that this tolerance has been exceeded. The notification preferably specifies the relevant initial reference point.

[0075] The range of acceptable values ​​for the parameter is preferably defined according to the following steps:

[0076] a) According to the present invention, a historical feature vector database or an indexed historical model database is generated for multiple historical intraoral organs, and for at least some historical intraoral organs, preferably each historical intraoral organ, it is determined whether the condition related to the relevant intraoral organ is acceptable.

[0077] b) Determine the range of “acceptable” values ​​for at least some parameters of historical feature vectors in a historical feature vector database or an indexed historical model database. Preferably, for each parameter, the value is considered acceptable if the probability that the dental condition associated with a value is acceptable with respect to step a) exceeds a probability threshold.

[0078] Based on the historical feature vector database generated according to the present invention and optionally merged into the index history database generated according to the present invention, individuals can also be identified, especially when the intraoral organs are teeth.

[0079] Therefore, the present invention relates to a method for identifying an individual, the method comprising the following steps:

[0080] i) Based on the generation method according to the present invention, an index history database or feature vector database is generated in the first instance; and

[0081] Each historical feature vector, index history, or feature vector database generated by the model processing intraoral organs (preferably teeth) is associated with an identifier of an individual having said intraoral organ.

[0082] ii) The second time after the first time:

[0083] - Based on the characterization method according to the present invention, the target intraoral organ of the target individual to be identified is characterized to determine the digital three-dimensional model or "target model" of the intraoral organ and the corresponding target feature vector; then

[0084] - Search for historical feature vectors corresponding to the target feature vector in the index history database or historical feature vector database, and associate the identifier with the historical feature vector.

[0085] Finally, the characterization method according to the invention allows for the provision of a simple three-dimensional representation of intraoral organs. This representation can be advantageously analyzed using neural networks.

[0086] Therefore, the present invention relates to a method for evaluating attributes based on a three-dimensional representation (referred to as an "evaluation representation") of an individual's (referred to as an "evaluation individual") intraoral organs (referred to as "intraoral evaluation organs," preferably teeth), the method comprising the following steps:

[0087] - Create a learning database containing more than 1000 historical structures, each including:

[0088] - A three-dimensional representation of the intraoral organs of a "historical" individual (referred to as "historical representation"); and

[0089] - A "history" descriptor that contains the "history" value of the attribute in relation to the historical individual;

[0090] - Train at least one neural network using a training database;

[0091] - The evaluation representation is submitted to the neural network, such that the neural network determines at least one "evaluation" value for the attribute based on the evaluation representation;

[0092] The method includes: characterizing historical intraoral organs and intraoral assessment organs based on the characterization method according to the invention, so as to generate historical feature vectors and "assessment feature vectors" respectively, as historical representations and assessment representations.

[0093] Any method according to the invention may also specifically include one or more of the following optional features:

[0094] -All the organs described in the mouth are teeth;

[0095] - All teeth have the same tooth number or the same type;

[0096] - The initial reference model represents a group of individuals;

[0097] -The reference model is a model of the orthodontic operation training mold;

[0098] - The number of initial reference points depends on the number or the type;

[0099] - In the characterization method according to the present invention, the number of initial reference points is greater than 10 and / or less than 10% of the number of points of the model to be characterized, preferably less than 1%;

[0100] - The number of initial reference points is greater than 10, 50, 100, 400 and / or less than 10000, 5000, 1000, 800;

[0101] - Each feature vector (history, test, target, analysis, evaluation, etc.) includes more than 10 and less than 1000 values;

[0102] - In step 3), the deformation algorithm completes the constrained deformation of the initial reference model;

[0103] - The set includes parameterization of interpolation functions that can generate a final basic surface around a final reference point;

[0104] - Preferably, in step 3), the deformation algorithm implements the interpolation function based on several pairs of initial reference points and associated final reference points, preferably based on all initial reference points and final reference points, in order to determine the "final" surface that extends substantially along the final reference point;

[0105] - Preferably, interpolation is performed using radial basis functions (RBF);

[0106] - In step 4), the feature vector includes only the parameterized values ​​of the interpolation function.

[0107] This invention also relates to:

[0108] - A computer program, particularly a dedicated application for a mobile phone, comprising program code instructions for performing one or more steps of any method according to the invention when the program is executed by a computer;

[0109] - A computer medium for storing such programs, such as a memory or CD-ROM.

[0110] This invention also relates to:

[0111] - A computer loaded with a program according to the invention; and

[0112] - A system comprising a scanner for modeling an intraoral organ to be characterized in the form of a model to be characterized, and a computer according to the invention, the computer being capable of implementing a characterization method according to the invention and optionally one or more other methods according to the invention based on the model to be characterized.

[0113] The method according to the invention can also be applied to oral organs other than teeth, such as the gums.

[0114] definition

[0115] The term "individual" refers to any person to whom the method according to the invention is applied, regardless of whether that person is ill.

[0116] "Orthodontic treatment" is all or part of treatment aimed at altering the structure of the dental arch (active orthodontic treatment) or maintaining the structure of the dental arch (passive orthodontic treatment).

[0117] According to international conventions, each tooth in a dental arch has a predetermined number. Figure 5 The table lists the tooth numbers defined by the convention.

[0118] The term "dental device" refers to any device designed to be worn by the dental arch, particularly orthodontic appliances, crowns, implants, bridges, or facets.

[0119] "Intraoral organs" are elements within an individual's oral cavity. In particular, intraoral organs can be teeth, soft tissues (especially gums), or dental devices placed in the oral cavity, such as orthodontic appliances or dental instruments, or parts thereof, especially orthodontic appliances, braces, orthodontic arches, or fillings, implants, or prostheses with arches and fasteners.

[0120] A "model" refers to a digital 3D model. A model consists of a set of voxels. A model typically includes a mesh of points connected by line segments, i.e., a combination of triangles.

[0121] A "teeth model" is a digital three-dimensional model of a tooth. The dental arch model can be segmented to define a teeth model for at least some teeth, preferably all teeth, represented within the dental arch model. Therefore, a teeth model is a model within the dental arch model.

[0122] A “dental arch model” is a model representing at least a portion of the dental arch (preferably at least 2, preferably at least 3, preferably at least 4 teeth).

[0123] "Cutting" a dental arch model into "tooth models" allows for the definition and autonomous manipulation of the representation of teeth within the dental arch model (the tooth model). Computational tools exist for manipulating the tooth models within the dental arch model. An example of software used to manipulate tooth models and create treatment scenarios is the Treat program, as described on the webpage https: / / en.wikipedia.org / wiki / Clear_aligners#cite_note-invisalignsystem-10.

[0124] A 3D scanner, or "scanner," is a well-known device that allows the acquisition of dental arch models or tooth models.

[0125] A model is said to be "constantly meshed" modified when the distance between adjacent points remains constant. In other words, the triangles that make up the model are not modified, so the arrangement of these points relative to each other is constant. This modification corresponds to the model being moved in space by translation and / or rotation without altering its shape. In effect, the relative positions of all points in the model are preserved.

[0126] Unlike modifications to a constant mesh model, "deformation" modifies the shape of the model, i.e., modifies the mesh. Algorithms for deforming the model are known, for example, from https: / / www.hilarispublisher.com / open-access / mesh-deformation-approaches--a-survey-2090-0902-1000181.pdf. These deformation algorithms can be used.

[0127] In particular, the deformation of the initial reference model can be calculated "using radial basis functions" in steps 3), 3"), or C). To perform such deformation, it is preferable to determine the corresponding final reference point on the surface of the model to be characterized by projection for each initial reference point, and then, using radial basis functions, determine the final reference surface around the final reference point (step 3)). The surface of the model to be characterized can be, in particular, the target surface (step 3") or the analysis surface (step C)).

[0128] Interpolation using radial basis functions (“radial basis function interpolation”) is described in: https: / / www.hilarispublisher.com / open-access / mesh-deformation-approaches--a-survey-2090-0902-1000181.pdf

[0129] The "parameterization" of a function consists of all the values ​​of its parameters. For example, for an affine function f(x,y,z) = a.x + b.y + cz, the parameters are a, b, and c, and the parameterization is, for example, (1; 2; 1.5), that is, a = 1, b = 2, and c = 1.5.

[0130] When the deformation algorithm uses a projection function to determine the position of the final reference point, the parameterization of this function is preferably common to all initial reference points. Alternatively, the parameterization of the function can be determined based on the initial reference points under consideration. For an initial reference point, the set consisting of the coordinates of the initial reference point and the parameterization of the function for that initial reference point constitutes an example of a "set of values ​​that determine the position of the final reference point".

[0131] When the deformation algorithm uses an interpolation function, such as a radial basis function, to generate the final basic surface associated with the final reference point, the set determined in step 3) may consist of, for example, a parameterization of the interpolation function, and optionally, may also consist of the coordinates if the coordinates of the initial reference point cannot be derived from the ordering of values ​​in the eigenvectors.

[0132] When the final basic surface is generated based on the final reference point, preferably based on the final reference point and generated by interpolation using a plurality of pairs consisting of each free initial reference point and an associated final reference point, preferably based on all said pairs, the final basic surface is considered to be "dependent on the final reference point" or "associated with the final reference point".

[0133] Generally, the set of values ​​for "determining the position of the final reference point" is complete enough to allow information to be obtained from that set of values ​​to determine the position of the final reference point. The set of values ​​for "determining the final base surface" is complete enough to allow information about the surface to be obtained from that set of values. Those skilled in the art will readily understand how to define such a set. Such a set includes, for example, the coordinates of the translation vector between the initial and final reference points, or parameterizations of functions used to project the initial reference point onto the surface to be characterized, or parameterizations of interpolation functions.

[0134] All values ​​in one of the sets of values ​​can be grouped into an eigenvector, but this grouping is not mandatory. Specifically, when the positions of the initial reference points are known and these points are ordered, the initial reference points corresponding to the parameters of the eigenvector can be determined based on the positions of the parameters in the eigenvector. For example, it is known that n first parameters are associated with a first initial reference point, the subsequent n parameters with a second initial reference point, and so on. Therefore, it is not necessary to incorporate the positions of the initial reference points into the eigenvector.

[0135] Furthermore, it is not worthwhile to group the parameters in the eigenvector into identical values ​​that are independent of the initial reference points being considered. For example, if the parameterization of the projection function used to determine the final reference point is the same for all initial reference points, it is not worthwhile to repeat it in the eigenvector for each initial reference point.

[0136] In one implementation, the feature vector includes only values ​​from the set that determine the final base surface.

[0137] To enable comparison of eigenvectors, each model to be represented (historical model, analytical model, test model, etc.) must be arranged in the same manner as the initial reference model before determining the final reference point. This standardization or "resetting" of the models to be represented produces a "standardized" configuration of the model to be represented relative to the initial reference model. Standardization methods are well-known. In particular, the Iterative Closest Point (ICP) search algorithm can be used. An example of the Iterative Closest Point search algorithm is described at https: / / fr.wikipedia.org / wiki / Iterative_Closest_Point. These methods also allow, if desired, to make the model to be represented the same size as the initial reference model.

[0138] The “match” or “fit” between two eigenvectors, particularly analyzing the “match” or “fit” between an eigenvector or target eigenvector and historical eigenvectors, is the reciprocal of the measure of the difference between these eigenvectors. A match is considered a maximum match (“best fit”) when the difference is minimized. Methods for evaluating the difference between eigenvectors and finding the minimum difference are well-known. For example, the difference could be the Euclidean distance between the eigenvectors to be compared.

[0139] Similarly, when the shapes of the first model and the second model are as close as possible, the first model and the second model are matched as closely as possible. Methods for measuring the shape differences between two models are well-known. For example, the cumulative distance between points in the first model and the surfaces of the second model is a measure of this difference.

[0140] The "cumulative" distance between the set of points and the surface is the sum of the distances between those points and the surface.

[0141] A "vector" is typically an ordered collection of values, where each value quantizes a corresponding parameter.

[0142] A "movement" vector defines the values ​​of its parameters, which define the movement of a point in space. Therefore, a movement vector provides the coordinates of the point's movement in space. The length of the vector, or its "norm," is the length of that movement.

[0143] A translation vector in three-dimensional space can be defined by a set of three corresponding values ​​for three parameters, such as the translation values ​​(x, y, z) in Cartesian coordinates, (ρ, θ, h) in cylindrical coordinates, or the translation values ​​in spherical coordinates.

[0144] For example, if point M1 with coordinates (x1, y1, z1) in the Oxyz coordinate system is moved to point M2 with coordinates (x2, y2, z2) in the same coordinate system, then the vector (x2-x1, y2-y1, z2-z1) is the movement vector.

[0145] The “feature” vector consists of model-specific values ​​and forms the model’s identifier or “signature”.

[0146] The methods used to generate the index history repository are thought to involve generating a historical feature vector database. However, the historical feature vector database can be generated without generating an index history repository (when the records do not include historical models).

[0147] The method according to the invention is implemented by a computer, preferably solely by a computer. The term "computer" refers to any electronic device, including a group of multiple machines with computer processing capabilities. Typically, a computer includes a processor, memory, and a human-machine interface, which usually includes a screen, and is connected via Wi-Fi. Or an internet communication module for a telephone network. Software configured to implement the methods of the present invention is loaded into the computer's memory. The computer may also be connected to a printer.

[0148] A computer can be a server located far from the user; for example, it can be the "cloud".

[0149] For clarity, the terms “history,” “analysis,” “test,” or “evaluation” are used to distinguish content related to individuals or to historical models, analytical models, test models, or evaluation models, respectively. For clarity, “to be represented” is also used. Depending on the context, the “to be represented” element can be a “historical” element, an “analysis” element, a “test” element, or an “evaluation” element.

[0150] The term "history" can be used to refer to a database that includes many objects, such as a historical feature vector database or a learning database composed of historical structures. Therefore, "history" can be used to define various objects depending on the context.

[0151] "History" can specifically limit the vectors in the index history library or the historical feature vector database. Unlike the index history library, the historical feature vector database does not contain historical models.

[0152] Unless otherwise stated, “including”, “contains”, or “has” must be interpreted in a non-restrictive manner. Attached Figure Description

[0153] Further features and advantages of the invention will become more apparent after reading the following detailed description and referring to the accompanying drawings, in which:

[0154] -[ Figure 1 ] Figure 1 The various steps of the generation method according to the present invention are illustrated schematically;

[0155] -[ Figure 2 ] Figure 2 The various steps of the correction method according to the present invention are illustrated schematically;

[0156] -[ Figure 3 ] Figure 3 An example view of a dental arch model is shown;

[0157] -[ Figure 4 ] Figure 4An example view of a tooth model is shown;

[0158] -[ Figure 5 ] Figure 5 The numbers of teeth used in the dental field are shown. Detailed Implementation

[0159] In general, the present invention relates to a method in which a model of an intraoral organ is "generalized" in the form of a feature vector. In computer memory, the size of the feature vector, in bytes, is preferably at most one-tenth, and more preferably at most one-hundredth, of the size of the model.

[0160] However, preferably, the feature vector includes information sufficient to substantially reconstruct the aforementioned model based on the initial reference model. Preferably, for a practical scale, the difference between the reconstructed model and the model already summarized by the feature vector is less than 1 mm, preferably less than 0.5 mm, preferably less than 0.2 mm, preferably less than 0.1 mm, preferably less than 0.05 mm, and / or greater than 0.01 mm.

[0161] The following description relates to teeth, to which the invention is particularly useful. However, the invention can be extended to any oral organ.

[0162] Therefore, the present invention specifically proposes a method for characterizing teeth, the method comprising the following steps:

[0163] 1) The tooth is modeled in the form of a digital three-dimensional model, the digital three-dimensional model including a mesh of points defining the surface;

[0164] 2) Set the model relative to a digital 3D model of a reference tooth, referred to as the “initial reference model,” in a standardized configuration, the initial reference model comprising a mesh of points referred to as “initial reference points”; then

[0165] 3) For each initial reference point,

[0166] The final reference point is determined using a deformation algorithm; then

[0167] A set of values ​​is determined, the set of values ​​determines the position of the final reference point and / or the final basic surface is determined based on the final reference point, the deformation algorithm determines the position of the final reference point such that the final reference model composed of the mesh of the final reference point matches the model to be represented as closely as possible;

[0168] 4) Generate feature vectors that group the values ​​determined for all the initial reference points in step 3) in an ordered manner.

[0169] Feature vectors are tools for characterizing the teeth, and they are advantageously simpler and faster to use than models of the teeth.

[0170] Methods for generating an index history database

[0171] The purpose of the generation method according to the present invention is to create an index history database that can be quickly browsed.

[0172] This invention specifically proposes a method for generating an index history model database (referred to as an "index history database"), wherein for each "historical" tooth in a set of teeth comprising more than 500 teeth, the method includes the following steps:

[0173] 1) The tooth is modeled in the form of a digital 3D model (referred to as the “history model”), which includes a mesh that defines points on the history surface (referred to as “history points”);

[0174] 2) Set the historical model relative to a digital 3D model of a reference tooth (referred to as the "initial reference model") in a standardized configuration, the initial reference model comprising a mesh of points referred to as "initial reference points"; then

[0175] 3) For each initial reference point,

[0176] The final reference point is determined using a deformation algorithm; then

[0177] A set of values ​​is determined, which determines the position of the final reference point, and is, for example, a parameter of a movement vector whose starting point is the initial reference point and whose ending point is the corresponding final reference point; and / or the set of values ​​determines the final base surface based on the final reference point, and the deformation algorithm determines the position of the final reference point such that the final reference model composed of the mesh of the final reference point matches the historical model as closely as possible.

[0178] 4) Generate a vector (referred to as the "historical feature vector") that groups the values ​​determined for all the initial reference points in step 3) in an ordered manner;

[0179] 5) A record is generated, including historical feature vectors and historical models; this record is called the "historical record"; then...

[0180] Add the historical records to the index history library.

[0181] For clarity, steps 1) through 4) are numbered the same as the corresponding steps in the characterization method, as they are identical to those steps but only apply to “historical” teeth.

[0182] The cycles of steps 1) through 5) can be implemented for multiple sets of teeth comprising more than 500 teeth, wherein the set of teeth processed during a cycle contains only teeth with the same number or the same type, and the reference teeth used during the cycle each have the number or the type.

[0183] Such an indexed history database may specifically include more than 1,000, preferably more than 5,000, more than 10,000, and / or less than 1,000,000, or even less than 500,000, or even less than 100,000, or even less than 50,000 historical records, where each historical record includes a historical model and an associated historical feature vector. Each historical model represents a different tooth.

[0184] Historical records may also include descriptions that provide information related to the teeth being modeled, such as the tooth number or type of tooth (i.e., its nature: "incisor", "canine", "molar").

[0185] Creating a dental history involves steps 1) through 5) above, such as... Figure 1 As shown. Therefore, these steps can be repeated for a set of teeth, which includes more than 1,000 teeth, preferably more than 5,000 teeth, more than 10,000 teeth, and / or less than 1,000,000 teeth, or even less than 500,000 teeth, or even less than 100,000 teeth, or even less than 50,000 teeth.

[0186] These teeth may belong to more than 1,000, preferably more than 5,000, more than 10,000 different individuals, and / or less than 1,000,000, or even less than 500,000, or even less than 100,000, or even less than 50,000 different individuals.

[0187] In step 1), the teeth are modeled using a 3D scanner, preferably according to any known technology, to obtain a historical model. This model (referred to as a "3D model") can be viewed from any angle. A model viewed from a defined angle and distance is called a "view". Figure 4 An example view of a tooth model is shown.

[0188] Historical models can be prepared based on measurements of an actual tooth or a mold of that tooth (e.g., a plaster casting). Historical models can also be generated by cutting a dental arch model into a tooth model. Figure 3 An example view of the dental arch model 10 is shown. The tooth model 12 has been separated by cutting the dental arch model.

[0189] like Figure 4As shown, the historical model includes a grid of historical points 14 connected by straight line segments 16. Typically, the historical model includes more than 200, preferably more than 500, more than 1,000, preferably more than 3,000, preferably more than 4,000 historical points, and / or less than 100,000, preferably less than 10,000, preferably less than 5,000 historical points.

[0190] Therefore, relying on the analysis of the shape of the historical model to search for the historical model in the index history database takes a long time to execute, even with a powerful computer, especially if it must be executed in less than 5 seconds, less than 3 seconds, or less than 1 second.

[0191] In step 2), in order to evaluate the shape difference between the historical model and the initial reference model, the initial reference model and the historical model are first set to a standardized configuration or "comparison configuration" as defined by the configuration rules.

[0192] The initial reference model can be any model. However, compared to the historical model, the initial reference model contains far fewer points (referred to as "initial reference points"). Preferably, the number of initial reference points is less than 20%, 10%, 5%, 1%, 0.1%, or 0.01% of the number of historical points.

[0193] Preferably, the number of initial reference points is greater than 10, 50, 100, or 400, and / or less than 10,000, 5,000, 1,000, or 800.

[0194] The initial reference point can be any reference point. However, if the initial reference points are distributed substantially uniformly on the surface of the initial reference model, the quality of the characterization is better. Preferably, the initial reference points form a uniform grid, where any point is equidistant from its neighbors.

[0195] Step 3) is even faster when the shape of the initial reference model is close to that of the historical model. Therefore, preferably, the initial reference model is a model of a reference tooth, and more preferably, a model of a reference tooth with the same number or type as the tooth in the historical model.

[0196] Initial reference models can be obtained, in particular, with the help of 3D scanners.

[0197] In one implementation, the initial reference model is obtained by simplifying, for example, an original digital 3D model generated using a 3D scanner, which has more than 1,000, 5,000, 10,000, or 50,000 points. In other words, points are removed from the original model to obtain the initial reference model.

[0198] In one implementation, the initial reference model is obtained by statistically processing a set of basic digital 3D models, which preferably model teeth with the same number or type. Thus, the initial reference model can represent a group of individuals, such as people grouped into an age group or suffering from the same disease.

[0199] The initial reference model can be a model of an orthodontic training mold.

[0200] Preferably, the number of initial reference points depends on the number or type of the reference teeth modeled by the initial reference model.

[0201] Aside from possible modifications to the scale, setting the initial reference model and historical model to a comparison configuration is a constant mesh modification.

[0202] The configuration rules for setting up a historical model relative to an initial reference model specifically stipulate that the initial reference model and the historical model have the same scale, are oriented in the same way in space, and are separated from each other by a predetermined distance. Preferably, the initial reference model and the historical model are centered on each other, such that in a standardized configuration, the historical model and the initial reference model overlap as closely as possible, wherein the initial reference model and the historical model are substantially superimposed on each other.

[0203] In a preferred embodiment, for each historical point, the points of the initial reference model, i.e. the points of the closest initial reference model, are associated with the historical point using a "best fit" type search algorithm. Then, the historical model is located without deforming it, so as to minimize the distance between the points of the historical model and the points associated with the points of the initial reference model.

[0204] In step 3), an algorithm called the "deformation algorithm" is first used to determine a final reference model whose shape is as close as possible to the shape of the historical model, based on the initial reference model and the historical model.

[0205] Specifically, the deformation algorithm is configured to associate the final reference point with each initial reference point.

[0206] Deformation algorithms can be based on physical analogies or interpolation.

[0207] In one implementation, the deformation algorithm deforms an initial reference model while adhering to constraints to arrive at a final reference model. The constraints of the deformation algorithm define the rules for moving the initial reference point, for example, by fixing the elasticity of the initial reference model. For instance, these constraints can specify deformations that become increasingly difficult to measure as the point moves away from its initial position.

[0208] In particular, the following transformation algorithm is known:

[0209] - Linear spring analogy, developed by Batina (Batina, JT (1990), Unsteady Euler Airfoil Solutions Using Unstructured Dynamic Meshes. AIAA Journal 28:1381-1388);

[0210] Or variations of this analogy, especially:

[0211] -A torsion spring analogy, which is described in the following literature: Farhat C., Degand C., Koobus B., Lesoinne M. (1998), “Torsional Springs for Two-Dimensional Dynamic Unstructured Fluid Meshes”, Computer Methods in Applied Mechanics and Engineering, 163:231-245;

[0212] - A semi-torsional spring analogy, which is described in the following literature: Blom F. (2000), "Considerations on the spring analogy", International Journal of Numerical Methods for Fluids 32:647-668;

[0213] - The ball-vertex analogy is described in the following literature: Bottasso CL., Detomi D., Serra R. (2005), “The ball-vertex method: a new simple analogy method for unstructured dynamic meshes”, Computer Methods in Applied Mechanics and Engineering, 194:4244-4264;

[0214] - An analogy to the ortho-semi-torsional (OST) spring method is described in the following literature: Markou GA., Mouroutis ZS., Charmpis DC., Papadrakakis M. (2007), “The ortho-semi-torsional (OST) spring analogy method for 3D mesh moving boundary problems”, Computational Methods in Applied Mechanics and Engineering, 196:747-765.

[0215] The deformation algorithm can also specify a linear elasticity for moving the initial reference point.

[0216] The deformation algorithm can also use the Laplace smoothing equation or a modified Laplace operator.

[0217] It can guide the search for the final reference model, minimizing the cumulative distance between the final reference point and the historical surface, taking constraints into account.

[0218] The deformation algorithm can implement any optimization method, preferably in conjunction with metaheuristic methods, and more preferably through simulated annealing.

[0219] The metaheuristic method can be selected, in particular, from the group consisting of:

[0220] - Evolutionary algorithms, preferably selected from:

[0221] Evolutionary strategies, genetic algorithms, differential evolution algorithms, distribution estimation algorithms, artificial immune systems, shuffling complex evolutionary paths, simulated annealing, ant colony optimization, particle swarm optimization, tabu search, and GRASP method;

[0222] - Kangaroo Algorithm;

[0223] - The Fletcher and Powell method;

[0224] - Sound method;

[0225] - Random tunnels;

[0226] - A ramp that restarts randomly;

[0227] - Cross-entropy method; and

[0228] - A hybrid approach combining the above metaheuristic methods.

[0229] Preferably, among multiple "test" reference models obtained through different deformations of the initial reference model (by moving only the initial reference point), the differences between the test reference model and the historical model are measured. The differences between the test reference model and the historical model can be evaluated by adding the distances between points on the test reference model and the historical surfaces, where each distance is weighted according to constraints specified by the deformation algorithm. Thus, the final reference model is the "test" reference model with the smallest measured difference.

[0230] Each initial reference point has a movement vector, which originates at the initial reference point and terminates at the final reference point. This vector comprises three values ​​that together define the movement within the space. Therefore, each of these values ​​represents a positioning parameter of the final reference point. These values ​​can specifically be distances between the initial reference point and the corresponding final reference point along each of the three axes of an orthogonal coordinate system.

[0231] Other positioning parameter values ​​are also possible, such as the length of the movement vector, the direction of the movement vector, the route of the movement vector, or the distance between an initial reference point and a corresponding final reference point in a predetermined direction. Preferably, at least three positioning parameter values ​​are determined such that they, together with the coordinates of the initial reference point, constitute a set of values ​​for determining the position of the final reference point.

[0232] The deformation algorithm can also specify the regeneration of the initial reference model.

[0233] In a preferred embodiment, the deformation algorithm defines a final reference point for each initial reference point by projecting an initial reference point onto the surface of the historical model. All possible projection methods are conceivable, such as the projection that minimizes the distance between the initial and final reference points. The projection method can specify that the final reference point is a point on the grid of the historical model.

[0234] Preferably, the deformation algorithm performs interpolation based on the final reference point, especially the final reference point obtained by projecting it onto the historical model.

[0235] Deformation algorithms, in particular, can be implemented as follows:

[0236] - Transboundary interpolation, which is described in the following literature: Thompson JF., Warsi ZUA., Mastin CW. (1985), “Numerical Grid Generation, Foundations and Applications”, Elsevier Science Publishing, New York;

[0237] - Algebraic damping method, which is described in the following literature: Zhao Y., Forhad AA. (2003), “General method for simulation of fluid flows with moving and compliant boundaryaries on unstructured grids”, Computer Methods in Applied Mechanics and Engineering, 192:4439-4466;

[0238] - The inverse distance weighting method, for example, is described in the following literature: Wittevien JAS., (2010), “Explicit and Robust Inverse Distance Weighting Mesh Deformation for CFD”, 48, AIAA Aerospace Science Conference, including the New Horizons Forum and Aerospace Expo, USA;

[0239] -Radial basis functions, or "RBF", as described in the following literature: Boer AD., van der Schoot MS., Bijl H. (2007), "Mesh Deformation Based on Radial Basis Function Interpolation", Computer & Structure Journal 45:784-795.

[0240] Of all the algorithms, a deformation algorithm implementing radial basis functions for performing interpolation based on all final reference points is preferred. Thus, for each initial reference point, the radial basis functions determine the final basic surface. The deformation algorithm can then combine all the final basic surfaces to obtain the surface of the final reference model. Interpolation advantageously allows this surface to closely approximate the surface of the historical model. The methods for combining the final basic surfaces are known to those skilled in the art. For example, one method of combination is described in: http: / / math.univ-lille1.fr / ~calgaro / TER_2019 / wa_files / challioui_makki.pdf.

[0241] The parameterization of the radial basis functions used for the initial reference point is determined based on information about the location of the final reference point. Therefore, the parameterized values ​​are "derived" from the values ​​used to determine the location of the final reference point.

[0242] Based on the parameterized values ​​of the radial basis functions associated with the initial reference point, the radial basis functions can actually be parameterized such that when applied to the initial reference point, they produce the corresponding final reference point.

[0243] Preferably, the same parameters are evaluated for all initial reference points.

[0244] The deformation algorithms listed above are not restrictive. In particular, control mesh methods are also known, including the Delaunay graph method, the RBFs-MSA (moving submesh method) hybrid method, and quaternion-based methods.

[0245] In step 4), a historical feature vector is determined, which contains at least some or even all of the values ​​determined in step 3, sorted according to a predetermined order.

[0246] The values ​​of the historical feature vectors are selected so that the final reference model can be reconstructed based on the initial reference model.

[0247] Preferably, the initial reference points are sorted according to a predetermined order, and in order to form the historical feature vector, the values ​​determined in step 3) are sorted according to the predetermined order of the initial reference points. Advantageously, as long as the order of the initial reference model and the initial reference points is known, the historical feature vector does not necessarily need to contain information about the initial reference points.

[0248] Each historical feature vector is invariably specific to the historical model. Historical feature vectors are therefore indices that allow for the identification of the historical model. They advantageously represent the shape of the historical model and allow for rapid comparison.

[0249] Historical feature vectors are used to synthesize and quantize the deformations of the initial reference model to derive the final reference model.

[0250] In step 5), a record (called "historical record") including historical feature vectors and historical models is formed, and then the record is added to the index history database.

[0251] Methods for calibrating analytical models

[0252] Vectorizing the model using a finite number of initial reference points allows for particularly efficient use of the model.

[0253] In particular, this is effective for calibrating analytical models.

[0254] A model of the teeth to be corrected (called the "analysis model") can be processed to determine "analysis eigenvectors" when determining historical eigenvectors. The analysis eigenvectors can then be compared with historical eigenvectors. This comparison allows for the determination of historical eigenvectors that best match the analysis eigenvectors, and thus allows for the finding of the "optimal" historical model, i.e., the "best" historical model that best matches the analysis model. This comparison of eigenvectors is advantageously much faster than comparing the analysis model with the historical model itself.

[0255] The optimal historical model can be advantageously used to correct or even replace analytical models, such as those within the dental arch model.

[0256] This invention particularly relates to a method for correcting an "analytical" dental model (referred to as an "analytical model"), the analytical model comprising a mesh of points called "analytical points" that define an "analytical" surface, the method comprising the following steps:

[0257] A) Prior to step E), an index history database is generated based on the generation method according to the present invention;

[0258] B) Set the analysis model to a standardized configuration relative to the initial reference model; then

[0259] C) For each initial reference point

[0260] The final reference point is determined using the same deformation algorithm implemented when generating the history database, and then...

[0261] A set of values ​​is determined, which determines the position of the final reference point, and is, for example, a parameter of a movement vector whose starting point is the initial reference point and whose ending point is the corresponding final reference point; and / or the set of values ​​determines the final basic surface based on the final reference point.

[0262] The deformation algorithm determines the position of the final reference point, so that the final reference model composed of the mesh of the final reference point matches the analysis model as closely as possible;

[0263] The final reference points are preferably determined such that they can be generated by constrained deformation of the initial reference model, or preferably calculated by projecting the initial reference points onto the analysis surface.

[0264] D) Generate a feature vector (called the “analysis feature vector”), which groups the values ​​determined in step C) according to the order in which the historical feature vectors were used;

[0265] E) Search the index history database for historical feature vectors that match the analysis feature vectors as closely as possible, and correct the analysis model with the historical model of the historical data, wherein correction may involve replacing the analysis model with the historical model.

[0266] Steps B), C), and D) are similar to steps 2), 3), and 4), respectively, but are applicable to analytical models rather than historical models.

[0267] In step E), comparing the analytical feature vector with various historical feature vectors allows for the determination of the historical feature vector closest to the analytical feature vector, thereby indirectly allowing for the determination of the historical model closest to the analytical model. The historical and analytical feature vectors advantageously include a finite number of values ​​because they involve only a finite number of movements of the initial reference points. Therefore, browsing the index history database is advantageously very fast.

[0268] Preferably, the index history database includes only historical models that model teeth with the same number or type as the teeth modeled by the analysis model. Preferably, the initial reference model also models teeth with the same number or type.

[0269] The analytical model corrected based on the correction method according to the invention can be used in particular to construct dental arch models of individuals with the teeth and / or fill white areas of the analytical model and / or remove any errors from the analytical model, and / or replace the portion of the analytical model representing intraoral organs with the surface of the historical model and / or replace the analytical model with the historical model.

[0270] Steps A through E will now be described in detail.

[0271] In the example described in detail, the analysis model is a dental model intended to be analyzed (e.g., for its correction). Like the historical model, the analysis model consists of a mesh of points (called "analysis points"). The analysis model can be generated using any known technique, preferably using a 3D scanner. The analysis model can be prepared based on measurements of an individual's teeth or a mold of those teeth (e.g., a plaster casting). Alternatively, the analysis model can be generated by cutting a dental arch model of an individual into a dental model.

[0272] Typically, the analysis model includes more than 1,000, preferably more than 5,000, preferably more than 10,000 analysis points, and / or less than 1,000,000, or even less than 500,000, or even less than 100,000, or even less than 50,000 analysis points.

[0273] In step A), the index history database is generated as described above.

[0274] Step A) can be performed at any time before the search in step E).

[0275] In step B), the initial reference model and analysis model are set up with a standardized configuration, that is, by following the configuration rules used in step 2). This step is similar to the steps performed on the historical model in step 2).

[0276] In step C), a deformation algorithm is used to determine a final reference model for each initial reference point based on the initial reference model and the analysis model. The shape of the final reference model is as close as possible to the shape of the analysis model.

[0277] The deformation algorithm is the deformation algorithm used for the historical model in step 2).

[0278] In one implementation, a constraint deformation of the initial reference model is performed to define the final reference model, preferably using an optimization method. The objective then involves minimizing the cumulative distance between the final reference points and the “analysis” surface of the analysis model. Thus, the set of final reference points corresponds as closely as possible to the analysis model. The deformation algorithm can implement any optimization method, preferably in conjunction with a metaheuristic method, and more preferably by simulated annealing, particularly the methods listed above. Advantageously, computing the constraint deformation of the initial reference model is fast.

[0279] Specifically, when the final reference point is determined by constrained deformation of the initial reference model, the set of values ​​for determining the position of the final reference point preferably includes the values ​​of parameters of a translation vector, the starting point of which is the initial reference point and the ending point of which is the corresponding final reference point. Preferably, the values ​​of several parameters of the translation vector are determined, for example, the distances between the initial reference point and the corresponding final reference point along each of the three axes of the coordinate system (e.g., an orthogonal coordinate system).

[0280] Preferably, the deformation algorithm calculates the position of the final reference point by projecting the initial reference point onto the surface of the analysis model, and then, more preferably, uses radial basis functions in a manner known per se in order to obtain a final reference model whose shape is as close as possible to the shape of the analysis model.

[0281] Then, for each initial reference point, a set of values ​​is determined that determines the position of the corresponding final reference point and / or the final basic surface is determined based on the final reference point.

[0282] In particular, when radial basis functions are applied to the initial reference point, the set preferably includes parameterized values ​​of the radial basis functions.

[0283] The parameters corresponding to the values ​​of the set are the same as those used for the historical model in step 3), so that the analytical feature vector can be comparable to the historical feature vector.

[0284] Preferably, the same parameters are evaluated for all initial reference points.

[0285] In step D), analytical feature vectors are determined for all initial reference points, grouping the values ​​determined in step C).

[0286] The values ​​determined in step C) are sorted according to the same order used in step 4) to construct the historical feature vector.

[0287] This allows us to obtain the feature vectors for analysis.

[0288] The eigenvectors are analyzed and the “deformation” of the initial reference model is quantified in order to derive the final reference model calculated in step C), which is as similar as possible to the analytical model.

[0289] The comparison rules used to determine the standardized configuration between the initial reference model and the analysis model, the parameters of the values ​​determined in step C), the parameters of the analysis feature vector in step D), and the order of the values ​​within the analysis feature vector are the same as the comparison rules used to determine the standardized configuration between the initial reference model and the historical model, the parameters of the values ​​determined in step 3), the parameters of the historical feature vector in step 4), and the order of the values ​​within the historical feature vector. Therefore, the analysis feature vector can be compared with the historical feature vectors of records in the index history database.

[0290] In step E), a historical model whose shape is closest to the analysis model is sought. The closeness or "matching degree" of the shapes between the analysis model and the historical model can be quantified by comparing the analysis feature vectors with the historical feature vectors of the historical model.

[0291] For example, a function can be used to evaluate one, several, or all the differences between the values ​​of the eigenvectors and the corresponding historical eigenvectors, such as the sum of squares of the differences between the values ​​of the eigenvectors and the corresponding historical eigenvectors, or, to correspond to Euclidean distance, the square root of the sum can be evaluated.

[0292] The function of the difference can be distance, specifically Manhattan distance, Euclidean distance, or Minkowski distance. The reciprocal of this function can be used to evaluate the degree of matching.

[0293] If the historical feature vector is (h1, h2, ..., h n The eigenvectors are (a1, a2, ..., a). n If the degree of matching between these vectors is such that, for example, the reciprocal of the Euclidean distance is 1 / [(h1-a1)], then the degree of matching between these vectors can be, for example, the reciprocal of the Euclidean distance: 1 / [(h1-a1)] 2 +(h2-a2) 2 +…+((h n -a n ) 2 ] 0.5 .

[0294] In the simplified version, the degree of matching between these vectors can be, for example: 1 / [(h1-a1)] 2 +(h2-a2) 2 ] 0.5 , or even 1 / [(h1-a1)] 2 ] 0.5 .

[0295] Comparing the analytical feature vector with various historical feature vectors is advantageously very fast and enables reliable identification of historical models containing shapes that are closest to the analytical model.

[0296] Then, the historical model that is closest to the analysis model (or the "optimal" historical model) can be used to correct the analysis model, for example, to fill in the white areas of the analysis model, or to remove any errors from the analysis model, or to replace parts of the analysis model that do not represent teeth, for example, because that part represents an orthodontic appliance or a part of an orthodontic appliance (e.g., an orthodontic bracket).

[0297] Correction can even involve replacing the analytical model with the optimal historical model.

[0298] In one embodiment, the correction method according to the invention is implemented to continuously correct multiple tooth models of a dental arch model. Therefore, the correction method according to the invention is particularly useful when the dental arch model must be corrected in less than 5 seconds, less than 3 seconds, or less than 1 second, especially when the dental arch model is used to provide information in real time, for example, via radio waves.

[0299] In one implementation, in step E), the following steps are performed:

[0300] e1) Define filters associated with one or more parameters of the historical feature vector, for example, filters associated with a single first parameter;

[0301] e2) Filter the index history database using the filter to retain a subset of the index history database;

[0302] e3) Modify the filter to make the filtration conditions more stringent, preferably by increasing the number of parameters involved in the filter and / or by enhancing the filtration conditions of the filter;

[0303] e4) Filter the subset using the modified filter to define a new subset, and repeat the loop of steps e3) and e4) until the subset obtained in step e3) includes fewer than 5, preferably fewer than 3, and preferably fewer than 2 historical records.

[0304] e5) Use one of the historical records obtained from step e4) to correct the analysis model.

[0305] This implementation method advantageously allows for even faster correction methods.

[0306] In one implementation, when the analysis model represents a tooth with a tooth number or tooth type, the index history database contains only historical records associated with teeth having the same tooth number or the same type. Advantageously, this accelerates the search for the optimal historical model. Such an index history database can be generated, in particular, by filtering a larger index history database by tooth number or tooth type, which includes historical records associated with teeth having different tooth numbers or different tooth types.

[0307] It is worth noting that, in this embodiment, the reference model preferably represents teeth with the same tooth number or the same type.

[0308] As is now apparent, this invention allows for very fast and efficient browsing of an indexed history database that may contain thousands of historical records, based on an analytical model. Therefore, this invention allows for the rapid finding of an “optimal” historical model, typically within less than 5 seconds or less than 1 second, which is comparable to the analytical model and thus provides information about it. Notably, the optimal historical model can provide information that allows for the correction of the analytical model, and in particular, allows for, for example, replacing the analytical model in a dental arch model.

[0309] The representation method facilitates searching in the index history and allows for other applications:

[0310] 3D model compression

[0311] The feature vectors (especially historical or analytical vectors) generated in step 4) together with the initial reference model form sufficient information to compute the final reference model, which is essentially the same as the "to be represented" model (historical or analytical model, respectively) generated in step 1).

[0312] In fact, each initial reference point of the initial reference model only needs to be moved according to the values ​​of the eigenvector parameters to reconstruct a model that approximates the model to be represented. This deformation operation is the same as the deformation operation performed in step 3), but instead of determining the eigenvector, it is based on the eigenvector to essentially reconstruct the model to be represented.

[0313] Therefore, a "reconstructed" database consisting of such reconstructed historical models is essentially equivalent to a database containing the historical models themselves. However, advantageously, the historical feature vectors and initial reference models that allow the generation of the reconstructed database occupy much less space in computer memory compared to a database consisting of a collection of historical models.

[0314] The database containing historical models may contain more than 1,000, preferably more than 5,000, more than 10,000 historical models, and / or less than 1,000,000, or even less than 500,000, or even less than 100,000, or even less than 50,000 historical models.

[0315] Therefore, the present invention also relates to a method for computerized compression of a historical model database by generating a historical feature vector database representing the historical model database in a "compressed" manner, for each intraoral organ in a set of intraoral organs including more than 500 intraoral organs, the method comprising steps 1) to 4) above, followed by the following step 5'):

[0316] 5') Add historical feature vectors to the historical feature vector database.

[0317] The preferred oral organ is the tooth.

[0318] The present invention also relates to a method for generating a reconstructed or "decompressed" historical model based on historical feature vectors contained in the historical feature vector database and the initial reference model. According to this method, each point of the reconstructed historical model is generated based on a corresponding initial reference point and historical feature vectors.

[0319] In one implementation, the generation corresponds to a constrained deformation of the initial reference model, wherein the initial reference point moves according to historical eigenvectors.

[0320] Preferably, the historical feature vector includes at least three values ​​for each initial reference point, wherein these three values ​​define a spatial translation vector between the initial reference point and the associated final reference point. For example, these three values ​​can define a three-dimensional translation vector in a cylindrical or spherical Cartesian coordinate system, such as a three-dimensional translation vector in an orthogonal coordinate system (x,y,z).

[0321] In this implementation, the historical feature vector includes a parameterized projection function. This function is parameterized using this parameterization and applied to an initial reference point to determine the points of the reconstructed historical model.

[0322] In a preferred embodiment, for each initial reference point, the historical feature vector includes a parameterization of an interpolation function (preferably a radial basis function). This function is parameterized by this parameterization and applied to the initial reference points to determine a final base surface for each initial reference point. The final base surfaces are then combined to form the final surface that defines the surface of the reconstructed historical model.

[0323] Biometrics

[0324] Each historical feature vector is preferably a signature of the historical model, i.e., a unique identifier for the historical model. The historical model can represent the teeth of a "historical" individual. However, the shape of an individual's teeth is specific to that individual. Therefore, the historical feature vector corresponding to an individual's teeth constitutes that individual's unique identifier. Thus, historical feature vectors can be used to identify individuals.

[0325] Therefore, the present invention also relates to a method for identifying an individual, the method comprising the following steps:

[0326] i) Based on the generation method according to the invention, an index history database or feature vector database is generated in the first time, wherein each feature vector generated by the historical model of processing teeth is associated with an identifier of an individual having said teeth;

[0327] ii) A second time following the first time, for example, more than one week or more than one month after the first time:

[0328] 1”) Model the target teeth of the target individual to be identified to form a digital three-dimensional model (called the “target model”), which includes a grid of points (called “target points”) defining the target surface;

[0329] 2) Set the target model relative to the initial reference model with a standardized configuration; then

[0330] 3”) For each initial reference point:

[0331] The final reference point is determined using a deformation algorithm; then

[0332] Determine the set of values ​​that determine the position of the final reference point, such as parameters of a movement vector whose starting point is the initial reference point and whose ending point is the corresponding final reference point; and / or the set of values ​​that determine the final basic surface based on the final reference point.

[0333] The deformation algorithm determines the position of the final reference point, so that the final reference model composed of the final reference point mesh matches the target model as closely as possible;

[0334] 4”) Generate a vector (referred to as the “target feature vector”) that groups the values ​​determined for all the initial reference points in step 3”) in an ordered manner;

[0335] 5) Search for historical feature vectors that match the target feature vector in the index history database or historical feature vector database, and extract the identifiers associated with the target feature vector.

[0336] Steps 1”) to 4”) are the same as steps 1) to 4) respectively, but are applicable to the target tooth.

[0337] Of course, the parameters are the same as those of historical feature vectors in the index history database or historical feature vector database. These parameters are ordered in the same way in both the target feature vector and the historical feature vector.

[0338] The transformation algorithm in step 3) is the same as that in step 3).

[0339] Therefore, the target model allows for the very rapid identification of the target individual.

[0340] The historical feature vector corresponding to the target feature vector is preferably the same as the target feature vector. However, the shape of the target tooth may have evolved between the first and second time points. Therefore, in one embodiment, the historical feature vector corresponding to the target feature vector is a historical feature vector that matches the target feature vector as closely as possible (“best fit”).

[0341] Detecting specific features of teeth

[0342] Each historical or analytical feature vector provides information about the shape differences between the historical or analytical model and the initial reference model.

[0343] When a historical or analytical model represents teeth of the same type or with the same number as the initial reference model, their shapes are substantially identical. In particular, the initial reference model can represent a group of individuals, for example, groups of people within an age group or those with the same disease. The initial reference model can also be a model for orthodontic manipulation training molds.

[0344] Therefore, significant differences in shape between a historical or analytical model and an initial reference model, or more generally, significant differences in shape between a historical or analytical model and one or more other models representing the same type or having the same number of teeth, can indicate atypical shapes or “shape anomalies,” such as shape anomalies caused by growth defects of the teeth (short teeth), fractures, caries, or wear of the historical or analytical teeth under consideration.

[0345] Advantageously, shape anomalies can be detected by simply examining the historical or analytical eigenvectors. If one or more values ​​in the historical or analytical eigenvectors are excessive, it means that the initial reference point has moved aberrantly during the deformation of the initial reference model. Therefore, comparing these values ​​to a range of acceptable values ​​allows for rapid detection of shape anomalies and also allows for the identification of the region of the relevant tooth.

[0346] Therefore, the present invention relates to a method for detecting shape abnormalities in an intraoral organ, preferably a tooth, to be tested, the method comprising analyzing one or more values ​​of a feature vector associated with a model of the intraoral organ to be tested. This feature vector is established based on a representation method according to the present invention.

[0347] This invention particularly relates to a method for detecting shape abnormalities in a tooth to be tested, the method comprising the following steps:

[0348] c) Characterize the tooth to be tested according to steps 1) to 4) in order to obtain the “test” feature vector;

[0349] d) For at least one parameter of the feature vector to be tested, compare the value of the parameter with a predetermined range of “acceptable” values;

[0350] e) If the value to be measured is not within the range, a notification indicating the presence of a shape abnormality is generated, the notification preferably specifying the area of ​​the tooth to be measured affected by the shape abnormality.

[0351] In step d), an inspection is performed to determine whether the values ​​are acceptable by checking whether the values ​​of the feature vectors to be tested fall within the corresponding “acceptable” range.

[0352] The range of "acceptable" values ​​for a parameter is the range of values ​​for which no notification will be generated. Therefore, the range of "acceptable" values ​​for a parameter defines the set of values ​​considered acceptable. This range can be defined by dental professionals, such as dentists or orthodontists, and optionally in conjunction with the individual with the tooth being tested.

[0353] Preferably, the range of "acceptable" values ​​for the parameter is statistically defined by the computer prior to step c), i.e., by analyzing the situation of multiple teeth. This determination can be performed by a dental care professional, such as a dentist or orthodontist. Preferably, this statistical analysis is performed by a computer.

[0354] Specifically, it can be done as follows:

[0355] a) According to the present invention, a historical feature vector database or an indexed historical model database is generated for multiple historical teeth, and for at least some historical teeth, preferably each historical tooth, it is determined whether the tooth condition is acceptable.

[0356] b) Determine the “acceptable” range of values ​​for at least some parameters of historical feature vectors in a historical feature vector database or an indexed historical model database. Preferably, for each parameter, the value is considered acceptable if the probability that the dental condition associated with a value is acceptable with respect to step a) exceeds a probability threshold.

[0357] In step a), the number of historical teeth preferably includes more than 1,000, more than 5,000, more than 10,000 historical teeth, and / or less than 1,000,000, or even less than 500,000, or even less than 100,000, or even less than 50,000 historical teeth.

[0358] The acceptability of a dental condition is preferably determined by a dental care professional, such as a dentist or orthodontist. For example, if they detect an abnormal shape of the tooth, such as a broken tooth, they may consider the dental condition unacceptable, or otherwise acceptable.

[0359] In step b), the range of each "acceptable" value associated with the parameter of the historical feature vector is defined by an upper and lower bound. Therefore, this range defines the set of values ​​considered acceptable for that parameter. This range can be defined based on the values ​​taken for that parameter under dental conditions considered acceptable in the previous step.

[0360] The probability threshold can be, for example, greater than 70%, greater than 80%, greater than 90%, or greater than 95%, preferably essentially 100%. The larger the number of historical teeth, the better the reliability.

[0361] For example, it can be seen that for the fifth parameter, the value of historical teeth associated with acceptable conditions is greater than 95% (probability threshold), ranging between 3 and 7, while the value of historical teeth associated with unacceptable conditions is greater than 95%, ranging from less than 3 to greater than 7. Therefore, 3 and 7 can be chosen to define the acceptable range of the fifth parameter.

[0362] Preferably, the regions of the teeth are associated with each parameter of the feature vector. Specifically, the parameters are preferably associated with points in the tooth model, such that identifying the parameter allows the association of a region of the tooth with that point. For example, a fifth parameter can provide the value of the movement vector of a third initial reference point along axis Ox. If the value of this parameter is "unacceptable," it can be inferred that the region affected by the shape anomaly is the region of that third point.

[0363] In step e), the notification can be used by a computer and / or sent to the individual, particularly in the form of a written or oral message. In one implementation, the notification is sent to the mobile phone of the individual with the tooth to be tested. This allows them to schedule a treatment appointment.

[0364] Detecting diseases or disease risks

[0365] The present invention also relates to a method for evaluating attributes based on a three-dimensional representation (referred to as an "evaluation representation") of an individual's (referred to as an "evaluation individual") intraoral organs (referred to as "intraoral evaluation organs"), the method comprising the following steps:

[0366] - Create a learning database containing more than 1,000, preferably more than 10,000, and / or less than 1,000,000 historical structures, each historical structure including:

[0367] - A three-dimensional representation of the intraoral organs of a "historical" individual (referred to as "historical representation"); and

[0368] - A "history" descriptor that contains "history" values ​​related to the historical individual, such as those related to the historical representation, for the attribute described above;

[0369] - Train at least one neural network using a training database;

[0370] - The evaluation representation is submitted to the neural network, such that the neural network determines at least one "evaluation" value for the attribute based on the evaluation representation.

[0371] PCT / EP2019 / 052127 provides useful information for performing the evaluation method according to the present invention.

[0372] This attribute is preferably related to disease, particularly to diseases related to teeth and / or gums. Thus, the value of this attribute can indicate whether the relevant individual (i.e., historically or in assessment, depending on the steps considered) is currently suffering from a specific disease, or whether they have already suffered from a specific disease, for example, whether they have suffered from a specific disease within 1 year, 6 months, or 1 month after the oral organs of the relevant individual have been represented.

[0373] This attribute can also relate to the intraoral organs of the individual, particularly to the orthodontic appliance or part of the appliance worn by the individual. Thus, the value of this attribute can be specified, for example, whether the individual's intraoral organs are functional or dysfunctional (e.g., no longer in use), or whether, for example, the intraoral organs have become dysfunctional within one year, six months, or one month after being represented.

[0374] Attributes can be particularly related to the following:

[0375] - Visible or invisible parts of oral organs (such as the dental arch);

[0376] - The position and / or orientation and / or calibration of the acquisition device (e.g., a scanner) used to acquire the relevant representation; and / or

[0377] - Characteristics of the relevant representation, especially those related to the brightness, contrast, or sharpness of the relevant representation.

[0378] Therefore, based on the assessment values, information related to the individual's disease and / or the intraoral organs present in the individual can be determined, and then the information is transmitted to the operator, wherein the information indicates, for example, the risk of disease occurrence or the risk of damage to the orthodontic appliance.

[0379] It is noteworthy that the assessment method according to the invention particularly allows for the detection of diseases or disease risks that are undetectable by humans, especially orthodontists. In fact, learning allows neural networks to associate the value of an attribute with any assessment representation without having to establish a causal relationship between that representation and the value that is understandable to humans.

[0380] Therefore, this assessment method opens up a new research area for the detection or prevention of diseases.

[0381] Specifically, within the scope of the assessment method according to the invention, the intraoral organ can be a single tooth, a collection of teeth, or preferably a dental arch. For training to be effective, the historical intraoral organ preferably has the same type and properties as the intraoral assessment organ. Preferably, they are both representations of a dental arch.

[0382] Neural networks can be selected from:

[0383] - Specialized networks used for image classification are called "CNNs" ("Convolutional Neural Networks"), for example:

[0384] -AlexNet (2012);

[0385] -ZF Net (2013);

[0386] -VGG Net (2014);

[0387] -GoogleNet (2015);

[0388] -Microsoft ResNet (2015);

[0389] -Caffe:BAIR Reference CaffeNet,BAIR AlexNet;

[0390] -Torch: VGG_CNN_S, VGG_CNN_M, VGG_CNN_M_2048, VGG_CN_M_1024, VGG-CNN_M_128, VGG_CNN_F, VGG ILSVRC-2014 16-layer, VGG IL SVRC-2014 19-layer, Network-in-Network (Imagenet & CIFAR-10);

[0391] -Google:Inception(V3, V4);

[0392] - Specialized networks used for locating and detecting objects in images, i.e., object detection networks, for example:

[0393] -R-CNN (2013);

[0394] -SSD (Single-Lens Multi-Box Detector: Object Detection Network), Faster R-CNN (Faster Region-Based Convolutional Network Approach: Object Detection Network);

[0395] -Faster R-CNN (2015);

[0396] -SSD(2015).

[0397] The above list is not restrictive.

[0398] Training a neural network typically involves taking a historical representation as input and outputting the corresponding historical descriptor as output. Thus, the neural network learns to determine a descriptor that has the same properties as the historical representation input into the neural network.

[0399] The assessment representation may be an "analytical representation" within the scope of the methods described in the international application No. PCT / EP2019 / 052127 by the applicant, Dental Monitoring Company.

[0400] The present invention characterizes intraoral organs by allowing the acquisition of feature vectors as simple three-dimensional representations of intraoral organs. These feature vectors may actually include values ​​for at least a portion of the surface of the intraoral organ in three dimensions.

[0401] In a preferred embodiment, the historical representation preferably includes a historical feature vector, and the evaluation representation is preferably an "evaluation" feature vector having the same structure as the historical feature vector. The historical feature vector and the evaluation feature vector are generated from the representation of historical intraoral organs and intraoral evaluation organs based on the characterization method according to the present invention, respectively.

[0402] Therefore, the evaluation method according to the invention allows for very rapid training of neural networks with three-dimensional representations that represent complex intraoral organs in a simple yet reliable manner. This greatly accelerates the implementation of the evaluation method according to the invention.

[0403] Each historical 3D representation can preferably include a set of historical feature vectors representing the same historical intraoral organ at different times. The evaluation representation preferably includes a set of "evaluation feature vectors" having the same structure as the set of historical feature vectors. Thus, the neural network can advantageously consider not only the shape of the intraoral organ in space, but also the temporal evolution of that shape.

[0404] Of course, the present invention is not limited to the embodiments described and shown above.

[0405] In particular, oral organs, especially teeth, are not necessarily human oral organs. The method according to the invention can be used on other animals. The individual can be alive or dead; preferably alive.

[0406] The method according to the invention can be implemented in the context of orthodontic treatment, but it can also be implemented outside of any orthodontic treatment, and even outside of any therapeutic treatment.

Claims

1. A method for characterizing an intraoral organ to be characterized, the method comprising the following steps: 1) The intraoral organ to be represented is modeled in the form of a digital three-dimensional model or a "model to be represented", wherein the model to be represented includes a mesh of points defining the surface; 2) The model to be represented is set up in a standardized configuration relative to a digital 3D model called the "initial reference model," which includes a mesh of points called "initial reference points," the number of which is less than 20% of the number of points in the model to be represented; Then 3) Use a deformation algorithm to determine the final reference point for each initial reference point, and then... A set of values ​​is determined, the set of values ​​determines the position of the final reference point and / or the final basic surface is determined based on the final reference point, and the deformation algorithm determines the position of the final reference point such that the final reference model composed of the mesh of the final reference point matches the model to be represented as closely as possible; 4) Generate feature vectors that group the values ​​determined for all the initial reference points in step 3) in an ordered manner. In step 3), the deformation algorithm is either a constrained deformation algorithm of the mesh of the initial reference model or an algorithm for projecting the initial reference point onto the surface of the model to be characterized.

2. The method as described in claim 1, wherein, The initial reference model is a digital three-dimensional model of the intraoral organs and / or a group representing an individual and / or a model of an orthodontic training mold.

3. The method as described in claim 1, wherein, The number of initial reference points is greater than 10 and / or less than 10% of the number of points in the model to be represented, or the number of initial reference points is less than 1% of the number of points in the model to be represented.

4. The method of claim 1, wherein, Each feature vector contains more than 10 but less than 1000 values.

5. The method of claim 1, wherein, The initial reference points are evenly distributed on the surface of the initial reference model.

6. The method of claim 1, wherein, The values ​​determined in step 3) are the values ​​of the following parameters: - Parameters of the movement vector, wherein the starting point of the movement vector is the initial reference point, and the ending point of the movement vector is the corresponding final reference point; or - Parameters of the function that determines the position of the final reference point based on the position of the initial reference point.

7. The method of claim 1, wherein, In step 4), the feature vector includes a parameterization of a determined interpolation function, such that the interpolation function parameterized by the parameterization generates a final basic surface for the initial reference point.

8. The method of claim 7, wherein, The interpolation function is a radial basis function, and / or, in step 4), the feature vector includes only the parameterized values ​​of the interpolation function.

9. A method for generating a feature vector database, the method comprising, for each intraoral organ to be characterized in a set of intraoral organs comprising more than 500 intraoral organs, characterizing the intraoral organ to be characterized according to the method of claim 1, and then the following step 5'): 5') Add the feature vector to the feature vector database.

10. The method of claim 9, wherein, The oral organs are teeth, or all of the oral organs are teeth and all teeth have the same tooth number or the same type.

11. A method for generating an intraoral organ model, the method comprising: The method according to claim 1 generates a feature vector, and then, based on the feature vector and the initial reference model, the initial reference model is deformed by moving the initial reference point according to the feature vector.

12. A method for detecting shape abnormalities of intraoral organs, the method comprising the following steps: c) The method for characterization according to claim 1 characterizes the intraoral organ to be tested to obtain a feature vector called the "feature vector to be tested"; d) For at least one parameter of the feature vector to be tested, compare the value of the parameter with a predetermined range of "acceptable" values; e) If the value of the parameter is not within the range, a notification indicating the presence of a shape anomaly is generated.

13. A method for generating an index history model database called an "index history repository," the method comprising, for each intraoral organ in a set of intraoral organs comprising more than 500 intraoral organs: - The method according to claim 1 characterizes the intraoral organs to determine a model or "historical model" of the intraoral organs and a feature vector called a "historical feature vector"; then - Form a record including the historical feature vector and the historical model, the record being called a "historical record"; then - Add the historical record to the index history library.

14. The method of claim 13, wherein the method is performed for a plurality of said intraoral organ sets, said intraoral organ being teeth, each set containing only teeth having the same number or the same type, and the initial reference model used in characterizing the teeth in said sets being a model of a reference tooth having said number or said type, respectively.

15. The method of claim 13, wherein, The intraoral organs are teeth, and the set contains teeth with different numbers or different types, and the initial reference model used to characterize the teeth of the set is the same, regardless of the teeth as the subject of the characterization.

16. A method for identifying an individual, the method comprising the steps of: i) At the first moment, for each intraoral organ in an intraoral organ set comprising more than 500 intraoral organs: - The method according to claim 1 characterizes the intraoral organs to determine a model or "historical model" of the intraoral organs and a feature vector called a "historical feature vector"; then - Add the historical feature vectors to the feature vector database, or - A record is generated that includes the historical feature vector and the historical model; this record is called a "historical record"; then the historical record is added to the index history database. Then, - Associate each historical feature vector in the index history database or the feature vector database with an identifier of the individual having the intraoral organ corresponding to the historical feature vector. ii) At a second time after the first time: - The method according to claim 1 characterizes the target intraoral organs of the target individual to be identified, to determine a digital three-dimensional model or "target model" of the intraoral organs and a corresponding target feature vector; then - Search the index history database or the feature vector database for historical feature vectors corresponding to the target feature vector, and associate the presentation of the identifier with the historical feature vectors corresponding to the target feature vector.

17. A method for calibrating a digital three-dimensional model called an "analysis model," the analysis model modeling intraoral organs called "intraoral analysis organs," the method comprising the steps of: - For each intraoral organ in a collection of more than 500 intraoral organs: - The method according to claim 1 characterizes the intraoral organs to determine a model or "historical model" of the intraoral organs and a feature vector called a "historical feature vector"; then - Form a record including the historical feature vector and the historical model, the record being called a "historical record"; then - The method according to claim 1 characterizes the intraoral analytical organ to generate the analytical model and a corresponding feature vector called an "analytical feature vector"; - Search the index history database for historical feature vectors that best match the analytical feature vectors, and use the historical model of the historical data to correct the analytical model, wherein the correction may involve replacing the analytical model with the historical model.

18. The method of claim 17, wherein, - The intraoral organs and the intraoral analytical organs in the set are teeth with the same number or the same type, and - The initial reference model used in characterizing the teeth in the set is a model of a reference tooth, which has the number or type respectively.

19. The method of claim 17, wherein, For the search, perform the following steps: e1) Define filters associated with one or more parameters of the historical feature vector; e2) Filter the index history database using the filter to retain a subset of the index history database; e3) Modify the filter by increasing the number of parameters involved in the filter and / or by making the filter conditions more stringent by enhancing the filter conditions; e4) Filter the subset using the modified filter to define a new subset, repeating the loop of steps e3) and e4) until the subset obtained in step e3) contains fewer than 5 historical records; e5) Correct the analysis model using one of the historical records obtained from step e4).

20. The method of claim 17, wherein, After the historical model is found in the historical records, the white areas of the analysis model are filled and / or errors in the analysis model are deleted, and / or a portion of the analysis model representing intraoral organs is replaced with the surface of the historical model, and / or the analysis model is replaced with the historical model.

21. A method for evaluating attributes based on a three-dimensional representation called an "evaluation representation" of an intraoral organ called an "intraoral evaluation organ" of an individual called an "evaluation individual," the method comprising the steps of: - Create a learning database containing more than 1000 historical structures, each including: - The three-dimensional representation of the "historical individual's" "historical intraoral organs" is called "historical representation"; and - "History" descriptor, which contains the "history" value of the attribute in relation to the historical individual; - Train at least one neural network using a training database; - The evaluation representation is submitted to the neural network, such that the neural network determines at least one "evaluation" value for the attribute based on the evaluation representation; The method includes: characterizing the historical intraoral organ and the intraoral assessment organ according to the method of claim 1, so as to generate a historical feature vector and an "assessment feature vector" respectively, as historical representation and assessment representation respectively.

Images