Method for generating dental arch model

Abstract

The invention relates to a method for generating a dental arch model, comprising: 1) before step 3), determining an active intermediate model according to an update time; (2) searching one or more representations of teeth which do not conform to the treatment scheme in updated images, the updated images being obtained at the updating time in the treatment process, and each updated image being an orthodontic appliance image representing a movable orthodontic appliance or an image representing a bare tooth without an orthodontic appliance; if one or more non-conforming teeth are detected, 3) identifying one or more tooth models representing the one or more non-conforming teeth in the active intermediate models; 4) deforming the active intermediate model until an updated model adapted to at least one updated image, referred to as a'deformed updated image ', is obtained, in which method a non-conformity of at least one non-conforming tooth is measured by comparing the updated image with a view of the active intermediate model, and then in step 4), in accordance with the measurement, a non-conformity of at least one non-conforming tooth is determined. And moving the at least one tooth model which does not conform to the teeth.

Description

[0001] This application is a divisional application of the invention patent application with the application date of May 20, 2020, application number "202080052662.7", and invention name "Method for generating a dental arch model". Technical Field

[0002] The invention relates to a method for generating a three-dimensional digital model of a dental arch.

[0003] The invention furthermore relates to a method for producing an orthodontic appliance by means of the production method according to the invention, in particular for the purpose of regulating an orthodontic treatment by means of such an appliance.

[0004] Finally, the invention relates to a computer system for implementing these methods. Background Art

[0005] like Figure 1 and Figure 2 As shown, the orthodontic appliance 10 is generally in the form of a removable device of unitary construction, generally made of a transparent polymer material and shaped to conform to the continuous teeth of the dental arch to which it is attached. The orthodontic appliance includes a groove 12, the overall shape of the groove 12 is U-shaped and is shaped so that a plurality of teeth of the dental arch, typically all of the teeth of the dental arch, can be accommodated therein.

[0006] The shape of the groove is determined to ensure that the appliance is attached to the teeth, but is also determined according to the desired target position of the teeth. More precisely, the shape is determined so that when the appliance is in its use position, it exerts a stress tending to move the treated teeth towards the target position.

[0007] Typically, at the beginning of orthodontic treatment, it is determined what shapes the various appliances must have at different times during the treatment, and then all the corresponding appliances are manufactured. To this end, it is known to carry out the following steps:

[0008] - at an initial time t1 (usually at the start of treatment), generating a three-dimensional digital model of the patient's dental arch (referred to as the "initial model"), said dental arch being in an initial configuration; and cutting the initial model into tooth models;

[0009] - determine the time suitable for moving the dental arch from the initial configuration to the intermediate configuration at each intermediate time t n Intermediate structure repair

[0010] Change to final time t N+1 Orthodontic treatment of the final structure, n is 2 to N;

[0011] - Deform the initial model to generate intermediate and final structures of the dental arch, respectively

[0012] Model and final model;

[0013] - Based on the initial model, the intermediate model and the final model, a series of N appliances is determined, the first appliance being intended to be worn until time t2 and the nth appliance being intended to be worn from time t n Worn until time t n+1 ;

[0014] - Manufacturing at least some of said appliances.

[0015] All manufactured aligners are then provided to the patient so that he can change the aligners at predetermined multiple intermediate times.

[0016] During treatment, the patient visits the orthodontist regularly for visual examinations, especially to verify that the movement of the teeth is as expected and that the appliances he is wearing are still appropriate for treatment.

[0017] In particular, the orthodontist can visually diagnose the unseat of the appliance. Specifically, the bottom 20 of the groove has a shape that is substantially complementary to the shape of the free end 22 of the tooth ( Figure 5 ). Thus, the contour of the groove bottom can be compared with the contour of the tooth D to assess the gap between the groove bottom and one or more free ends of the tooth.

[0018] If a dislocation is detected, the orthodontist takes a new impression of the teeth, or equivalently a new scan of the teeth, and then repeats the above process to design and manufacture a new series of appliances.

[0019] Having to visit the orthodontist is a burden on the patient. The patient’s confidence in their orthodontist may also decrease. Ultimately, this leads to additional costs. Therefore, the number of visits to the orthodontist must be limited.

[0020] Additionally, ill-fitting aligners may also be unsightly.

[0021] In order to solve these problems, the applicant proposed in EP 3 412 245 a method for evaluating the shape of an orthodontic appliance worn by a patient.

[0022] The method advantageously allows remote detection of dislocation of an appliance. It greatly facilitates the assessment of the suitability of an appliance treatment. In particular, it can be implemented based on simple images, in particular on photographs or films, for example, taken by the patient without special measures. Thus, the number of appointments with the orthodontist can be limited.

[0023] However, when a tooth that does not comply with treatment is detected, in particular in the case of dislocation between the appliance and the tooth, an appointment must be made with the orthodontist in order to produce a new set of appliances. In addition to the inconvenience this causes to the patient, such appointments also lead to delays in treatment. In particular, treatment must be interrupted during the time between the detection of dislocation and the receipt of the new appliance.

[0024] A technical solution to these problems is needed.

[0025] It is an object of the present invention to at least partially meet this need. Summary of the invention SUMMARY OF THE INVENTION

[0027] The present invention provides a method for generating a three-dimensional digital model called an "updated model" for a patient's dental arch during treatment of the dental arch using an orthodontic appliance called a "removable appliance", in particular when treatment is performed using a series of orthodontic appliances intended to be successively attached to the dental arch, the treatment having been simulated by a treatment plan generated, for example, at an initial time t1 at the beginning of the treatment, the treatment plan comprising a plurality of intermediate models, each of which is a three-dimensional digital model of the dental arch, the intermediate models being cut into tooth models and determined to represent the dental arch at a corresponding intermediate time after the initial time.

[0028] The generation method according to the present invention comprises the following steps:

[0029] 1) at an update time during treatment, acquiring at least one update image, each update image being an image of the appliance representing the removable appliance attached to the dental arch in a use position, or an image representing the bare teeth of the dental arch without the appliance;

[0030] 2) before step 4), preferably before step 3), determining the intermediate model or "active intermediate model" according to the update time;

[0031] 3) searching in an updated image, referred to as an "analysis updated image," for one or more representations of teeth that do not conform to the treatment plan;

[0032] If one or more non-conforming teeth are detected, then

[0033] 4) identifying one or more tooth models that respectively represent one or more non-conforming teeth in the active intermediate model;

[0034] 5) deforming the active intermediate model until an updated model matching at least one of the update images (referred to as the "updated deformed image") is obtained.

[0035] The invention is based on the fact that an intermediate model of a treatment plan designed before the start of treatment correctly simulates the dental arch at a corresponding intermediate time (if non-compliant teeth are neglected). In particular, if the appliance is not dislocated from a tooth, this indicates that the treatment of this tooth is progressing as planned. Therefore, at an intermediate time close to the update time, the corresponding intermediate model corresponds to the reality of the "compliant" teeth (usually almost all teeth). Therefore, it is possible to use such an active intermediate model, which is usually initially generated for the manufacture of the appliance, as a starting point for the creation of an update model.

[0036] In particular, all parts of the active intermediate model that are relevant to the conforming teeth can be immediately utilized.The updated model can thus be constructed based on the active intermediate model, thereby finding only the actual positions of the tooth model that do not conform to the teeth.

[0037] Advantageously, the patient therefore no longer needs to perform a new scan to adapt the appliance to a negative progression of his treatment. The treatment plan and one or preferably more than one updated image are sufficient. This greatly simplifies and accelerates the treatment.

[0038] Furthermore, the determination of the updated model is simplified, not only because the number of tooth models to be moved is very limited (typically 1 to 5 tooth models represent non-compliant teeth), but also because these movements are constrained by the tooth models that are compliant.

[0039] The method may include, before step 1), generating a treatment plan including the plurality of intermediate models. The treatment plan is generated after the treatment itself has been determined. The treatment plan may also be generated while the treatment itself is determined.

[0040] The method according to the invention may also include one or more of the following optional features:

[0041] - at least some or even all of the intermediate models of the treatment plan represent the dental arch in the intended configuration at intermediate times marking appliance changes;

[0042] - in step 1), a reminder is sent to the patient, preferably to his mobile phone, in order for him to take at least one update image, preferably at least one image of the appliance and preferably at least one image of the bare tooth;

[0043] - in step 1), acquiring more than two updated images, preferably more than four updated images;

[0044] - in step 1), at least two updated images are acquired under different acquisition conditions, in particular at different orientations of the acquisition device, the angle of the optical axis of the acquisition device and the frontal plane of the patient preferably varying by more than 20°, more than 30°, more than 45°, more than 60°, or even more than 90° between the acquisition of the at least two updated images;

[0045] - In step 1), at least one, preferably each, updated image is a photograph or an image extracted from a film;

[0046] - In step 1), at least one, preferably each, updated image is an extraoral image;

[0047] - in step 1), at least one, preferably each, update image is acquired with a mobile phone, preferably by the patient himself, and optionally after the dental retractor has been installed;

[0048] - In one embodiment, in step 1), at least one, preferably each, update image is acquired using the mobile phone, preferably by the patient himself, after the mobile phone and the dental retractor have been attached to the bracket and then the dental retractor has been placed in the patient's mouth;

[0049] - the support is in the form of a housing that opens exclusively to the retractor and the mobile phone;

[0050] - acquiring all updated images at time intervals of a duration of less than 5 days, preferably less than 1 day, preferably less than 1 hour, and preferably less than 10 minutes;

[0051] - the patient sends one or more updated images to a computer, preferably via a mobile phone, which acquires an image of the appliance;

[0052] - the computer is configured to receive and process updated images of a plurality of patients, preferably more than 100, more than 1000 or more than 10000 patients;

[0053] - before step 2), an intermediate model of the dental arch is generated based on the intermediate model of the treatment plan, and then added to the treatment plan as an intermediate model;

[0054] - in step 2), the active intermediate model is determined manually by an operator, preferably by a dental professional, more preferably by an orthodontist, preferably using a computer allowing him to view the treatment plan, or the determination is performed automatically by a computer, preferably having received and processed updated images of a plurality of patients, preferably by comparing the update times with the intermediate times of the intermediate models of the treatment plan;

[0055] - in step 2), the active intermediate model is an intermediate model whose intermediate time differs from the update time by less than 4 weeks, less than 2 weeks, preferably less than 1 week;

[0056] - in step 2), the active intermediate model is the intermediate model whose intermediate time is closest to the update time;

[0057] - in step 3), the search for representations of non-conforming teeth in the analyzed updated images is performed manually by an operator, preferably by a dental professional, more preferably by an orthodontist, preferably using a computer allowing him to view one or more updated images, or preferably automatically by a computer, preferably having received and processed updated images of a plurality of patients, the computer preferably implementing a deep learning device, preferably a neural network;

[0058] - in step 3), if, when the analysis updated image is superimposed on the view of the active intermediate model matched to said analysis updated image, the analysis updated image has the same scale as the view of the active intermediate model and has a true scale (the size of the actual tooth is the same as the size of its representation), at least one point of the representation of the tooth in the analysis updated image is separated from the corresponding point in said view by a distance greater than 1 / 10 mm, 3 / 10 mm, 5 / 10 mm or 1 mm and preferably less than 7 mm or 5 mm, then said representation is considered not to comply with the treatment plan;

[0059] - in step 4), the tooth models that do not conform to the teeth in the active intermediate model are identified manually by an operator, preferably by a dental professional, more preferably by an orthodontist, preferably using a computer that allows him to view the active intermediate model, or the identification is performed automatically by a computer, preferably having received and processed updated images of a plurality of patients, the computer preferably implementing a deep learning device, preferably a neural network;

[0060] - in step 3), the analysis update image is analyzed independently of the active intermediate model in order to detect a not conforming tooth, in which case the active intermediate model may be determined after step 3);

[0061] - alternatively, in step 3), the updated image is compared with a view of an active intermediate model for detecting non-compliant teeth, in which case the active intermediate model must be determined before step 3);

[0062] - In step 3), in order to detect non-compliant teeth,

[0063] finding a position, orientation and calibration of the virtual acquisition device that allows it to have a view on the active intermediate model that is as close as possible to the analysis update image, i.e. a view of the active intermediate model that has the greatest degree of match (i.e. the best fit) with the analysis update image; then

[0064] comparing the view with the analysis update image, or comparing an update map representing the determination information of the analysis update image with a reference map representing the determination information in the view;

[0065] - In step 3), in order to detect non-compliant teeth,

[0066] analyzing and updating the image into an image of an orthodontic appliance, determining a contour of at least one tooth and a contour of the orthodontic appliance in the analyzed and updated image, and then comparing the contours;

[0067] - If in the appliance image, a tooth is dislocated from the appliance beyond a threshold, the tooth is considered noncompliant;

[0068] - before step 5), processing the active intermediate model to improve its accuracy;

[0069] - in step 5), the deformation of the active intermediate model comprises, preferably consists of, a movement of the tooth model of said active intermediate model;

[0070] - In step 5), continue to move the tooth model until the positioning error of each tooth model relative to the deformed updated image is less than 1 mm, preferably less than 5 / 10 mm, preferably less than 3 / 10 mm, preferably less than 2 / 10 mm, and preferably less than 1 / 10 mm;

[0071] - in step 5), the tooth model that does not conform to the teeth is moved manually by an operator, preferably by a dental professional, more preferably by an orthodontist, preferably using a computer that allows him to view the live intermediate model, or automatically by a computer, preferably having received and processed updated images of a plurality of patients, the computer preferably implementing a deep learning device, preferably a neural network, or an optimization method, preferably a metaheuristic optimization method;

[0072] - in step 5), the movement of the non-conforming tooth model is constrained by the conforming tooth model (which remains stationary);

[0073] - In step 5), the movement of the tooth model that does not conform to the tooth is an iterative process, so that in each iteration,

[0074] - moving one or more of the tooth models that do not conform to the teeth to obtain a dental arch model to be tested, and then

[0075] - testing the model to be tested by evaluating the degree of match between the model and the deformed updated image, in particular the image of the bare tooth;

[0076] The updated model is the one that provides the highest fit among all the models tested;

[0077] - before said iterative process, finding the position, orientation and calibration of the virtual acquisition device or "constrained virtual acquisition conditions" that allow observing the active intermediate model to generate a view in which the conforming representation of the teeth can be superimposed and registered with said conforming representation of the teeth in the deformed updated image, then,

[0078] During the iterative process, in each iteration, evaluating the degree of fit between the model under test and the deformed updated image by comparing the deformed updated image with the view of the model under test obtained under the constrained virtual acquisition conditions;

[0079] - if the number of iterations exceeds a predetermined number or if the value of the degree of match exceeds a predetermined threshold, interrupting the iteration loop;

[0080] - in step 5), during said deformation of the active intermediate model, the only tooth model that moves is the tooth model that does not conform to the teeth;

[0081] - in step 3) or 4), preferably at the end of step 4), measuring the non-conformity of one or more non-conforming teeth, in particular the movement of one or more dislocated teeth, based on at least one updated image, preferably by comparing said analyzed updated image with the active intermediate model, and then

[0082] In step 5), according to the measurement, one or more tooth models that do not conform to the teeth are moved, preferably until the positioning error of each tooth model relative to the deformed updated image is less than 1 mm, preferably less than 5 / 10 mm, preferably less than 3 / 10 mm, preferably less than 2 / 10 mm, and preferably less than 1 / 10 mm;

[0083] - In step 5), the deformation of the active intermediate model comprises a movement of one or more tooth models that do not conform to the teeth, the amplitude and / or direction of said movement being determined according to a measurement of the non-conformity of said one or more non-conforming teeth, said measurement being based on at least one updated image, in particular an image of the appliance, preferably by comparing said updated image with the active intermediate model.

[0084] The present invention also relates to a method for manufacturing an orthodontic appliance, the method comprising steps 1) to 5) and then comprising the following steps:

[0085] 6) designing, based on the updated model and the final model representing the dental arch in the theoretical final configuration, an “updated” appliance suitable for modifying the dental arch in the use position from the actual configuration at the time of the update towards said theoretical final configuration;

[0086] 7) Manufacturing the updated appliance and providing the updated appliance to the patient.

[0087] The theoretical final configuration planned for the dental arch at the final time after the last intermediate time is generally the configuration of the target dental arch at the end of treatment.

[0088] The method according to the invention may be partly computer-implemented, in particular with regard to the steps of modifying the model, and computing or exploring the model in particular to search for constrained virtual acquisition conditions, or analyzing images or maps (for example finding contours).

[0089] The present invention also relates to:

[0090] - a computer program comprising program code instructions for performing one or more, preferably all, of steps 2), 3), 4), 5) and even step 6) when said program is executed by a computer,

[0091] - a computer-readable storage medium, such as a memory or a CD-ROM, on which such a program is stored, and

[0092] - A computer loaded with such a program.

[0093] The present invention also relates to a system, comprising:

[0094] - a personal device, preferably a mobile phone, configured to acquire one or more update images of step 1),

[0095] - a computer loaded with a program comprising means for performing one or more, preferably all, of steps 2) to 5) and preferably step 6) when said program is executed by the computer, i.e. "configured

[0096] Program code instructions for "performing these steps";

[0097] - Optionally, a computer loaded with a program configured to manufacture the appliance in step 7).

[0098] definition

[0099] "Patient" or "user" refers to any person on whom the methods according to the invention are performed, whether or not that person is ill.

[0100] "Plurality of teeth" refers to a group of teeth in a dental arch.

[0101] A “dental professional” is anyone qualified to provide dental care, and this includes, in particular, orthodontists and dentists.

[0102] An appliance that a patient uses at some time during treatment is called an "active appliance." In treatment using multiple appliances, it is expected that each appliance will be active in succession.

[0103] A 3D scanner or "scanner" is a device that allows obtaining a model of the dental arch.

[0104] The "position of use" is the position of the appliance when it has been attached to the dental arch to treat the dental arch. Typically, the patient can detach the appliance by simple traction.

[0105] When the appliance is attached to the dental arch in the position of use, teeth that are not properly located in the appliance and teeth that are properly located in the appliance are referred to as "luxated teeth" and "non-luxated teeth", respectively. An orthodontist is well able to distinguish between luxated and non-luxated teeth. This distinction can also be made by computer, in particular by evaluating the distance between the tooth and the bottom of the groove of the appliance attached thereto.

[0106] More generally, a tooth is said to be "compliant" or "noncompliant" when, at the update time, it is or is not, respectively, in the position planned in the treatment plan. A luxated tooth is an example of a noncompliant tooth.

[0107] The "update time" is the time during which the updated image is acquired. The length of this time is short enough that the structure of the tooth does not change significantly during this period.

[0108] When the structure of the dental arch is the structure that the patient's dental arch actually has, the structure is called an "actual" structure. When the structure of the dental arch is a "simulated" structure of the patient's dental arch or a structure "planned" for a future time, the structure is called a "theoretical" structure.

[0109] A "model" refers to a three-dimensional digital model. A model consists of a set of voxels. A "dental arch model" is a model representing at least a portion of a dental arch, preferably at least 2 teeth, preferably at least 3 teeth, preferably at least 4 teeth. Figure 3 An example of a view showing a dental arch model.

[0110] A "tooth model" is a three-dimensional digital model of the teeth of a patient's dental arch. The dental arch model can be cut to define tooth models for at least some of the teeth, preferably all of the teeth, represented in the dental arch model. Thus, the tooth model is a model within the dental arch model. Figure 4 An example of a view showing a cut dental arch model. Computational tools exist for manipulating the tooth models of a dental arch model. These tools allow constraints to be set, in particular in order to limit the movement of the tooth models to actual movements, for example to prevent adjacent tooth models from penetrating each other.

[0111] A "regimen" is a series of dental arch models representing successive dental arch structures. In particular, a "treatment regimen" or "treatment plan" includes models representing the dental arch structure at different times during the treatment of the dental arch. These times are typically an initial time before the start of treatment, intermediate times during treatment, and a final time at the end of treatment. Each model of the regimen representing the intended structure of the dental arch at an intermediate time is called an "intermediate model". Figure 7 Examples of treatment regimens are shown.

[0112] The arch configurations at the intermediate and final times are theoretical, as they are derived from simulations of future times. They are therefore anticipated or "planned" and may therefore differ from reality at the intermediate times. A model showing the scheme in chronological order allows simulation of the effects of treating the arch.

[0113] An example of a software package for manipulating tooth models and creating treatment plans is the program Treat, which is described on the web page https: / / en.wikipedia.org / wiki / Clear_aligners#cite_note-invisalignsystem-10. US5975893A also describes the creation of a treatment plan.

[0114] An "image" is a two-dimensional picture, such as a photograph or an image extracted from a film. An image is made up of pixels.

[0115] The "acquisition conditions" specify the spatial position, spatial orientation and calibration (e.g., values ​​of aperture aperture and / or exposure time and / or focal length and / or sensitivity) of:

[0116] - the actual equipment used to acquire the images in relation to the patient's dental arch (actual acquisition conditions), or

[0117] - A virtual device for acquiring images related to the patient's dental arch model (virtual acquisition conditions).

[0118] The "calibration" of an acquisition device includes all values ​​of calibration parameters. Calibration parameters are parameters intrinsic to the acquisition device (as distinct from its position and its orientation) whose values ​​affect the acquired image. For example, the aperture opening is a calibration parameter that modifies the depth of field. The exposure time is a calibration parameter that modifies the brightness (or "exposure") of the image. The focal length is a calibration parameter that modifies the angle of view (i.e. the amount of "zoom"). "Sensitivity" is a calibration parameter that modifies the reaction of the sensor of a digital acquisition device to incident light.

[0119] Preferably, the calibration parameters are selected from the group formed by aperture opening, exposure time, focal length and sensitivity.

[0120] The model observed at a determined angle and a determined distance under determined virtual acquisition conditions (in particular through calibration of the virtual acquisition device) is referred to as a "view".

[0121] An "image of a dental arch", "view of a dental arch", "representation of a dental arch", "scan of a dental arch" or "model of a dental arch" refers to an image, view, representation, scan or model of all or part of the dental arch.

[0122] The model is "matched" to an image when there is a view of the model of the patient's dental arch that corresponds to the image (i.e., such that the representations of the teeth in the view are positioned relative to each other as the representations of the teeth in the image). Thus, the contours of the model of the teeth represented in the view are substantially in register with the contours of the representations of the teeth in the image.

[0123] This view of the model may also be defined as being "matchable" or "registerable" with the image.

[0124] Deep learning algorithms are deep learning devices well known to those skilled in the art. They include "neural networks" or "artificial neural networks".

[0125] A person skilled in the art will be able to select a neural network according to the task to be performed. In particular, the neural network may be selected in particular from:

[0126] -Networks called Convolutional Neural Networks (CNNs) that are specialized for image classification, such as:

[0127] -AlexNet (2012)

[0128] -ZF Net (2013)

[0129] -VGG Net (2014)

[0130] -GoogleNet (2015)

[0131] - Microsoft ResNet (2015)

[0132] -Caffe: BAIR Reference (reference) CaffeNet, BAIR AlexNet

[0133] -Torch: VGG_CNN_S, VGG_CNN_M, VGG_CNN_M_2048, VGG_CNN_M_1024, VGG_CNN_M_128, VGG_CNN_F, 16-layer (layer)

[0134] VGG ILSVRC-2014, 19-layer (layer) CGG ILSVRC-2014, Network-in-Network (Network in Network) (Imagenet & CIFAR-10)

[0135] -Google: Inception (V3, V4)

[0136] -Networks specialized for localizing and detecting objects in images (object detection networks), such as:

[0137] - R-CNN (2013)

[0138] -SSD(Single Shot MultiBox Detector: Object Detection Network), Faster(Faster)

[0139] R-CNN (Faster Region-based Convolutional Networks: Object Detection Networks)

[0140] - Faster R-CNN (2015)

[0141] -SSD (2015)

[0142] -RCF (Richer Convolutional Features for Edge Detection) (2017)

[0143] -Networks specialized for image generation, such as:

[0144] - Cycle-Consistent Adversarial Networks (2017)

[0145] -Enhanced CycleGAN (2018)

[0146] - Deep Photo Style Transfer (2017)

[0147] -FastPhotoStyle(2018)

[0148] -pix2pix (2017)

[0149] - Style-based Generator Architectures for GANs (2018)

[0150] -SRGAN (2018).

[0151] The above list is non-limiting.

[0152] Training of the neural network involves comparing it to a training database containing information about the two types of objects that the neural network must learn to "correspond," ie, to be associated with each other.

[0153] The training may be performed based on a training database consisting of the following records: each record includes a first object of a first type and a corresponding second object of a second type.

[0154] Alternatively, training can be performed based on a training database consisting of the following records: each record includes a first object of a first type or a second object of a second type, however, each record includes information related to the type of object it contains. For example, Zhu, Jun-Yan et al. describe such a training technique in the article "Unpaired image-to-image translation using cycle-consistent adversarial networks"

[0155] Training a neural network with these records teaches it to pass from any object of the first type to the corresponding object of the second type.

[0156] The quality of the analysis performed by the neural network directly depends on the number of records in the training database. Preferably, the training database contains more than 10,000 records.

[0157] In order to evaluate the "positioning error" of the tooth model, the distance between each representation point of the tooth modeled by the tooth model in the deformed updated image and the corresponding point in the view of the active intermediate model (when the deformed updated image in 1:1 scale is superimposed on the view of the active intermediate model matched with the deformed updated image in 1:1 scale) is measured. The positioning error is the maximum of these distances, taking into account all points of the representation that have a corresponding point in the view. In 1:1 scale it is meant that the representation of the tooth is in true scale, the deformed updated image and the view of the active intermediate model represent the tooth with actual size.

[0158] Unless otherwise indicated, "comprising," "including," or "having" must be interpreted in a broad, non-limiting manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0159] Other features and advantages of the present invention will become more apparent upon reading the following detailed description and examining the accompanying drawings, in which:

[0160] -[ Figure 1 ] shows a perspective view of an orthodontic appliance;

[0161] -[ Figure 2 ] shows Figure 1 A top view of an orthodontic appliance in;

[0162] -[ Figure 3 ] shows an example of an initial model (intermediate models, final models and updated models may take similar forms);

[0163] -[ Figure 4 ] shows an example of a model from which a tooth model has been cut (only the tooth model is shown);

[0164] -[ Figure 5 ] schematically illustrates a supporting dental arch carrying an orthodontic appliance;

[0165] -[ Figure 6 ] shows a system suitable for implementing the method according to the invention;

[0166] -[ Figure 7 ] shows the treatment regimen;

[0167] -[ Figure 8 ] schematically illustrates a method according to the present invention;

[0168] -[ Fig. 9 ] shows an example of an appliance image;

[0169] -[ Fig.10 ] schematically illustrates the acquisition of an appliance image and / or a bare tooth image;

[0170] -[ Fig.11 ] schematically shows the Fig.10 A retractor for use with the acquisition kit shown in ;

[0171] -[ Fig.12 ] schematically illustrates a first method for detecting dislocation in an image;

[0172] -[ Fig.13 ] schematically illustrates a second method for detecting dislocation in an image. DETAILED DESCRIPTION

[0173] In one embodiment, the method according to the present invention comprises the following steps before carrying out steps 1) to 5) Figure 8 ):

[0174] At the initial time t1, usually at the start of treatment,

[0175] a) generating a three-dimensional digital model of the patient's dental arch, referred to as an "initial model", in its actual initial configuration; and cutting the initial model into tooth models;

[0176] b) determining the time required for moving the dental arch from the initial configuration through the intermediate time t n A dental arch treatment in which a theoretical intermediate configuration (n is 2 to N) is modified to a theoretical final configuration at a final time (usually at the end of treatment);

[0177] c) deforming the initial model to generate a treatment plan including a final model and an intermediate model representing the dental arch in a final structure and an intermediate structure, respectively;

[0178] d) Design a series of orthodontic appliances based on the initial model, intermediate model and final model;

[0179] e) Manufacturing one or more appliances and providing these appliances to the patient.

[0180] In one embodiment of the invention, the method further comprises steps 1) to 5) of the method for generating an updated model according to the invention during treatment, and then preferably comprises the following steps:

[0181] 6) designing an update appliance based on the update model and the final model, the update appliance being shaped to modify the dental arch from the actual structure at the time of the update toward the final structure;

[0182] 7) Manufacture an updated appliance and provide the updated appliance to the patient.

[0183] In step a), an initial model is generated at an initial time t1 before the start of orthodontic treatment with the orthodontic appliance, and this initial time t1 is preferably less than 6 months, preferably less than 3 months or less than 1 month or less than 2 weeks before the start of treatment.

[0184] The initial model may be prepared based on measurements of the patient's teeth or a physical model of their teeth (eg, a plaster model).

[0185] The initial model is preferably created by professional means, such as by a 3D scanner, which is preferably performed by a dental professional, such as an orthodontist or an orthodontic laboratory. In an orthodontic clinic, a physical model of the patient or his teeth can advantageously be placed in precise positions and the professional means can be state-of-the-art. Thus, a very accurate initial model can be obtained. The initial model preferably provides information about the tooth positions with an error of less than 5 / 10 mm, preferably less than 3 / 10 mm, and preferably less than 1 / 10 mm.

[0186] The initial model is for example of point cloud type or one of the following types: STL, OBJ, 3D DXF, IGES, STEP, VDA. Advantageously, such a so-called "3D" model can be viewed from any angle.

[0187] The initial model is typically viewable and manipulable via a computer. The initial model is then cut to define the tooth model.

[0188] Cutting a three-dimensional model into a dental model is a conventional operation, wherein the model is cut to define a representation of one or more teeth in the initial model. Other elements of the dental arch, such as the gums, may also be modeled.

[0189] The initial model may be cut manually by an operator using a computer, or automatically by a computer that preferably implements a deep learning device, preferably a neural network.

[0190] In particular, the tooth model may be defined as described, for example, in international patent application PCT / EP2015 / 074896.

[0191] Figure 4 An example of an initial model is shown from which a tooth model 32 has been cut (only the tooth model is shown; they have been shaded differently for easier identification).

[0192] After cutting, the tooth model can be moved. Thus, using a computer, the initial cutting model can be deformed by movement of the tooth model without modifying the tooth model to simulate the movement of the tooth from an initial time to a final time (at which the tooth is in its final configuration), which final time can specifically mark the end of treatment.

[0193] In step b), the treatment through which one or more teeth are moved from an initial structure through an intermediate structure to a final structure is determined.

[0194] This group allows visualization of the dental arch model at each stage of treatment to form a treatment plan. The computer is used to display the treatment plan and store the deformed initial model to simulate the dental arch structure at various intermediate times.

[0195] Typically, for a given treatment, there are multiple possible potential solutions. In one embodiment, the computer determines the potential solutions and selects a treatment solution from them. In another embodiment, the computer determines the potential solutions, presents them to the dental professional, and the dental professional selects a treatment solution from them. In another preferred embodiment, the dental professional, preferably an orthodontist, determines the potential solutions and selects a treatment solution from them. The computer advantageously allows him to view a simulation of the effects of the potential solutions on the dental arch.

[0196] In step c), the initial model is deformed to generate a final model representing the dental arch in a theoretical final configuration and an intermediate model marking a stage between the initial model and the final model.

[0197] Depending on the orthodontic treatment, the deformation may be determined by a dental professional, preferably an orthodontist, preferably using a computer allowing him to see the effect of the envisaged treatment on the dental arch, or automatically by a computer, preferably implementing a deep learning device, preferably a neural network.

[0198] Preferably, steps b) and c) are simultaneous. In particular, the treatment plan is determined as a result of one or more simulations of deforming the initial model into the structure of the final model by moving the tooth model. After the treatment has been selected, in order to generate an intermediate model, it is sufficient to simulate the treatment and save the deformed initial model at an intermediate time.

[0199] Fig. 9 An example of a treatment plan including an initial model, two intermediate models, and a final model is shown.

[0200] In step d), according to the treatment plan, a series of orthodontic appliances are designed to deform the dental arch by moving the teeth.

[0201] In step e), one or more first appliances of the series are manufactured. Typically, all appliances of the series are manufactured. These appliances are provided to the patient so that he can start treatment.

[0202] The method comprising steps a) to e) is well known and is commonly used for designing and manufacturing a range of orthodontic appliances.

[0203] Typically, the patient is monitored by his orthodontist. As explained in the introduction, the orthodontist regularly checks the fit of the appliances and, in case of non-compliance with the treatment, in particular in case of dislocation, the orthodontist generates a new model of the dental arch using a scanner and then repeats steps b) to e), replacing the initial model with this new model, in order to produce a new series of appliances for the remainder of the treatment.

[0204] According to the invention, the method comprises steps 1) to 5) and preferably steps 1) to 7) during treatment.

[0205] In step 1), at an update time, an update image, called an “analysis” update image, is acquired which allows detecting non-compliance, in particular dislocation, of the teeth relative to the appliance worn at the update time.

[0206] The update time may be, for example, greater than 2 weeks, greater than 4 weeks, greater than 8 weeks, or greater than 12 weeks after the initial time.

[0207] Preferably, at least one reminder is sent to the patient before the update time, for example less than 2 weeks before the update time, informing the patient that an analysis update image needs to be taken. The reminder may be in paper form, or preferably in electronic form, such as in the form of an email, an automatic reminder from a dedicated mobile application, or SMS. For example, such a reminder may be sent by a clinic or orthodontic laboratory or dentist or through a dedicated application on the patient's mobile phone.

[0208] Analysis Update The images are taken with an image acquisition device, preferably a personal acquisition device, preferably a mobile phone, a so-called "connected" camera, a so-called "smart" watch, a tablet or a (desktop or portable) personal computer comprising an image acquisition system (e.g. a webcam or a camera).

[0209] The analysis update image is preferably extraoral.

[0210] In one embodiment, if Fig.10 and Fig.11 As shown, a photographing kit 15 is used. Preferably, such a kit comprises a stand 17, a dental retractor 19 and an image acquisition device, preferably a mobile phone 21. The dental retractor 19 and the acquisition device, preferably a mobile phone, are preferably attached to the stand 17 in a removable manner.

[0211] The retractor 19 may have the features of a conventional retractor.

[0212] like Fig.11 As shown (where the retractor has been separated from the bracket), the retractor preferably comprises a channel 23 extending around a retractor opening having an axis X and arranged so that the patient's lips can be placed therein while allowing the patient's teeth to be seen through the retractor opening.

[0213] The retractor 14 may be attached to the bracket by one or more clips 27a and 27b, such as magnetic clips.

[0214] Preferably, the retractor includes lugs 26a and 26b for moving the cheeks apart so that an acquisition device attached to the bracket can acquire a photograph of the vestibular surface of teeth located at the back of the mouth, such as molars, through the retractor opening.

[0215] Preferably, the support is in the form of a housing that opens specifically to the retractor opening and the mobile phone. The mobile phone thus observes the patient's dental arch through the housing. Advantageously, the support allows presetting the position of the mobile phone relative to the dental arch.

[0216] The acquisition is preferably performed by the patient or a friend or relative of the patient, but can also be performed by any other person, in particular a dentist or orthodontist, preferably without requiring a precise positioning of the image acquisition device relative to the teeth.

[0217] Preferably, the analysis update image is a photograph or an image extracted from a film. It is preferably a color image, and preferably a true color image. More preferably, the analysis update image is a photograph (unlike a tomographic image or a panoramic image obtained by X-ray) showing the actual dental arch (as perceived by the human eye).

[0218] Preferably, the analysis update image is then sent to a central computer, preferably via a mobile phone which acquires the analysis update image.

[0219] Preferably, a dedicated application is loaded into the mobile phone to preferably verbally and / or visually guide the patient through the various procedures to be performed and transmit analysis update images.

[0220] In one embodiment, the analysis update image is an appliance image, which advantageously allows the detection of non-compliance, in particular dislocation, by analyzing only this image. This detection may also result from a comparison of this image with an active intermediate model.

[0221] In one embodiment, the analysis update image is an image of the bare tooth, which advantageously allows detecting non-compliance with good accuracy and in particular without being hindered by the representation of the appliance, although this image then needs to be compared with the active intermediate model.

[0222] Preferably, in step 1), an image of the appliance and an image of the bare teeth are acquired in order to benefit from two complementary analyses.

[0223] Preferably, in step 1), at least one naked tooth image is acquired to be used as the deformation update image in step 5).

[0224] In step 2), the intermediate model or "active intermediate model" is preferably determined by a computer according to the update time. If in step 3), the updated images are analyzed independently of the active intermediate model for detecting non-compliant teeth, step 2) can follow step 3).

[0225] From the treatment plan, preferably by computer, an intermediate model is selected that should best represent the dental arch at the update time according to the plan. Under the assumption that treatment is ongoing according to the treatment plan, the intermediate model whose intermediate time is closest to the update time can be selected. This intermediate model is called "active".

[0226] Preferably, the active intermediate model is an intermediate model whose intermediate time is close to the update time and preferably differs from the update time by less than 4 weeks, less than 2 weeks, preferably less than 1 week. Preferably, the intermediate time of the active intermediate model is before the update time.

[0227] In a preferred embodiment, the active intermediate model is not modified before step 5). Therefore, the movement of the teeth is evaluated by comparative analysis of the updated images and the intermediate model designed at the initial time.

[0228] In a preferred embodiment, the active intermediate model can be corrected in a rough manner, for example manually, in order to take into account modifications of the patient's dental arch between the initial time and the update time that are not caused by deviations during the treatment process, for example to remove the dental model of teeth that have fallen out or been extracted.

[0229] In step 3), representations of teeth that do not conform to the treatment plan, in particular representations of teeth that have been dislocated from the orthodontic appliance, are sought in the analysis update image.

[0230] Preferably, the central computer is programmed to automatically detect non-conforming teeth and identify those teeth.

[0231] Detecting non-compliance with active intermediate models

[0232] The computer can specifically use an analysis update image in the form of a bare tooth image and compare it to the active intermediate model.

[0233] Preferably, a position, orientation and calibration of the virtual acquisition device (these position, orientation and calibration are collectively referred to as "virtual acquisition conditions") are sought that best corresponds ("best fit") to the actual acquisition conditions of the analyzed update images.

[0234] The view under the virtual acquisition condition is then compared with the analysis updated image.

[0235] Comparison of said views with the analysis updated image allows detection of non-compliant teeth, ie not only luxated teeth, but also non-luxated teeth whose position does not correspond to the treatment plan.

[0236] The comparison of the view and analysis update images may result from a comparison of corresponding maps associated with the discriminant information (eg, contours representing teeth). The comparison process described below for comparing the test map and the update map may be used.

[0237] The search for the virtual acquisition conditions and the comparison can be performed in particular according to the teaching of PCT / EP2015 / 074896.

[0238] In a preferred embodiment, if at least one point of the representation of a tooth in the analysis updated image at a 1:1 scale is separated from the corresponding point in said view at a 1:1 scale (i.e., the point representing the same point of the tooth) by a distance greater than 1 / 10 mm, 3 / 10 mm or 5 / 10 mm and preferably less than 7 mm or 5 mm, the representation is considered not to comply with the treatment plan.

[0239] Distances can also be measured in pixels, which advantageously avoids the need to establish a scale bar.

[0240] The non-conformity is therefore advantageously one that allows detection of a deviation in the treatment execution. In particular, an excessively large distance is considered not to be related to a deviation in the treatment, but to an anomaly, for example because a tooth was incorrectly determined, for example because it was masked in the analysis update image.

[0241] Detect dislocation in appliance images without relying on active intermediate models

[0242] For the detection of dislocations by computer, the updated image in the form of an image of the appliance can be analyzed, in particular to determine the contours of the groove bases of the appliance and of the free ends of the teeth.

[0243] Those skilled in the art know how to process an image or view to separate contours. Such processing includes, for example, applying well-known masks or filters provided with image processing software. For example, such processing operations allow detection of high contrast areas.

[0244] These treatment operations include in particular one or more of the following known and preferred methods:

[0245] - Apply the Canny filter, in particular search for edges using the Canny algorithm;

[0246] - Applying a Sobel filter, in particular computing derivatives via the extended Sobel operator;

[0247] - Apply the Laplace filter to calculate the Laplacian of the image;

[0248] - Detect blobs in an image (“Blob detector”);

[0249] - Apply Threshold to apply a set threshold to each element of a vector;

[0250] - Resize using relations between pixel areas ("Resize(Area)"), or bicubic interpolation in pixel context;

[0251] - Erode the image by a specific structure element;

[0252] - Extend the image by specific structural elements;

[0253] - Especially use the area near the restoration area for trimming;

[0254] - Apply a bilateral filter;

[0255] -Apply Gaussian blur;

[0256] - Apply Otsu's method to find the threshold that minimizes the within-class variance;

[0257] - Apply A* filter to find paths between points;

[0258] - Apply Adaptive Threshold to apply an adaptive threshold to the vector;

[0259] - in particular, applying an equalization filter to the histogram of a grayscale image;

[0260] - Detect blur (“BlurDetection”) to calculate the entropy of the image using its Laplacian operator;

[0261] - Detect contours in binary images (“FindContour”);

[0262] - Color filling ("FloodFill"), in particular to fill the associated elements with a certain color. The following non-limiting method, although not preferred, can also be implemented:

[0263] - Apply the "MeanShift" filter in order to find objects in the projection of the image;

[0264] -Apply the "CLAHE" filter, CLAHE stands for "Contrast Limited Adaptive Histogram Equalization";

[0265] - Apply a "Kmeans" filter to determine the centers of clusters and the centers of groups of samples around the clusters;

[0266] - Apply DFT filters to perform discrete, forward or inverse Fourier transforms on vectors;

[0267] - Calculate torque;

[0268] - Apply the "HuMoments" filter to compute Hu moment invariants;

[0269] - Calculate the integral of the image;

[0270] - Applying the Scharr filter, allowing the calculation of the derivative of the image by implementing the Scharr operator;

[0271] - Convex hull of the search points ("ConvexHull");

[0272] - Search for convex points of the contour ("ConvexityDefects");

[0273] - Compare shapes ("MatchShapes");

[0274] - Check if the point is on the contour ("PointPolygonTest");

[0275] - Apply Harris corner detection (“CornerHarris”);

[0276] - Search for the minimum eigenvalue of the gradient matrix to detect corners (“CornerMinEigenVal”);

[0277] - Apply the Hough transform to find circles in a grayscale image (“HoughCircles”);

[0278] - “Active contour modeling” (tracking the contours of an object based on a potentially “noisy” 2D image);

[0279] - Compute a force field ("Gradient Vector Flow" or GVF) in a portion of the image;

[0280] - Apply a cascade classifier (“CascadeClassification”).

[0281] The determination of the tooth contour may be optimized according to the teaching of PCT / EP2015 / 074900.

[0282] In one embodiment, the contour of the groove bottom and the contour of the free end of the tooth are divided so as to define segments of these contours for each tooth. The segments of the contour of the groove bottom and the segments of the contour of the free end of the tooth are respectively referred to as "external tooth contour" 24 i and "Internal tooth profile" 30 i ( Fig.12 and Fig.13 ). The adjectives "inside" and "outside" are used here only for clarity. Fig.12 In , the dashed lines show the boundaries of consecutive segments.

[0283] This comparison can then be performed in any manner, in particular using the technique described in EP 3 412 245 for comparing internal and external tooth profiles.

[0284] In particular, for each of the plurality of teeth for which the inner tooth profile and the outer tooth profile have been determined, the following steps may be performed:

[0285] i) determining the distance between the inner tooth profile and the outer tooth profile;

[0286] ii) determining a distance threshold, preferably based on the distance determined in step i);

[0287] iii) for each of the teeth, determining a distance score according to:

[0288] - the distance between the inner tooth profile and the outer tooth profile, and

[0289] -Distance threshold.

[0290] In step i), the distance d between the inner tooth profile and the outer tooth profile is determined for each of the teeth ( Fig.13 ).

[0291] The distance between the inner tooth contour and the outer tooth contour of a tooth may be, for example, an average distance or a maximum distance between pixels of the contours corresponding to the same point of the tooth.

[0292] The distance is preferably measured in pixels, which advantageously avoids the need to establish a scale.

[0293] In step ii), a distance threshold Sd is determined, preferably based on the distance determined in step i).

[0294] Preferably, in step ii), the distance threshold Sd is substantially equal to the minimum distance (d min ). Usually, at least one of the treated teeth is in contact with the bottom of the groove into which it is inserted. The distance between the inner tooth profile and the outer tooth profile of this tooth is then equal to the minimum distance d corresponding to the normal situation. min Therefore, it can be used as a criterion for evaluating the distance between the inner tooth profile and the outer tooth profile of other teeth in step iii).

[0295] In step iii), for each tooth, a score called the "distance score" S(d, Sd) is determined according to:

[0296] - the distance d between the inner tooth profile and the outer tooth profile, and

[0297] -Distance threshold Sd.

[0298] Preferably, the distance score of a tooth is equal to (d-Sd), that is, equal to the difference between the distance between the inner tooth profile and the outer tooth profile of the tooth and the distance threshold. The higher the distance score, the more the affected tooth is dislocated from the groove.

[0299] Fig.12 An example of implementing steps i) to iii) is shown, in which tooth D1 has been dislocated from the bottom of the appliance to dd min >Sd point.

[0300] For each of the plurality of teeth whose inner tooth profile and outer tooth profile have been determined, the following steps may also be performed:

[0301] i') for each pair consisting of a left tooth adjacent to a right tooth in a triplet having at least a first tooth, a second tooth and a third tooth adjacent to each other (the inner tooth profile and the outer tooth profile of each tooth being determined, the first tooth and the third tooth being adjacent to the second tooth),

[0302] Determine the offset between the inner tooth profile of the left tooth and the inner tooth profile of the right tooth (referred to as

[0303] and determining the offset between the outer tooth profile of the left tooth and the outer tooth profile of the right tooth (called the “outer offset”),

[0304] Then

[0305] Determine the difference between the inner offset and the outer offset, which is called the "offset difference";

[0306] ii') determining an offset difference threshold, preferably based on the offset difference determined in step i');

[0307] iii') determining at least one offset score for at least one tooth of the triad, preferably for each tooth, based on the offset difference to the adjacent tooth and an offset difference threshold.

[0308] In step i'), at least one triplet consisting of a first tooth D1, a second tooth D2 and a third tooth D3 is considered, the first tooth and the third tooth being adjacent to the second tooth, ie the first tooth, the second tooth and the third tooth being arranged in sequence along the dental arch.

[0309] The inner tooth profiles 301 , 302 and 303 and the outer tooth profiles 241 , 242 and 243 of the teeth D1 , D2 and D3 are determined respectively.

[0310] The inner or outer "offset" refers to the distance between the inner tooth profile or the outer tooth profile, respectively, of two adjacent teeth.

[0311] Determine the following:

[0312] - An offset Δ between the inner tooth profile 301 of the first tooth and the inner tooth profile 302 of the second tooth 1-2 i, this offset is called the “first inner offset”;

[0313] - An offset Δ between the inner tooth profile 302 of the second tooth and the inner tooth profile 303 of the third tooth 2-3 i, this offset is called the “second inner offset”;

[0314] - an offset Δ between the outer tooth profile 241 of the first tooth and the outer tooth profile 242 of the second tooth 1-2 e, this offset is called the “first outer offset”;

[0315] - an offset Δ between the outer tooth profile 242 of the second tooth and the outer tooth profile 243 of the third tooth 2-3 e, this offset is called the “second outer offset”.

[0316] The inner offset between the inner tooth profiles of two adjacent teeth is preferably equal to the maximum distance between the inner tooth profiles of the two teeth.

[0317] The outer offset between the outer tooth profiles of two adjacent teeth is preferably equal to the maximum distance between the outer tooth profiles of these two teeth.

[0318] The insets and outsets are preferably measured in pixels, which advantageously avoids the need to establish a scale.

[0319] Then determine the following:

[0320] -First inner offset Δ 1-2 i and the first external offset Δ 1-2 The difference between the two is called the "first offset difference" Δ 1-2 (=Δ 1-2 i-Δ 1-2 e);

[0321] - Second inner offset Δ 2-3 i and the second external offset Δ 2-3 The difference between the two values ​​is called the “second offset difference” Δ 2-3 (=Δ 2-3 i-Δ 2-3 e).

[0322] exist Fig.12 In the example, Δ 1-2 Much smaller than Δ 2-3 .

[0323] In step ii'), a first offset difference Δ is determined, preferably based on the first offset difference Δ determined in step i'). 1-2 and the second offset difference Δ 2-3 To determine the offset difference threshold SΔ.

[0324] Preferably, in step ii'), the offset threshold is substantially equal to the smallest of the offset differences determined in step i').

[0325] Typically, at least two adjacent treated teeth are in contact with the bottom of the grooves into which they are inserted. The offset difference between the two treated teeth is then substantially zero. Since this zero offset difference corresponds to a normal situation, it can be used as a criterion for evaluating the offset difference between adjacent treated teeth.

[0326] exist Fig.13 In FIG. 5 , the offset difference between the two teeth D1 and D2 is substantially zero.

[0327] In step iii'), for each pair of teeth in the triad, at least one score (called "offset score") is determined according to the offset difference between a tooth and its adjacent tooth and an offset difference threshold.

[0328] In particular, the offset difference between the first tooth and the second tooth may be compared with an offset difference threshold SΔ, which may be, for example, zero. In particular, the offset difference threshold may be subtracted from the offset difference between the first tooth and the second tooth to determine the offset scores of the first tooth and the second tooth.

[0329] The offset score, if it is positive for example, indicates that one or each of the first tooth and the second tooth tends to have dislocated from the bottom of the groove.

[0330] exist Fig.13 In the figure, the offset difference between the two teeth D2 and D3 is positive, which indicates that the second tooth or the third tooth has been dislocated.

[0331] exist Fig.13 In FIG. 1 , the offset difference between the two teeth D1 and D2 is substantially zero, so a positive offset difference between the two teeth D2 and D3 indicates that the third tooth has been dislocated.

[0332] Typically, when the first offset scores of the first tooth and the second tooth indicate that one of the two teeth is dislocated, a second offset score is determined for the second tooth and a third tooth adjacent to the second tooth. If the second offset score is less than the first offset score, it is likely that the first tooth has been dislocated from the bottom of the groove. Otherwise, it is likely that the second tooth has been dislocated.

[0333] Alternatively, the identification of non-conforming teeth, in particular dislocated teeth, can be performed by an operator, preferably a dental professional, more preferably an orthodontist, by simply observing the image of the appliance displayed on a computer screen.

[0334] Analysis through deep learning devices

[0335] A deep learning device, preferably a neural network, may be used to identify non-compliant teeth in the analysis update image.

[0336] In particular, the analysis may be performed as described in EP 3 432 218 .

[0337] Preferably, the following steps are performed:

[0338] I. creating a training database comprising more than 1000, preferably more than 5000, preferably more than 10000, preferably more than 30000, preferably more than 50000, preferably more than 100000 historical records, each historical record comprising a historical image and a historical description, each historical image comprising one or more regions (each region representing a tooth) or "historical tooth regions", the associated historical description specifying a tooth attribute value of at least one tooth attribute for each historical tooth region;

[0339] II. Training at least one deep learning device, preferably a neural network, through a training database;

[0340] III. Submitting the analysis update image, preferably the appliance image, to the at least one deep learning device so that it determines at least one probability associated with the attribute value of at least one tooth in a region (or

[0341] "Tooth Analysis Area"), the attribute value is related to the conformity of the tooth represented;

[0342] IV. Based on the probability, determine whether the tooth of the dental arch exists at the position represented by the tooth analysis area, and the attribute value of the tooth.

[0343] The deep learning device may in particular be a neural network specialized for localizing and detecting objects in images (i.e. an object detection network), and in particular selected from the above-mentioned examples of such networks.

[0344] It allows recognition analysis not only to update the representation of the teeth in the image, but also to determine whether they are in conformity.

[0345] In step I, tooth attributes are attributes whose values ​​are specific to each tooth.

[0346] For example, the tooth attribute "dislocation" will have a value of "conform" or "conform" depending on whether the tooth in question appears to be normally or abnormally positioned relative to the appliance, respectively.

[0347] In step II, a history record may be input to the deep learning device, each history record comprising a history image and a description, preferably describing, for each tooth represented in the history image, the contour of the tooth and any non-conformities. Thus, the deep learning device gradually learns to recognize patterns in the image and associates them with tooth regions and tooth attribute values ​​related to non-conformities, in particular tooth attribute values ​​related to tooth dislocation.

[0348] In step III, the deep learning device identifies the pattern in the analysis update image. In particular, it can determine the probabilities associated with:

[0349] - at a position in said analysis image there is a region at least partially representing a tooth ("tooth analysis region"),

[0350] - attribute values ​​of the teeth represented in the tooth analysis area.

[0351] For example, in an appliance image, it can be determined that there is a 99.5% probability that the shape in the analysis image represents a tooth, and a 99% probability that the tooth has been dislocated from the appliance.

[0352] Preferably, the deep learning device analyses the entire analysis update image and determines probabilities for all tooth analysis regions it has identified.

[0353] In step IV, based on the probabilities determined in step III, it is determined, preferably by computer, for each tooth represented in the analysis update image, whether the represented tooth must be considered as being in compliance or not in compliance. For example, if the probability that a tooth is not in compliance is greater than a threshold value, such as 98%, then the tooth can be considered to have been dislocated from the appliance.

[0354] Analysis using combined images of appliances and bare teeth

[0355] Preferably, non-conforming teeth are detected via a comparison of the bare tooth image with the active intermediate model and via a detection performed solely by analyzing the appliance image. The end result is thereby improved.

[0356] Specifically, if the analysis update image is an appliance image, some dislocations may be difficult to detect by analyzing only that image. Therefore, non-conforming teeth may be considered conforming based on the analysis of only the appliance image. Using the bare tooth image and the active intermediate model, non-conformities can be detected that cannot be detected in the appliance image.

[0357] If no non-conforming teeth are detected, the treatment can be continued without modifying the treatment plan. Thus, the appliance designed in step d) is always suitable for the treatment. Preferably, an information message is sent to the patient to inform him that the treatment is proceeding correctly. Preferably, an information message is also sent to the orthodontist who is caring for the patient.

[0358] If one or more non-conforming teeth are detected, the process continues with step 4).

[0359] In step 4), a tooth model is identified which represents the tooth identified in step 3) as a movable intermediate model that does not conform to the tooth, in particular the luxated tooth.

[0360] This identification can be performed by any computer that has access to the active intermediate model and the identifiers of the non-conforming teeth, in particular the dislocated teeth determined in step 3). These identifiers can be transmitted to a computer that has access to the active intermediate model, and vice versa, or the active intermediate model can be transmitted to a computer that has access to these identifiers.

[0361] Preferably, in step 3), the active intermediate model is transmitted to a central computer, which analyses the analysis update image and detects non-conforming teeth, in particular dislocated teeth, and which performs step 4).

[0362] As a variant, step 4) can be performed by an operator having access to the active intermediate model.

[0363] In step 5), the active intermediate model is deformed until a structure matching the update image taken at the update time is found. Preferably, the update image (referred to as a "deformed update image") is an image of multiple bare teeth without orthodontic appliances, i.e., an image of a bare dental arch. This facilitates the deformation of the active intermediate model.

[0364] The deformation update image may be the same as or different from the analysis update image. Preferably, the deformation update image is acquired and transmitted as described above with respect to the analysis update image, preferably using a mobile phone, preferably using an acquisition kit comprising a mobile phone and a bracket to which the dental retractor is attached. Preferably, the deformation update image is extraoral.

[0365] The deformation achieved by acting on all the tooth models of the active intermediate model takes a long time to achieve. Above all, the inventors have observed that it does not allow obtaining a model that correctly represents the real structure of the dental arch at the time of updating. In particular, it is observed that if all the tooth models are movable during deformation, the extraoral updated images do not always allow determining the precise position of the teeth in the posterior part of the mouth, in particular the molars. Therefore, in areas where the active intermediate model correctly represents the real position of the teeth, this deformation may lead to significant deviations from this model. Therefore, the orthodontic appliances manufactured based on these deformed models may not be usable.

[0366] In one embodiment of the invention, the active intermediate model is not deformed except in areas that do not conform to the tooth models of the teeth, in particular the dislocated teeth. The general shape of the active intermediate model is thus preserved and only these tooth models are moved.

[0367] The deformation of the active intermediate model is therefore limited to the movement of the active intermediate model representing the tooth model which does not conform to the teeth.

[0368] In particular, deforming the intermediate model by moving only the tooth models of non-conforming teeth makes it possible to very accurately assess the relative deviation of said non-conforming teeth at the update time with respect to the active intermediate model before the deformation.

[0369] In a preferred embodiment, the movement of the tooth models continues until the positioning error of each tooth model with respect to the deformed updated image is less than 1 mm, preferably less than 5 / 10 mm, preferably less than 3 / 10 mm, preferably less than 2 / 10 mm, and preferably less than 1 / 10 mm.

[0370] constraint

[0371] Preferably, a virtual acquisition condition corresponding to the actual acquisition condition of the deformed update image and allowing observation of the active intermediate model such that the view of the model is as close as possible to the virtual acquisition condition of the deformed update image is sought. If the deformed update image is an analysis update image, this search may have been performed in step 3).

[0372] In one embodiment, these virtual acquisition conditions are determined taking into account all tooth models. Therefore, non-compliant teeth will have a negative impact on the accuracy of the virtual acquisition conditions. However, the number of these teeth is small and their impact is usually small.

[0373] Preferably, the virtual acquisition conditions are determined without taking into account the tooth model of non-compliant teeth, in particular luxated teeth. Thus, a "constrained" virtual acquisition condition can be determined that corresponds exactly to the actual acquisition conditions of the deformed updated image and allows observing the active intermediate model so that the view of the model (called "constrained view") is very close to the deformed updated image. The obtained constrained view is very accurate especially for compliant teeth, because it is not degraded by the influence of non-compliant teeth.

[0374] In particular in the latter embodiment, the representations of the teeth in the deformation updated image can advantageously be used as reference objects whose positions in the active intermediate model are known. In particular, in the view obtained under the virtual acquisition conditions, these reference objects have the same relative positions as they have in the deformation updated image.

[0375] Preferably, at least three misaligned points are used as reference objects - for example cusps of aligned teeth, in particular cusps of non-luxated teeth. Analyzing the distances between the representations of these reference objects in the deformed updated image then allows the virtual acquisition conditions to be evaluated by simple calculations.

[0376] Advantageously, the virtual acquisition conditions can then be used to test all positions of the tooth model that do not conform to teeth, in particular luxated teeth.

[0377] Move tooth models that do not conform to the teeth by optimization

[0378] The active intermediate model is preferably deformed by an optimization algorithm.

[0379] Preferably, an iterative process is implemented, whereby in each iteration one or more tooth models of non-conforming teeth, in particular mallocated teeth, are moved, and then the degree of match between the active intermediate model thus modified and the deformed updated image is evaluated, the iterations being continued until a match between the modified model and the deformed updated image is found.

[0380] The number of iterations may be, for example, greater than 10, greater than 100, greater than 1000, greater than 10,000, and / or less than 1,000,000.

[0381] Preferably the following steps are performed:

[0382] A) analyzing the deformation update image and generating an update map associated with the discriminative information;

[0383] B) searching, based on the update graph, for an updated model corresponding to the position that the teeth had when the deformed update image was acquired by moving tooth models that do not conform to the teeth, in particular the dislocated teeth, the search is preferably performed by a metaheuristic method, preferably an evolutionary metaheuristic method, and preferably by simulated annealing.

[0384] After step A), the deformed update image is analyzed to generate an update map associated with at least one form of discriminative information.

[0385] "Discriminative information" is characteristic information (ie, image features) that can be extracted from an image, typically by a computer processing the image.

[0386] The discriminant information can have a variable number of values. For example, contour information can be equal to 1 or 0, depending on whether the pixel belongs to a contour or not. Gloss information can take a large number of values. Image processing allows the discriminant information to be extracted and quantified.

[0387] The update map represents discriminant information in the reference frame of the deformed update image. The discriminant information is preferably selected from the group consisting of contour information, color information, density information, distance information, gloss information, saturation information, information about reflection and combinations of these forms of information. The discriminant information is preferably contour information.

[0388] The purpose of step B) is to modify the active intermediate model until an updated model corresponding to the deformed updated image is obtained. Ideally, the updated model is therefore a three-dimensional digital dental arch model, based on which the deformed updated image could have been taken if the model were real, ignoring the appliance if it was represented in the deformed updated image.

[0389] Therefore, various models are tested successively, the selection of the model to be tested preferably being based on the level of correspondence of the previously tested model with the deformed updated image. The selection is preferably performed using a known optimization method, in particular an optimization method selected from meta-heuristic optimization methods, preferably evolutionary meta-heuristic optimization methods, in particular an optimization method selected from simulated annealing methods.

[0390] Preferably, the metaheuristic optimization method is selected from the group formed by:

[0391] - an evolutionary algorithm, preferably selected from:

[0392] Evolutionary strategies, genetic algorithms, differential evolution algorithms, distribution estimation algorithms, artificial immune systems, path relinking, hybrid complex evolution, simulated annealing, ant colony optimization, particle swarm optimization, tabu search and GRASP methods;

[0393] - Kangaroo algorithm,

[0394] -Fletcher-Powell method,

[0395] - Noise perturbation method,

[0396] - Random tunnels,

[0397] - Randomly restart the hill climbing,

[0398] - the cross entropy method, and

[0399] A hybrid approach that combines the above metaheuristics.

[0400] Preferably, step B) comprises the following steps:

[0401] B1) In the active intermediate model, the model to be tested is defined by moving the tooth model that does not conform to the teeth, especially the dislocated teeth;

[0402] B3) generating a view of the model to be tested under the constrained virtual acquisition condition;

[0403] B4) processing the view to generate at least one test pattern, the test pattern at least partially representing the discrimination information;

[0404] B5) testing the model to be tested by comparing the update graph with the test graph, so as to measure the difference between the update graph and the test graph, which is also called "matching degree" or "consistency";

[0405] B6) Based on the difference, for example, if the difference is less than a threshold value,

[0406] - modifying the model to be tested by moving one or more tooth models that do not conform to the teeth, in particular the dislocated teeth, and then returning to step B3); or

[0407] -Define the updated model as the tested model whose test graph has the smallest difference with the updated graph.

[0408] The difference is measured according to the discrimination information used. The difference can be measured, for example, by the ratio of the number of points belonging to both the contour of the test graph and the contour of the update graph to the total number of points of the contour of the update graph, or by the product of the inverse of the average distance between the contours represented in the update graph and the test graph and the length of the contour represented in the update graph.

[0409] The updated model obtained at the end of step B) is therefore a model resulting from the successive modification of the active intermediate model and is very accurate, since it results from the deformation of the initial model. Advantageously, the updated model is therefore very accurate even if it is obtained on the basis of a simple photograph or film image taken without special measures.

[0410] The tooth models cannot penetrate each other. Therefore, the movement of the tooth model of the dislocated tooth (especially in step 5)) is limited by the tooth model of the non-dislocated tooth that remains stationary. This further speeds up the search for an updated model.

[0411] Moving the tooth model of the active intermediate model allows obtaining an updated model that can be observed under the condition that the view of the updated model matches the deformed updated image. In other words, the view can be overlapped and registered with the deformed updated image so that the teeth represented in the view and the deformed updated image are substantially completely overlapped.

[0412] The update of the active intermediate model can be refined by repeating the previous operation using a plurality of update images as analysis update images and / or deformation update images. This results in an updated model that represents the tooth substantially in its actual structure at the update time. Alternatively, or in addition to using an optimization method, the search for the update model relies on a deep learning device, preferably a neural network.

[0413] Move the non-conforming teeth into the tooth model based on the non-conformity assessment

[0414] In one embodiment, analysis of the analysis update image allows for quantification of non-compliance via a "non-compliance level."

[0415] In particular, in one embodiment, analysis of the analysis update image allows for measuring dislocation, and preferably measuring changes in dislocation along the edges of the dislocated teeth.

[0416] Preferably, in step 5), this information about the non-conformity is used to move the tooth model of the non-conforming teeth, in particular the dislocated teeth. Fig.12 The measurement of d can be used to move tooth D1 downward, for example by translating dd min .

[0417] This is particularly useful for the movement of the tooth model as an initial operation of step 5) according to the nature and magnitude of the non-conformity assessed based on the analysis of the updated image. Preferably, this coarse movement operation is followed by a fine movement operation, preferably performed by optimization or using a deep learning device, preferably a neural network.

[0418] In step 6), at least one update appliance is designed that is adapted to modify the dental arch from its actual configuration at the time of the update toward the final configuration. By "towards said final configuration" is meant that the update appliance is shaped to modify the configuration of the dental arch to bring it closer to the final configuration. However, multiple new appliances may be required to reach the final configuration.

[0419] Preferably, a new series of appliances are designed taking into account the updated model instead of the initial model. Preferably, the above steps a) to d) are used to perform this operation, the initial time and the initial model being replaced by the updated time and the updated model respectively.

[0420] In step 7), at least one or more first appliances of the series of new appliances are manufactured, preferably as described in step e).

[0421] The appliance may be manufactured, for example, according to the teaching of EP1835864.

[0422] These new aligners are provided to the patient, such as by mail.

[0423] The patient then continues treatment with these new appliances.

[0424] system

[0425] The method according to the invention is at least partially, preferably completely, implemented by a computer. Any computer can be envisaged, in particular a PC, a server or a tablet.

[0426] Typically, a computer includes, among other things: a processor; a memory; a human-computer interface, usually including a keyboard, a screen, and a mouse; a computer for communicating with the Internet, via WiFi, or via a or a module for communicating via a telephone network; and a communication bus. The memory typically includes ROM and RAM memory. A software package configured to implement some of the methods of the invention discussed is loaded into the memory of the computer.

[0427] The computer can also be connected to a printer, scanner, CD-ROM drive, DVD drive, hard drive, disk burner, or speakers.

[0428] A communication bus is a device that ensures wired or remote communication between other components of a computer.

[0429] Computers can be used, automatically or with the assistance of an operator, to:

[0430] - in step a), displaying and manipulating the initial model, in particular modifying the viewpoint of the initial model;

[0431] - In step a), the initial model is cut;

[0432] -Determine the final model;

[0433] - Identify and / or display potential options for a given treatment;

[0434] - displaying and / or determining a treatment plan, thereby determining and storing an intermediate model;

[0435] - In step 2), the active intermediate model is determined;

[0436] - in step 3), analyzing the updated image to detect representations of non-conforming teeth, in particular representations of dislocated teeth;

[0437] - in step 4), identifying tooth models that do not conform to teeth, in particular tooth models of dislocated teeth, in the active intermediate model;

[0438] - In step 5), the tooth model of the active intermediate model is moved;

[0439] - In optional steps 6) and d), the appliance is designed.

[0440] The operator may be a dental professional, preferably an orthodontist. The computer may implement one or more deep learning devices, preferably neural networks.

[0441] Figure 6 A system in a preferred embodiment of the present invention is shown.

[0442] The system comprises a plurality of mobile phones, for example more than 1000, preferably more than 10,000 mobile phones 21 belonging to the patient P, each of which communicates with a central computer 50 .

[0443] Preferably, the central computer 50 is configured to receive and process updated images, in particular appliance images Ig and bare tooth images Id, of a plurality of patients, preferably more than 100, more than 1000 or more than 10000 patients. The central computer 50 may be configured to receive and process updated images of patients who are all under the care of the same orthodontist. Preferably, the central computer 50 is configured to receive and process updated images of patients who are under the care of a plurality of different orthodontists, for example more than 10, more than 100 or more than 1000 orthodontists.

[0444] The central computer 50 includes communication modules, which are used to communicate with each other via WiFi, Communication is carried out via optical fiber or via a telephone network with a plurality of local computers 52, for example more than 10 or more than 100 local computers, which are preferably located in the orthodontic clinic.

[0445] The system also includes a plurality of scanners 54 that communicate with one or more local computers, preferably each scanner communicates with a single local computer, such as via a wired link, via WiFi, via a wireless network, or via a wireless network. Communications are made via optical fiber or via a telephone network.Preferably, each scanner 54 is co-located with a respective local computer, and preferably in the same orthodontic clinic.

[0446] The system also includes a manufacturing unit equipped with a communication channel with a central computer and / or a local computer, for example, via a wired link, via WiFi, via Manufacturing computer 56 communicates via fiber optics or via the telephone network.

[0447] Example

[0448] In step a), an initial model M0 is generated at an initial time with a scanner 54. It is then transferred to a local computer 52. The software package loaded into the local computer preferably allows automatic cutting of the initial model to create a tooth model.

[0449] After examining the patient, the orthodontist defines the final model M by moving the tooth models using a local computer. f .

[0450] Preferably, the local computer 52 is programmed to determine one or more modifications of the dental arch in steps b) and c) to achieve a final structure corresponding to the final model. Preferably, the local computer also allows the orthodontist to view potential solutions and select a treatment plan. Preferably, the orthodontist is also able to create or modify the plan proposed by the local computer 52.

[0451] The intermediate times and the corresponding intermediate models can be defined by the local computer or presented by the local computer 52 to the orthodontist so that he can verify and / or modify them.

[0452] The local computer 52 sends the treatment plan, in particular at least the intermediate model M i As well as the initial model M0 and the final model M f Transmitted to the central computer 50.

[0453] In step d), the software package loaded into the central computer 50 determines the shape of the appliance to be manufactured based on these models and then transmits this information I0 to the manufacturing unit with the views to manufacture the appliance G0 in step e).

[0454] Alternatively, the initial model, the intermediate model and the final model can be transmitted from the central computer 50 to a production computer 56 which determines the shape of the N orthodontic appliances in step d) and controls the production.

[0455] Alternatively again, step d) may be performed by the local computer 52. The local computer then transmits the information required for manufacturing the appliance to the manufacturing unit.

[0456] Appliance G0 is sent to the patient to start treatment.

[0457] Preferably, the patient receives a reminder to take one or more update images, preferably at least one image of the appliance and preferably one or more images of the bare teeth.

[0458] In step 1), at the update time, the patient takes pictures with and without the active appliance that he needs to wear at that time. To do this, the patient uses his mobile phone 21 and transmits these pictures to the central computer 50.

[0459] In step 2), the central computer identifies the active intermediate model according to the update time of the acquired photo. Preferably, it identifies the intermediate model whose intermediate time is closest to the update time as the active intermediate model. More preferably, it identifies the intermediate model used to design the active appliance worn by the patient at the update time as the active intermediate model.

[0460] In step 3), the central computer 50 analyses the photographs, in particular the photographs showing the appliance in use position, to automatically detect non-conformities, in particular the dislocation of teeth, and identifies non-conforming teeth, in particular dislocated teeth. If no non-conforming teeth, in particular dislocated teeth, are detected, it transmits a message to the patient and / or orthodontist to inform him that the treatment is proceeding normally.

[0461] Preferably, the central computer 50 programs the teeth.

[0462] Otherwise, in step 4), the central computer 50 (possibly with the assistance of an operator) identifies tooth models that do not conform to teeth, in particular tooth models of luxated teeth.

[0463] In step 5), the central computer modifies the active intermediate model by moving the tooth models that do not match the teeth, especially the tooth models of the dislocated teeth, until an updated model M is found that matches the photo taken at the update time. a .

[0464] In step 6), the central computer designs one or more new appliances G1 taking into account the updated model and then transmits the information required to manufacture them to the manufacturing computer 56 of the manufacturing unit.

[0465] In step 7), the manufacturing unit manufactures new appliances G1. The new appliances are then delivered to the patient so that he can continue the treatment with these new appliances.

[0466] Variants

[0467] In one embodiment, the method according to the present invention uses the appliance image at the initial time to update the determined active intermediate model to represent the intermediate time t at which the appliance is replaced. i Therefore, the intermediate model of the treatment plan represents the corresponding intermediate time t at which the patient is required to change the orthodontic appliance. n The patient will therefore start wearing the first appliance in a series of appliances at the start of treatment (i.e., essentially at the initial time t1 when the initial model is generated), and will then change appliances at intermediate times t2, t3, etc. i is the time at which the patient is expected to replace the (i-1)th orthodontic appliance in the series with the i-th orthodontic appliance in the series, i being greater than or equal to 2. For example, if treatment is planned with 30 appliances (N=30), he will therefore replace the (i-1)th orthodontic appliance in the series with the i-th orthodontic appliance in the series at time t 30 Start wearing the 30th aligner and will wear it until the final time t 31 .

[0468] The number N of orthodontic appliances may be greater than 5, greater than 10, greater than 20, or greater than 30 and / or less than 60, preferably less than 50.

[0469] The time interval between two consecutive replacements of the orthodontic appliance (ie between two consecutive intermediate times) may be greater than 7 days, or greater than 15 days and / or less than 60 days, preferably less than 30 days.

[0470] In one embodiment, the treatment plan is not limited to a series of intermediate models for the replacement of the aligner, but includes other intermediate models whose intervening times do not mark the replacement of the aligner. Preferably, the treatment plan is a substantially continuous series of intermediate models. Thus, the treatment plan is similar to a movie that allows viewing the progression from the initial model to the final model.

[0471] Advantageously, the suitability of the active appliance can thus be checked at any time. An update image can be acquired at any update time during the treatment. According to the treatment plan, an active intermediate model corresponding to the update time is then selected, i.e. an active intermediate model representing the dental arch in the expected configuration for the update time according to the treatment plan.

[0472] In one embodiment, an additional intermediate model can be generated from the intermediate model of the treatment plan and added to the treatment plan. In particular, if the update time is between two intermediate times t i and t i+1 If the time is between i and t i+1The additional intermediate model is created from the intermediate model of the present invention (for example by interpolation) as an active intermediate model. In particular, the additional intermediate model can be generated during the treatment.

[0473] As should now be clear, the method according to the invention makes it possible to obtain a very accurate updated model corresponding to the actual structure of the dental arch at the time of the update based on a simple photograph or a simple film, without having to perform a new scan. The method can therefore be implemented without having to make an appointment with an orthodontist.

[0474] Of course, the invention is not limited to the embodiments described above and shown.

[0475] Orthodontic treatment can be therapeutic and / or aesthetic.

[0476] A plurality of update images may be used in step 3) and / or step 5).

[0477] Positioning errors in tooth models can be used to detect drift in orthodontic treatment, i.e. to detect situations where the progression of tooth position does not follow the treatment plan.

[0478] Finally, the patient is not necessarily a human being. In particular, the method according to the invention can be used in another animal.

Claims

1. A non-transitory computer medium storing computer program instructions for executing a method for generating a three-dimensional digital model of a dental arch, called an "updated model", during treatment of a dental arch of a patient using an orthodontic appliance, called an "active appliance", the treatment having been simulated by a treatment plan generated at an initial time (t1) and comprising a plurality of intermediate models (M i ), each intermediate model is a three-dimensional digital model of the dental arch, the intermediate model is cut into tooth models and is determined to represent a corresponding intermediate time (t i ), the method comprising the following steps: 1) Before step 3), the intermediate model or "active intermediate model" is determined according to the update time; 2) searching for one or more representations of teeth that do not comply with the treatment plan in update images called "analysis update images", the update images being acquired at update times during the treatment process, each update image being an appliance image (Ig) representing a removable appliance attached to the dental arch in a use position or an image (Id) representing bare teeth of the dental arch without an appliance; If one or more non-conforming teeth are detected, then 3) identifying one or more tooth models that respectively represent the one or more non-conforming teeth in the active intermediate model; 4) deforming the active intermediate model until an updated model is obtained that matches at least one update image called the “deformed update image”, In the method, at least one non-conforming tooth non-conformity is measured by comparing the updated image with the view of the active intermediate model, and then in step 4), the at least one non-conforming tooth model is moved according to the measurement, In step 2), the non-compliant teeth are identified by determining whether the teeth of the dental arch exist at the position represented by the tooth analysis area and the attribute value of the teeth according to at least one probability, The at least one probability is determined by submitting the analysis update image to at least one deep learning device and is related to a property value of at least one tooth represented in a region at least partially representing the tooth in the analysis update image or in the tooth analysis region, The attribute value is related to the conformity of the tooth represented, The at least one deep learning device is a neural network and is trained using a training database, The training database includes more than 1000 historical records, each historical record includes a historical image and a historical description, each historical image includes one or more regions each representing a tooth or "historical tooth regions", and the associated historical description specifies a tooth attribute value of at least one tooth attribute for each historical tooth region.

2. The non-transitory computer medium of claim 1, wherein: Step 1) precedes step 2).

3. The non-transitory computer medium of claim 1, wherein: In step 2), in order to detect non-compliant teeth, - finding a position, orientation and calibration of a virtual acquisition device that allows it to have a view as close as possible to the active intermediate model of the analysis update image; Then comparing the view with the analysis update image, or comparing an update graph representing the discrimination information of the analysis update image with a reference graph representing the discrimination information in the view; and / or The analysis and updating image is an appliance image, in which the contour of at least one tooth and the contour of the active appliance are determined, and then the contour of the at least one tooth is compared with the contour of the active appliance.

4. The non-transitory computer medium of claim 1, wherein: During said deformation of said active intermediate model in step 4), the only tooth models that move are the ones that do not conform to the teeth.

5. The non-transitory computer medium of claim 1, wherein: In step 4), the movement of the tooth model that does not conform to the teeth is an iterative process, whereby in each iteration, - moving one or more of the tooth models to obtain a dental arch model to be tested, and then - testing the dental arch model to be tested by evaluating the degree of match between the dental arch model to be tested and the deformed updated image, The updated model is the model that provides the highest fit among all the models tested.

6. The non-transitory computer medium of claim 5, wherein: Prior to said iterative process, a position, orientation and calibration of a virtual acquisition device or "constrained virtual acquisition conditions" are found that allow viewing of said active intermediate model so as to generate a view in which a conforming representation of the teeth can be superimposed and registered with said conforming representation of the teeth in said deformed updated image, then, During the iterative process, in each iteration, the degree of matching between the measured dental arch model and the deformed updated image is evaluated by comparing the deformed updated image with the view of the measured dental arch model under the constrained virtual acquisition condition.

7. The non-transitory computer medium of claim 1, wherein: The update image is a photograph or an image extracted from a film and is acquired using a mobile phone.

8. The non-transitory computer medium of claim 1, wherein: Before step 4), the activity intermediate model is processed to improve its accuracy.

9. The non-transitory computer medium of claim 1, wherein: In step 4), the deformation of the active intermediate model comprises movement of the tooth models of the active intermediate model, and wherein the movement continues until the positioning error of the image updated according to the deformation for each tooth model is less than 1 mm.

10. The non-transitory computer medium of claim 9, wherein: The movement continues until the positioning error of the updated image according to the deformation is less than 5 / 10 mm for each tooth model.

11. The non-transitory computer medium of claim 9, wherein: The movement continues until the positioning error of the updated image according to the deformation is less than 3 / 10 mm for each tooth model.

12. The non-transitory computer medium of claim 9, wherein: The movement continues until the positioning error of the updated image according to the deformation is less than 2 / 10 mm for each tooth model.

13. The non-transitory computer medium of claim 9, wherein: The movement continues until the positioning error of the updated image for each tooth model according to the deformation is less than 1 / 10 mm.

14. The non-transitory computer medium of claim 1, wherein: In step 4), the deformed updated image is an image of a naked tooth.

15. The non-transitory computer medium of claim 1, wherein: The active intermediate model is selected so that the median time of the active intermediate model differs from the update time by less than two weeks.

16. The non-transitory computer medium of claim 1, wherein: The intermediate model of the treatment plan represents the dental arch in the intended configuration at an intermediate time marking an appliance change.

17. The non-transitory computer medium of claim 1, wherein: The treatment regimen is a series of essentially continuous intermediate models.

18. The non-transitory computer medium of claim 1, wherein: Prior to step 1), an intermediate model of the dental arch is generated from the intermediate model of the treatment plan and then added to the treatment plan as an intermediate model.

19. The non-transitory computer medium of claim 1, wherein: In step 2), if, when the analysis update image is superimposed and registered on the view of the active intermediate model matched with the analysis update image, the analysis update image has the same scale as the view of the active intermediate model and is in actual scale, and at least one point of the representation of the tooth in the analysis update image is separated from the corresponding point in the view by a distance greater than 1 / 10 mm and less than 7 mm, then the representation does not comply with the treatment plan.

20. The non-transitory computer medium of claim 1, wherein: If in the appliance image a tooth is dislocated from the appliance beyond a threshold, the tooth is considered noncompliant.

21. The non-transitory computer medium of claim 1 further storing computer program instructions for executing, after steps 1) to 4), a method for manufacturing an orthodontic appliance, the method comprising the steps of: 5) designing, based on the updated model and the final model representing the dental arch in the theoretical final structure, an updated appliance suitable for modifying the dental arch from the actual structure at the updating time toward the theoretical final structure; 6) Manufacturing the updated appliance and providing the updated appliance to the patient.

Images