Method for tracking movement of tooth row

Abstract

To disclose a method for training a neural network intended to analyze the dental situation of the patient being the latest.SOLUTION: A method includes A) creating a history learning database related to tooth bodies and space attributes associated with the tooth bodies, wherein the history learning database includes 1000 history records or more, each history record is related to an individual history patient, and each history record includes one set of history images depicting the whole tooth bodies (called as a "history tooth body") and one set of values for the space attributes in the history patient, and includes one item space information (called "history space information") for the history patent, and B) a step for training a neural network by providing the neural network with the one set of history images as an input and the history space information as an output.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method for training a neural network intended to analyze a patient's dental situation, to a method for analyzing a patient's dental situation implementing the neural network so trained, and to a method for determining the amount of dental body movement, in particular a method for monitoring the activity of an active orthodontic appliance or the loss of effectiveness of a passive orthodontic appliance. [Background technology]

[0002] The applicant has developed methods for remotely monitoring a patient's dental status before, during or after orthodontic treatment, which methods rely on comparing a photograph taken by the patient with their mobile phone at the most recent moment with a view of a three-dimensional digital model of at least one dentition of the patient's dental arch.

[0003] More specifically, an initial model of the alveolar arch is conventionally created at an initial time point with a 3D scanner and then segmented into tooth models. After the patient takes a photograph, the initial model is deformed by moving the tooth model to best match the photograph. A comparison of the initial model and the deformed model thus obtained provides information about tooth movement since the initial time point. Because the initial model is very accurate, the same is advantageously true for the deformed model, and therefore for the information resulting from the comparison.

[0004] These methods are described in particular in International Patent Application No. PCT / EP2015 / 074868 or International Patent Application No. PCT / EP2015 / 074859.

[0005] In particular, to animate tooth models, they require significant computational resources: typically, several hours of computer processing are required to evaluate a dental situation.

[0006] Moreover, these methods require the creation of the initial model, thus requiring the patient to visit an orthodontist and use a 3D scanner, a procedure that is expensive and uncomfortable for the patient.

[0007] Therefore, there is a need for a method that allows for remote monitoring of a patient and that does not have the disadvantages mentioned above. Summary of the Invention [Problem to be solved by the invention]

[0008] It is an object of the present invention to at least partially address these needs. [Means for solving the problem]

[0009] The present invention proposes a method for training a neural network intended to analyze the dental situation of a current patient, said method comprising: A) creating a historical learning database for a tooth body, e.g., tooth number 13, and spatial attributes associated with the tooth body, e.g., the centroid of tooth number 13, wherein; The historical learning database includes over 1,000 historical records, each historical record relating to an individual historical patient, each historical record comprising: a set of historical images, where all of the historical images depict the dental body in the historical patient (referred to as the "historical dental body"); and one item of spatial information for the historical patient (called "historical spatial information") containing a set of values ​​for the spatial attributes (i.e., values ​​that determine the location of the centroid of tooth 13 for the historical patient in the considered example); Including, B) training the neural network by providing the set of historical images as input and the historical spatial information as output to the neural network. The process includes the steps of:

[0010] Thus, after training, the neural network is able to determine the most recent items of spatial information for a set of recent images that are compatible with the set of historical images in the historical learning database and that are input to the neural network as inputs.

[0011] Preferably, the historical learning database is specialized for precise tooth bodies and for spatial attributes with limited variables, thereby improving its performance capabilities. Preferably, the number of historical images in the historical record is nevertheless limited, thereby allowing the specialized neural network to be used subsequently without having to feed it many images.

[0012] The training method according to the invention preferably has one or more of the following optional features: the neural network is convolutional (CNN, "Convolutional Neural Network"); The spatial attribute is the position of one or more notable points of said dental body in a three-dimensional reference frame, for example the position of one or more notable points of said dental body in a three-dimensional reference frame fixed relative to the alveolar arch of the patient under consideration or relative to a tooth point of the patient under consideration; and / or one or more vectors between characteristic points of the tooth body and / or between a characteristic point of the tooth body and another point, in particular a characteristic point of the patient under consideration, preferably a characteristic point of the oral cavity of the patient under consideration, Define; - the tooth body is one tooth or a set of less than five teeth, preferably less than four teeth; - the history tooth body is a set of teeth including one tooth having a predetermined number or one tooth having a predetermined number and one or two teeth adjacent to the one tooth; - the spatial attributes comprise less than 30, preferably less than 20, preferably less than 10 variables; - the set of historical images for any historical record comprises more than 2, preferably more than 3, preferably more than 5, preferably more than 6, and / or less than 30, preferably less than 20, preferably less than 15, preferably less than 10 historical images; - each historical image of said set of historical images for any historical record is acquired at an angle selected from a group of potential angles including more than 2, preferably more than 4, preferably more than 4, and / or less than 30, preferably less than 20, preferably less than 10 potential angles, or standard angles; - at least three history images of any one set of history images, preferably all history images of any one set of history images, are acquired at different angles; - the historical images are true color photographs depicting dental scenes as they can be perceived by the human eye and / or do not depict dental retractors and / or are extraoral; The history images are acquired at various distances from the history patient and then cropped around the tooth body that the history image depicts.

[0013] The present invention therefore relates to a method for analyzing the dental situation of a patient at the most recent moment (called the "current patient"), said method comprising: a) prior to said latest instant, training the neural network according to the training method according to the invention; b) acquiring, by an image acquisition device, a set of latest images that fit the neural network and depict the dental bodies of the latest patient at the latest moment in time (referred to as "current dental bodies"); c) analyzing the set of most recent images with the neural network to obtain one item of spatial information for the most recent patient (referred to as "most recent spatial information"), the item including a set of values ​​for the spatial attributes; The process includes the steps of:

[0014] An image of a dental scene, e.g., a photograph, is the result of projecting the dental scene onto a plane. Analysis of a representation of a dental body of the dental scene on the image can provide spatial information if the shape of the dental body is known, or if the shape of another object in the dental scene that is depicted and linked to the dental body is known. However, the shape of a patient's dental bodies, particularly teeth, is generally unknown. Therefore, simple analysis of an image of a dental scene generally does not allow reliable spatial information to be determined for a dental body.

[0015] According to the principles of 3D scanners, analyzing several images of the same dental scene can provide accurate spatial information about the dental bodies even if the shape of the dental bodies in the dental scene is unknown. However, recognizing the dental bodies on the images and comparing the images requires lengthy and costly operations.

[0016] As will be seen in more detail throughout the remainder of this specification, and quite unexpectedly, the inventors have discovered that the use of neural networks makes it possible to obtain spatial information with very good accuracy from simple images, particularly from photographs. In particular, it was surprising to discover that errors in spatial information can be less than 1 mm, less than 0.5 mm, or even less than 0.3 mm, without resorting to a three-dimensional model of the patient's alveolar arch. Such accuracy, in fact, was thought, prior to the present invention, to be impossible to achieve from images alone without complex processing, such as processing using a 3D scanner. In particular, it was thought to be impossible to achieve using photographs, such as simple photographs taken by the patient themselves.

[0017] Tests have shown that the current spatial information is reliable even when images are taken with a simple mobile phone without any special precautions, even when the current patient is not wearing a dental retractor and the mobile phone does not need to be fixed on a support, e.g., on a tripod.

[0018] Furthermore, whereas previously several hours of computer processing were required to assess each dental situation, the method according to the present invention allows the latest item of spatial information to be obtained in a matter of seconds.

[0019] Finally, modern patients no longer need to visit a dental professional, and therefore the analysis method can be performed by anyone with a mobile phone, independent of contact with a dental professional.

[0020] The analysis method according to the invention preferably has one or more of the following optional features: - the image capture device is a mobile phone; - the latest image is a true color photograph, the true color photograph depicting a dental scene as can be perceived by the human eye; - the latest patient did not wear a dental retractor in step b); - The most recent image is extraoral; The latest images are taken at various distances from the latest patient and then cropped around the tooth body that the latest image depicts.

[0021] Preferably, in step a), multiple neural networks are trained with individual historical learning databases that differ in that they relate to different tooth bodies and / or different spatial attributes, to obtain multiple specialized neural networks; In step b), a latest image is acquired, and for each specialized neural network of the plurality of specialized neural networks, a set of latest images is created from the latest images; In step c), each of the sets of recent images is analyzed by a corresponding specialized neural network to obtain a plurality of recent items of spatial information, which can generally be referred to as static information.

[0022] Preferably, each historical learning database relates to a tooth or group of teeth having a number that is unique to that historical learning database, where the numbers of the teeth of the group are unique to that historical learning database.

[0023] Preferably, the spatial attributes are the same for all of the historical learning databases.

[0024] The analysis method allows multiple tooth bodies to be analyzed using a neural network specific to each tooth body, so the up-to-date spatial information is accurate and substantial.

[0025] Therefore, the present invention opens up a very wide range of applications.

[0026] In a particularly advantageous manner, the analysis method according to the invention can be carried out several times at different time instants.

[0027] The invention particularly relates to a method for determining the amount of movement between a "previous" most recent instant and a "later" most recent instant after said previous most recent instant, said method comprising: 1) performing the analysis method according to the invention at the most recent previous moment in time to obtain the most recent "previous" item of spatial information, or more generally, a static "previous" item of information; 2) performing the analysis method according to the invention at a later latest instant in time to obtain a "later" latest item of spatial information, or more generally, a static "later" item of information; 3) comparing the previous most recent spatial information with the subsequent most recent spatial information, more generally, comparing the static previous most recent spatial information with the static latest spatial information to obtain the amount of movement between the previous most recent moment and the subsequent most recent moment, wherein the comparison includes in particular finding the difference between the previous most recent spatial information and the subsequent most recent spatial information (or more generally, between the previous static information and the subsequent static information), and optionally dividing the difference by the time interval between the previous most recent moment and the subsequent most recent moment; 4) Presenting the amount of movement, preferably to the current patient and / or dental care professional, for example on a personal computer or mobile phone screen. The process includes the steps of:

[0028] The method of determining according to the invention preferably has one or more of the following optional features: the amount of movement is of one or more points on the latest dental body and / or of one or more vectors connecting a plurality of points on the latest dental body, or of one or more vectors connecting one or more points on the latest dental body with one or more other points in the oral cavity of the latest patient, before The latest moment and rear define the amplitude and / or velocity of the translational and / or rotational movement between the most recent instant of time and the In step 3), the amount of movement is compared with a threshold and, depending on the difference between the amount of movement and the threshold, the following is determined: - the activity index of the most recent patient-worn orthodontic appliance; and / or - a match index of the dental situation of the latest patient with a situation predetermined by orthodontic treatment performed by the latest patient, or a situation resulting from orthodontic treatment performed by the latest patient, or a situation defined by a dental professional independently of orthodontic treatment, for example a defined situation that is normal for the latest patient; - the validity index and / or the fit index are presented in the form of a graph; The validity index and / or the fitness index are presented on a screen, for example on the screen of a computer or mobile phone.

[0029] Preferably, as described above, multiple neural networks are specialized for each tooth or group of teeth, where each neural network is specialized for a tooth, e.g., with a unique number for that tooth.

[0030] The method of determination according to the invention therefore makes it possible to analyze the evolution of several tooth bodies over time using a neural network specific for each tooth body, so that the amount of movement is accurate and substantial. All these amounts of movement can generally be qualified as dynamic information.

[0031] In a preferred embodiment, the specialization is by tooth type: for example, one neural network can be specialized for canines, another for incisors, a third for molars, etc.

[0032] The specialization can be by tooth number: for example, one neural network can be specialized for tooth number 13, another for tooth number 14, a third for tooth number 15, etc.

[0033] The analysis method and the determination method according to the invention in particular comprise: - detecting or assessing the tooth position or shape and / or the evolution of the tooth position or shape and / or the speed of the evolution of the tooth position or shape; in particular - in the pre-treatment period, i.e., the period prior to orthodontic treatment; - in particular to monitor tooth eruption, or to detect repositioning or abnormal positions of teeth, or to detect tooth wear, for example wear caused by bruxism, or to monitor the opening or closing of a space between two or more teeth, in particular between two adjacent teeth, or to monitor the stability or correction of occlusion, when no orthodontic treatment has been carried out; in the context of orthodontic treatment, in particular for monitoring the movement of teeth into a predetermined position, in particular towards an improved position of the teeth, or for monitoring the eruption of teeth, or for monitoring the opening or closing of a space between two or more teeth, in particular between two adjacent teeth, for example to create a suitable space for the attachment of a dental implant; and / or - detecting or assessing the position or shape of an orthodontic appliance, in particular an abnormal position or shape of an orthodontic appliance, such as a delaminating ring or a delaminating orthodontic aligner, and / or the evolution of the position or shape of the orthodontic appliance, and / or the rate of evolution of the position or shape of the orthodontic appliance For this reason, in particular, - Optimizing appointment dates with orthodontists or dentists; and / or - assessing the effectiveness of active orthodontic treatment; and / or - measuring the activity of active orthodontic appliances; and / or - Measuring the loss of effectiveness of passive orthodontic appliances; and / or -Dentistry; and / or measuring the evolution of the shape of the patient's teeth between two dates, for example between two dates before treatment, in particular between two dates separated by an event that is likely to have changed the position and / or shape of at least one tooth, for example between two dates separated by the occurrence of an impact on said tooth, or between two dates separated by the use of a dental device that may have undesirable effects, for example a dental device intended for the treatment of sleep apnea, or between two dates separated by the occurrence of an implant in the patient's oral cavity, in particular a periodontal implant, in particular a gingival implant It can be used for:

[0034] Steps A) and B), and / or steps a) and c), and / or steps 1) to 3) (excluding step b) are preferably performed by a computer. a computer program comprising program code instructions for implementing steps A) and B), and / or steps a) and c), and / or steps 1) to 3) (excluding step b); a computer medium on which such a program is recorded, such as a memory or a CD-ROM; and Computers on which such programs are loaded Regarding.

[0035] definition

[0036] The term "patient" is understood to mean any person on whom the method according to the invention is carried out, whether this person is ill or undergoing orthodontic treatment.

[0037] "Orthodontic treatment" means all or part of a treatment aimed at changing the configuration of the alveolar arch (active orthodontic treatment) or maintaining the configuration of the alveolar arch, especially one performed after the completion of active orthodontic treatment (passive orthodontic treatment).

[0038] An "orthodontic appliance" is an appliance worn by a patient or intended to be worn by a patient. Orthodontic appliances can be intended not only for curative or preventive treatment, but also for aesthetic treatment. Orthodontic appliances can be, in particular, arch-type orthodontic appliances, or orthodontic aligners, or Carrier Motion-type auxiliaries. Such aligners extend along successive teeth of the alveolar arch to which they are attached. They generally define a "U"-shaped channel. The configuration of an orthodontic appliance can be determined not only to ensure its attachment to the teeth, but also as a function of the desired target position of the teeth. More specifically, the shape is determined so that, in the use position, the orthodontic appliance exerts a force tending to move the treated teeth toward their target position (active orthodontic appliance) or a force tending to retain the teeth in this target position (passive orthodontic appliance, or "retainer").

[0039] A "dental situation" defines a set of features related to a patient's alveolar arch at a certain moment, such as the position of the teeth at this moment, their shape, the position of orthodontic appliances, etc. These features may also relate to the general shape of the alveolar arch and / or its position relative to the patient's other alveolar arches, especially in a "closed mouth" position.

[0040] The term "arcade" or "dental arcade" is understood to mean all or part of an alveolar arch. Accordingly, the term "image of an alveolar arch" is understood to mean a two-dimensional representation of all or part of said alveolar arch.

[0041] According to the international convention of the International Dental Federation, each tooth in the alveolar arch has a predetermined number. The tooth numbers defined by this convention are shown in FIG. 7.

[0042] A "scene" is formed by a set of elements that can be viewed simultaneously. A "dental scene" is a scene that includes at least one tooth body in a patient's oral cavity.

[0043] A "noteworthy point" is a point in a dental scene from which one can identify, for example, the apex or cusp of a tooth, an interdental contact point, i.e., a tooth with an adjacent tooth, e.g., a midpoint or distal point of the incisal edge of a tooth, or the center point of a tooth crown, i.e., the "barycenter."

[0044] The term "dental body" is understood to mean an element that can be identified in the oral cavity, such as a tooth, a set of teeth, for example a pair or set of three adjacent teeth, a device intended to be supported by the gums or alveolar arch, in particular an orthodontic appliance, a crown, an implant, a bridge, or a facet. The dental body can also be a subset of the aforementioned elements, such as a tooth with a specific number, or a set of two, three, or more than three adjacent teeth, where one tooth of the set has a specific number.

[0045] A tooth body position is "abnormal" if it does not conform to therapeutic or aesthetic criteria.

[0046] The term "image" is understood to mean a two-dimensional digital representation, for example an image taken from a photograph or film. An image is made up of a number of pixels.

[0047] "Angulation" is the orientation of the optical axis of the image acquisition device relative to the patient when acquiring an image. By extension, if an image is acquired at this angle, then it "exhibits" or "has" or "is associated with" that angle.

[0048] The "tooth zone" of an image is the portion of the image that exclusively depicts a tooth, i.e., that portion of the image that follows the profile of the tooth. In other words, the depiction of the tooth on the image depicts substantially 100% of the tooth zone.

[0049] The term "model" is understood to mean a digital three-dimensional model, also called a "3D" model. A model is made up of a set of voxels. A "tooth model" is a three-dimensional digital model of one tooth. The terms "image of the alveolar arch" or "model of the alveolar arch" are understood to mean a representation of all or part of the alveolar arch in two or three dimensions, respectively.

[0050] 3D scanners or "scanners" are well-known devices for obtaining models of teeth or alveolar arches, which conventionally use structured light and can create 3D models based on different images and matching specific points on those images.

[0051] The method according to the present invention is implemented by a computer, preferably exclusively by a computer, except for image acquisition.The term "computer" is understood to mean any electronic device, including a set of machines, that has computing power.The computer can be a server that is remote from the user, for example, the computer can be "cloud" based.Preferably, the computer is a mobile phone.

[0052] Conventionally, a computer includes, among other things, a processor, memory, a human-machine interface (conventionally including a screen), and communication via the internet, Wi-Fi, Bluetooth, etc. 登録商標 The computer includes a module for communicating via a telephone network or via a telephone line. Software configured to implement the method of the present invention is loaded into the memory of the computer. The computer can also be connected to a printer.

[0053] The terms "primary", "secondary", "updated", "historical", "forward", "backward", "static", and "dynamic" are used for clarity.

[0054] "Before" and "after" refer to successive moments in time.

[0055] A "current" patient is one whose dental condition is intended to be evaluated. A "historical" patient is one who has a corresponding historical record.

[0056] "Spatial attribute" is a generic term that specifies the structure of an item of spatial information. It defines a three-dimensional reference frame, e.g., an orthonormal reference frame, e.g., for tooth 14, an ordered sequence of variables (abscissa of the centroid; ordinate of the centroid; applicate of the centroid). The three-dimensional reference frame is determined relative to the considered patient, e.g., relative to the center of the patient's oral cavity. The three-dimensional reference frame is preferably fixed relative to the considered patient or relative to a part of the considered patient. The origin of the reference frame can in particular be the center of the oral cavity.

[0057] In particular, the three-dimensional reference frame is independent of the position and orientation of the image capture device when capturing the image.

[0058] One or more values ​​of an item of spatial information can always be zero. For example, if the spatial attribute is used to determine the horizontal axis of a dental scene feature along the X axis of the three-dimensional reference frame, only the value of this horizontal axis is non-zero. Alternatively, the value is always non-zero.

[0059] An item of spatial information cannot be inferred solely from the observation of an image where the acquisition conditions are unknown, in particular from the observation of a single acquired image by an image acquisition device whose orientation and distance relative to the dental body under consideration is unknown, e.g. acquired with a mobile phone that is not fixed on a support at a predetermined distance from the patient, e.g. a mobile phone that is supported by the patient, or a mobile phone that is fixed on such a support but whose orientation can be changed.

[0060] Thus, an image provides "surface" information in the plane of the image, for example the position of a particular point of a tooth depicted on the image in the two-dimensional reference frame of the image. Items of spatial information provide depth information relative to the plane of the image. In the example of the position of a particular point of a tooth depicted on the image, the spatial information provides the coordinates of this point that allow this point to be located not only on the image, but also in the depth direction perpendicular to the plane of the image.

[0061] The historical spatial information can be relative, for example, if the historical spatial information represents the distance between two notable points of a tooth, or the distance between a notable point of a tooth and a notable point of another tooth, e.g., a notable point of an adjacent tooth.

[0062] An item of spatial information is the occurrence of a spatial attribute. For example, (2;3;1.5) may define the location of the centroid of tooth 14 of patient "Mr. Martin." This is said to be "updated" or "historical" depending on whether it is associated with the current patient or with history.

[0063] "Before" and "after" are consecutive moments in time.

[0064] "Tooth body" is a generic term, for example, designating tooth number 14. "Current tooth body" and "historical tooth body" are occurrences of the tooth body in the current patient and historical patient, respectively. For example, tooth number 14 of patient "Mr. Martin" can be the current tooth body.

[0065] "Vertical", "horizontal", "right", "left", "horizontal", "in front of" or "frontal", "posterior", "superior", and "inferior" refer to a patient standing upright in a vertical orientation.

[0066] The information is "static" or "dynamic" depending on whether it is the result of analysis at one recent instant or at a series of recent instants.

[0067] "Including" or "including" or "having" is to be understood in an open-ended manner unless otherwise specified.

[0068] Further features and advantages of the present invention will become more clearly apparent upon reading the following detailed description of the invention and upon reference to the accompanying drawings. [Brief explanation of the drawings]

[0069] [Figure 1] FIG. 1 shows schematically how the method for training a neural network according to the present invention can be performed multiple times to obtain a set of neural networks, each specialized for an item of tooth body and / or spatial information. [Figure 2] FIG. 2 shows diagrammatically how the analysis method according to the invention can be performed multiple times, each time with a different specialized neural network. [Figure 3] FIG. 3 shows a schematic representation of the various steps of the method for determining the amount of current tooth movement. [Figure 4] FIG. 4 shows schematically that the method for determining the amount of movement according to the present invention is performed multiple times at two different moments in time, using different specialized neural networks each time, to obtain complex and accurate dynamic information. [Figure 5] FIG. 5 is an example of a set of recent images after processing to isolate the tooth regions. [Figure 6] FIG. 6 is a graph illustrating the accuracy of the most recent item of spatial information obtained in accordance with the present invention. [Figure 7] FIG. 7 shows the tooth numbers used in dentistry. [Figure 8] Figure 8 is a graphical representation of the fit indices.

[0070] Throughout the various figures, the same reference numbers are used to denote similar or identical objects. DETAILED DESCRIPTION OF THE INVENTION

[0071] A training method according to the present invention is illustrated in FIG.

[0072] Neural Networks In step a), the neural network is preferably specialized for image classification.

[0073] Preferably, the neural network is a Convolutional Neural Network (CNN), preferably selected from the following neural networks: -AlexNet (2012); -ZF Net (2013); -VGG Net (2014); -GoogleNet (2015); -Microsoft ResNet (2015); -Caffe:BAIR Reference CaffeNet, BAIR AlexNet; -Torch; -VGG_CNN_S,VGG_CNN_M,VGG_CNN_M_2048,VGG_CNN_M_1024,VGG_CNN_M_128,VGG_CNN_F,VGG ILSVRC-2014 16-layer,VGG ILSVRC-2014 19-layer,Network-in-Network (Imagenet & CIFAR-10); -Google: Inception (V3, V4).

[0074] Preferably, a "squeeze-and-excitation" (SE) processing block is added to the CNN convolutional operator, as described by Jie Hu et al., "Squeeze-and-Excitation Networks," arXiv:1709.01507v4 [cs.CV], 16 May 2019. More preferably, the neural network is of the VGG type with an SE block.

[0075] To operate, an orientation neural network must be trained, conventionally using a learning process called "deep learning," based on a historical learning database tailored to the desired function.

[0076] History Learning Database Training a neural network is a process well known to those skilled in the art and involves presenting the neural network with a historical learning database containing a number of historical records, each of which contains input data and output data.

[0077] Thus, the neural network learns to "match" or stitch together the input and output data.

[0078] To learn to evaluate from a set of recent images the most recent item of spatial information relating to the most recent tooth bodies depicted on those images, e.g., spatial information relating to tooth configurations depicted on those images, the historical learning database preferably consists of a set of historical records each containing: - a set of "historical" images depicting all of the "historical" dental bodies of a "historical" patient, preferably photographs depicting at least one tooth of this patient; - the set of historical descriptors of the historical image, where the descriptors include items of historical spatial information relating to the historical dental bodies depicted in the historical image.

[0079] Preferably, the historical learning database contains more than 1,000, more than 5,000, preferably more than 10,000, preferably more than 30,000, preferably more than 50,000, preferably more than 100,000 historical records. The greater the number of records, the better the analytical ability of the neural network. The number of historical records is conventionally less than 10,000,000 or less than 1,000,000.

[0080] The historical records are each associated with an individual historical patient.

[0081] Historical images All sets of historical images, regardless of the historical record considered, contain the same number of historical images, which is preferably greater than 1, 2, 4, 5 and / or less than 100, 50, 20, 15 or 10, preferably between 5 and 15.

[0082] In one embodiment, this number is three, where the three historical images have different angles.

[0083] The historical images are captured by an image capture device, preferably selected from a mobile phone, a "connected" camera, a "smart" watch, a tablet, or a personal, fixed or portable computer, the image capture device being equipped with an image capture system, e.g. a webcam or camera.

[0084] The historical image is preferably a photograph taken with a mobile phone.

[0085] Preferably, each historical image is a photograph or an image extracted from film. Preferably, it is in color, preferably true color. Preferably, it depicts the dental scene substantially as seen by an operator of the image acquisition device, and in particular in the same colors.

[0086] The historical images are preferably extraoral images, ie, no equipment is introduced into the oral cavity of the historical patient to obtain these images.

[0087] More preferably, the device for acquiring historical images is more than 5 cm, more than 8 cm, or more than 10 cm away from the mouth of the patient, which avoids condensation of water vapor on the optics of the image acquisition device and facilitates focusing. Moreover, preferably, the image acquisition device, in particular a mobile phone, is not equipped with any specific optics for acquiring historical images, which is particularly possible due to the separation between the image acquisition device and the mouth of the patient being acquired.

[0088] In one embodiment, the history patient wears a dental retractor to better expose the history patient's teeth. The retractor can have the features of a conventional retractor. Preferably, it has a rim extending around the retractor opening and is positioned so that the history patient's lips can rest on it while exposing the history patient's teeth through the retractor opening. Preferably, the retractor has brackets to separate the cheeks so that the device for acquiring history images can acquire photographs of the vestibular surfaces of teeth located at the bottom of the oral cavity, such as molars, through the retractor opening.

[0089] In a preferred embodiment, no dental retractors are used. In fact, tests have shown that photographs taken without retractors are generally sufficient for carrying out the method according to the invention. Of course, if necessary, the patient may have to separate their cheeks or lips, for example with their fingers or a spoon.

[0090] All of the historical images in a historical record at least partially depict the same tooth body of the historical patient associated with the record, for example, the incisor of the historical patient.

[0091] The historical image may depict, in particular, one or more teeth, and preferably, several teeth and at least a portion of the gums, or the lips or nose of the historical patient.

[0092] The historical images in a history record all depict the same historical tooth body, but preferably at different angles, i.e., the history images were taken at different orientations of the image capture device relative to the oral cavity of the historical patient. For example, a history record may each include six historical images depicting the same tooth as seen in a "front view," a "right front view," a "right view," a "left front view," a "left view," and a "bottom view."

[0093] The angle of the historical image of a historical record is preferably substantially identical for all historical records relating to the same historical tooth body, e.g., the same type of tooth, e.g., for all historical records relating to incisors, e.g., for the same tooth number, e.g., for all historical records relating to upper right incisors, regardless of the historical record being considered.

[0094] A history record can contain one or more history images at the same angle.

[0095] When acquiring historical images as well as current images in the method according to the present invention, angles can be defined with unlimited accuracy. However, tests have shown that angles do not need to be defined very precisely. Therefore, advantageously, acquiring these images does not require prior training of the operator of the image acquisition device. The historical images, as well as current images, can be acquired with minimal care, for example, using a simple mobile phone.

[0096] In a preferred embodiment, the angle is defined very generally, for example, the angle can be selected from a group of potential angles consisting of "front view", "right view", "right front view", "left view", "left front view", "bottom view", "front bottom view", "top view", and "front top view" of the angle.

[0097] The angle can be defined relative to a "natural" frame of reference, i.e., as a function of how the patient perceives the image acquisition device, in which the angle is selected from a group of potential angles consisting of, for example, the following angles: On the occlusal surface, - "Front view" when the optical axis of the image capture device is substantially coincident with a straight line at the intersection of the occlusal plane and the midsagittal plane; - "right side view" when the optical axis of the image capture device is substantially in the occlusal plane and perpendicular to the midsagittal plane and the image capture device is on the right side of the patient; - "left-sided view" when the optical axis of the image capture device is substantially in the occlusal plane and perpendicular to the midsagittal plane and the image capture device is on the patient's left side; In the midsagittal plane, - "Top view" when the optical axis of the image capture device is substantially in the midsagittal plane and perpendicular to the occlusal plane and the image capture device is above the patient; - "Bottom view" when the optical axis of the image capture device is substantially in the midsagittal plane and perpendicular to the occlusal plane and the image capture device is below the patient.

[0098] The angles can be defined more precisely. In particular, for each of the above angles, the optical axis of the image acquisition device lies in the occlusal plane or in the midsagittal plane. For example, the following angles can be added: - in one of two planes, "right anterior" and "left anterior", inclined at 45° relative to the midsagittal plane and including the line of intersection of the occlusal plane with the midsagittal plane, in the right and left anterior views, the right and left anterior views, the left and right anterior views, and the left and left anterior views; or - occlusal upper and frontal views in one of the "occlusal upper" and "occlusal bottom" planes, which includes a line inclined at 45° relative to the occlusal plane and passing through the center of the oral cavity at the intersection of the occlusal plane with a plane parallel to the frontal plane; or - when the optical axis is at the intersection of a first surface selected from one of two surfaces, the "right anterior" and the "left anterior" surface, and a second surface selected from one of two surfaces, the "occlusal top" surface and the "occlusal bottom" surface.

[0099] Preferably, the angle is selected from one group of potential angles consisting of the angles listed above.

[0100] Preferably, the angle is nevertheless determined as a function of the tooth body, for example, if the tooth body is one tooth or a group of teeth, the angle is preferably fixed as a function of the number of the tooth or teeth, as in practice the configuration of the mouth does not always allow the same angle to be used for all teeth.

[0101] However, there may be cases where the same angle is used for two different tooth bodies, for example an incisor and a canine.

[0102] Precise positioning of the acquisition device is not required when acquiring the history images. Thus, the history images can be acquired at different distances from the mouth. Tests show that the history dental bodies, e.g., teeth, can be depicted at different scales, depending on the history image considered and / or depending on the history record considered, without the performance capabilities of the trained neural network being substantially affected in a significant manner.

[0103] Preferably, however, the image acquisition device is fixed on a support that is placed in a bearing position on the patient's body and is preferably introduced into the patient's mouth. If the support is rigid, it advantageously imposes a predetermined distance between the image acquisition device and the patient's mouth, thereby improving the performance capabilities of the neural network.

[0104] Preferably, the support supports a conventional dental retractor. Such dental retractors conventionally include a rim extending around a retractor opening and are positioned so that the patient's lips can rest thereon while exposing the patient's teeth through the retractor opening.

[0105] Preferably, an image capture device such as that described in European Patent Application No. 17,306,361.1, filed October 10, 2017, is used.

[0106] Moreover, the history image is preferably cropped before being incorporated into the history record. "Cropping" or "re-cutting" is a conventional operation that involves trimming an image to isolate relevant portions therefrom, followed by normalizing the dimensions. Preferably, the history image is cropped to isolate the history tooth bodies, i.e., to depict substantially only the history tooth bodies. Re-cutting can be performed manually or, as described below, by a computer, specifically by a neural network trained for this purpose. Re-cutting the history image significantly improves the performance capabilities of the trained neural network.

[0107] Testing has also shown that the angle does not need to be set precisely as shown above.

[0108] History and specialization The dental body can in particular be a single tooth, or a set of teeth, or an alveolar arch. Preferably, the dental body is selected to limit the variability of its shape among different history records. The dental body is preferably a specific type of tooth or a specific number of teeth.

[0109] A "historical" dental body is a dental body in the particular case of the historical patient considered. In other words, in the historical learning database, every historical dental body is a particular occurrence of a dental body associated with the learning database.

[0110] Preferably, the training database, and therefore the neural network, is specialized only for teeth with specific numbers. For example, they may be specialized for upper right incisors. Then, all historical images of a given history record depict the same historical tooth, the historical spatial information for this history record relates to this historical tooth, and all historical teeth in the training database use the same number. Specializing the neural network for tooth numbers in this way greatly improves its efficiency.

[0111] In this way, several specialized neural networks are preferably trained, where each neural network is trained using a historical learning database dedicated to a tooth type or tooth number. Thus, advantageously, the analysis of the dental situation can implement a specialized neural network for each tooth of the current patient, for the corresponding tooth type or tooth number.

[0112] History Descriptor A set of historical descriptions of a historical image includes items of historical spatial information, i.e., a set of values ​​for variables of spatial attributes, where these values ​​relate to the historical tooth body depicted in the historical image. By definition, a spatial attribute includes at least three variables corresponding to the three dimensions of space. Therefore, the spatial information is a set of at least three values ​​for at least three individual variables of the spatial attribute.

[0113] The spatial attributes are the same not only for all historical images in the set, but also for all historical records.

[0114] The spatial attributes are in particular: - one or more locations in the space of one or more tooth bodies and / or one or more tooth body parts; and / or one or more orientation directions in the space of one or more tooth bodies and / or one or more parts of one tooth body; and / or - one or more orientation courses in the space of one or more tooth bodies and / or in the space of one or more tooth body parts can be defined as:

[0115] One of the portions of one of the tooth bodies may be, for example, one or more points, for example, if the tooth body is a single tooth, a contact point with an adjacent tooth or the center of gravity of the tooth; and / or - one or more rows, for example, if the tooth body is a single tooth, a separation row between adjacent teeth or an edge of a cusp of the tooth; and / or - one or more surfaces of this tooth body It can be.

[0116] In particular, the spatial attribute (x A ,y A ,z A ,x B ,y B and z B) But two points (x A ,y A and z A) and (x B ,y B and z B ), respectively, in an ordered fashion (the position of point A before the position of point B), it indirectly defines an azimuth direction, an azimuth course, and a distance. If it defines positions for three points, and these positions are ordered, it indirectly defines three azimuth directions, and hence the angle between pairs of these directions, the azimuth course along each of these directions, and three distances.

[0117] For example, when defining the centroid position of a first tooth and the centroid position of the tooth immediately to the left of the first tooth, the spatial attributes not only directly define the positions of these centroids, but also indirectly define the azimuth direction of the line connecting these points, the azimuth course from the first centroid to the second centroid, and the distance between these centroids.

[0118] A spatial attribute can define an absolute position in a three-dimensional reference frame fixed relative to the patient, for example, with its origin at the center of the oral cavity of the patient under consideration, with the horizontal axis oriented horizontally and toward the front, the vertical axis oriented horizontally and toward the right, and the applicate axis oriented vertically and upward. It can also define a vector between two points, i.e., the relative position of one point relative to another. For example, the spatial attribute can be defined as (x B -x A ,y B -y A ,z B -z A ), i.e., the relative position of point B relative to point A.

[0119] The spatial information may therefore define absolute or relative positions, and / or azimuthal directions and / or azimuthal courses, and / or distances in a particular dental situation.

[0120] Determining the historical spatial information of the historical record does not pose any problems: it can be determined by any means, for example, manual or computer, in particular by taking measurements from the historical patient, or from a plaster cast of the teeth of the historical patient, or from a digital three-dimensional model of the alveolar arch supporting the teeth under consideration.

[0121] Preferably, the spatial attributes define less than 30 variables, preferably less than 20 variables, preferably less than 10 variables, preferably less than 5 variables, preferably less than 4 variables. The effectiveness of the trained neural network is improved.

[0122] In one embodiment, the spatial attribute defines the location in space of one, preferably more than one, preferably more than two, and / or less than five, preferably less than four, notable points of the tooth body, preferably of a tooth having a particular number. By limiting the historical spatial information related to teeth to a limited number of points, the efficiency of the neural network is significantly improved.

[0123] In step b), the neural network is trained using a historical learning database by sequentially presenting historical records, more specifically, sets of historical images as input and historical spatial information as output.

[0124] Thus, from a set of images similar to a set of historical images presented as input, the neural network learns to provide as output corresponding items of spatial information. In particular, after being trained in this way, the neural network can provide "up-to-date" spatial information regarding the "up-to-date" dental body of the "up-to-date" patient at the "up-to-date" moment. Therefore, the latest spatial information can be used alone or in combination with other information to analyze the dental situation of the latest patient. The analysis method according to the present invention includes steps a) to c), as shown in FIG. 2.

[0125] Thus, the neural network so trained can determine the most recent item of spatial information for the most recent set of images taken on any recent patient and presented to it as input, compatible with the set of historical images in the historical learning database.

[0126] Step a) includes carrying out steps A) and B) above.

[0127] In step b), a set of current images depicting the current dental bodies, preferably the current teeth, of the current patient are acquired at the latest instant by the image acquisition device.

[0128] The latest moment is - allowing each patient to monitor their dental situation at any time, independent of any orthodontic treatment, for example via their mobile phone; - Under active orthodontic treatment; -After passive orthodontic treatment, especially during passive orthodontic treatment It can be.

[0129] The analysis method can particularly be performed during active orthodontic treatment to monitor the progress of the active orthodontic treatment, where the most recent moment is preferably less than three months, less than two months, and / or more than one week, preferably more than two weeks, after wearing an active orthodontic appliance, such as an orthodontic aligner (i.e., "aligner") or an orthodontic arch, intended to correct the position of the patient's teeth.

[0130] The analysis method can also be performed after orthodontic treatment to ensure that tooth positions do not progress adversely ("relapse"). The latest moment is then preferably less than 3 months, less than 2 months, and / or more than 1 week, preferably more than 2 weeks, after passive orthodontic appliances (called "retainers") intended to hold teeth in a certain position after the completion of active orthodontic treatment are worn.

[0131] The latest image is preferably extraoral.

[0132] Preferably, the latest images are images extracted from a photograph or film. They are preferably in color, preferably true color. Preferably, they depict the alveolar arch substantially as seen by an operator of the image acquisition device.

[0133] In one embodiment, the latest patient wears a dental retractor to better expose teeth.However, preferably, no dental retractor is used.Of course, if necessary, the latest patient may need to separate cheeks or lips, for example, by using fingers or by using spoons or other instruments suitable for this purpose.

[0134] The latest image may depict one or more teeth, and preferably depicts at least a portion of the patient's gums, or even lips or nose.

[0135] All of the most recent images must be compatible with the neural network, i.e., they must be images that could have been used for the historical record.

[0136] Preferably, the number of most recent images is the same as the number of historical images in the historical record.

[0137] The latest image must depict the same tooth as the historical image, for example tooth 14.

[0138] The angle of the most recent image is preferably similar to or close to the angle of the historical images in any historical record.

[0139] Generally, the latest images should be similar to the historical images used to train the neural network. For example, if these historical images are extraoral photographs typically taken of the historical patient themselves at variable distances, e.g., 10-50 cm from their mouth, and at approximate angles (e.g., as "front view" or "right view"), then preferably the latest images are also photographs taken under these conditions. If the historical images depict views taken with a dental retractor, then preferably the same is true for the latest images.

[0140] The latest image is acquired by an image acquisition device, which may be the same or different from the image acquisition device used to acquire the historical image, and is preferably the same type of image acquisition device.

[0141] Preferably, the image acquisition system is selected from among a mobile phone, a "connected" camera, a "smart" watch, a tablet, or a personal, fixed or portable computer, including a webcam or camera. Preferably, the image acquisition device is a mobile phone. Preferably, the image acquisition device, in particular a mobile phone, does not have any specific optics for acquiring current images.

[0142] More preferably, to acquire the latest image, the device for acquiring the latest image is located more than 5 cm, more than 8 cm, or even more than 10 cm, and / or less than 50 cm, away from the latest patient's mouth. Advantageously, this distance does not need to be set precisely.

[0143] Preferably, however, the latest image, like the historical images, is "cropped" (or "recut") before being incorporated into the set of latest images that are input to a neural network trained using a historical learning database. Preferably, the latest image is cropped to isolate the latest tooth body, i.e., to depict substantially only the latest tooth body. Recutting can be performed manually or, preferably, by a computer, using a neural network trained specifically for this purpose.

[0144] In particular, a neural network can be trained to identify tooth bodies on an image, for example, to identify tooth regions. Such a neural network for "identifying tooth bodies" is described below in this specification. To crop a current image, the current tooth bodies simply need to be identified by the neural network, and the smallest rectangle that can contain the identified tooth bodies on the image needs to be defined to retain only the interior of the rectangle to define a cut. Then, the dimensions of the cut need to be normalized to define the current image to be incorporated into the set of current images. It is advantageous if the aspect ratio of the rectangle is substantially always the same, regardless of the current image, since the conditions for acquiring the images are likely to be similar. Normalizing the dimensions of the cut includes adjusting these dimensions so that all current images have the same dimensions. Preferably, the operation of re-cutting the historical images is similar to that of the current images, so that the dimensions and pixel counts of these images are similar or substantially identical.

[0145] Recutting the historical images and the most recent images significantly improves the performance of the trained neural network.

[0146] The image acquisition device is preferably used by an operator who is the most recent patient or a close relative of the most recent patient, but this can be any other person, in particular a dentist or orthodontist or a caregiver. Preferably, the most recent image is acquired by the most recent patient.

[0147] Preferably, the latest images are acquired without the use of supports, resting on the ground and without fixing the image acquisition device, in particular without a tripod.

[0148] However, in one embodiment, the image acquisition device is attached to a support that is supported and positioned on the body of the historical patient, and is preferably partially introduced into the mouth of the historical patient. If the support is rigid, it advantageously imposes a predetermined distance between the image acquisition device and the mouth of the current patient, thereby improving the performance capabilities of the neural network.

[0149] Preferably, the support supports a conventional dental retractor, which conventionally includes a rim extending around a retractor opening and positioned so that the latest patient's lips can rest thereon while exposing the latest patient's teeth through the retractor opening.

[0150] Preferably, an image capture device such as that described in European Patent Application No. 17,306,361.1, filed October 10, 2017, is used.

[0151] The set of most recent images being acquired More preferably, the operator is guided during step b), preferably in real time, to orient the image acquisition device at a predetermined angle and / or preferably to take a predetermined number of images at different angles and / or to orient the image acquisition device at the required angle.

[0152] For this reason, it is preferred that an application is loaded into the image acquisition device to provide on-board control during step b), i.e. to ensure that the number and / or angle and / or quality of the most recent images are satisfactory.

[0153] The application may in particular implement fool-proofing means that facilitate approximate positioning of the image acquisition device relative to the current patient before acquiring the current image.

[0154] The foolproof means may in particular include a display on the screen of the image acquisition device and a criterion, e.g., a superimposition criterion, that the operator must match with the latest patient part displayed on the screen, e.g., the teeth, gums, lip profile or face profile.

[0155] The fiducials can be, for example, a geometric shape, such as a point, one or more lines, such as parallel lines, a star, a circle, an ellipse, a regular polygon, in particular a square, a rectangle or a rhombus, or a combination of any one or more of these shapes. The one or more fiducials are preferably "fixed" on the screen, i.e. they do not move on the screen when the image acquisition device is moving.

[0156] The reference can consist, for example, of a horizontal line intended to coincide with the general orientation of an on-screen representation of the horizontal joint between the upper and lower teeth when the teeth are clamped together by the current patient, and / or a vertical line intended to coincide with a representation of the vertical joint between the two upper incisors. The reference can be formed, for example, by two circles to be placed over the on-screen representation of the current patient's eyes. The reference can be formed, for example, by an ellipse to be placed around the on-screen representation of the current patient's mouth or face.

[0157] A "part of the patient" can also be a support worn by the current patient, for example, supported by a dental retractor or by a part bitten by the current patient.

[0158] The application can also help the operator to change the angle of the image acquisition device by announcing or displaying on the screen a message, for example "Take the right photo", "Higher", "Lower", or by emitting a series of beeps whose frequency increases as the orientation of the image acquisition device is improved. For this purpose, the image displayed on the screen of the image acquisition device must be analyzed in real time to determine in particular whether tooth bodies are depicted and, preferably, to check whether the angle is appropriate.

[0159] Algorithms for detecting objects in images are well known to those skilled in the art and can be used to search for tooth bodies in the displayed image. Preferably, for example, a neural network is used to identify tooth bodies, preferably selected from among the following "Object Detection Networks": -R-CNN (2013); -SSD (Single Shot MultiBox Detector:Object Detection Network),Faster R-CNN (Faster Region-based Convolutional Network method:Object Detection Network); -Faster R-CNN (2015); -SSD (2015); -RCF (Richer Convolutional Features for Edge Detection) (2017).

[0160] It is no problem for a person skilled in the art to train a neural network to detect tooth bodies in an image, e.g., teeth with a given number, in particular by providing the neural network with an image as input and with information relating to the presence or absence of tooth bodies as output.

[0161] The following articles deal specifically with detection and segmentation: https: / / arxiv.org / pdf / 1405.0312.pdf and https: / / arxiv.org / pdf / 1703.06870.pdf.

[0162] In one embodiment, the neural network is trained using a training database consisting of a set of more than 1,000, preferably more than 10,000, records, each record comprising: - an image comprising a zone depicting a tooth body, for example comprising at least one tooth zone relating to the tooth having the determined number; - a description of the image that identifies the area on the image in which the tooth body is depicted Includes.

[0163] During training, each image is provided as an input to the neural network, while the associated descriptor is provided as an output from the neural network.

[0164] Thus, upon completion of the training, the neural network is able to determine regions depicting tooth bodies, e.g., tooth regions, in the images provided to the neural network as input.

[0165] Figure 5 shows a set of recent images taken at different angles after being processed by the neural network trained in this way.

[0166] The angle can also be identified by a neural network trained for this purpose, preferably chosen among CNNs, where the final layer of the neural network operates a regression.

[0167] The neural network is trained using a training database consisting of a set of more than 1,000, preferably more than 10,000, records, each record being: - an image comprising a zone depicting a tooth body, for example comprising at least one tooth zone relating to the tooth having the determined number; - a descriptor of the image that identifies an angle on the image Includes.

[0168] During training, each image is provided as an input to the neural network, while the associated descriptor is provided as an output from the neural network.

[0169] Thus, upon completion of the training, the neural network can determine the angle of the images that are fed to the neural network as input.

[0170] Preferably, the application defines a set of predetermined angles and the number of latest images to be taken for each predetermined angle. When the application is launched, in step b), the application preferably calculates, in real time: - analyzing the image displayed on the screen of the image acquisition device at the current moment, i.e. the "current image", to determine at which angle the tooth body is depicted and preferably at which angle the image acquisition device, i.e. the "current angle"; - triggering the acquisition by an operator or automatically if the tooth body is depicted, preferably if the current angle complies with the current requirements, i.e. if it is still necessary at the current moment to acquire a new image at the current angle; otherwise, preferably by informing the operator to correct the angle of the image acquisition device, or by terminating with step b) if no further up-to-date images need to be acquired. The process is carried out.

[0171] In a preferred embodiment, the acquisition is triggered automatically, i.e., without operator action, as soon as the displayed image depicts the tooth body and the angle is accepted by the image acquisition device.

[0172] Written and / or voice messages can be emitted by the image capture device to guide the operator, for example, the image capture device can announce "take a frontal photo" and signal the operator that the orientation is acceptable, or conversely, the image capture device can signal the operator that the photo needs to be retaken.

[0173] The end of the capture process may be announced verbally by the image capture device or by an on-screen display.

[0174] Guiding the operator in acquiring the most recent images advantageously enables a set of recent images to be formed that is immediately suitable for a trained neural network.

[0175] Composition of the set of most recent images after acquisition The set of recent images can also be formed in part or in whole by acquiring the recent images and then selecting a sufficient number of the recent images that depict the recent tooth bodies at the desired angles.

[0176] Determining whether a current image can belong to a set of current images specialized for a current tooth body, e.g., for a tooth number, includes checking whether this current image depicts this current tooth body, e.g., the tooth with this number, and preferably checking whether the angle is appropriate.

[0177] The selection can be made manually, especially if the desired angle is approximate: for example, it is easy to take a front view photo, a right view photo with the mouth open, a left view photo with the mouth open, a right view photo with the mouth closed, and a left view photo with the mouth closed.

[0178] The latest image can also be selected by a computer. A neural network, such as the neural network described above for acquiring the image, can be used. Identification of tooth bodies and / or determination of angles can also be performed using conventional image analysis, although such analysis is time consuming.

[0179] If the most recent image depicts the tooth body at an angle and the set of most recent images still requires an additional most recent image for this angle, the most recent image is added to the set.

[0180] If the most recent image depicts the tooth body at the desired angle but is redundant, e.g., is of higher quality because it depicts more surface area of ​​the tooth body than the other most recent image, it can also replace other most recent images present in the specialized set.

[0181] In step c), the set of most recent images formed in step b) is input into the neural network trained in step a).

[0182] In response, the neural network provides up-to-date spatial information about the current tooth body.

[0183] Multiple runs with different neural networks The neural network is preferably specialized for a limited number of tooth bodies, e.g., a determined number of teeth, and the spatial attributes preferably include a limited number of variables.The analysis method is then preferably performed multiple times at the latest instant by modifying each time the considered tooth bodies and / or the considered spatial attributes.

[0184] All of the determined up-to-date spatial information is therefore referred to as "static information", which allows a detailed analysis of the patient's current dental situation to be obtained by applying it to specialized neural networks.

[0185] In a preferred embodiment, in step a), several specialized neural networks are trained, each specialized for a tooth body, preferably for an individual tooth type or tooth number. Preferably, then, in step b), a sufficient number of recent images are acquired to form a set of "specialized" recent images suitable for each of the specialized neural networks.

[0186] Preferably, the latest images are all acquired substantially simultaneously. Optionally, the batch of acquired latest images is analyzed to identify one or more specialized sets to which they may belong.

[0187] The static information can be enhanced. For example, for each of three sets of adjacent teeth consisting of a central tooth, a left tooth, and a right tooth, if the central tooth is located between the left and right teeth, the analysis method can be performed to determine the position of the tooth's centroid in a reference frame fixed relative to the alveolar arch supporting these teeth. This set of three positions is the static information. To enhance this static information, it is possible to determine the angle formed by two straight line segments passing through the centroids of the left and right teeth, respectively, with the centroid of the central tooth as the common origin, from the three positions. The distance between the centroid of the central tooth and the centroid of the left tooth or the centroid of the right tooth can also be determined.

[0188] The analysis method can be performed multiple times by changing the spatial attributes considered each time. For example, the analysis method can be performed to determine the location of the contact point between a tooth and a first adjacent tooth, and then to determine the location of the contact point between the tooth and a second adjacent tooth. The set of coordinates defining these two locations is static information.

[0189] The analysis method is preferably performed multiple times by modifying the considered tooth body or the considered spatial attribute each time. For example, if a triplet of adjacent teeth consisting of a central tooth, a left tooth, and a right tooth is considered, where the central tooth is located between the left and right teeth, the analysis method can be performed successively to determine, in a reference frame fixed relative to the alveolar arch supporting these teeth, the position of the centroid of the left tooth, then the position of the centroid of the right tooth, then the position of the contact point of the central tooth with the left tooth, then the position of the contact point of the central tooth with the right tooth.

[0190] To enhance the static information thus obtained, for example, the angle between two planes perpendicular to the line connecting the two centroids of the left and right teeth and to the line connecting the two contact points of the central tooth with the left and right teeth, respectively, can be measured, thus providing information related to the orientation of the central tooth relative to the left and right teeth.

[0191] Use of static information The static information can be used to assess the current dental status of the patient.

[0192] In particular, it can be used to evaluate whether the current dental body is in a position that belongs to a predetermined spatial region, in particular, for example, a region that defines a set of predefined positions that are considered acceptable. For example, such a region can be defined around a tooth or a notable point on a tooth, so that if the tooth is even partially out of this region or if the notable point is out of this region, the dental situation is considered abnormal. Such a region can also be defined around an orthodontic appliance or a notable point on the orthodontic appliance, so that if the orthodontic appliance is even partially out of this region or if the notable point is out of this region, the dental situation is considered abnormal.

[0193] The static information can be used to evaluate whether the current tooth body has an orientation that belongs to a predetermined set of orientations, in particular an orientation that belongs to a set of orientations that define a set of orientations that are considered acceptable. In particular, the orientation of a tooth can be defined by the angle formed by two straight lines passing through a notable point of the tooth, for example the center of gravity of the tooth, where a first of the two straight lines passes through a notable point, for example the center of gravity, of the right side of the tooth, preferably the tooth adjacent to the tooth, and a second of the two straight lines passes through a notable point, for example the center of gravity, of the left side of the tooth, preferably the tooth adjacent to the tooth.

[0194] The static information can also be used to evaluate the distance between a notable point on the current tooth body and another notable point on the alveolar arch that supports the current tooth body, for example, to evaluate if the distance between the center of gravity of a tooth and the center of gravity of a tooth adjacent to that tooth belongs to a predetermined distance range, in particular a set of distances that are considered acceptable.

[0195] The area limits defined for an item of static information or the tolerance limits defined for an item of static information, ie, "static constraints," are preferably defined by dental professionals.

[0196] The static information can be used, inter alia, to determine if the current patient's dental situation has become abnormal. For example, to detect relapse, the static constraints can correspond to the dental situation at the completion of orthodontic treatment, and can have an optional tolerance margin.

[0197] The static information can be used to determine the position of the current patient's first alveolar arch relative to the current patient's second alveolar arch, and in particular to detect and / or assess the presence of vertical or horizontal overhang, especially if the overhang is abnormal.

[0198] In the case of monitoring orthodontic treatment, the static constraints may correspond to the dental situation expected at the latest moment or at the completion of the orthodontic treatment, and may optionally have a tolerance margin.

[0199] If no treatment is performed, the static constraints may define a set of dental conditions that are considered normal.

[0200] In one embodiment, the static constraints are current patient independent, i.e., applicable to any patient in a patient group, and therefore form a standard, i.e., a "standard set-up."

[0201] Preferably, the standard is specific to a pathology and / or a type of orthodontic treatment and / or to a group of patients sharing a common characteristic, for example, a group of patients belonging to the same age group and / or the same sex, and the standard can in particular determine the arcade shape.

[0202] The standard may in particular define the dental situation at the time of completion of treatment or at the latest moment.

[0203] In particular, if static constraints are not respected, e.g., if the position, orientation and / or distance determined from the static information is not acceptable, a message can be sent to the most recent patient and / or dental professional to inform them.

[0204] The static information can be displayed in the form of a graph.

[0205] For example, the static information can be presented on a computer screen or on a mobile phone screen.

[0206] For example, the graph may represent teeth with colors that depend on a compatibility index that represents the compatibility of each tooth position with a predefined position, e.g., with the predefined position at the latest moment, e.g., the darker the tooth, the further its position is from the predefined position.

[0207] The static information in particular comprises: - detecting or assessing the tooth position or shape and / or the evolution of the tooth position or shape and / or the speed of the evolution of the tooth position or shape: in particular - in the pre-treatment period, i.e., the period prior to orthodontic treatment; - in particular to monitor tooth eruption, or to detect repositioning or abnormal positions of teeth, or to detect tooth wear, for example wear caused by bruxism, or to monitor the opening or closing of a space between two or more teeth, in particular between two adjacent teeth, or to monitor the stability or correction of occlusion, when no orthodontic treatment has been carried out; in the context of orthodontic treatment, in particular for monitoring the movement of teeth into a predetermined position, in particular towards an improved position of the teeth, or for monitoring the eruption of teeth, or for monitoring the opening or closing of a space between two or more teeth, in particular between two adjacent teeth, for example to create a suitable space for the attachment of a dental implant; and / or - detecting or assessing the position or shape of an orthodontic appliance, in particular an abnormal position or shape of an orthodontic appliance, such as a delaminating ring or a delaminating orthodontic aligner, and / or the evolution of the position or shape of the orthodontic appliance, and / or the rate of evolution of the position or shape of the orthodontic appliance For this reason, in particular, - Optimizing appointment dates with orthodontists or dentists; and / or - assessing the effectiveness of active orthodontic treatment; and / or - measuring the activity of active orthodontic appliances; and / or - Measuring the loss of effectiveness of passive orthodontic appliances; and / or -Dentistry; and / or measuring the evolution of the shape of the patient's teeth between two dates, for example between two dates before treatment, in particular between two dates separated by an event that is likely to have changed the position and / or shape of at least one tooth, for example between two dates separated by the occurrence of an impact on said tooth, or between two dates separated by the use of a dental device that may have undesirable effects, for example a dental device intended for the treatment of sleep apnea, or between two dates separated by the occurrence of an implant in the patient's oral cavity, in particular a periodontal implant, in particular a gingival implant It can be used for:

[0208] If the static information is used to evaluate the evolution, it can be compared with a situation defined at a moment before the latest moment without resorting to the method according to the invention. For example, if the static information is used to evaluate the evolution of a tooth position or shape, it can be compared with this predefined tooth position or shape without carrying out steps a) to c), where the position or shape was predefined, for example, at the start of the treatment or at an intermediate moment of the treatment.

[0209] Multiple runs with different recent moments: amount of movement The analysis method according to the invention can also be performed one or more times at different "before" and "after" latest moments, as shown in FIG.

[0210] By comparing the "previous" latest spatial information (i.e., static information) obtained at the previous latest instant with the "later" latest spatial information (i.e., each with static information) obtained at the later latest instant, it is possible that the evolution can be determined between these two latest instants. This evolution is brought about in the time interval between the two latest instants, and it is possible that the speed of this evolution can be determined.

[0211] The term "amount of movement" refers to the information that results directly or indirectly from such a comparison.

[0212] The amount of movement can in particular be the movement of a notable point between the two most recent moments, or by dividing this movement by the time interval between these two most recent moments, it can be the average movement speed between these most recent moments.

[0213] Accordingly, the present invention relates to a method for determining the amount of current movement of a patient's teeth, comprising steps 1) to 3), and optionally step 4).

[0214] In step 1), the analysis method according to the invention is performed at the most recent previous moment to obtain previous spatial information regarding the most recent tooth body.

[0215] Applicable before The most recent moment can be, for example, less than 3 months, less than 2 months, less than 1 month, less than 1 week, or less than 2 days after wearing an active or passive orthodontic appliance, such as an orthodontic aligner, orthodontic arch, or retainer.

[0216] The analysis method according to the invention can be performed multiple times as a function of the previous static information desired, as described above.

[0217] In step 2), the same analysis method according to the present invention is performed at a later, most recent time instant to obtain later spatial information, which therefore relates to the same latest tooth body and the same spatial attributes as the analysis method performed at the previous time instant.

[0218] The later most recent moment is preferably more than 2 weeks, more than 1 month, more than 2 months or more than 6 months, and / or less than 5 years, less than 3 years or less than 1 year later than the previous most recent moment.

[0219] In step 2), the same one or more neural networks as in step 1) are used. Therefore, step a) is not necessary.

[0220] If the analysis method according to the invention is performed multiple times in step 1), then similarly in step 2), subsequent items of static information equivalent to the previous static information are obtained.

[0221] In step 3), the previous and subsequent spatial information are compared to determine the amount of motion.

[0222] The previous and subsequent spatial information are values ​​that can be compared with each other, for example in terms of differences.

[0223] For example, if the previous and subsequent spatial information consist of the coordinates (x1, y1, and z1) and (x2, y2, and z2) of the tooth centroids at the most recent previous moment t1 and the most recent subsequent moment t2 in a fixed reference frame relative to the alveolar arch, then (x2-x1) 2 +(y2-y1) 2 +(z2-z1) 2 The square root of allows the distance covered by this centroid to be evaluated between the latest previous instant t1 and the latest subsequent instant t2.

[0224] The result of the comparison can consist of one or more values. For example, consider a situation where the spatial attributes are (x';y';z';x'';y'';z''), where (x';y';z') and (x'';y'';z'') are respectively the positions of two salient points of one tooth, namely P' and P'' in a three-dimensional reference frame, for example an orthonormal reference frame, i.e. (Ox; Oy; Oz), and the origin is for example the center of the alveolar arch supporting this tooth. If the anterior and posterior spatial information are denoted as (x1';y1';z1';x1'';y1'';z1'') and (x2';y2';z2';x2'';y2'';z2''), respectively, then: -x1',y1',z1' are the coordinates x, y and z of point P' at the most recent previous moment (e.g., in mm); -x2', y2', z2' are the coordinates x, y and z of point P'' at the latest moment later (e.g., in mm); -x1'', y1'', z1'' are the coordinates x, y and z of point P'' at the most recent previous moment (e.g., in mm); -x2'', y2'', z2'' are the coordinates x, y and z of point P'' at the latest moment in time (e.g., in mm).

[0225] Next, the comparison result is, for example, before The latest moment of t1 and rear It can be the distance covered by point P' and the distance covered by point P'' between the most recent instant t1 and the most recent instant t2 (the comparison results in a binary value), or the arithmetic mean of these two distances (the comparison results in a single value).

[0226] In step 4), preferably, the amount of movement is present, for example, on the current patient's and / or dental care professional's personal computer screen or mobile phone screen.

[0227] As shown in FIG. 4, steps 1) and 2) can be performed multiple times by modifying the spatial attributes of the latest tooth body and / or analysis method each time. For example, they can each be repeated for multiple teeth of the latest patient. In step 3), the latest pre-spatial information and the latest post-spatial information obtained for each set of steps 1) and 2) can be compared. The latest pre-spatial information and the latest post-spatial information obtained for different sets of steps 1) and 2) can also be combined to obtain improved information.

[0228] The term "dynamic information" generally refers to all information that results directly or indirectly from one or more performances of steps 1) through 3), where each performance is at the most recent previous instant t1 and the most recent subsequent instant t2.

[0229] Use of Dynamic Information The dynamic information can be used to assess the evolution of the patient's dental situation over time.

[0230] It can be used to evaluate whether the translational or rotational movement speed of a particular point of the current tooth body is within a range of values ​​that define a predefined set of speeds, for example a set of speeds that are considered acceptable.

[0231] The dynamic information can be used to assess whether the dynamics of an active orthodontic treatment comply with the expected dynamics, i.e., whether the teeth move at a rate consistent with the orthodontic treatment.

[0232] The tolerance limits defined for the dynamic information, or "dynamic constraints," are preferably defined by dental professionals.

[0233] The dynamic information can be displayed in the form of a graph.

[0234] For example, the dynamic information can be presented on a computer screen or on a mobile phone screen.

[0235] Preferably, the graph summarizes all information resulting directly or indirectly from one or more performances of steps 1) to 3), where each performance is the most recent previous instant t1 and the most recent subsequent instant t2.

[0236] For example, in FIG. 8, teeth would be stained as a function of a fit index representing whether the speed of movement of each tooth matches a predetermined speed in the context of orthodontic treatment undergone by a current patient.

[0237] Using this dynamic information is generally easier than using static information: for example, determining whether a notable point on a tooth has moved abnormally is generally easier than determining whether the position of this point in space is abnormal.

[0238] For example, to detect relapse, the dynamic constraint can correspond to an allowed movement margin for the center of gravity of a tooth. To monitor orthodontic treatment, the dynamic stress can correspond to the course of movement of one tooth relative to another (a decrease or increase in the distance between these teeth) to ensure that the two teeth move toward or away from each other. The dynamic constraint can also correspond to a threshold value for the speed of movement of an orthodontic appliance or the speed of movement of the tooth point on which the orthodontic appliance acts to ensure the activity of the orthodontic appliance.

[0239] The dynamic information can also be used to measure the evolution of the shape of a tooth or set of teeth.

[0240] Thus, the dynamic information may in particular include: - detecting or assessing the evolution of the tooth position or shape and / or the speed of the evolution of the tooth position or shape; in particular - in the pre-treatment period, i.e., the period prior to orthodontic treatment; - in particular to monitor tooth eruption, or to detect repositioning or abnormal positions of teeth, or to detect tooth wear, for example wear caused by bruxism, or to monitor the opening or closing of a space between two or more teeth, in particular between two adjacent teeth, or to monitor the stability or correction of occlusion, when no orthodontic treatment has been carried out; in the context of orthodontic treatment, in particular for monitoring the movement of teeth into a predetermined position, in particular towards an improved position of the teeth, or for monitoring the eruption of teeth, or for monitoring the opening or closing of a space between two or more teeth, in particular between two adjacent teeth, for example to create a suitable space for the attachment of a dental implant; and / or - detecting or assessing the position or shape of an orthodontic appliance, in particular an abnormal position or shape of an orthodontic appliance, such as a delaminating ring or a delaminating orthodontic aligner, and / or the evolution of the position or shape of the orthodontic appliance, and / or the rate of evolution of the position or shape of the orthodontic appliance For this reason, in particular, - Optimizing appointment dates with orthodontists or dentists; and / or - assessing the effectiveness of active orthodontic treatment; and / or - measuring the activity of active orthodontic appliances; and / or - Measuring the loss of effectiveness of passive orthodontic appliances; and / or -Dentistry; and / or measuring the evolution of the shape of the patient's teeth between two dates, for example between two dates before treatment, in particular between two dates separated by an event that is likely to have changed the position and / or shape of at least one tooth, for example between two dates separated by the occurrence of an impact on said tooth, or between two dates separated by the use of a dental device that may have undesirable effects, for example a dental device intended for the treatment of sleep apnea, or between two dates separated by the occurrence of an implant in the patient's oral cavity, in particular a periodontal implant, in particular a gingival implant It can be used for:

[0241] In particular, if the dynamic constraints are not respected, e.g. if the amplitude and / or speed and / or direction of movement of a tooth feature determined from the dynamic information is not permissible, a message can be sent to the current patient and / or dental professional to inform them.

[0242] In one embodiment, the static information and / or the dynamic information is used to assess whether a goal has been achieved and / or to measure the difference between the patient's current dental condition at the most recent moment and the achievement of the goal.

[0243] The objective is preferably selected from the following objectives: - The current patient achieves a Class 1 occlusion for the canines; - The current patient achieves a Class 1 occlusion for the molars; - the anterior sector of the latest patient is closed; - The space resulting from the patient's most recent tooth extraction is closed; - the current patient has a normal horizontal overhang, i.e. "overjet", preferably a normal horizontal overhang of 1 to 3 mm; - the current patient has a normal vertical overhang, i.e. an "overbite", preferably a normal vertical overhang of 1-3 mm; - the current patient's inter-incisal sectors of the upper and lower arcades are not offset; - the current patient has no lateral offset of the inferior and / or superior cisternal arch relative to the sagittal plane of the patient's head; - the current patient does not have a lateral offset of the superior cisternal arch relative to the inferior cisternal arch; - the orthodontic appliances worn by the current patient, such as orthodontic arches and / or orthodontic aligners and / or auxiliaries, are passive, i.e. no longer act to change the position of the current patient's teeth; - no or only limited detection of the patient's most recent tooth movement has been detected during the last two checks of the upper and / or lower alveolar arch, preferably no detection of the patient's most recent tooth movement; - All primary teeth of the most recent patient have been lost; -Lack of lateral open bite (closure of lateral open bite); -Lack of posterior open bite; -Lack of anterior open bite; -Lack of anterior crossbite; -Lack of posterior crossbite; -Improvement of crowding; -Dent stability; -Closed interdental space; -Absence of mucosal irregularities.

[0244] Example

[0245] In one example, the dental bodies under consideration are a set of two teeth having a determined type or number, for example, tooth number 13 (upper right canine) and the adjacent tooth number 14 (upper right first premolar). The spatial attributes are a triplet of coordinates, i.e., parameters (X, Y, Z), for the vector connecting the centroid of tooth number 13 and the centroid of tooth number 14. Therefore, the spatial information is formed by the triplet of values ​​of these coordinates.

[0246] The spatial information is measured relative to a Cartesian reference frame (Ox;Oy;Oz), the origin O of which is the center of the upper alveolar arch of the patient under consideration.

[0247] In step a), a historical learning database containing 100,000 historical records is created.

[0248] All the historical images were extraoral photographs taken in true color, without retractors, and then cropped.

[0249] The photographs were most often taken with a personal camera, typically a cell phone, but sometimes with a dental professional's camera. They were taken at various distances from the historical patient and then preferably cropped so that teeth 13 and 14 were substantially the same size, regardless of the historical image considered.

[0250] Any historical record includes a set of four historical images depicting all of the historical patient's teeth 13 and 14, as well as their angles (in the occlusal plane) as "front view," "top view," "right view," and "right front view."

[0251] The historical spatial information of a historical record consists of the vector (Xi, Yi, Zi). In other words, from the centroid of tooth 13, we reach the centroid of tooth 14 by moving along axis Ox to value Xi, along axis Oy to value Yi, and then along axis Oz to value Zi. The historical spatial information of each record is determined manually from a 3D model of the alveolar arch of the historical patient created with a 3D scanner.

[0252] In step b), a neural network CNN, such as GoogleNet (2015), is trained using the historical learning database so that it can determine the vector between the centroids of teeth 13 and 14 based on a set of recent images similar to the set of historical images used for training.

[0253] In step b), a current patient, e.g., a patient who is not scheduled for any orthodontic treatment and does not wear a retainer, is considered at the most recent previous instant t1. The current patient has a mobile phone onto which an application capable of performing steps b) and c) has been downloaded. The current patient wants to check the dental status of teeth 13 and 14 of the current patient, and the patient launches the application.

[0254] The application activates the face-phone camera and prompts the current patient to take four photos of these two teeth at the above angles: "front view," "top view," "right view," and "right front view."

[0255] The application then crops these photographs, taken at various distances from the most recent patient, so that the size of teeth 13 and 14 is substantially the same regardless of the most recent image and is substantially identical to the size of these teeth in the historical images.

[0256] Next, the application inputs all four latest images into the trained neural network. This trained neural network can be integrated within the application or can be mounted on a computer remote from the mobile phone. In the latter case, the mobile phone transmits the four latest images to the remote computer to input the four latest images into the trained neural network.

[0257] In step c), based only on these four latest images, the trained neural network preferably provides the latest item of spatial information in less than 120 seconds, less than 60 seconds, less than 40 seconds, less than 20 seconds, less than 10 seconds, or less than 5 seconds. The latest item of spatial information is a vector (Xa, Ya, Za), and in an orthonormal reference frame (Ox; Oy; Oz) whose origin O is at the center of the latest patient's upper dental arch, it enables the center of gravity of the 13th tooth of the latest patient to connect to the center of gravity of the 14th tooth of the latest patient.

[0258] The vector (Xa, Ya, Za) is a static item of information that can be compared with preset static constraints. For example, it is possible to check whether |Xa| < Sx, and / or |Ya| < Sy, and / or |Za| < Sz, where Sx, Sy, and Sz are threshold values, for example, 0.5 mm, 0.7 mm, and 0.3 mm. For example, it is also possible to check whether |Xa| + |Ya| + |Za| < S, where S is a threshold value. If the constraints are not respected because, for example, |Xa| > Sx, a message is sent to notify the latest patient of them. For example, a message asking the latest patient to contact a dental healthcare worker is displayed on the screen of their mobile phone.

[0259] If the neural network is in a remote computer, the computer transmits the latest spatial information and / or the message to the application.

[0260] The latest patient can, for example, execute the same process (launching the application, taking a photo, and inputting them into a trained neural network) at the previous latest moment t2, for example, one month after the previous latest moment.

[0261] For this second implementation of the present invention, the application can not only analyze the static information obtained at the later latest moment, i.e., the vector (Xa', Ya', Za'), in the same way as the previous latest moment, but also compare it with the static information obtained at the previous latest moment, i.e., the vector (Xa, Ya, Za). It can, for example, determine the following amounts of movement: |Xa'-Xa|, |Ya'-Ya|, |Za'-Za|, |Xa'-Xa|+|Ya'-Ya|+|Za'-Za|, |Xa'-Xa| / (t2-t1), |Ya'-Ya| / (t2-t1), |Za'-Za| / (t2-t1), or (|Xa'-Xa|+|Ya'-Ya|+|Za'-Za|) / (t2-t1).

[0262] These amounts of movement form dynamic information that advantageously teaches the temporal development of the dental situation regarding the 13th and 14th teeth, more specifically, the relative movement of the 13th tooth with respect to the 14th tooth. Therefore, these amounts of movement above complement the static information.

[0263] The dynamic information can be compared with predefined dynamic constraints. For example, it is possible to check whether |Xa'-Xa| / (t2-t1)<Vx, |Ya'-Ya| / (t2-t1)<Vy, |Za'-Za| / (t2-t1)<Vz, or (|Xa'-Xa|+|Ya'-Ya|+|Za'-Za|) / (t2-t1)<V, where Vx, Vy, Vz, and V represent, for example, the threshold values of 0.1 mm / month, 0.2 mm / month, 0.1 mm / month, and 0.3 mm / month respectively. If the constraints are not respected, a message is sent to notify the latest patient. For example, the message is displayed on the screen of the latest patient's mobile phone, prompting the latest patient to contact the dental practitioner.

[0264] If the neural network is in a remote computer, the computer sends all or part of the latest spatial information and / or the dynamic information and / or the message to the application.

[0265] Preferably, the current patient performs the above-described operations for all pairs of teeth in his or her alveolar arch (tooth 1 and tooth 2, tooth 2 and tooth 3, etc.). For each pair of teeth, the application preferably executes a specialized procedure to provide sufficient photographs taken at different predefined angles for the considered pair of teeth. The photographs for the determined pair of teeth are input into a neural network specialized for this pair of teeth.

[0266] In one embodiment, the dynamic information is used to measure the effectiveness, or "activity," of an active orthodontic appliance, i.e., its ability to affect the alveolar arch at the most recent instant. For example, the rate of tooth movement can be used to determine whether the orthodontic appliance will remain effective (if these rates are greater than a threshold, e.g., 0.1 mm / month) and therefore whether the orthodontic appliance needs to be changed or modified and / or whether an appointment with a dental professional is necessary. If appropriate, written or verbal messages are sent to the current patient and / or dental professional.

[0267] Example 6 illustrates an example of an implementation of the method according to the invention. The date is the horizontal axis. The vertical axis provides the cumulative movement in millimeters in all directions for all teeth in the patient's alveolar arch.

[0268] The solid curve represents the "real" development. To determine this curve, a digital three-dimensional model of the patient's alveolar arch is first generated with a 3D scanner. It is then deformed to correspond to the tooth arrangement observed at different moments by implementing the method described in International Patent Application No. PCT / EP2015 / 074859. Each point on this curve requires several hours of computer processing.

[0269] The dashed curve represents the evolution determined according to the present invention, each point on this curve requiring only a few seconds of computer processing.

[0270] It is quite surprising to see that the dashed curve follows the solid curve very well, and is therefore very quickly computed yet realistically represents the actual development.

[0271] As should now be apparent, the present invention provides a solution for determining position, distance, azimuth direction, or azimuth course in the volume of a patient's oral cavity that is fast, reliable, and requires limited computational resources.

[0272] In particular, the application can be downloaded to a mobile phone and implemented within seconds.

[0273] Additionally, it provides accurate information, typically with an accuracy of 0.3 mm or less.

[0274] Finally, it can be performed effortlessly and without the need for any pre-created 3D models, from simple extraoral photographs taken by the patient with their own mobile phone.

[0275] Thus, the present invention allows the dental situation of any person to be assessed not only during active or passive orthodontic treatment, but also when no orthodontic treatment is being performed, even if the person has never previously seen a dental professional. The present invention may be configured as follows. [Section 1] 1. A method for training a neural network intended to analyze a current patient's dental situation, comprising: A) creating a historical learning database of tooth bodies and spatial attributes associated with said tooth bodies, wherein: The historical learning database includes over 1,000 historical records, each historical record relating to an individual historical patient, each historical record comprising: a set of historical images, where all of the historical images depict the dental body in the historical patient (referred to as the "historical dental body"); and One item of spatial information for the historical patient (called "historical spatial information"), including a set of values ​​for the spatial attributes. Including, B) training the neural network by providing said set of historical images as input and said historical spatial information as output to said neural network, wherein said spatial attributes define an ordered sequence of a plurality of variables in a three-dimensional reference frame; The training method includes the steps of: [Section 2] Item 1, wherein the neural network is a convolutional neural network. [Section 3] The spatial attribute is the location of one or more notable points of said tooth body in a three-dimensional reference frame; and / or One or more vectors between distinctive points of the tooth body and / or between one distinctive point of the tooth body and another distinctive point of the tooth body The training method according to item 1 or 2, wherein [Section 4] the dental body is one tooth or a set of less than five teeth; and / or the spatial attribute comprises fewer than 30 variables; and / or the set of historical images for any historical record includes more than 2 and less than 30 historical images; and / or each historical image in said set of historical images for any historical record has an angle selected from a group of potential angles including more than 2 and less than 30 potential angles; and / or At least three history images in any one set of history images have different angles; Item 3. The training method according to any one of Items 1 to 3. [Section 5] Item 5: The training method according to any one of items 1 to 4, wherein the history tooth body is a set of teeth including a tooth having a predetermined number or a tooth having a predetermined number and one or two teeth adjacent to the tooth. [Section 6] 6. The training method according to any one of items 1 to 5, wherein the historical images are true color photographs, and / or do not depict dental retractors, and / or are extraoral. [Section 7] 1. A method for analyzing the dental situation of a patient at a most recent moment (referred to as the "current patient"), comprising: a) training the neural network prior to said latest moment according to the training method described in any one of items 1 to 6; b) acquiring, by an image acquisition device, a set of latest images that fit the neural network and depict the dental bodies of the latest patient at the latest moment in time (referred to as "current dental bodies"); c) analyzing the set of most recent images with the neural network to obtain one item of spatial information for the most recent patient (referred to as "most recent spatial information"), the item including a set of values ​​for the spatial attributes; The method for analyzing includes the steps of: [Section 8] Item 8. The method of analysis according to item 7, wherein the latest image is a true color photograph and / or is extraoral, and / or in step b), the image acquisition device is a mobile phone, and / or the latest patient is not wearing a dental retractor. [Section 9] Item 9. The method of analysis according to item 7 or 8, wherein in step b), the latest patient is wearing a dental retractor. [Section 10] In step a), multiple neural networks are trained with individual historical learning databases that differ in that they relate to different tooth bodies and / or different spatial attributes, to obtain multiple specialized neural networks; In step b), a latest image is acquired, and for each specialized neural network of the plurality of specialized neural networks, a set of latest images is created from the latest images; In step c), each of the plurality of sets of recent images is analyzed by a corresponding specialized neural network to obtain a plurality of recent items of spatial information. Item 10. The analysis method according to any one of Items 7 to 9. [Section 11] Item 11. The analysis method described in item 10, wherein in step a), each historical learning database relates to a tooth or group of teeth having a number unique to the historical learning database, and the numbers of the teeth in the group are unique to the historical learning database. [Section 12] The latest spatial information is used to assess whether an objective has been achieved and / or to measure a difference between the dental situation of the latest patient at the latest moment and the achievement of the objective, wherein the objective is selected from the following objectives: The latest patient achieves a Class 1 occlusion for the canines; The latest patient achieves a Class 1 occlusion for the molars; The anterior portion of the space in the latest patient is closed; The space resulting from the extraction of the patient's most recent tooth has closed; The latest patient has a normal horizontal overhang; The latest patient has normal vertical overhang; the inter-incisor sectors of the upper and lower alveolar arches of the latest patient are not offset; the latest patient has no lateral offset of the inferior cisternal arch and / or superior cisternal arch relative to the sagittal plane of the patient's head; the latest patient does not have a lateral offset of the superior cisternal arch relative to the inferior cisternal arch; orthodontic appliances worn by the current patient no longer operate to alter the position of the current patient's teeth; No or limited tooth movement has been detected in the latest patient between the last two checks of the upper and / or lower alveolar arch; The most recent patient has lost all of their primary teeth; absence of lateral open bite; absence of posterior open bite; Absence of an anterior open bite; Lack of anterior crossbite; Lack of posterior crossbite; Improvement of crowding; dent stabilization; Closed interdental spacing; Absence of mucosal irregularities, Item 12. The analysis method according to any one of Items 7 to 11. [Section 13] 1. A method for determining an amount of motion between a most recent previous instant and a most recent subsequent instant after said most recent previous instant, comprising: 1) performing the analysis method described in any one of items 7 to 10 at the most recent previous moment to obtain the most recent previous item of spatial information; 2) performing the analysis method described in any one of items 7 to 10 at a later latest moment to obtain a later latest item of spatial information; 3) comparing the most recent previous spatial information with the most recent subsequent spatial information to obtain an amount of motion between the most recent previous moment and the most recent subsequent moment; 4) Optionally, presenting the amount of movement to the current patient and / or dental care professional. The method, comprising the steps of: [Section 14] 14. The method of claim 13, wherein the amount of movement defines the amplitude and / or speed of translational and / or rotational movement of one or more points of the latest dental body and / or of one or more vectors connecting a plurality of points of the latest dental body, or of one or more vectors connecting one or more points of the latest dental body with one or more other points of the oral cavity of the latest patient, between the most recent previous time instant and the most recent subsequent time instant. [Section 15] In step 3), the amount of movement is compared with a threshold, and depending on the difference between the amount of movement and the threshold, the effectiveness index of the most recent patient-worn orthodontic appliance; and / or a match index of the latest patient's dental condition with a condition predetermined by orthodontic treatment performed by the latest patient, or a condition resulting from orthodontic treatment performed by the latest patient, or a condition determined by a dental care professional independent of orthodontic treatment; Item 15. The method for determining according to Item 13 or 14, wherein: [Section 16] Item 16. The method of any one of items 13 to 15, wherein the effectiveness index and / or the compatibility index are presented in the form of a graph. [Section 17] Detecting or assessing tooth position or shape and / or the evolution of tooth position or shape and / or the rate of evolution of tooth position or shape; and / or Detecting or assessing the position or shape of an orthodontic appliance and / or the evolution of the position or shape of an orthodontic appliance and / or the rate of evolution of the position or shape of an orthodontic appliance; and / or Measuring the evolution of the patient's tooth shape between two dates; and / or dentistry A method for using the analysis method according to any one of Items 7 to 12 or the determination method according to any one of Items 13 to 16 for the purpose of: [Section 18] Monitoring tooth eruption; and / or Detecting recurrent or abnormal tooth positions; and / or Detecting tooth wear; and / or monitoring the opening and closing of spaces between two or more teeth; and / or Monitoring the stability or correction of dental occlusion; and / or monitoring the movement of teeth into position; and / or Detecting or assessing ring delamination or orthodontic aligner delamination; Optimizing appointment dates with orthodontists or dentists; and / or To evaluate the effectiveness of active orthodontic treatment; and / or Measuring the activity of active orthodontic appliances; and / or Measuring the loss of effectiveness of passive orthodontic appliances; and / or Measuring the evolution of the shape of a patient's teeth between two dates separated by the occurrence of an impact on the teeth, or between two dates separated by the use of a dental device intended for the treatment of sleep apnea syndrome, or between two dates separated by the occurrence of an implant in the patient's oral cavity. The method of use described in paragraph 17 for.

Claims

1. 1. A method for training a neural network intended to analyze a current patient's dental situation, comprising: training the neural network by providing the neural network with a set of historical images as input and historical spatial information as output, wherein the spatial attributes define an ordered sequence of a plurality of variables in a three-dimensional reference frame; wherein all of said set of historical images depict dental bodies in a historical patient (referred to as "historical dental bodies"); the historical spatial information is an item of spatial information for the historical patient that includes a set of values ​​for the spatial attributes; and wherein the set of historical images and the historical spatial information are included in each historical record associated with the historical patient in a historical learning database, the historical learning database including more than 1,000 historical records associated with the dental bodies and the spatial attributes associated with the dental bodies. The method.

2. 2. The method of claim 1, wherein the neural network is convolutional.

3. The spatial attribute is the location of one or more notable points of said tooth body in a three-dimensional reference frame; and / or One or more vectors between distinctive points of the tooth body and / or between one distinctive point of the tooth body and another distinctive point of the tooth body. The method of training according to claim 1 or 2, further comprising defining:

4. the tooth body is one tooth or a set of less than five teeth; and / or the spatial attribute comprises fewer than 30 variables; and / or the set of historical images for any historical record includes more than 2 and less than 30 historical images; and / or each historical image in said set of historical images for any historical record has an angle selected from a group of potential angles including more than 2 and less than 30 potential angles; and / or At least three of any set of history images have different angles; A training method according to any one of claims 1 to 3.

5. The training method according to any one of claims 1 to 4, wherein the history tooth body is a set of teeth including a tooth having a predetermined number or a tooth having a predetermined number and one or two teeth adjacent to the plurality of teeth.

6. 6. A method of training according to any one of claims 1 to 5, wherein said historical images are true-color photographs and / or do not show dental retractors and / or are extraoral.

7. 1. A method for analyzing the dental situation of a patient at a most recent moment (called the "current patient"), comprising: analyzing a set of recent images with a neural network to obtain one item of spatial information for the recent patient (referred to as "recent spatial information"), the item including a set of values ​​for the spatial attributes; wherein the neural network was trained according to the training method of claim 1 prior to the most recent time instant, the set of latest images is acquired by an image acquisition device and fitted to the neural network, depicting the dental bodies of the latest patient (referred to as "current dental bodies"); The method for analyzing.

8. 8. The method of claim 7, wherein the most recent image is a true color photograph, and / or is extraoral, and / or the image acquisition device is a mobile phone, and / or the most recent patient is not wearing a dental retractor.

9. The method of claim 7 or 8, wherein the latest patient is wearing a dental retractor.

10. Multiple neural networks are trained with individual historical learning databases that differ in that they relate to different tooth bodies and / or different spatial attributes, resulting in multiple specialized neural networks; latest images are acquired, and for each specialized neural network of the plurality of specialized neural networks, a set of latest images is created from the latest images; Each of the plurality of sets of recent images is analyzed by a corresponding specialized neural network to obtain a plurality of recent items of spatial information. The analysis method according to any one of claims 7 to 9.

11. The method of claim 10, wherein each historical learning database relates to a tooth or group of teeth having a number that is unique to the historical learning database, and the numbers of the teeth in the group are unique to the historical learning database.

12. The latest spatial information is used to assess whether an objective has been achieved and / or to measure a difference between the latest dental situation of the patient at the latest moment and the achievement of the objective, wherein the objective is selected from the following objectives: The current patient achieves a Class 1 occlusion for the canines; The current patient achieves a Class 1 occlusion for the molars; The anterior portion of the space in the latest patient is closed; the space resulting from the extraction of the patient's most recent tooth is closed; The latest patient has a normal horizontal overhang; The latest patient has normal vertical overhang; the inter-incisor sectors of the upper and lower alveolar arches of the latest patient are not offset; the latest patient has no lateral offset of the inferior cisternal arch and / or superior cisternal arch relative to the sagittal plane of the patient's head; the latest patient does not have a lateral offset of the superior cisternal arch relative to the inferior cisternal arch; orthodontic appliances worn by the current patient no longer operate to alter the position of the current patient's teeth; No or limited tooth movement has been detected in the latest patient between the last two checks of the upper and / or lower alveolar arch; All of the latest patient's primary teeth have been lost; absence of lateral open bite; Absence of posterior open bite; Absence of an anterior open bite; Lack of anterior crossbite; Absence of posterior crossbite; Improvement of crowding; dent stability; Closed interdental spacing; Absence of mucosal irregularities, The analysis method according to any one of claims 7 to 11.

13. 1. A method for determining an amount of motion between a most recent previous instant and a most recent subsequent instant after said most recent previous instant, comprising: 1) performing the analysis method according to any one of claims 7 to 10 at the most recent previous moment to obtain the most recent previous item of spatial information; 2) performing the analysis method according to any one of claims 7 to 10 at a later latest instant to obtain a later latest item of spatial information; 3) comparing the most recent previous spatial information with the most recent subsequent spatial information to obtain an amount of motion between the most recent previous moment and the most recent subsequent moment; 4) Optionally, presenting the amount of movement to the current patient and / or dental care professional. The method, comprising the steps of:

14. 14. The method of claim 13, wherein the amount of movement defines the amplitude and / or velocity of translational and / or rotational movement of one or more points of the most recent dental body and / or of one or more vectors connecting a plurality of points of the most recent dental body or one or more vectors connecting one or more points of the most recent dental body with one or more other points of the oral cavity of the most recent patient, between the most recent previous time instant and the most recent subsequent time instant.

15. In step 3), the amount of movement is compared with a threshold, and depending on the difference between the amount of movement and the threshold, the effectiveness index of the most recent patient-worn orthodontic appliance; and / or a match index of the latest patient's dental condition with a condition predetermined by orthodontic treatment performed by the latest patient, or a condition resulting from orthodontic treatment performed by the latest patient, or a condition determined by a dental care professional independent of orthodontic treatment; 15. The method of claim 13 or 14, wherein:

16. 16. The method of claim 15, wherein the validity index and / or the fit index are presented in the form of a graph.

17. Detecting or assessing tooth position or shape and / or the evolution of tooth position or shape and / or the rate of evolution of tooth position or shape; and / or Detecting or assessing the position or shape of the orthodontic appliance and / or the evolution of the position or shape of the orthodontic appliance and / or the rate of evolution of the position or shape of the orthodontic appliance; and / or measuring the evolution of the patient's tooth shape between two dates; and / or dentistry A method for using the analysis method according to any one of claims 7 to 12 or the determination method according to any one of claims 13 to 16 for the purpose of

18. monitoring tooth eruption; and / or Detecting recurrent or abnormal positions of teeth; and / or Detecting tooth wear; and / or monitoring the opening and closing of spaces between two or more teeth; and / or Monitoring the stability or correction of dental occlusion; and / or monitoring the movement of the teeth into position; and / or Detecting or assessing ring delamination or orthodontic aligner delamination; Optimizing appointment dates with orthodontists or dentists; and / or To assess the effectiveness of active orthodontic treatment; and / or Measuring the activity of active orthodontic appliances; and / or Measuring the loss of effectiveness of passive orthodontic appliances; and / or Measuring the evolution of the shape of the patient's teeth between two dates separated by the occurrence of an impact on the teeth, or between two dates separated by the use of a dental device intended for the treatment of sleep apnea syndrome, or between two dates separated by the occurrence of an implant in the patient's oral cavity. The use according to claim 17 for:

Images