Manufacturing of retainers

Abstract

The present invention relates to a method for manufacturing a retainer, comprising the steps of: a) acquiring, at an updated time, by means of an image acquisition device, at least one two-dimensional image of at least a portion of a user's dental arch, referred to as an "updated image", after completion of orthodontic treatment; b) generating, based on the updated image, a digital three-dimensional model representative of the user's dental arch, referred to as an "updated model", c) designing, preferably by computer, based on the updated model, a retainer suitable for holding the teeth in the updated configuration; and d) manufacturing the retainer.

Description

[Technical field]

[0001] The present invention relates to methods and apparatus for the design and manufacture of orthodontic retainers. [Background technology]

[0002] After orthodontic treatment, the position of the teeth may change. This undesirable change is commonly referred to as "relapse." Traditionally, the patient visits the orthodontist periodically for an examination. The patient may also visit a dentist, who can detect imperfections in the positioning of the teeth.

[0003] However, many patients do not undergo examinations that would allow for the detection of possible relapses, and thus the teeth may revert to a maloccluded position, requiring further orthodontic treatment to correct, which may be as extensive as the initial treatment.

[0004] Visits to the dentist are burdensome for the patient and stressful for the orthodontist. US Patent Application Publication No. 2018 / 0204332 describes a method that partially solves this problem. However, at the end of the orthodontic treatment, it is necessary to obtain at least one digital three-dimensional reference model of the patient's dental arch. Thus, the patient is forced to visit the orthodontist.

[0005] To prevent undesirable changes in tooth position and to limit the number of appointments, patients may be asked to wear retainers to hold the teeth in the positions achieved at the end of orthodontic treatment.

[0006] A retainer is typically an orthodontic retention tray or archwire, consisting of a wire that forms an arch and is fixed to the teeth, usually bonded to the lingual surfaces from canine to canine. The shape of the orthodontic appliance is usually determined by the orthodontist by obtaining a 3D scan of the patient's dental arch after orthodontic treatment has been completed.

[0007] Obtaining a 3D scan of a patient's dental arch is time consuming and requires the patient to visit the clinic.

[0008] There is a need to facilitate the determination and manufacture of orthodontic retainers following orthodontic treatment.

[0009] One object of the present invention is to at least partially meet this need. Summary of the Invention

[0010] Disclosure of the Invention The present invention aims to meet some or all of these needs and, according to one of its aspects, achieves this by a method for manufacturing a retainer comprising the following steps. a) preferably during, at the end or after an update of orthodontic treatment, acquiring, using an image acquisition device, at least one two-dimensional image of at least a portion of the user's dental arch, referred to as an "update image", and preferably several update images are acquired; b) generating a digital three-dimensional model representing the user's dental arch, known as the "updated model", preferably from the updated images; c) designing a retainer based on the updated model, the retainer designed to keep the teeth in their updated configuration when positioned over the teeth; d) manufacturing a retainer.

[0011] Preferably, in step b), the updated model is obtained by deforming a digital three-dimensional model of the user's dental arch, known as the "reference model", obtained before step a), on the basis of the updated image. The reference model is deformed starting from the updated image until an updated model is obtained which represents the dental arch in a configuration that matches the configuration of the dental arch as represented on the updated image.

[0012] As will be seen in more detail in the following description, updated images can be taken outside of the orthodontic practice and reference models may be generated before or during orthodontic treatment. In particular, no intraoral scan is required after orthodontic treatment to manufacture a retainer. In this way, a retainer can be determined and manufactured after orthodontic treatment is completed without the user having to visit a dental care professional or even meet with a dental care professional at all.

[0013] Thus, the method according to the invention can advantageously limit the number of appointments with a dental care professional.

[0014] The method according to the invention may include one or more of the following optional features. the retainer is an archwire or an orthodontic retention tray, preferably an archwire; A plurality of update images are acquired, preferably three, preferably four, preferably five, - a plurality of updated images are acquired at different acquisition angles, the updated images preferably including at least one image to the right of the user, at least one image to the left of the user, and at least one image facing the user; The update image is taken by the user or someone close to the user, - the acquisition in step a) is carried out outside an orthodontic or dental practice, preferably at the user's home, - the image capture device is a mobile phone, - a dental retractor and / or an image acquisition kit, as described for example in EP 3,391,810, is used in step a), the image acquisition kit taking the form of a box in which the image acquisition device is mounted, the housing is in the form of a tube having a first opening designed to be positioned in the mouth of a user and a second opening designed to be attached to an image capture device through the openings to capture an image of at least one of the user's teeth; - in step b), an updated model is obtained by deforming a digital three-dimensional model of the user's dental arch, known as the "reference model", obtained before step a), on the basis of the updated images, -The reference model is generated from a 3D digital model, Preferably, it is generated from a digital three-dimensional model, acquired for example by a 3D scanner, before the start of orthodontic treatment or during orthodontic treatment, the reference model is an updated model resulting from the implementation of the method according to the invention, called the "historical updated model", which is preferably stored in a database after step b), the reference model is a three-dimensional digital model of the user's dental arch generated at the beginning of the orthodontic treatment, in particular for modeling the dental arch at or during a stage of the orthodontic treatment, for example more than two weeks or more than one month before step b), Preferably, the reference model represents the user's dental arch in a desired theoretical configuration at an intermediate point in orthodontic treatment or at the end of orthodontic treatment, Preferably, the reference model represents a dental arch in a final configuration corresponding to a theoretical configuration desired at the end of the orthodontic treatment, the reference model is transformed using metaheuristic methods or neural networks, in step b), preferably before step b), preferably in step a), more preferably before step a), the reference model is cut into a tooth model, the deformation comprising a displacement of the tooth model equal to a difference in tooth positioning between the tooth configuration in the reference model and the tooth configuration in the updated image, in step b), or preferably before step b), preferably in step a), even more preferably before step a), at least one neural network is trained from a learning base to determine an update model at the output of the neural network based on a reference model and an update image provided at the input of the neural network, the training base being composed of a history image and a first history model similar to the update image and the reference model respectively, and a second history model similar to the expected update model at the output of the neural network, - in step b), the updated model is obtained by assembling a selected tooth model, in particular on the basis of the updated image, without using a reference model, In step c), the design of the orthodontic appliance involves determining a curvature of the archwire; Preferably, the archwire is designed to conform to the shape of the teeth to which it is applied; archwires are typically designed to be attached to the inner surfaces of a user's teeth; - steps b) to c) are carried out by a computer, preferably automatically, i.e. without operator intervention or under the supervision of an operator, e.g. a technician or dental care professional, - step d) is carried out using a 3D printer, The method according to the invention further comprises, after step d), the following step e): e) Shipping the retainer to the user or dental professional.

[0015] In one embodiment, the updated model is obtained by implementing a dental arch modeling method, preferably by a user's mobile phone, the modeling method comprising the following steps: 1) analyzing the updated image to determine at least one tooth region and at least one tooth attribute value associated with the tooth region, preferably the tooth type or number to which the tooth region belongs; 2) for each tooth region determined in step 1), searching in a historical library of over 1000 tooth models, called "historical tooth models", associated with tooth attribute values ​​to find a historical tooth model that is closest to the tooth region, called the "best tooth model"; 3) Arranging the set of optimal tooth models to create a model that best matches the updated image, i.e. the updated model.

[0016] The modeling method preferably includes one or more of the following features: steps 1) and 2) of the modelling method are carried out on a number of updated images, preferably in steps 2) and 3) seeking to obtain an optimal tooth model and a model assembled so as to maximally correspond to the set of updated images, the tooth attributes are selected from tooth type, tooth number, tooth thickness, tooth crown height, mesial-palatal width of the tooth, mesial and distal deviation index relative to the incisal edge of the tooth, tooth color, tooth shape, tooth contour, clinical condition of the tooth, e.g. "carious", "broken", relative position of the tooth to another tooth, or relative position of the tooth to a fixed reference frame for the arch, and combinations of these attributes; A plurality of attributes are associated with the tooth region and the historical tooth model; in step 2), the proximity is a shape proximity evaluated by the difference between the shape of the tooth represented on the tooth region and the shape of the historical tooth model, preferably by the shape difference between the tooth region and a view of the historical tooth model having a shape closest to the shape of the tooth region, The view of the historical tooth model having the shape closest to the shape of the tooth region is searched for using a metaheuristic method, preferably an evolutionary method, preferably a simulated annealing method, or a neural network.

[0017] In one embodiment, the reference model that is transformed to obtain the updated model is a model generated by a user, preferably using the user's mobile phone.

[0018] In particular, at a point prior to the update time, e.g., during the orthodontic treatment stage, the user can use the modeling method described above to define a "previous" update model from the "previous" update images and use this previous update model as a reference model.

[0019] To generate an updated model, or a so-called reference model, the user can also scan the dental arch using a portable scanner, for example of the LIDAR type, preferably a mobile phone LIDAR, or by analyzing a series of images representing luminous marks. In particular, the user, preferably with a mobile phone, can carry out a method for generating a digital three-dimensional model of a dental arch comprising the following steps: I) projecting at least one light beam onto a dental arch of a user to place at least one light mark on the dental arch; II) simultaneously with step I), moving the dental arch through the beam and acquiring a series of "marked" images of the dental arch during the movement, each showing a representation of a projected mark or "projection"; III) Identifying projections on each marked image and then generating a digital three-dimensional model that maximally matches the set of projections.

[0020] In one embodiment, the generation method is performed prior to an update time to generate a reference model, and the update image is used to update the reference model.

[0021] In one embodiment, the generation method is performed at update time to generate a reference model, and the update images are used to correct any errors in the reference model.

[0022] Preferably, steps b) to e) are performed by a computer, preferably automatically, ie without operator intervention or under the supervision of an operator, e.g. a technician or dental care professional.

[0023] Preferably, if the retainer is an archwire, the retainer is sent to a dental professional.

[0024] When the retainer is an archwire, the method preferably comprises: - in a design step c), designing an orthodontic transfer tray for receiving the retainer and the adhesive material based on the updated model, such that when the orthodontic transfer tray is positioned over the user's teeth, the retainer is positioned and secured over the user's teeth in a position that keeps the user's teeth in the updated configuration; - in step d) of manufacturing the archwire, manufacturing an orthodontic transfer tray.

[0025] The adhesive can be made of a composite material. Preferably, the adhesive comprises a combination of silane-coated inorganic particles and a dimethacrylic resin, for example of the "ENAMEL plus HRI Flow HF" type. The adhesive may also comprise a composite containing "Bis GMA" and / or "Bis EMA", for example "Braceplaste".

[0026] In step e), the archwire may be delivered inside an orthodontic transfer tray, preferably to a dental care professional or user, preferably to a dental professional.

[0027] The invention further relates to a computer program product comprising instructions which, when said program is executed by a computer, cause the computer to carry out the following steps: a') receiving at least one two-dimensional image representative of the user's dental arch after completion of orthodontic treatment, referred to as an "update image"; b) generating from the updated images a digital three-dimensional model representative of the user's dental arch, known as an "updated model", preferably the generation of the updated model is performed by transforming from the updated images a digital three-dimensional model of the user's dental arch, obtained before step a'), known as a "reference model", c) designing a retainer based on the updated model. The updated image received in step a') is captured by an image capture device.

[0028] Preferably, the computer program product comprises after step c) step d) which comprises transmitting the manufacturing instructions, preferably via digital communication, to an orthodontic appliance manufacturer or directly to a manufacturing device, for example a 3D printer.

[0029] The present invention further relates to the following: a computer program product according to the present invention, - a data carrier on which such a product is recorded, e.g. a memory or a CD-ROM; - The computer on which such product is loaded.

[0030] The medium may be a computer memory, a hard drive or a CD-ROM.

[0031] The present invention further comprises: an image acquisition device, preferably a mobile phone; a computer program product according to the invention, preferably recorded on a medium, loaded into a computer and optionally comprising code for the steps of transmitting manufacturing instructions, The image capture device is in communication with the computer to transmit updated captured images to the computer.

[0032] The computer program product may be a dedicated application installed on an image capture device, which is preferably a mobile phone.

[0033] The kit may further include an apparatus for manufacturing orthodontic appliances, such as a 3D printer configured to receive the retainer manufacturing instructions from the computer executing the computer program product.

[0034] The kit may further include a shipping vehicle designed to transport the retainer from its manufacturing location to a distribution location for the retainer, which is preferably the user's home and / or the user's dental care professional's office.

[0035] definition The term "user" refers to any person on whom the method according to the invention is carried out, whether that person is ill or not.

[0036] The term "dental care professional" refers to any person qualified to provide dental care, including, among others, orthodontists and dentists.

[0037] "Orthodontic treatment" is any treatment, in whole or in part, intended to correct the configuration of a patient's teeth. Treatment intended to maintain the configuration of the teeth (retentive treatment) is not considered orthodontic treatment.

[0038] A 3D scanner or "scanner" is a device that creates a 3D model of the dental arch.

[0039] An "update time" is a point in time at which one or more update images are acquired. The duration of this point in time is short enough that the tooth configuration does not change significantly during this time.

[0040] The term "model" refers to a three-dimensional digital model. A model is composed of a set of voxels. An "arch model" is a model representing at least a portion of a dental arch, preferably at least two, preferably at least three, most preferably at least four teeth. Figure 3 is an example of an arch model.

[0041] A "tooth model" is a three-dimensional digital model of the teeth in a user's dental arch. The arch model can be cut to define tooth models for at least some, and preferably all, of the teeth represented in the arch model. Thus, the tooth model is a model within the arch model. Figure 4 shows an example of an arch model divided into tooth models.

[0042] A "scenario" is a series of arch models representing successive arch configurations. In particular, a "treatment scenario" or "treatment plan" includes models representing the configuration of the arch at different points in treatment. These points are typically an initial point before the start of treatment, intermediate points during treatment, and a final point at the end of treatment. Each scenario model representing the arch in a predicted configuration at an intermediate point is called an "intermediate model." Figure 2 shows an example of a treatment scenario.

[0043] The measure of the difference between two objects is called the “match” or “fit.” The “best fit” is the result of an optimization to minimize the difference described above.

[0044] In particular, the updated image maximally matches the model when a view of the model provides an image that maximally matches the updated image.

[0045] A "tooth region" is a region of an image that at least partially represents a tooth.

[0046] The term "historical tooth model" refers to a richly descriptive tooth model.

[0047] "Proximity" is a measure of one or more differences between the historical tooth model and the tooth region. These differences may include not only differences in shape but also other differences such as translucency or color. The closest proximity can be found by minimizing several differences successively, or by minimizing a combination of these differences, for example a weighted sum of these differences.

[0048] Thus, "proximity" is a broader concept than "match", which only measures proximity with respect to shape.

[0049] The arch configurations at intermediate and final time points are theoretical because they are obtained from simulations for future time points. Therefore, they are expected or "projected" and may differ from the reality at the intermediate or final time points. Viewing the model in a scenario simulates the effect of dental arch treatment over time.

[0050] An arch configuration is said to be "realistic" if it is the configuration of an actual user's arch. An arch configuration is said to be "theoretical" when it is the configuration of a user's arch that is "simulated," "predicted," or "desired" for a future time point.

[0051] An example of software for manipulating tooth models and creating treatment scenarios is the program Treat described at https: / / en.wikipedia.Org / wiki / Clear_aligners#cite_note-invisalignsystem-10. US Patent No. 5,975,893(A) also describes creating treatment scenarios.

[0052] "Image" refers to a two-dimensional image such as a photograph or a video frame. Images are made up of pixels.

[0053] "Acquisition conditions" include the position in space, the orientation in space, and the calibration, e.g. aperture and / or exposure time and / or focal length and / or sensitivity values. a realistic image acquisition device (realistic acquisition conditions) for the user's dental arch, or - Specify a virtual image acquisition device (virtual acquisition conditions) for the user's dental arch.

[0054] The "calibration" of an acquisition device consists of all its calibration parameter values. Calibration parameters are parameters that are specific to the acquisition device (as opposed to position and orientation) and whose values ​​affect the acquired image. For example, aperture is a calibration parameter that modifies the depth of field. Exposure time is a calibration parameter that modifies the brightness (or "exposure") of the image. Focal length is a calibration parameter that modifies the angle of view, or the degree of "zoom". "Sensitivity" is a calibration parameter that modifies the response of the digital acquisition device's sensor to incident light.

[0055] Preferably, the calibration parameters are selected from the group formed by aperture, exposure time, focal length and sensitivity.

[0056] A view of a model at a particular angle and distance under particular virtual acquisition conditions, particularly with a calibration of the virtual acquisition device, is called a "view."

[0057] The terms "dental arch image," "dental arch view," "dental arch representation," "dental arch scan," or "dental arch model" mean an image, view, representation, scan, or model of all or a portion of a dental arch.

[0058] A model of the user's dental arch is "matched" to an image when a view of the model corresponding to the image is present, i.e., such that the tooth representations in the view are positioned relative to one another as are the tooth representations on the image. Thus, the contours of the tooth model depicted on the view can be aligned and substantially superimposed with the contours of the tooth representation on the image.

[0059] This view of the model may also be described as "matched" or "registered and overlayable" with the image.

[0060] The term "light mark" refers to the interaction between a light beam and the dental arch. A light mark can be a dot, a line, a stripe or a set of dots and / or lines and / or stripes. Thus, a line can be continuous, locally interrupted or composed of several lines. A line can be of constant or variable width. The representation of a light mark, especially in the update image, changes depending on the direction from which the mark is viewed.

[0061] The term "light" includes all electromagnetic radiation from infrared to ultraviolet.

[0062] "Metaheuristic" methods are well-known optimization methods. The metaheuristic methods are preferably selected from the group formed by: an evolutionary algorithm, preferably an evolutionary algorithm selected from the following evolutionary strategies: a genetic algorithm, a differential evolution algorithm, a distribution estimation algorithm, an artificial immune system, a shuffled composite evolutionary path relinking, a simulated annealing, an ant colony algorithm, a particle swarm optimization algorithm, a tabu search, and a GRASP method; -Kangaroo algorithm, Fletcher-Powell method, noise method, stochastic tunneling, random restart hill climbing, cross entropy method, - Hybrid methods between the metaheuristic methods mentioned above.

[0063] Deep learning algorithms are well known to those skilled in the art and include "neural networks" or "artificial neural networks."

[0064] Those skilled in the art know how to select a neural network depending on the task at hand. In particular, the neural network can be chosen from: -Networks specialized in image classification, called "CNN" (Convolutional Neural Network), e.g. AlexNet(2012), ZF Net(2013), VGG Net(2014), GoogleNet(2015), Microsoft ResNet(2015), Caffe:BAIR Reference Caff eNet, BAIR AlexNet, Torch VGG CNN S, VGG CNN M, VGG_CNN_M_2048, VGG_CNN_M_1024, VGG CNN M 128, VGG CNN F, VGG ILSVRC-2014 16-layer, VGG ILSVRC-2014 19-layer, Network-in-Network(Imagenet & CIFAR-10), Google:Inception(V3, V4), - Object detection networks, which are networks specialized in searching and detecting objects in images, such as R-CNN (2013), SSD (Single Shot MultiBox Detector: Object Detection network), Faster R-CNN (Faster Region-based Convolutional Network method: Object Detection network), Faster R-CNN (2015), SSD (2015), RCF (Richer Convolutional Features for Edge Detection) (2017), SPP-Net (2014), OverFeat (Sermanet et al.) (2013), GoogleNet (Szegedy et al.) (2015), VGGNet (Simonyan and Zisserman) (2014), R-CNN (Girshick et al.) (2014), Fast R-CNN (Girshick et al.) (2015), ResNet (He et al.) (2016), Faster R-CNN (Ren et al.) (2017), al.) (2016), FPN (Lin et al.) (2016), YOLO (Redmon et al.) (2016), SSD (Liu et al.) (2016), ResNet v2 (He et al.) (2016), R-FCN (Dai et al.) (2016), ResNeXt (Lin et al. al.) (2017), DenseNet (Huang et al.) (2017), DPN (Chen et al.) (2017), YOL09000 (Redmon and Farhadi) (2017), Hourglass (Newell et al.) (2016), MobileNet (Howard et al.) (2017), DCN (Dai et al. al.) (2017), RetinaNet (Lin et al.) (2017), Mask R-CNN (He et al. al.) (2017), RefineDet (Zhang et al.) (2018), Cascade RCNN (Cai et al.) (2018), NASNet (Zoph et al.)(2019), ComerNet(Law and Deng)(2018), FSAF(Zhu et al.)(2019), SENet(Hu et al.)(2018), ExtremeNet(Zhou et al.)(2019), NAS-FPN(Ghiasi et al.)(2019), Detnas(Chen et al.)(2019), FCOS(Tian et al. al.)(2019), CenterNet(Duan et al.)(2019), EfficientNet(Tan and Le)(2019), AlexNet(Krizhevsky et al.)(2012),. -Networks specialized in image generation, e.g., Cycle-Consistent Adversarial Networks(2017), Augmented CycleGAN(2018), Deep Photo Style Transfer(2017), FastPhotoStyle(2018), pix2pix(2017), Style-Based Generator Architecture for GANs(2018), SRGAN(2018).

[0065] The above list is not exhaustive.

[0066] Training a neural network consists of refining it with a training base that contains information about two types of objects that the neural network must learn to "match", i.e., to associate with each other.

[0067] The training can be based on a learning base made up of recordings, each including a first object of a first type and a corresponding second object of a second type.

[0068] Alternatively, training can be performed using a learning base composed of records, each of which includes either a first object of a first type or a second object of a second type, but each record includes information about the type of object it contains. Such training techniques are described, for example, in the paper "Unpaired image-to-image translation using cycle-consistent adversarial networks" by Zhu, Jun-Yan, et al., 2017 IEEE International Conference on Computer Vision (ICCV).

[0069] Training a neural network with these records teaches it to provide, for any object of the first type, a corresponding object of the second type.

[0070] The quality of the analysis performed by the neural network depends directly on the number of records in the training database. Preferably, the learning base contains more than 10,000 records.

[0071] By "computer" is meant a computer processing unit, including a set of several machines with computer processing capabilities. In particular, this unit may be integrated into a mobile phone or may be a PC-type computer or a server, for example a server remote from the user, for example a computer located in the "cloud" or at the dental care professional's premises. The image acquisition device and the computer are then equipped with communication means for exchanging information between them, in particular for sending updated images.

[0072] Typically, the computer comprises a processor, a memory, a human machine interface, typically comprising a screen, and a communication module via the Internet, WIFI, Bluetooth, or a telephone network. Software configured to implement the method of the invention is loaded into the memory of the computer. The computer can also be connected to a printer, such as a 3D printer.

[0073] The terms "comprising," "including," and "having" are to be interpreted broadly and without limitation, unless otherwise specified. [Brief description of the drawings]

[0074] Further features and advantages of the invention will become apparent from the following detailed description and from a consideration of the accompanying drawings, which are provided for illustrative and non-limiting purposes and in which: [Figure 1] 1 illustrates diagrammatically a manufacturing method according to the invention; [Diagram 2] Treatment scenarios are presented. [Diagram 3] 1 shows an example of a dental arch model. [Figure 4] 1 shows an example of an arch model divided into teeth models. [Diagram 5] 1 shows a kit according to the present invention. [Figure 6] 4 shows a variant of the reference model according to the invention; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0075] 1 shows an example of a production method according to the present invention. This production method includes steps a) to d), preferably steps a) to e).

[0076] In step a), an updated image 22, preferably a photograph, representative of the user's dental arch is acquired by the image acquisition device 20 for updating a reference model 32 representing the user's dental arch.

[0077] Alternatively, the updated images may be obtained by video capture, where the user's dental arch is photographed using an image capture device, and the updated images may be extracted from this video.

[0078] Step a) is performed at an update time, the update time being after orthodontic treatment. Preferably, the acquisition is performed less than 6 months, preferably less than 3 months, more preferably less than 1 month, even more preferably immediately after completion of the orthodontic treatment.

[0079] The acquisition in step a) may be performed by the user, a relative or a dental professional, preferably the acquisition is performed by the user himself.

[0080] The image capture device 20 may be a mobile phone, a tablet, a camera, or a computer, preferably the image capture device is a mobile phone.

[0081] Preferably, multiple updated images 22 are acquired. The updated images may include a front view image, a left view image, and a right view image. The updated images may be taken of the upper dental arch, the lower dental arch, or both. The updated images may be taken with the user's mouth open or closed. Preferably, the updated images are in color, preferably in realistic color.

[0082] In step b), an updated model 42 is generated based on the updated images acquired in step a), the model being generated without the need for a user to be physically present at the oral surgery clinic, preferably when the user is not physically present at the oral surgery clinic.

[0083] Several methods are available.

[0084] In one embodiment, the updated model is generated directly from the updated images, preferably using at least one neural network.

[0085] At least one neural network is pre-trained such that the neural network learns to output an updated model from one or more updated input images.

[0086] The training is performed using a training base including a historical image and a historical model, where the historical image can be compared to the updated image and the historical model can be compared to the updated model. The historical model can be acquired by a 3D scanner when a substantially corresponding historical image is acquired or can be generated manually. The records in the training database can include the historical image as a first object and the historical model as a second object. The neural network can be told that for a first object including the historical image, it must match a second object including the historical model. In particular, to train the neural network, a set of first objects and a set of second objects can be provided and it can learn to match the first object with the second object. Once trained, the neural network can determine the updated model as an output when the updated image is given as an input.

[0087] Preferably, the first object comprises a plurality of updated images, and the at least one neural network is trained to generate as output an updated model from the plurality of updated images provided as input.

[0088] At least one of the neural networks may be a reconfiguration neural network, i.e., an occupational neural network.

[0089] The occupancy neural network is trained to output a 3D model of at least a portion of the dental arch constructed from a plurality of updated images representative of at least a portion of the dental arch. Preferably, the occupancy neural network takes as input only the updated images representative of at least a portion of the user's dental arch.

[0090] In another embodiment, the reference model 32 is determined. The reference model can be selected manually, for example from a set of 3D models stored in the database 30, which represent the user's dental arch. Preferably, the reference model is determined automatically, i.e. without human intervention, preferably by a computer.

[0091] The selected reference model 32 is preferably an intermediate or final model created when determining an orthodontic treatment scenario. An orthodontic treatment scenario is illustrated in Fig. 2. The reference model 32 preferably corresponds to a 3D model acquired by a 3D scanner before the start of the orthodontic treatment (also known as "initial model"), which has been modified to represent a theoretical configuration of the teeth at an intermediate point in the orthodontic treatment (also known as "intermediate model") or at the end of the orthodontic treatment (also known as "final model"). Preferably, the reference model corresponds to the final model.

[0092] The reference model can be selected from a database. Alternatively, the reference model can be transmitted by digital communication, for example by a dental care professional or technician, preferably via a dedicated application.

[0093] The reference model 32 is then deformed to obtain an updated model 42 that represents the user's dental configuration at the time of the update.

[0094] 6 shows the deformation of a reference model using a transformation algorithm, which receives as input a reference model 32 and an update image 22, and then deforms the reference model according to the update image to determine an updated model 42 that fits with the update image.

[0095] The transformation algorithm may be a metaheuristic method, preferably including the following steps: - searching for a position, orientation and calibration of a virtual acquisition device, collectively referred to as "virtual acquisition conditions", that best matches the actual acquisition conditions of the updated image; - comparing a view of the reference model under virtual acquisition conditions with the updated image; - Deforming the reference model such that the positioning of the teeth observed on the views under virtual acquisition conditions corresponds to the positioning of the teeth observed on the acquired images.

[0096] Therefore, deformations are performed to update the reference model at the time an acquired image is acquired.

[0097] Preferably, the reference model is pre-cut into a tooth model, as shown in Figure 4, and the transformation includes moving the tooth model. In one embodiment, the transformation involves moving the tooth model whose shape remains constant.

[0098] In particular, the search and comparison of virtual acquisition conditions can be carried out according to the teachings of International Application PCT / EP2015 / 074896.

[0099] Alternatively, the transformation algorithm includes at least one neural network.

[0100] At least one neural network is pre-trained such that the neural network learns to output an updated model based on an updated image and a reference model, both of which are provided as inputs.

[0101] The training is performed using a training base including a historical image, a first historical model, and a second historical model, where the historical image is similar to the updated image, the first historical model is similar to the reference model, and the second historical model is similar to the updated model. Preferably, each second historical model is determined from the historical image and the corresponding first historical model, for example manually, or is acquired by a 3D scanner at the time the substantially corresponding historical image is acquired. The records in the learning database may include the historical image and the first historical model as a first object and the second historical model as a second object. The neural network may be told that for a first object including the historical image and the first historical model, it must be matched with a second object including the second historical model. In particular, to train the neural network, a set of first objects and a set of second objects may be provided and it may learn to match the first object with the second object. Once trained, the neural network may determine the updated model as an output when given the updated image and the reference model as input.

[0102] In particular, if some updated images are available, it is possible to use a PPO type reinforcement network, as described on the website: https: / / openai.com / blog / openai-baselines-ppo / .

[0103] The PPO neural network can be configured to iteratively modify tooth positions of a 3D model representing at least a portion of the user's dental arch based on at least two updated images provided as input to the neural network, the 3D model, when modified, maximally matching the image provided as input, the at least two updated images being images representing at least a portion of the user's dental arch acquired at different acquisition angles. Providing multiple updated images acquired at different acquisition angles allows for the calculation of tooth movements in three dimensions using principles of triangulation.

[0104] Advantageously, such a neural network does not require any data other than the updated image as input to modify the 3D model available to the neural network.

[0105] To train the neural network, it is given the expected final positions of the teeth as output. The neural network can then learn to correct the results it outputs (the "output" 3D model) by comparing the positions of the teeth in this "output" model with the expected final positions of the teeth. The deviation between the expected tooth positions and the tooth positions in the output model provided by the neural network may be calculated, for example, in millimeters or degrees, or may be a surface tolerance or Euclidean LL distance.

[0106] Compared to the PPO neural network, the occupancy neural network may be less accurate, especially when at least some specific parts of the dental arch are not represented on the multiple updated images provided as input to the neural network. Advantageously, a previously acquired 3D model of the user's dental arch can be used to reconstruct the missing parts. The previously acquired 3D model of the user's dental arch can also provide dimensional information, for example in millimeters, that cannot be estimated from the multiple updated images provided as input alone.

[0107] In another embodiment of step b), the updated model is generated based on the update images, preferably from a set of several update images and from a library of tooth models based on a modeling method, the method comprising, for each update image, the following steps: 1) analyzing the updated image to determine at least one tooth region and at least one tooth attribute value associated with the tooth region, preferably the tooth type or number to which the tooth region belongs; 2) for each tooth region determined in step 1), searching in a historical library containing over 1000 tooth models, called "historical tooth models", associated with tooth attribute values ​​to find a historical tooth model, called the "best tooth model", that is closest to the tooth region; 3) Arranging the set of optimal tooth models to create a model that maximally matches the updated image.

[0108] A library of tooth models is preferably created prior to step a) of the manufacturing method according to the invention. Preferably, the library of tooth models comprises more than 2,000, preferably more than 5,000, most preferably more than 10,000 historical tooth models. The higher the number of historical tooth models, the more accurate the assembled model will be.

[0109] The tooth attributes are used to filter the historical tooth model to limit the search to teeth that have the same tooth attribute values ​​as the tooth represented by the tooth region in the update image.

[0110] In particular, one or more tooth attributes may be associated with the tooth model selected from the number of teeth, tooth type, tooth shape parameters such as tooth width, in particular mesial-palatal width, thickness, crown height, mesial and distal deflection index relative to the incisal edge, or wear level, tooth appearance parameters, in particular translucency index or color parameters, parameters relating to the condition of the tooth such as "worn", "broken", "corroded" or "fit" (i.e. in contact with an orthodontic appliance), the age of the user, or a combination of these attributes.

[0111] Preferably, the modelling method comprises the following optional steps: 4) Optionally, replacing at least one optimal tooth model with another tooth model, a historical tooth model, and repeating step 3) to maximize the match between the assembled model and the analysis image. 5) Optionally, repeating step 1) with another update image, searching for the maximum match with the set of update images used in step 3) and / or step 4).

[0112] The accuracy of the constructed model can be improved if several update images are processed.

[0113] In step 1), the updated image is preferably analyzed by subjecting it to a neural network trained to determine at least one tooth region and at least one tooth attribute value associated with the tooth region.

[0114] At the end of step 1), an updated image is obtained and for each tooth region is enriched with a description providing at least one tooth attribute, e.g. a tooth attribute value for a tooth number.

[0115] In step 2), for each tooth region determined in the previous step, the historical library is searched to find the historical tooth model that is closest to the tooth region. This tooth model is called the "best tooth model."

[0116] Preferably, historical tooth models are determined that have the same value for at least one tooth attribute as the tooth region. In particular, the tooth attribute may relate to a tooth type or a tooth number. In other words, the historical tooth models are filtered to examine in more detail only those that relate to the same tooth type as the tooth represented on the tooth region.

[0117] Alternatively, or preferably, in addition to comparing the attribute values, the tooth shape represented on the tooth region may be compared to the shape of a historical tooth model, which is evaluated, preferably by a metaheuristic, preferably evolutionary, preferably simulated annealing method.

[0118] For this purpose, the historical tooth model to be evaluated is viewed from different angles. Each view thus obtained is compared with the updated image, preferably the tooth region, to establish a "distance" between the view and the updated image or preferably the tooth region. The distance thus measures the difference between the view and the tooth region.

[0119] For each tested historical tooth model, the view that provides the minimum distance to the updated image or tooth region is determined. Thus, each examined historical tooth model is associated with a particular minimum distance, which measures the formal proximity to the tooth region.

[0120] The best historical tooth model is the one that is deemed to be closest to the tooth region based on the comparisons performed.

[0121] The minimum distances obtained for the different tested tooth models are then compared and the tooth model with the smallest minimum distance is used to define the optimal tooth model, which is thus the closest match to the updated image.

[0122] The search for the maximum match is preferably carried out using a metaheuristic method, preferably an evolutionary method, preferably a simulated annealing method.

[0123] In a preferred embodiment, a first evaluation of the historical tooth models is performed successively by comparing the value of at least one tooth attribute, e.g. tooth number, with the corresponding value of the tooth analysis area, followed by a second evaluation by shape comparison. The first quick evaluation advantageously allows filtering of the historical tooth models, so that only those historical tooth models retained by the first evaluation can be subjected to a second, slower evaluation.

[0124] Preferably, several first evaluations are performed before the second evaluations.

[0125] At the end of step 2), an optimal tooth model is associated to each tooth region.

[0126] In step 3), an assembled model is created by positioning the optimal tooth models.

[0127] A first rough placement can be established by considering the tooth attribute values ​​of the best fit tooth models, for example, if the tooth numbers of the best fit tooth models are the tooth numbers of the canines and incisors, then these tooth models can be placed in the arch that conventionally corresponds to the regions of the arch that carry these types of teeth.

[0128] This arch shape can be refined according to other tooth attribute values.

[0129] The order of the optimal tooth models is the order of the corresponding tooth regions.

[0130] Furthermore, the closest proximity or "minimum distance" associated with the optimal tooth model results from viewing the tooth model along the observation point. In other words, the teeth modeled by this model are likely to be viewed in the updated image as well. Therefore, preferably, the optimal tooth model is oriented such that the surfaces corresponding to each tooth region on the updated image are viewed from the same observation point.

[0131] In this way, an optimal initial configuration of the tooth model can be defined.

[0132] Preferably, the first configuration of the optimal tooth model is then iteratively modified to provide the maximum match with the updated image.

[0133] In another embodiment of step b), the updated model (or the reference model that is intended to be transformed to generate the updated model) is generated by implementing a method for generating a digital three-dimensional model of a dental arch comprising the following steps: I) projecting at least one light beam onto a dental arch of a user to place at least one light mark on the dental arch; II) simultaneously with step I), moving the dental arch through the beam and acquiring a series of "marked" images of the dental arch during the movement, each showing a representation of the projected marks or an "updated projection"; III) Identifying the projections on each marked image and then generating a digital three-dimensional model that maximally matches the updated set of projections.

[0134] Preferably, in step I) a structured light beam is projected.

[0135] In step III), an updated model can be searched for using optimization methods and / or one or more neural networks.

[0136] In step III), it is possible to modify the defined model according to user characteristics or a model from a database, for example the initial model, the intermediate model or the final model, in order to generate an updated model.

[0137] In step III), the updated model may be determined as the test model obtained at the end of the following cycle of steps i) to iii): i) creating a model to be tested; and then ii) determining a metric that represents the difference between the updated set of projections and the model to be tested; and then iii) if the representative distance exceeds a predefined tolerance threshold, modifying the model being tested and returning to step i).

[0138] The representative distances can be evaluated from basic distances, each basic distance being determined by evaluating, for each updated projection, the difference between the updated projection and an optimal reference projection, the reference projection being a representation on a reference image representing a view of the model to be tested of a virtual light mark resulting from the projection, onto the model to be tested, of a virtual light beam of the same shape as the light beam projected onto the arch in step I), and the optimal reference image being a reference image showing a reference projection that has the smallest distance from the updated projection.

[0139] The model to be tested may be segmented to define a tooth model, and modifying the model to be tested may include moving and / or deforming the tooth model.

[0140] The first test model to be modified may be a dental arch model, preferably selected according to user characteristics, or a model from a database, for example an initial model, an intermediate model, or a final model.

[0141] In step III) at least one, preferably two of the following methods are used: optimization methods, artificial intelligence methods, size evaluation methods involving stereoscopic vision, size evaluation methods involving analyzing the shape of the light mark, and size evaluation methods involving distance analysis.

[0142] In step c), the retainer 44 is designed based on the updated model 42. The design can be done manually, for example, by an engineer or dental care professional.

[0143] Alternatively, the design is performed by software, preferably automatically, i.e., without human intervention. The retainer is designed to hold the user's teeth in their updated configuration.

[0144] To ensure this, the shape of the orthodontic appliances is traditionally adapted to the shape of the patient's dental arch when updated, and the orthodontic appliances do not exert any effect that tends to move the teeth when updated.

[0145] Preferably, the retainer is an archwire that is designed to be fixed, e.g., bonded, to the user's teeth. Alternatively, the retainer is an orthodontic retention tray.

[0146] The design of the retainer involves determining the shape of the appliance to hold the teeth in their updated configuration.

[0147] In step d), the retainer 52 is manufactured.

[0148] In one embodiment, the manufacturing occurs from an impression made on a part that physically represents the renewal model. In such a case, steps c) and d) are combined. For example, an archwire can be created by pressing a metal wire onto a plaster or polymer cast that physically represents the renewal model. A polymer tray can also be molded onto such a part.

[0149] The retainer can also be made directly based on the updated model, for example by designing a 3D model of the retainer and then manufacturing it, for example by printing it with a 3D printer.

[0150] Preferably, when the retainer is an archwire, a transfer tray is also manufactured, which allows the retainer to be easily positioned on the user's teeth. First, the archwire is positioned in the orthodontic transfer tray with the aid of an adhesive substance. In a second step, the orthodontic transfer tray is placed in a service position on the user's teeth. This facilitates the transport and the positioning and fixing of the archwire. In particular, the arch can be positioned and fixed in a single step. The orthodontic transfer tray is based on a three-dimensional model of the user's dental arch, so that the archwire can be accurately positioned. An example of a method for manufacturing an orthodontic transfer tray is described in the article "Methode simple et rapide de contention indirecte" by Christine Muller et al., Orthod Fr 2009;80:233-238, EDP Sciences, SFODF, 2009 DOI:10.105 l / orthodfr / 2009006.

[0151] In optional step e, the manufactured retainer is preferably sent to a dental care professional or to a user. Instructions on how to wear and / or position the retainer can be provided with the retainer. Advantageously, the user does not need to visit a dental care professional to obtain the 3D model using a 3D scanner or to receive the retainer and / or to position it in their mouth.

[0152] Preferably, steps b) and c) are performed by a computer, preferably without human intervention. Steps d) and / or e) may also be performed without human intervention. After step c), the computer performing step c) may communicate design instructions to manufacturing equipment 50 for manufacturing the retainer.

[0153] FIG. 5 shows diagrammatically a design kit 1 according to the invention.

[0154] The design kit 1 comprises an image acquisition device 20, preferably a mobile phone, and a computing system 40, for example a computer, that includes a computer program product configured to perform step b) of the manufacturing method according to the invention, or better to perform steps b) to c) of the manufacturing method according to the invention. The computer program product may also be configured to transmit the design instructions 64 directly to the manufacturing device 50.

[0155] The image capture device 20 can be in communication with the computing system to send updated captured images 60 to the computing system 40 via digital communication. In a particular embodiment, the capture device and computing system are integrated into a mobile phone.

[0156] The reference model may be obtained from a database 30 accessible to the computing system via digital communication. The reference model may be transmitted to the computing system, in particular by being downloaded (62). The database may be on a remote server.

[0157] The design kit 1 may include a manufacturing device, which is preferably configured to communicate with the computing system 40. Once the retainer is designed, the computing system 40 can send manufacturing instructions 64 to the manufacturing device 50. EXAMPLES

[0158] The updated image 22 is preferably obtained remotely by a user, preferably from a dental or orthodontic practice, using a mobile phone 20 .

[0159] The updated image 22 is sent to a computing system 40. The computing system may or may not be integrated into the mobile phone. Preferably, the computing system is a dedicated application that can be installed on the mobile phone.

[0160] The computing system 40 can preferably communicate with a database 30 that includes intermediate, initial and final models from one or more orthodontic treatment scenarios that represent configurations of a user's dental arch. The intermediate and final models represent theoretical configurations of the user's dental arch. The initial models are traditionally models acquired by a dental professional, preferably by means of a 3D scanner. In step b), the reference model is determined by selecting a model from the database 30.

[0161] Alternatively, in step b), the reference model 32 is manually selected and transmitted to the computing system 40, for example by a dental care professional or technician.

[0162] On the basis of the updated image 22 and the reference model 32, the computing system 40 is able to determine an updated model 42, as described in step c) of the method according to the invention.

[0163] The computing system 40 can then use the updated model 42 to design the retainer 52, optionally with the assistance of an operator or dental care professional, as described in step d) of the method according to the invention. Determining the updated model (step c)) and designing the retainer (step d)) can be performed remotely from each other. The computing system may include multiple processors.

[0164] Preferably, the mobile phone comprises a dedicated application adapted to carry out steps a) to c), preferably steps a) to d), according to the invention.

[0165] Step d) of the retainer design makes it possible to determine the manufacturing instructions for the retainer 52. The retainer 52 may be an archwire or an orthodontic retention tray. Advantageously, these instructions are transmitted to the manufacturing device 50.

[0166] A manufacturing device manufactures the retainer from the instructions. The manufacturing device may be a 3D printer. The 3D printer may be located at the user's site.

[0167] Once manufactured, the retainer can be sent directly to the user. Preferably, if the retainer is an archwire, it is shipped in a transfer tray that contains, in addition to the archwire, an adhesive for securing the archwire to the user's teeth when the transfer tray is positioned over the user's teeth.

[0168] As has been seen by now, the method according to the invention allows the manufacture of a retainer on the basis of one or more updated images, preferably photographs taken by the user, and a reference model obtained when the teeth have not yet reached their final configuration, and thus this manufacturing method can be carried out without the intervention of an orthodontist specifically for this purpose, obviating the need for the user to make an appointment with the orthodontist.

[0169] Thus, by acquiring simple extraoral images of the user's teeth, i.e., images acquired by an image acquisition device located outside the user's mouth at the time of acquisition, it is possible for the user to design and manufacture a retainer without the need for the user to physically consult with their dental care professional. In particular, the method according to the invention makes it possible to manufacture an archwire intended to be positioned on the inner surfaces of the user's teeth from the acquisition of extraoral images specifically representative of the outer surfaces of the user's teeth.

[0170] Naturally, the invention is not limited to the embodiments described and illustrated above.

Claims

1. 1. A method of modeling a patient retainer, comprising: - before step a), generating a reference model representing the user's dental arch during orthodontic treatment; a) during update, at or after completion of orthodontic treatment, acquiring by an image acquisition device at least one two-dimensional image of at least a portion of the user's dental arch, referred to as an "update image"; b) generating a digital three-dimensional model representing the user's dental arch, known as an "updated model", based on the updated images, said updated model being obtained by transforming the reference model from the updated images; c) designing, preferably by computer, a retainer suitable for holding the teeth in said updated configuration based on said updated model; d) manufacturing the retainer.

2. The method of claim 1 , wherein the obtaining of the updated image is performed by the user, preferably outside an orthodontic or dental practice.

3. The method according to claim 1 or 2, wherein the retainer is an archwire or an orthodontic retention tray, preferably an archwire.

4. 4. The method of claim 1, wherein the reference model represents the user's dental arch in a desired theoretical configuration at an intermediate point in the orthodontic treatment or at the end of the orthodontic treatment, referred to as an "intermediate model" or a "final model", respectively.

5. The method according to claim 1 , wherein the reference model is a final model.

6. In step b), The reference model is obtained by implementing a dental arch modeling method using a LIDAR integrated in the user's mobile phone or preferably by the user's mobile phone, the modeling method comprising: 1) analyzing an updated image representative of the dental arch of the user to determine at least one tooth region, at least one tooth attribute value associated with said tooth region, and preferably a tooth type or tooth number to which said tooth region belongs; 2) for each tooth region determined in step 1), searching in a historical library of over 1000 tooth models, called "historical tooth models", associated with said tooth attribute values ​​to find a historical tooth model that is closest to said tooth region, called the "best tooth model"; 6. The method according to claim 1, further comprising: 3) arranging the set of optimal tooth models to create a model that maximally matches the images analyzed in step 1), the images being the update images to obtain the updated model or images obtained before the update time to obtain the reference model.

7. In step b), said reference model is obtained by implementing a method for generating a digital three-dimensional model of a dental arch, I) projecting at least one light beam onto the user's dental arch to place at least one light mark on the arch; II) simultaneously with step I), moving the dental arch through the beam and acquiring a series of "marked" images of the arch during said movement, each showing a representation of the projected mark or "projection"; III) identifying said projections on each marked image and then generating a digital three-dimensional model that maximally matches said set of projections.

8. After step d), The method according to any one of claims 1 to 7, comprising the step of: e) shipping said retainer, preferably to said user or to a dental care professional.

9. The method of claim 1 , wherein the acquisition device is a mobile phone and / or the retainer is manufactured using a 3D printer.

Images